sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*
   4  * Copyright 2001 Wasabi Systems, Inc.
   5  * All rights reserved.
   6  *
   7  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. All advertising materials mentioning features or use of this software
  18  *    must display the following acknowledgement:
  19  *      This product includes software developed for the NetBSD Project by
  20  *      Wasabi Systems, Inc.
  21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  22  *    or promote products derived from this software without specific prior
  23  *    written permission.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35  * POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 /*
  39  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  40  * All rights reserved.
  41  *
  42  * Redistribution and use in source and binary forms, with or without
  43  * modification, are permitted provided that the following conditions
  44  * are met:
  45  * 1. Redistributions of source code must retain the above copyright
  46  *    notice, this list of conditions and the following disclaimer.
  47  * 2. Redistributions in binary form must reproduce the above copyright
  48  *    notice, this list of conditions and the following disclaimer in the
  49  *    documentation and/or other materials provided with the distribution.
  50  *
  51  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  52  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  53  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  54  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  55  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  56  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  57  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  59  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  60  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  61  * POSSIBILITY OF SUCH DAMAGE.
  62  *
  63  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  64  */
  65
  66 /*
  67  * Network interface bridge support.
  68  *
  69  * TODO:
  70  *
  71  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  72  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  73  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
  74  *        consider heterogenous bridges).
  75  */
  76
  77 #include <sys/cdefs.h>
  78 __FBSDID("$FreeBSD$");
  79
  80 #include "opt_inet.h"
  81 #include "opt_inet6.h"
  82 #include "opt_carp.h"
  83
  84 #include <sys/param.h>
  85 #include <sys/mbuf.h>
  86 #include <sys/malloc.h>
  87 #include <sys/protosw.h>
  88 #include <sys/systm.h>
  89 #include <sys/time.h>
  90 #include <sys/socket.h> /* for net/if.h */
  91 #include <sys/sockio.h>
  92 #include <sys/ctype.h>  /* string functions */
  93 #include <sys/kernel.h>
  94 #include <sys/random.h>
  95 #include <sys/syslog.h>
  96 #include <sys/sysctl.h>
  97 #include <vm/uma.h>
  98 #include <sys/module.h>
  99 #include <sys/priv.h>
 100 #include <sys/proc.h>
 101 #include <sys/lock.h>
 102 #include <sys/mutex.h>
 103
 104 #include <net/bpf.h>
 105 #include <net/if.h>
 106 #include <net/if_clone.h>
 107 #include <net/if_dl.h>
 108 #include <net/if_types.h>
 109 #include <net/if_var.h>
 110 #include <net/pfil.h>
 111
 112 #include <netinet/in.h> /* for struct arpcom */
 113 #include <netinet/in_systm.h>
 114 #include <netinet/in_var.h>
 115 #include <netinet/ip.h>
 116 #include <netinet/ip_var.h>
 117 #ifdef INET6
 118 #include <netinet/ip6.h>
 119 #include <netinet6/ip6_var.h>
 120 #endif
 121 #ifdef DEV_CARP
 122 #include <netinet/ip_carp.h>
 123 #endif
 124 #include <machine/in_cksum.h>
 125 #include <netinet/if_ether.h> /* for struct arpcom */
 126 #include <net/bridgestp.h>
 127 #include <net/if_bridgevar.h>
 128 #include <net/if_llc.h>
 129 #include <net/if_vlan_var.h>
 130
 131 #include <net/route.h>
 132 #include <netinet/ip_fw.h>
 133 #include <netinet/ip_dummynet.h>
 134
 135 /*
 136  * Size of the route hash table.  Must be a power of two.
 137  */
 138 #ifndef BRIDGE_RTHASH_SIZE
 139 #define BRIDGE_RTHASH_SIZE              1024
 140 #endif
 141
 142 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 143
 144 /*
 145  * Maximum number of addresses to cache.
 146  */
 147 #ifndef BRIDGE_RTABLE_MAX
 148 #define BRIDGE_RTABLE_MAX               100
 149 #endif
 150
 151 /*
 152  * Timeout (in seconds) for entries learned dynamically.
 153  */
 154 #ifndef BRIDGE_RTABLE_TIMEOUT
 155 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 156 #endif
 157
 158 /*
 159  * Number of seconds between walks of the route list.
 160  */
 161 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 162 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 163 #endif
 164
 165 /*
 166  * List of capabilities to possibly mask on the member interface.
 167  */
 168 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 169 /*
 170  * List of capabilities to disable on the member interface.
 171  */
 172 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 173
 174 /*
 175  * Bridge interface list entry.
 176  */
 177 struct bridge_iflist {
 178         LIST_ENTRY(bridge_iflist) bif_next;
 179         struct ifnet            *bif_ifp;       /* member if */
 180         struct bstp_port        bif_stp;        /* STP state */
 181         uint32_t                bif_flags;      /* member if flags */
 182         int                     bif_savedcaps;  /* saved capabilities */
 183         uint32_t                bif_addrmax;    /* max # of addresses */
 184         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 185         uint32_t                bif_addrexceeded;/* # of address violations */
 186 };
 187
 188 /*
 189  * Bridge route node.
 190  */
 191 struct bridge_rtnode {
 192         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 193         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 194         struct bridge_iflist    *brt_dst;       /* destination if */
 195         unsigned long           brt_expire;     /* expiration time */
 196         uint8_t                 brt_flags;      /* address flags */
 197         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 198         uint16_t                brt_vlan;       /* vlan id */
 199 };
 200 #define brt_ifp                 brt_dst->bif_ifp
 201
 202 /*
 203  * Software state for each bridge.
 204  */
 205 struct bridge_softc {
 206         struct ifnet            *sc_ifp;        /* make this an interface */
 207         LIST_ENTRY(bridge_softc) sc_list;
 208         struct mtx              sc_mtx;
 209         struct cv               sc_cv;
 210         uint32_t                sc_brtmax;      /* max # of addresses */
 211         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 212         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 213         struct callout          sc_brcallout;   /* bridge callout */
 214         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 215         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 216         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 217         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 218         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 219         uint32_t                sc_rthash_key;  /* key for hash */
 220         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 221         struct bstp_state       sc_stp;         /* STP state */
 222         uint32_t                sc_brtexceeded; /* # of cache drops */
 223 };
 224
 225 static struct mtx       bridge_list_mtx;
 226 eventhandler_tag        bridge_detach_cookie = NULL;
 227
 228 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 229
 230 uma_zone_t bridge_rtnode_zone;
 231
 232 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 233 static void     bridge_clone_destroy(struct ifnet *);
 234
 235 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 236 static void     bridge_mutecaps(struct bridge_softc *);
 237 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 238                     int);
 239 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 240 static void     bridge_init(void *);
 241 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 242 static void     bridge_stop(struct ifnet *, int);
 243 static void     bridge_start(struct ifnet *);
 244 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 245 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 246                     struct rtentry *);
 247 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 248                     struct mbuf *);
 249 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 250
 251 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 252                     struct mbuf *m);
 253
 254 static void     bridge_timer(void *);
 255
 256 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 257                     struct mbuf *, int);
 258 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 259
 260 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 261                     uint16_t, struct bridge_iflist *, int, uint8_t);
 262 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 263                     uint16_t);
 264 static void     bridge_rttrim(struct bridge_softc *);
 265 static void     bridge_rtage(struct bridge_softc *);
 266 static void     bridge_rtflush(struct bridge_softc *, int);
 267 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 268                     uint16_t);
 269
 270 static int      bridge_rtable_init(struct bridge_softc *);
 271 static void     bridge_rtable_fini(struct bridge_softc *);
 272
 273 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 274 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 275                     const uint8_t *, uint16_t);
 276 static int      bridge_rtnode_insert(struct bridge_softc *,
 277                     struct bridge_rtnode *);
 278 static void     bridge_rtnode_destroy(struct bridge_softc *,
 279                     struct bridge_rtnode *);
 280 static void     bridge_rtable_expire(struct ifnet *, int);
 281 static void     bridge_state_change(struct ifnet *, int);
 282
 283 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 284                     const char *name);
 285 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 286                     struct ifnet *ifp);
 287 static void     bridge_delete_member(struct bridge_softc *,
 288                     struct bridge_iflist *, int);
 289 static void     bridge_delete_span(struct bridge_softc *,
 290                     struct bridge_iflist *);
 291
 292 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 293 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 294 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 295 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 296 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 297 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 298 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 323 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 324                     int);
 325 static int      bridge_ip_checkbasic(struct mbuf **mp);
 326 #ifdef INET6
 327 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 328 #endif /* INET6 */
 329 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 330                     struct ether_header *, int, struct llc *);
 331
 332 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 333 #define VLANTAGOF(_m)   \
 334     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 335
 336 static struct bstp_cb_ops bridge_ops = {
 337         .bcb_state = bridge_state_change,
 338         .bcb_rtage = bridge_rtable_expire
 339 };
 340
 341 SYSCTL_DECL(_net_link);
 342 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 343
 344 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 345 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 346 static int pfil_member = 1; /* run pfil hooks on the member interface */
 347 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 348 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 349 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 350                                    locally destined packets */
 351 static int log_stp   = 0;   /* log STP state changes */
 352 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 353     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 354 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 355     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 356 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 357     &pfil_bridge, 0, "Packet filter on the bridge interface");
 358 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 359     &pfil_member, 0, "Packet filter on the member interface");
 360 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 361     &pfil_local_phys, 0,
 362     "Packet filter on the physical interface for locally destined packets");
 363 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 364     &log_stp, 0, "Log STP state changes");
 365
 366 struct bridge_control {
 367         int     (*bc_func)(struct bridge_softc *, void *);
 368         int     bc_argsize;
 369         int     bc_flags;
 370 };
 371
 372 #define BC_F_COPYIN             0x01    /* copy arguments in */
 373 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 374 #define BC_F_SUSER              0x04    /* do super-user check */
 375
 376 const struct bridge_control bridge_control_table[] = {
 377         { bridge_ioctl_add,             sizeof(struct ifbreq),
 378           BC_F_COPYIN|BC_F_SUSER },
 379         { bridge_ioctl_del,             sizeof(struct ifbreq),
 380           BC_F_COPYIN|BC_F_SUSER },
 381
 382         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 383           BC_F_COPYIN|BC_F_COPYOUT },
 384         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 385           BC_F_COPYIN|BC_F_SUSER },
 386
 387         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 388           BC_F_COPYIN|BC_F_SUSER },
 389         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 390           BC_F_COPYOUT },
 391
 392         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 393           BC_F_COPYIN|BC_F_COPYOUT },
 394         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 395           BC_F_COPYIN|BC_F_COPYOUT },
 396
 397         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 398           BC_F_COPYIN|BC_F_SUSER },
 399
 400         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 401           BC_F_COPYIN|BC_F_SUSER },
 402         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 403           BC_F_COPYOUT },
 404
 405         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 406           BC_F_COPYIN|BC_F_SUSER },
 407
 408         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 409           BC_F_COPYIN|BC_F_SUSER },
 410
 411         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 412           BC_F_COPYOUT },
 413         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 414           BC_F_COPYIN|BC_F_SUSER },
 415
 416         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 417           BC_F_COPYOUT },
 418         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 419           BC_F_COPYIN|BC_F_SUSER },
 420
 421         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 422           BC_F_COPYOUT },
 423         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 424           BC_F_COPYIN|BC_F_SUSER },
 425
 426         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 427           BC_F_COPYOUT },
 428         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 429           BC_F_COPYIN|BC_F_SUSER },
 430
 431         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 432           BC_F_COPYIN|BC_F_SUSER },
 433
 434         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 435           BC_F_COPYIN|BC_F_SUSER },
 436
 437         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 438           BC_F_COPYIN|BC_F_SUSER },
 439         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 440           BC_F_COPYIN|BC_F_SUSER },
 441
 442         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 443           BC_F_COPYOUT },
 444
 445         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 446           BC_F_COPYOUT },
 447
 448         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 449           BC_F_COPYIN|BC_F_COPYOUT },
 450
 451         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 452           BC_F_COPYIN|BC_F_SUSER },
 453
 454         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 455           BC_F_COPYIN|BC_F_SUSER },
 456
 457         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 458           BC_F_COPYIN|BC_F_SUSER },
 459
 460 };
 461 const int bridge_control_table_size =
 462     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 463
 464 LIST_HEAD(, bridge_softc) bridge_list;
 465
 466 IFC_SIMPLE_DECLARE(bridge, 0);
 467
 468 static int
 469 bridge_modevent(module_t mod, int type, void *data)
 470 {
 471
 472         switch (type) {
 473         case MOD_LOAD:
 474                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 475                 if_clone_attach(&bridge_cloner);
 476                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 477                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 478                     UMA_ALIGN_PTR, 0);
 479                 LIST_INIT(&bridge_list);
 480                 bridge_input_p = bridge_input;
 481                 bridge_output_p = bridge_output;
 482                 bridge_dn_p = bridge_dummynet;
 483                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 484                     ifnet_departure_event, bridge_ifdetach, NULL,
 485                     EVENTHANDLER_PRI_ANY);
 486                 break;
 487         case MOD_UNLOAD:
 488                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 489                     bridge_detach_cookie);
 490                 if_clone_detach(&bridge_cloner);
 491                 uma_zdestroy(bridge_rtnode_zone);
 492                 bridge_input_p = NULL;
 493                 bridge_output_p = NULL;
 494                 bridge_dn_p = NULL;
 495                 mtx_destroy(&bridge_list_mtx);
 496                 break;
 497         default:
 498                 return (EOPNOTSUPP);
 499         }
 500         return (0);
 501 }
 502
 503 static moduledata_t bridge_mod = {
 504         "if_bridge",
 505         bridge_modevent,
 506         0
 507 };
 508
 509 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 510 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 511
 512 /*
 513  * handler for net.link.bridge.pfil_ipfw
 514  */
 515 static int
 516 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 517 {
 518         int enable = pfil_ipfw;
 519         int error;
 520
 521         error = sysctl_handle_int(oidp, &enable, 0, req);
 522         enable = (enable) ? 1 : 0;
 523
 524         if (enable != pfil_ipfw) {
 525                 pfil_ipfw = enable;
 526
 527                 /*
 528                  * Disable pfil so that ipfw doesnt run twice, if the user
 529                  * really wants both then they can re-enable pfil_bridge and/or
 530                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 531                  * layer2 type.
 532                  */
 533                 if (pfil_ipfw) {
 534                         pfil_onlyip = 0;
 535                         pfil_bridge = 0;
 536                         pfil_member = 0;
 537                 }
 538         }
 539
 540         return (error);
 541 }
 542 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 543             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 544
 545 /*
 546  * bridge_clone_create:
 547  *
 548  *      Create a new bridge instance.
 549  */
 550 static int
 551 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 552 {
 553         struct bridge_softc *sc, *sc2;
 554         struct ifnet *bifp, *ifp;
 555         u_char eaddr[6];
 556         int retry;
 557
 558         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 559         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 560         if (ifp == NULL) {
 561                 free(sc, M_DEVBUF);
 562                 return (ENOSPC);
 563         }
 564
 565         BRIDGE_LOCK_INIT(sc);
 566         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 567         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 568
 569         /* Initialize our routing table. */
 570         bridge_rtable_init(sc);
 571
 572         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 573
 574         LIST_INIT(&sc->sc_iflist);
 575         LIST_INIT(&sc->sc_spanlist);
 576
 577         ifp->if_softc = sc;
 578         if_initname(ifp, ifc->ifc_name, unit);
 579         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 580         ifp->if_ioctl = bridge_ioctl;
 581         ifp->if_start = bridge_start;
 582         ifp->if_init = bridge_init;
 583         ifp->if_type = IFT_BRIDGE;
 584         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 585         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 586         IFQ_SET_READY(&ifp->if_snd);
 587
 588         /*
 589          * Generate a random ethernet address with a locally administered
 590          * address.
 591          *
 592          * Since we are using random ethernet addresses for the bridge, it is
 593          * possible that we might have address collisions, so make sure that
 594          * this hardware address isn't already in use on another bridge.
 595          */
 596         for (retry = 1; retry != 0;) {
 597                 arc4rand(eaddr, ETHER_ADDR_LEN, 1);
 598                 eaddr[0] &= ~1;         /* clear multicast bit */
 599                 eaddr[0] |= 2;          /* set the LAA bit */
 600                 retry = 0;
 601                 mtx_lock(&bridge_list_mtx);
 602                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 603                         bifp = sc2->sc_ifp;
 604                         if (memcmp(eaddr, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 605                                 retry = 1;
 606                 }
 607                 mtx_unlock(&bridge_list_mtx);
 608         }
 609
 610         bstp_attach(&sc->sc_stp, &bridge_ops);
 611         ether_ifattach(ifp, eaddr);
 612         /* Now undo some of the damage... */
 613         ifp->if_baudrate = 0;
 614         ifp->if_type = IFT_BRIDGE;
 615
 616         mtx_lock(&bridge_list_mtx);
 617         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 618         mtx_unlock(&bridge_list_mtx);
 619
 620         return (0);
 621 }
 622
 623 /*
 624  * bridge_clone_destroy:
 625  *
 626  *      Destroy a bridge instance.
 627  */
 628 static void
 629 bridge_clone_destroy(struct ifnet *ifp)
 630 {
 631         struct bridge_softc *sc = ifp->if_softc;
 632         struct bridge_iflist *bif;
 633
 634         BRIDGE_LOCK(sc);
 635
 636         bridge_stop(ifp, 1);
 637         ifp->if_flags &= ~IFF_UP;
 638
 639         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 640                 bridge_delete_member(sc, bif, 0);
 641
 642         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 643                 bridge_delete_span(sc, bif);
 644         }
 645
 646         BRIDGE_UNLOCK(sc);
 647
 648         callout_drain(&sc->sc_brcallout);
 649
 650         mtx_lock(&bridge_list_mtx);
 651         LIST_REMOVE(sc, sc_list);
 652         mtx_unlock(&bridge_list_mtx);
 653
 654         bstp_detach(&sc->sc_stp);
 655         ether_ifdetach(ifp);
 656         if_free_type(ifp, IFT_ETHER);
 657
 658         /* Tear down the routing table. */
 659         bridge_rtable_fini(sc);
 660
 661         BRIDGE_LOCK_DESTROY(sc);
 662         free(sc, M_DEVBUF);
 663 }
 664
 665 /*
 666  * bridge_ioctl:
 667  *
 668  *      Handle a control request from the operator.
 669  */
 670 static int
 671 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 672 {
 673         struct bridge_softc *sc = ifp->if_softc;
 674         struct thread *td = curthread;
 675         union {
 676                 struct ifbreq ifbreq;
 677                 struct ifbifconf ifbifconf;
 678                 struct ifbareq ifbareq;
 679                 struct ifbaconf ifbaconf;
 680                 struct ifbrparam ifbrparam;
 681                 struct ifbropreq ifbropreq;
 682         } args;
 683         struct ifdrv *ifd = (struct ifdrv *) data;
 684         const struct bridge_control *bc;
 685         int error = 0;
 686
 687         switch (cmd) {
 688
 689         case SIOCADDMULTI:
 690         case SIOCDELMULTI:
 691                 break;
 692
 693         case SIOCGDRVSPEC:
 694         case SIOCSDRVSPEC:
 695                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 696                         error = EINVAL;
 697                         break;
 698                 }
 699                 bc = &bridge_control_table[ifd->ifd_cmd];
 700
 701                 if (cmd == SIOCGDRVSPEC &&
 702                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 703                         error = EINVAL;
 704                         break;
 705                 }
 706                 else if (cmd == SIOCSDRVSPEC &&
 707                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 708                         error = EINVAL;
 709                         break;
 710                 }
 711
 712                 if (bc->bc_flags & BC_F_SUSER) {
 713                         error = priv_check(td, PRIV_NET_BRIDGE);
 714                         if (error)
 715                                 break;
 716                 }
 717
 718                 if (ifd->ifd_len != bc->bc_argsize ||
 719                     ifd->ifd_len > sizeof(args)) {
 720                         error = EINVAL;
 721                         break;
 722                 }
 723
 724                 bzero(&args, sizeof(args));
 725                 if (bc->bc_flags & BC_F_COPYIN) {
 726                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 727                         if (error)
 728                                 break;
 729                 }
 730
 731                 BRIDGE_LOCK(sc);
 732                 error = (*bc->bc_func)(sc, &args);
 733                 BRIDGE_UNLOCK(sc);
 734                 if (error)
 735                         break;
 736
 737                 if (bc->bc_flags & BC_F_COPYOUT)
 738                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 739
 740                 break;
 741
 742         case SIOCSIFFLAGS:
 743                 if (!(ifp->if_flags & IFF_UP) &&
 744                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 745                         /*
 746                          * If interface is marked down and it is running,
 747                          * then stop and disable it.
 748                          */
 749                         BRIDGE_LOCK(sc);
 750                         bridge_stop(ifp, 1);
 751                         BRIDGE_UNLOCK(sc);
 752                 } else if ((ifp->if_flags & IFF_UP) &&
 753                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 754                         /*
 755                          * If interface is marked up and it is stopped, then
 756                          * start it.
 757                          */
 758                         (*ifp->if_init)(sc);
 759                 }
 760                 break;
 761
 762         case SIOCSIFMTU:
 763                 /* Do not allow the MTU to be changed on the bridge */
 764                 error = EINVAL;
 765                 break;
 766
 767         default:
 768                 /*
 769                  * drop the lock as ether_ioctl() will call bridge_start() and
 770                  * cause the lock to be recursed.
 771                  */
 772                 error = ether_ioctl(ifp, cmd, data);
 773                 break;
 774         }
 775
 776         return (error);
 777 }
 778
 779 /*
 780  * bridge_mutecaps:
 781  *
 782  *      Clear or restore unwanted capabilities on the member interface
 783  */
 784 static void
 785 bridge_mutecaps(struct bridge_softc *sc)
 786 {
 787         struct bridge_iflist *bif;
 788         int enabled, mask;
 789
 790         /* Initial bitmask of capabilities to test */
 791         mask = BRIDGE_IFCAPS_MASK;
 792
 793         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 794                 /* Every member must support it or its disabled */
 795                 mask &= bif->bif_savedcaps;
 796         }
 797
 798         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 799                 enabled = bif->bif_ifp->if_capenable;
 800                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 801                 /* strip off mask bits and enable them again if allowed */
 802                 enabled &= ~BRIDGE_IFCAPS_MASK;
 803                 enabled |= mask;
 804
 805                 bridge_set_ifcap(sc, bif, enabled);
 806         }
 807
 808 }
 809
 810 static void
 811 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 812 {
 813         struct ifnet *ifp = bif->bif_ifp;
 814         struct ifreq ifr;
 815         int error;
 816
 817         bzero(&ifr, sizeof(ifr));
 818         ifr.ifr_reqcap = set;
 819
 820         if (ifp->if_capenable != set) {
 821                 IFF_LOCKGIANT(ifp);
 822                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 823                 IFF_UNLOCKGIANT(ifp);
 824                 if (error)
 825                         if_printf(sc->sc_ifp,
 826                             "error setting interface capabilities on %s\n",
 827                             ifp->if_xname);
 828         }
 829 }
 830
 831 /*
 832  * bridge_lookup_member:
 833  *
 834  *      Lookup a bridge member interface.
 835  */
 836 static struct bridge_iflist *
 837 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 838 {
 839         struct bridge_iflist *bif;
 840         struct ifnet *ifp;
 841
 842         BRIDGE_LOCK_ASSERT(sc);
 843
 844         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 845                 ifp = bif->bif_ifp;
 846                 if (strcmp(ifp->if_xname, name) == 0)
 847                         return (bif);
 848         }
 849
 850         return (NULL);
 851 }
 852
 853 /*
 854  * bridge_lookup_member_if:
 855  *
 856  *      Lookup a bridge member interface by ifnet*.
 857  */
 858 static struct bridge_iflist *
 859 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 860 {
 861         struct bridge_iflist *bif;
 862
 863         BRIDGE_LOCK_ASSERT(sc);
 864
 865         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 866                 if (bif->bif_ifp == member_ifp)
 867                         return (bif);
 868         }
 869
 870         return (NULL);
 871 }
 872
 873 /*
 874  * bridge_delete_member:
 875  *
 876  *      Delete the specified member interface.
 877  */
 878 static void
 879 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 880     int gone)
 881 {
 882         struct ifnet *ifs = bif->bif_ifp;
 883
 884         BRIDGE_LOCK_ASSERT(sc);
 885
 886         if (!gone) {
 887                 switch (ifs->if_type) {
 888                 case IFT_ETHER:
 889                 case IFT_L2VLAN:
 890                         /*
 891                          * Take the interface out of promiscuous mode.
 892                          */
 893                         (void) ifpromisc(ifs, 0);
 894                         break;
 895
 896                 case IFT_GIF:
 897                         break;
 898
 899                 default:
 900 #ifdef DIAGNOSTIC
 901                         panic("bridge_delete_member: impossible");
 902 #endif
 903                         break;
 904                 }
 905                 /* reneable any interface capabilities */
 906                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
 907         }
 908
 909         if (bif->bif_flags & IFBIF_STP)
 910                 bstp_disable(&bif->bif_stp);
 911
 912         ifs->if_bridge = NULL;
 913         BRIDGE_XLOCK(sc);
 914         LIST_REMOVE(bif, bif_next);
 915         BRIDGE_XDROP(sc);
 916
 917         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 918         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 919         KASSERT(bif->bif_addrcnt == 0,
 920             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 921
 922         BRIDGE_UNLOCK(sc);
 923         bstp_destroy(&bif->bif_stp);    /* prepare to free */
 924         BRIDGE_LOCK(sc);
 925         free(bif, M_DEVBUF);
 926 }
 927
 928 /*
 929  * bridge_delete_span:
 930  *
 931  *      Delete the specified span interface.
 932  */
 933 static void
 934 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
 935 {
 936         BRIDGE_LOCK_ASSERT(sc);
 937
 938         KASSERT(bif->bif_ifp->if_bridge == NULL,
 939             ("%s: not a span interface", __func__));
 940
 941         LIST_REMOVE(bif, bif_next);
 942         free(bif, M_DEVBUF);
 943 }
 944
 945 static int
 946 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
 947 {
 948         struct ifbreq *req = arg;
 949         struct bridge_iflist *bif = NULL;
 950         struct ifnet *ifs;
 951         int error = 0;
 952
 953         ifs = ifunit(req->ifbr_ifsname);
 954         if (ifs == NULL)
 955                 return (ENOENT);
 956         if (ifs->if_ioctl == NULL)      /* must be supported */
 957                 return (EINVAL);
 958
 959         /* If it's in the span list, it can't be a member. */
 960         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
 961                 if (ifs == bif->bif_ifp)
 962                         return (EBUSY);
 963
 964         /* Allow the first Ethernet member to define the MTU */
 965         if (ifs->if_type != IFT_GIF) {
 966                 if (LIST_EMPTY(&sc->sc_iflist))
 967                         sc->sc_ifp->if_mtu = ifs->if_mtu;
 968                 else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
 969                         if_printf(sc->sc_ifp, "invalid MTU for %s\n",
 970                             ifs->if_xname);
 971                         return (EINVAL);
 972                 }
 973         }
 974
 975         if (ifs->if_bridge == sc)
 976                 return (EEXIST);
 977
 978         if (ifs->if_bridge != NULL)
 979                 return (EBUSY);
 980
 981         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
 982         if (bif == NULL)
 983                 return (ENOMEM);
 984
 985         bif->bif_ifp = ifs;
 986         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
 987         bif->bif_savedcaps = ifs->if_capenable;
 988
 989         switch (ifs->if_type) {
 990         case IFT_ETHER:
 991         case IFT_L2VLAN:
 992                 /*
 993                  * Place the interface into promiscuous mode.
 994                  */
 995                 error = ifpromisc(ifs, 1);
 996                 if (error)
 997                         goto out;
 998                 break;
 999
1000         case IFT_GIF:
1001                 break;
1002
1003         default:
1004                 error = EINVAL;
1005                 goto out;
1006         }
1007
1008         ifs->if_bridge = sc;
1009         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1010         /*
1011          * XXX: XLOCK HERE!?!
1012          *
1013          * NOTE: insert_***HEAD*** should be safe for the traversals.
1014          */
1015         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1016
1017         /* Set interface capabilities to the intersection set of all members */
1018         bridge_mutecaps(sc);
1019 out:
1020         if (error) {
1021                 if (bif != NULL)
1022                         free(bif, M_DEVBUF);
1023         }
1024         return (error);
1025 }
1026
1027 static int
1028 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1029 {
1030         struct ifbreq *req = arg;
1031         struct bridge_iflist *bif;
1032
1033         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1034         if (bif == NULL)
1035                 return (ENOENT);
1036
1037         bridge_delete_member(sc, bif, 0);
1038
1039         return (0);
1040 }
1041
1042 static int
1043 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1044 {
1045         struct ifbreq *req = arg;
1046         struct bridge_iflist *bif;
1047         struct bstp_port *bp;
1048
1049         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1050         if (bif == NULL)
1051                 return (ENOENT);
1052
1053         bp = &bif->bif_stp;
1054         req->ifbr_ifsflags = bif->bif_flags;
1055         req->ifbr_state = bp->bp_state;
1056         req->ifbr_priority = bp->bp_priority;
1057         req->ifbr_path_cost = bp->bp_path_cost;
1058         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1059         req->ifbr_proto = bp->bp_protover;
1060         req->ifbr_role = bp->bp_role;
1061         req->ifbr_stpflags = bp->bp_flags;
1062         req->ifbr_addrcnt = bif->bif_addrcnt;
1063         req->ifbr_addrmax = bif->bif_addrmax;
1064         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1065
1066         /* Copy STP state options as flags */
1067         if (bp->bp_operedge)
1068                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1069         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1070                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1071         if (bp->bp_ptp_link)
1072                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1073         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1074                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1075         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1076                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1077         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1078                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1079         return (0);
1080 }
1081
1082 static int
1083 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1084 {
1085         struct ifbreq *req = arg;
1086         struct bridge_iflist *bif;
1087         struct bstp_port *bp;
1088         int error;
1089
1090         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1091         if (bif == NULL)
1092                 return (ENOENT);
1093         bp = &bif->bif_stp;
1094
1095         if (req->ifbr_ifsflags & IFBIF_SPAN)
1096                 /* SPAN is readonly */
1097                 return (EINVAL);
1098
1099         if (req->ifbr_ifsflags & IFBIF_STP) {
1100                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1101                         error = bstp_enable(&bif->bif_stp);
1102                         if (error)
1103                                 return (error);
1104                 }
1105         } else {
1106                 if ((bif->bif_flags & IFBIF_STP) != 0)
1107                         bstp_disable(&bif->bif_stp);
1108         }
1109
1110         /* Pass on STP flags */
1111         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1112         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1113         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1114         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1115
1116         /* Save the bits relating to the bridge */
1117         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1118
1119         return (0);
1120 }
1121
1122 static int
1123 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1124 {
1125         struct ifbrparam *param = arg;
1126
1127         sc->sc_brtmax = param->ifbrp_csize;
1128         bridge_rttrim(sc);
1129
1130         return (0);
1131 }
1132
1133 static int
1134 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1135 {
1136         struct ifbrparam *param = arg;
1137
1138         param->ifbrp_csize = sc->sc_brtmax;
1139
1140         return (0);
1141 }
1142
1143 static int
1144 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1145 {
1146         struct ifbifconf *bifc = arg;
1147         struct bridge_iflist *bif;
1148         struct ifbreq breq;
1149         char *buf, *outbuf;
1150         int count, buflen, len, error = 0;
1151
1152         count = 0;
1153         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1154                 count++;
1155         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1156                 count++;
1157
1158         buflen = sizeof(breq) * count;
1159         if (bifc->ifbic_len == 0) {
1160                 bifc->ifbic_len = buflen;
1161                 return (0);
1162         }
1163         BRIDGE_UNLOCK(sc);
1164         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1165         BRIDGE_LOCK(sc);
1166
1167         count = 0;
1168         buf = outbuf;
1169         len = min(bifc->ifbic_len, buflen);
1170         bzero(&breq, sizeof(breq));
1171         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1172                 if (len < sizeof(breq))
1173                         break;
1174
1175                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1176                     sizeof(breq.ifbr_ifsname));
1177                 /* Fill in the ifbreq structure */
1178                 error = bridge_ioctl_gifflags(sc, &breq);
1179                 if (error)
1180                         break;
1181                 memcpy(buf, &breq, sizeof(breq));
1182                 count++;
1183                 buf += sizeof(breq);
1184                 len -= sizeof(breq);
1185         }
1186         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1187                 if (len < sizeof(breq))
1188                         break;
1189
1190                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1191                     sizeof(breq.ifbr_ifsname));
1192                 breq.ifbr_ifsflags = bif->bif_flags;
1193                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1194                 memcpy(buf, &breq, sizeof(breq));
1195                 count++;
1196                 buf += sizeof(breq);
1197                 len -= sizeof(breq);
1198         }
1199
1200         BRIDGE_UNLOCK(sc);
1201         bifc->ifbic_len = sizeof(breq) * count;
1202         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1203         BRIDGE_LOCK(sc);
1204         free(outbuf, M_TEMP);
1205         return (error);
1206 }
1207
1208 static int
1209 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1210 {
1211         struct ifbaconf *bac = arg;
1212         struct bridge_rtnode *brt;
1213         struct ifbareq bareq;
1214         char *buf, *outbuf;
1215         int count, buflen, len, error = 0;
1216
1217         if (bac->ifbac_len == 0)
1218                 return (0);
1219
1220         count = 0;
1221         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1222                 count++;
1223         buflen = sizeof(bareq) * count;
1224
1225         BRIDGE_UNLOCK(sc);
1226         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1227         BRIDGE_LOCK(sc);
1228
1229         count = 0;
1230         buf = outbuf;
1231         len = min(bac->ifbac_len, buflen);
1232         bzero(&bareq, sizeof(bareq));
1233         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1234                 if (len < sizeof(bareq))
1235                         goto out;
1236                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1237                     sizeof(bareq.ifba_ifsname));
1238                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1239                 bareq.ifba_vlan = brt->brt_vlan;
1240                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1241                                 time_uptime < brt->brt_expire)
1242                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1243                 else
1244                         bareq.ifba_expire = 0;
1245                 bareq.ifba_flags = brt->brt_flags;
1246
1247                 memcpy(buf, &bareq, sizeof(bareq));
1248                 count++;
1249                 buf += sizeof(bareq);
1250                 len -= sizeof(bareq);
1251         }
1252 out:
1253         BRIDGE_UNLOCK(sc);
1254         bac->ifbac_len = sizeof(bareq) * count;
1255         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1256         BRIDGE_LOCK(sc);
1257         free(outbuf, M_TEMP);
1258         return (error);
1259 }
1260
1261 static int
1262 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1263 {
1264         struct ifbareq *req = arg;
1265         struct bridge_iflist *bif;
1266         int error;
1267
1268         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1269         if (bif == NULL)
1270                 return (ENOENT);
1271
1272         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1273             req->ifba_flags);
1274
1275         return (error);
1276 }
1277
1278 static int
1279 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1280 {
1281         struct ifbrparam *param = arg;
1282
1283         sc->sc_brttimeout = param->ifbrp_ctime;
1284         return (0);
1285 }
1286
1287 static int
1288 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1289 {
1290         struct ifbrparam *param = arg;
1291
1292         param->ifbrp_ctime = sc->sc_brttimeout;
1293         return (0);
1294 }
1295
1296 static int
1297 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1298 {
1299         struct ifbareq *req = arg;
1300
1301         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1302 }
1303
1304 static int
1305 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1306 {
1307         struct ifbreq *req = arg;
1308
1309         bridge_rtflush(sc, req->ifbr_ifsflags);
1310         return (0);
1311 }
1312
1313 static int
1314 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1315 {
1316         struct ifbrparam *param = arg;
1317         struct bstp_state *bs = &sc->sc_stp;
1318
1319         param->ifbrp_prio = bs->bs_bridge_priority;
1320         return (0);
1321 }
1322
1323 static int
1324 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1325 {
1326         struct ifbrparam *param = arg;
1327
1328         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1329 }
1330
1331 static int
1332 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1333 {
1334         struct ifbrparam *param = arg;
1335         struct bstp_state *bs = &sc->sc_stp;
1336
1337         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1338         return (0);
1339 }
1340
1341 static int
1342 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1343 {
1344         struct ifbrparam *param = arg;
1345
1346         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1347 }
1348
1349 static int
1350 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1351 {
1352         struct ifbrparam *param = arg;
1353         struct bstp_state *bs = &sc->sc_stp;
1354
1355         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1356         return (0);
1357 }
1358
1359 static int
1360 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1361 {
1362         struct ifbrparam *param = arg;
1363
1364         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1365 }
1366
1367 static int
1368 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1369 {
1370         struct ifbrparam *param = arg;
1371         struct bstp_state *bs = &sc->sc_stp;
1372
1373         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1374         return (0);
1375 }
1376
1377 static int
1378 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1379 {
1380         struct ifbrparam *param = arg;
1381
1382         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1383 }
1384
1385 static int
1386 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1387 {
1388         struct ifbreq *req = arg;
1389         struct bridge_iflist *bif;
1390
1391         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1392         if (bif == NULL)
1393                 return (ENOENT);
1394
1395         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1396 }
1397
1398 static int
1399 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1400 {
1401         struct ifbreq *req = arg;
1402         struct bridge_iflist *bif;
1403
1404         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1405         if (bif == NULL)
1406                 return (ENOENT);
1407
1408         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1409 }
1410
1411 static int
1412 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1413 {
1414         struct ifbreq *req = arg;
1415         struct bridge_iflist *bif;
1416
1417         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1418         if (bif == NULL)
1419                 return (ENOENT);
1420
1421         bif->bif_addrmax = req->ifbr_addrmax;
1422         return (0);
1423 }
1424
1425 static int
1426 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1427 {
1428         struct ifbreq *req = arg;
1429         struct bridge_iflist *bif = NULL;
1430         struct ifnet *ifs;
1431
1432         ifs = ifunit(req->ifbr_ifsname);
1433         if (ifs == NULL)
1434                 return (ENOENT);
1435
1436         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1437                 if (ifs == bif->bif_ifp)
1438                         return (EBUSY);
1439
1440         if (ifs->if_bridge != NULL)
1441                 return (EBUSY);
1442
1443         switch (ifs->if_type) {
1444                 case IFT_ETHER:
1445                 case IFT_GIF:
1446                 case IFT_L2VLAN:
1447                         break;
1448                 default:
1449                         return (EINVAL);
1450         }
1451
1452         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1453         if (bif == NULL)
1454                 return (ENOMEM);
1455
1456         bif->bif_ifp = ifs;
1457         bif->bif_flags = IFBIF_SPAN;
1458
1459         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1460
1461         return (0);
1462 }
1463
1464 static int
1465 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1466 {
1467         struct ifbreq *req = arg;
1468         struct bridge_iflist *bif;
1469         struct ifnet *ifs;
1470
1471         ifs = ifunit(req->ifbr_ifsname);
1472         if (ifs == NULL)
1473                 return (ENOENT);
1474
1475         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1476                 if (ifs == bif->bif_ifp)
1477                         break;
1478
1479         if (bif == NULL)
1480                 return (ENOENT);
1481
1482         bridge_delete_span(sc, bif);
1483
1484         return (0);
1485 }
1486
1487 static int
1488 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1489 {
1490         struct ifbropreq *req = arg;
1491         struct bstp_state *bs = &sc->sc_stp;
1492         struct bstp_port *root_port;
1493
1494         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1495         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1496         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1497
1498         root_port = bs->bs_root_port;
1499         if (root_port == NULL)
1500                 req->ifbop_root_port = 0;
1501         else
1502                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1503
1504         req->ifbop_holdcount = bs->bs_txholdcount;
1505         req->ifbop_priority = bs->bs_bridge_priority;
1506         req->ifbop_protocol = bs->bs_protover;
1507         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1508         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1509         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1510         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1511         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1512         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1513
1514         return (0);
1515 }
1516
1517 static int
1518 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1519 {
1520         struct ifbrparam *param = arg;
1521
1522         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1523         return (0);
1524 }
1525
1526 static int
1527 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1528 {
1529         struct ifbpstpconf *bifstp = arg;
1530         struct bridge_iflist *bif;
1531         struct bstp_port *bp;
1532         struct ifbpstpreq bpreq;
1533         char *buf, *outbuf;
1534         int count, buflen, len, error = 0;
1535
1536         count = 0;
1537         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1538                 if ((bif->bif_flags & IFBIF_STP) != 0)
1539                         count++;
1540         }
1541
1542         buflen = sizeof(bpreq) * count;
1543         if (bifstp->ifbpstp_len == 0) {
1544                 bifstp->ifbpstp_len = buflen;
1545                 return (0);
1546         }
1547
1548         BRIDGE_UNLOCK(sc);
1549         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1550         BRIDGE_LOCK(sc);
1551
1552         count = 0;
1553         buf = outbuf;
1554         len = min(bifstp->ifbpstp_len, buflen);
1555         bzero(&bpreq, sizeof(bpreq));
1556         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1557                 if (len < sizeof(bpreq))
1558                         break;
1559
1560                 if ((bif->bif_flags & IFBIF_STP) == 0)
1561                         continue;
1562
1563                 bp = &bif->bif_stp;
1564                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1565                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1566                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1567                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1568                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1569                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1570
1571                 memcpy(buf, &bpreq, sizeof(bpreq));
1572                 count++;
1573                 buf += sizeof(bpreq);
1574                 len -= sizeof(bpreq);
1575         }
1576
1577         BRIDGE_UNLOCK(sc);
1578         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1579         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1580         BRIDGE_LOCK(sc);
1581         free(outbuf, M_TEMP);
1582         return (error);
1583 }
1584
1585 static int
1586 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1587 {
1588         struct ifbrparam *param = arg;
1589
1590         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1591 }
1592
1593 static int
1594 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1595 {
1596         struct ifbrparam *param = arg;
1597
1598         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1599 }
1600
1601 /*
1602  * bridge_ifdetach:
1603  *
1604  *      Detach an interface from a bridge.  Called when a member
1605  *      interface is detaching.
1606  */
1607 static void
1608 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1609 {
1610         struct bridge_softc *sc = ifp->if_bridge;
1611         struct bridge_iflist *bif;
1612
1613         /* Check if the interface is a bridge member */
1614         if (sc != NULL) {
1615                 BRIDGE_LOCK(sc);
1616
1617                 bif = bridge_lookup_member_if(sc, ifp);
1618                 if (bif != NULL)
1619                         bridge_delete_member(sc, bif, 1);
1620
1621                 BRIDGE_UNLOCK(sc);
1622                 return;
1623         }
1624
1625         /* Check if the interface is a span port */
1626         mtx_lock(&bridge_list_mtx);
1627         LIST_FOREACH(sc, &bridge_list, sc_list) {
1628                 BRIDGE_LOCK(sc);
1629                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1630                         if (ifp == bif->bif_ifp) {
1631                                 bridge_delete_span(sc, bif);
1632                                 break;
1633                         }
1634
1635                 BRIDGE_UNLOCK(sc);
1636         }
1637         mtx_unlock(&bridge_list_mtx);
1638 }
1639
1640 /*
1641  * bridge_init:
1642  *
1643  *      Initialize a bridge interface.
1644  */
1645 static void
1646 bridge_init(void *xsc)
1647 {
1648         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1649         struct ifnet *ifp = sc->sc_ifp;
1650
1651         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1652                 return;
1653
1654         BRIDGE_LOCK(sc);
1655         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1656             bridge_timer, sc);
1657
1658         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1659         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1660
1661         BRIDGE_UNLOCK(sc);
1662 }
1663
1664 /*
1665  * bridge_stop:
1666  *
1667  *      Stop the bridge interface.
1668  */
1669 static void
1670 bridge_stop(struct ifnet *ifp, int disable)
1671 {
1672         struct bridge_softc *sc = ifp->if_softc;
1673
1674         BRIDGE_LOCK_ASSERT(sc);
1675
1676         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1677                 return;
1678
1679         callout_stop(&sc->sc_brcallout);
1680         bstp_stop(&sc->sc_stp);
1681
1682         bridge_rtflush(sc, IFBF_FLUSHDYN);
1683
1684         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1685 }
1686
1687 /*
1688  * bridge_enqueue:
1689  *
1690  *      Enqueue a packet on a bridge member interface.
1691  *
1692  */
1693 static void
1694 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1695 {
1696         int len, err = 0;
1697         short mflags;
1698         struct mbuf *m0;
1699
1700         len = m->m_pkthdr.len;
1701         mflags = m->m_flags;
1702
1703         /* We may be sending a fragment so traverse the mbuf */
1704         for (; m; m = m0) {
1705                 m0 = m->m_nextpkt;
1706                 m->m_nextpkt = NULL;
1707
1708                 /*
1709                  * If underlying interface can not do VLAN tag insertion itself
1710                  * then attach a packet tag that holds it.
1711                  */
1712                 if ((m->m_flags & M_VLANTAG) &&
1713                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1714                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1715                         if (m == NULL) {
1716                                 if_printf(dst_ifp,
1717                                     "unable to prepend VLAN header\n");
1718                                 dst_ifp->if_oerrors++;
1719                                 continue;
1720                         }
1721                         m->m_flags &= ~M_VLANTAG;
1722                 }
1723
1724                 if (err == 0)
1725                         IFQ_ENQUEUE(&dst_ifp->if_snd, m, err);
1726         }
1727
1728         if (err == 0) {
1729
1730                 sc->sc_ifp->if_opackets++;
1731                 sc->sc_ifp->if_obytes += len;
1732
1733                 dst_ifp->if_obytes += len;
1734
1735                 if (mflags & M_MCAST) {
1736                         sc->sc_ifp->if_omcasts++;
1737                         dst_ifp->if_omcasts++;
1738                 }
1739         }
1740
1741         if ((dst_ifp->if_drv_flags & IFF_DRV_OACTIVE) == 0)
1742                 (*dst_ifp->if_start)(dst_ifp);
1743 }
1744
1745 /*
1746  * bridge_dummynet:
1747  *
1748  *      Receive a queued packet from dummynet and pass it on to the output
1749  *      interface.
1750  *
1751  *      The mbuf has the Ethernet header already attached.
1752  */
1753 static void
1754 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1755 {
1756         struct bridge_softc *sc;
1757
1758         sc = ifp->if_bridge;
1759
1760         /*
1761          * The packet didnt originate from a member interface. This should only
1762          * ever happen if a member interface is removed while packets are
1763          * queued for it.
1764          */
1765         if (sc == NULL) {
1766                 m_freem(m);
1767                 return;
1768         }
1769
1770         if (PFIL_HOOKED(&inet_pfil_hook)
1771 #ifdef INET6
1772             || PFIL_HOOKED(&inet6_pfil_hook)
1773 #endif
1774             ) {
1775                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1776                         return;
1777                 if (m == NULL)
1778                         return;
1779         }
1780
1781         bridge_enqueue(sc, ifp, m);
1782 }
1783
1784 /*
1785  * bridge_output:
1786  *
1787  *      Send output from a bridge member interface.  This
1788  *      performs the bridging function for locally originated
1789  *      packets.
1790  *
1791  *      The mbuf has the Ethernet header already attached.  We must
1792  *      enqueue or free the mbuf before returning.
1793  */
1794 static int
1795 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1796     struct rtentry *rt)
1797 {
1798         struct ether_header *eh;
1799         struct ifnet *dst_if;
1800         struct bridge_softc *sc;
1801         uint16_t vlan;
1802
1803         if (m->m_len < ETHER_HDR_LEN) {
1804                 m = m_pullup(m, ETHER_HDR_LEN);
1805                 if (m == NULL)
1806                         return (0);
1807         }
1808
1809         eh = mtod(m, struct ether_header *);
1810         sc = ifp->if_bridge;
1811         vlan = VLANTAGOF(m);
1812
1813         BRIDGE_LOCK(sc);
1814
1815         /*
1816          * If bridge is down, but the original output interface is up,
1817          * go ahead and send out that interface.  Otherwise, the packet
1818          * is dropped below.
1819          */
1820         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1821                 dst_if = ifp;
1822                 goto sendunicast;
1823         }
1824
1825         /*
1826          * If the packet is a multicast, or we don't know a better way to
1827          * get there, send to all interfaces.
1828          */
1829         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1830                 dst_if = NULL;
1831         else
1832                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1833         if (dst_if == NULL) {
1834                 struct bridge_iflist *bif;
1835                 struct mbuf *mc;
1836                 int error = 0, used = 0;
1837
1838                 bridge_span(sc, m);
1839
1840                 BRIDGE_LOCK2REF(sc, error);
1841                 if (error) {
1842                         m_freem(m);
1843                         return (0);
1844                 }
1845
1846                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1847                         dst_if = bif->bif_ifp;
1848
1849                         if (dst_if->if_type == IFT_GIF)
1850                                 continue;
1851                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1852                                 continue;
1853
1854                         /*
1855                          * If this is not the original output interface,
1856                          * and the interface is participating in spanning
1857                          * tree, make sure the port is in a state that
1858                          * allows forwarding.
1859                          */
1860                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1861                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1862                                 continue;
1863
1864                         if (LIST_NEXT(bif, bif_next) == NULL) {
1865                                 used = 1;
1866                                 mc = m;
1867                         } else {
1868                                 mc = m_copypacket(m, M_DONTWAIT);
1869                                 if (mc == NULL) {
1870                                         sc->sc_ifp->if_oerrors++;
1871                                         continue;
1872                                 }
1873                         }
1874
1875                         bridge_enqueue(sc, dst_if, mc);
1876                 }
1877                 if (used == 0)
1878                         m_freem(m);
1879                 BRIDGE_UNREF(sc);
1880                 return (0);
1881         }
1882
1883 sendunicast:
1884         /*
1885          * XXX Spanning tree consideration here?
1886          */
1887
1888         bridge_span(sc, m);
1889         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1890                 m_freem(m);
1891                 BRIDGE_UNLOCK(sc);
1892                 return (0);
1893         }
1894
1895         BRIDGE_UNLOCK(sc);
1896         bridge_enqueue(sc, dst_if, m);
1897         return (0);
1898 }
1899
1900 /*
1901  * bridge_start:
1902  *
1903  *      Start output on a bridge.
1904  *
1905  */
1906 static void
1907 bridge_start(struct ifnet *ifp)
1908 {
1909         struct bridge_softc *sc;
1910         struct mbuf *m;
1911         struct ether_header *eh;
1912         struct ifnet *dst_if;
1913
1914         sc = ifp->if_softc;
1915
1916         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1917         for (;;) {
1918                 IFQ_DEQUEUE(&ifp->if_snd, m);
1919                 if (m == 0)
1920                         break;
1921                 ETHER_BPF_MTAP(ifp, m);
1922
1923                 eh = mtod(m, struct ether_header *);
1924                 dst_if = NULL;
1925
1926                 BRIDGE_LOCK(sc);
1927                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1928                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
1929                 }
1930
1931                 if (dst_if == NULL)
1932                         bridge_broadcast(sc, ifp, m, 0);
1933                 else {
1934                         BRIDGE_UNLOCK(sc);
1935                         bridge_enqueue(sc, dst_if, m);
1936                 }
1937         }
1938         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
1939 }
1940
1941 /*
1942  * bridge_forward:
1943  *
1944  *      The forwarding function of the bridge.
1945  *
1946  *      NOTE: Releases the lock on return.
1947  */
1948 static void
1949 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
1950     struct mbuf *m)
1951 {
1952         struct bridge_iflist *dbif;
1953         struct ifnet *src_if, *dst_if, *ifp;
1954         struct ether_header *eh;
1955         uint16_t vlan;
1956         uint8_t *dst;
1957         int error;
1958
1959         src_if = m->m_pkthdr.rcvif;
1960         ifp = sc->sc_ifp;
1961
1962         ifp->if_ipackets++;
1963         ifp->if_ibytes += m->m_pkthdr.len;
1964         vlan = VLANTAGOF(m);
1965
1966         if ((sbif->bif_flags & IFBIF_STP) &&
1967             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1968                 goto drop;
1969
1970         eh = mtod(m, struct ether_header *);
1971         dst = eh->ether_dhost;
1972
1973         /* If the interface is learning, record the address. */
1974         if (sbif->bif_flags & IFBIF_LEARNING) {
1975                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
1976                     sbif, 0, IFBAF_DYNAMIC);
1977                 /*
1978                  * If the interface has addresses limits then deny any source
1979                  * that is not in the cache.
1980                  */
1981                 if (error && sbif->bif_addrmax)
1982                         goto drop;
1983         }
1984
1985         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
1986             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
1987                 goto drop;
1988
1989         /*
1990          * At this point, the port either doesn't participate
1991          * in spanning tree or it is in the forwarding state.
1992          */
1993
1994         /*
1995          * If the packet is unicast, destined for someone on
1996          * "this" side of the bridge, drop it.
1997          */
1998         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1999                 dst_if = bridge_rtlookup(sc, dst, vlan);
2000                 if (src_if == dst_if)
2001                         goto drop;
2002         } else {
2003                 /*
2004                  * Check if its a reserved multicast address, any address
2005                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2006                  * bridge.
2007                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2008                  */
2009                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2010                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2011                     dst[4] == 0x00 && dst[5] <= 0x0f)
2012                         goto drop;
2013
2014                 /* ...forward it to all interfaces. */
2015                 ifp->if_imcasts++;
2016                 dst_if = NULL;
2017         }
2018
2019         /*
2020          * If we have a destination interface which is a member of our bridge,
2021          * OR this is a unicast packet, push it through the bpf(4) machinery.
2022          * For broadcast or multicast packets, don't bother because it will
2023          * be reinjected into ether_input. We do this before we pass the packets
2024          * through the pfil(9) framework, as it is possible that pfil(9) will
2025          * drop the packet, or possibly modify it, making it difficult to debug
2026          * firewall issues on the bridge.
2027          */
2028         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2029                 ETHER_BPF_MTAP(ifp, m);
2030
2031         /* run the packet filter */
2032         if (PFIL_HOOKED(&inet_pfil_hook)
2033 #ifdef INET6
2034             || PFIL_HOOKED(&inet6_pfil_hook)
2035 #endif
2036             ) {
2037                 BRIDGE_UNLOCK(sc);
2038                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2039                         return;
2040                 if (m == NULL)
2041                         return;
2042                 BRIDGE_LOCK(sc);
2043         }
2044
2045         if (dst_if == NULL) {
2046                 bridge_broadcast(sc, src_if, m, 1);
2047                 return;
2048         }
2049
2050         /*
2051          * At this point, we're dealing with a unicast frame
2052          * going to a different interface.
2053          */
2054         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2055                 goto drop;
2056
2057         dbif = bridge_lookup_member_if(sc, dst_if);
2058         if (dbif == NULL)
2059                 /* Not a member of the bridge (anymore?) */
2060                 goto drop;
2061
2062         /* Private segments can not talk to each other */
2063         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2064                 goto drop;
2065
2066         if ((dbif->bif_flags & IFBIF_STP) &&
2067             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2068                 goto drop;
2069
2070         BRIDGE_UNLOCK(sc);
2071
2072         if (PFIL_HOOKED(&inet_pfil_hook)
2073 #ifdef INET6
2074             || PFIL_HOOKED(&inet6_pfil_hook)
2075 #endif
2076             ) {
2077                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2078                         return;
2079                 if (m == NULL)
2080                         return;
2081         }
2082
2083         bridge_enqueue(sc, dst_if, m);
2084         return;
2085
2086 drop:
2087         BRIDGE_UNLOCK(sc);
2088         m_freem(m);
2089 }
2090
2091 /*
2092  * bridge_input:
2093  *
2094  *      Receive input from a member interface.  Queue the packet for
2095  *      bridging if it is not for us.
2096  */
2097 static struct mbuf *
2098 bridge_input(struct ifnet *ifp, struct mbuf *m)
2099 {
2100         struct bridge_softc *sc = ifp->if_bridge;
2101         struct bridge_iflist *bif, *bif2;
2102         struct ifnet *bifp;
2103         struct ether_header *eh;
2104         struct mbuf *mc, *mc2;
2105         uint16_t vlan;
2106         int error;
2107
2108         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2109                 return (m);
2110
2111         bifp = sc->sc_ifp;
2112         vlan = VLANTAGOF(m);
2113
2114         /*
2115          * Implement support for bridge monitoring. If this flag has been
2116          * set on this interface, discard the packet once we push it through
2117          * the bpf(4) machinery, but before we do, increment the byte and
2118          * packet counters associated with this interface.
2119          */
2120         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2121                 m->m_pkthdr.rcvif  = bifp;
2122                 ETHER_BPF_MTAP(bifp, m);
2123                 bifp->if_ipackets++;
2124                 bifp->if_ibytes += m->m_pkthdr.len;
2125                 m_freem(m);
2126                 return (NULL);
2127         }
2128         BRIDGE_LOCK(sc);
2129         bif = bridge_lookup_member_if(sc, ifp);
2130         if (bif == NULL) {
2131                 BRIDGE_UNLOCK(sc);
2132                 return (m);
2133         }
2134
2135         eh = mtod(m, struct ether_header *);
2136
2137         bridge_span(sc, m);
2138
2139         if (m->m_flags & (M_BCAST|M_MCAST)) {
2140                 /* Tap off 802.1D packets; they do not get forwarded. */
2141                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2142                     ETHER_ADDR_LEN) == 0) {
2143                         m = bstp_input(&bif->bif_stp, ifp, m);
2144                         if (m == NULL) {
2145                                 BRIDGE_UNLOCK(sc);
2146                                 return (NULL);
2147                         }
2148                 }
2149
2150                 if ((bif->bif_flags & IFBIF_STP) &&
2151                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2152                         BRIDGE_UNLOCK(sc);
2153                         return (m);
2154                 }
2155
2156                 /*
2157                  * Make a deep copy of the packet and enqueue the copy
2158                  * for bridge processing; return the original packet for
2159                  * local processing.
2160                  */
2161                 mc = m_dup(m, M_DONTWAIT);
2162                 if (mc == NULL) {
2163                         BRIDGE_UNLOCK(sc);
2164                         return (m);
2165                 }
2166
2167                 /* Perform the bridge forwarding function with the copy. */
2168                 bridge_forward(sc, bif, mc);
2169
2170                 /*
2171                  * Reinject the mbuf as arriving on the bridge so we have a
2172                  * chance at claiming multicast packets. We can not loop back
2173                  * here from ether_input as a bridge is never a member of a
2174                  * bridge.
2175                  */
2176                 KASSERT(bifp->if_bridge == NULL,
2177                     ("loop created in bridge_input"));
2178                 mc2 = m_dup(m, M_DONTWAIT);
2179                 if (mc2 != NULL) {
2180                         /* Keep the layer3 header aligned */
2181                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2182                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2183                 }
2184                 if (mc2 != NULL) {
2185                         mc2->m_pkthdr.rcvif = bifp;
2186                         (*bifp->if_input)(bifp, mc2);
2187                 }
2188
2189                 /* Return the original packet for local processing. */
2190                 return (m);
2191         }
2192
2193         if ((bif->bif_flags & IFBIF_STP) &&
2194             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2195                 BRIDGE_UNLOCK(sc);
2196                 return (m);
2197         }
2198
2199 #ifdef DEV_CARP
2200 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2201         || ((iface)->if_carp \
2202             && carp_forus((iface)->if_carp, eh->ether_dhost))
2203 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2204         || ((iface)->if_carp \
2205             && carp_forus((iface)->if_carp, eh->ether_shost))
2206 #else
2207 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2208 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2209 #endif
2210
2211 #ifdef INET6
2212 #   define OR_PFIL_HOOKED_INET6 \
2213         || PFIL_HOOKED(&inet6_pfil_hook)
2214 #else
2215 #   define OR_PFIL_HOOKED_INET6
2216 #endif
2217
2218 #define GRAB_OUR_PACKETS(iface) \
2219         if ((iface)->if_type == IFT_GIF) \
2220                 continue; \
2221         /* It is destined for us. */ \
2222         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2223             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2224             ) {                                                         \
2225                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2226                         ETHER_BPF_MTAP(iface, m);                       \
2227                         iface->if_ipackets++;                           \
2228                         /* Filter on the physical interface. */         \
2229                         if (pfil_local_phys &&                          \
2230                             (PFIL_HOOKED(&inet_pfil_hook)               \
2231                              OR_PFIL_HOOKED_INET6)) {                   \
2232                                 if (bridge_pfil(&m, NULL, ifp,          \
2233                                     PFIL_IN) != 0 || m == NULL) {       \
2234                                         BRIDGE_UNLOCK(sc);              \
2235                                         return (NULL);                  \
2236                                 }                                       \
2237                         }                                               \
2238                 }                                                       \
2239                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2240                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2241                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2242                         if (error && bif->bif_addrmax) {                \
2243                                 BRIDGE_UNLOCK(sc);                      \
2244                                 m_freem(m);                             \
2245                                 return (NULL);                          \
2246                         }                                               \
2247                 }                                                       \
2248                 m->m_pkthdr.rcvif = iface;                              \
2249                 BRIDGE_UNLOCK(sc);                                      \
2250                 return (m);                                             \
2251         }                                                               \
2252                                                                         \
2253         /* We just received a packet that we sent out. */               \
2254         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2255             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2256             ) {                                                         \
2257                 BRIDGE_UNLOCK(sc);                                      \
2258                 m_freem(m);                                             \
2259                 return (NULL);                                          \
2260         }
2261
2262         /*
2263          * Unicast.  Make sure it's not for the bridge.
2264          */
2265         do { GRAB_OUR_PACKETS(bifp) } while (0);
2266
2267         /*
2268          * Give a chance for ifp at first priority. This will help when the
2269          * packet comes through the interface like VLAN's with the same MACs
2270          * on several interfaces from the same bridge. This also will save
2271          * some CPU cycles in case the destination interface and the input
2272          * interface (eq ifp) are the same.
2273          */
2274         do { GRAB_OUR_PACKETS(ifp) } while (0);
2275
2276         /* Now check the all bridge members. */
2277         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2278                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2279         }
2280
2281 #undef OR_CARP_CHECK_WE_ARE_DST
2282 #undef OR_CARP_CHECK_WE_ARE_SRC
2283 #undef OR_PFIL_HOOKED_INET6
2284 #undef GRAB_OUR_PACKETS
2285
2286         /* Perform the bridge forwarding function. */
2287         bridge_forward(sc, bif, m);
2288
2289         return (NULL);
2290 }
2291
2292 /*
2293  * bridge_broadcast:
2294  *
2295  *      Send a frame to all interfaces that are members of
2296  *      the bridge, except for the one on which the packet
2297  *      arrived.
2298  *
2299  *      NOTE: Releases the lock on return.
2300  */
2301 static void
2302 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2303     struct mbuf *m, int runfilt)
2304 {
2305         struct bridge_iflist *dbif, *sbif;
2306         struct mbuf *mc;
2307         struct ifnet *dst_if;
2308         int error = 0, used = 0, i;
2309
2310         sbif = bridge_lookup_member_if(sc, src_if);
2311
2312         BRIDGE_LOCK2REF(sc, error);
2313         if (error) {
2314                 m_freem(m);
2315                 return;
2316         }
2317
2318         /* Filter on the bridge interface before broadcasting */
2319         if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2320 #ifdef INET6
2321             || PFIL_HOOKED(&inet6_pfil_hook)
2322 #endif
2323             )) {
2324                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2325                         goto out;
2326                 if (m == NULL)
2327                         goto out;
2328         }
2329
2330         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2331                 dst_if = dbif->bif_ifp;
2332                 if (dst_if == src_if)
2333                         continue;
2334
2335                 /* Private segments can not talk to each other */
2336                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2337                         continue;
2338
2339                 if ((dbif->bif_flags & IFBIF_STP) &&
2340                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2341                         continue;
2342
2343                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2344                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2345                         continue;
2346
2347                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2348                         continue;
2349
2350                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2351                         mc = m;
2352                         used = 1;
2353                 } else {
2354                         mc = m_dup(m, M_DONTWAIT);
2355                         if (mc == NULL) {
2356                                 sc->sc_ifp->if_oerrors++;
2357                                 continue;
2358                         }
2359                 }
2360
2361                 /*
2362                  * Filter on the output interface. Pass a NULL bridge interface
2363                  * pointer so we do not redundantly filter on the bridge for
2364                  * each interface we broadcast on.
2365                  */
2366                 if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2367 #ifdef INET6
2368                     || PFIL_HOOKED(&inet6_pfil_hook)
2369 #endif
2370                     )) {
2371                         if (used == 0) {
2372                                 /* Keep the layer3 header aligned */
2373                                 i = min(mc->m_pkthdr.len, max_protohdr);
2374                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2375                                 if (mc == NULL) {
2376                                         sc->sc_ifp->if_oerrors++;
2377                                         continue;
2378                                 }
2379                         }
2380                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2381                                 continue;
2382                         if (mc == NULL)
2383                                 continue;
2384                 }
2385
2386                 bridge_enqueue(sc, dst_if, mc);
2387         }
2388         if (used == 0)
2389                 m_freem(m);
2390
2391 out:
2392         BRIDGE_UNREF(sc);
2393 }
2394
2395 /*
2396  * bridge_span:
2397  *
2398  *      Duplicate a packet out one or more interfaces that are in span mode,
2399  *      the original mbuf is unmodified.
2400  */
2401 static void
2402 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2403 {
2404         struct bridge_iflist *bif;
2405         struct ifnet *dst_if;
2406         struct mbuf *mc;
2407
2408         if (LIST_EMPTY(&sc->sc_spanlist))
2409                 return;
2410
2411         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2412                 dst_if = bif->bif_ifp;
2413
2414                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2415                         continue;
2416
2417                 mc = m_copypacket(m, M_DONTWAIT);
2418                 if (mc == NULL) {
2419                         sc->sc_ifp->if_oerrors++;
2420                         continue;
2421                 }
2422
2423                 bridge_enqueue(sc, dst_if, mc);
2424         }
2425 }
2426
2427 /*
2428  * bridge_rtupdate:
2429  *
2430  *      Add a bridge routing entry.
2431  */
2432 static int
2433 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2434     struct bridge_iflist *bif, int setflags, uint8_t flags)
2435 {
2436         struct bridge_rtnode *brt;
2437         int error;
2438
2439         BRIDGE_LOCK_ASSERT(sc);
2440
2441         /* Check the source address is valid and not multicast. */
2442         if (ETHER_IS_MULTICAST(dst) ||
2443             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2444              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2445                 return (EINVAL);
2446
2447         /* 802.1p frames map to vlan 1 */
2448         if (vlan == 0)
2449                 vlan = 1;
2450
2451         /*
2452          * A route for this destination might already exist.  If so,
2453          * update it, otherwise create a new one.
2454          */
2455         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2456                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2457                         sc->sc_brtexceeded++;
2458                         return (ENOSPC);
2459                 }
2460                 /* Check per interface address limits (if enabled) */
2461                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2462                         bif->bif_addrexceeded++;
2463                         return (ENOSPC);
2464                 }
2465
2466                 /*
2467                  * Allocate a new bridge forwarding node, and
2468                  * initialize the expiration time and Ethernet
2469                  * address.
2470                  */
2471                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2472                 if (brt == NULL)
2473                         return (ENOMEM);
2474
2475                 if (bif->bif_flags & IFBIF_STICKY)
2476                         brt->brt_flags = IFBAF_STICKY;
2477                 else
2478                         brt->brt_flags = IFBAF_DYNAMIC;
2479
2480                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2481                 brt->brt_vlan = vlan;
2482
2483                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2484                         uma_zfree(bridge_rtnode_zone, brt);
2485                         return (error);
2486                 }
2487                 brt->brt_dst = bif;
2488                 bif->bif_addrcnt++;
2489         }
2490
2491         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2492             brt->brt_dst != bif) {
2493                 brt->brt_dst->bif_addrcnt--;
2494                 brt->brt_dst = bif;
2495                 brt->brt_dst->bif_addrcnt++;
2496         }
2497
2498         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2499                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2500         if (setflags)
2501                 brt->brt_flags = flags;
2502
2503         return (0);
2504 }
2505
2506 /*
2507  * bridge_rtlookup:
2508  *
2509  *      Lookup the destination interface for an address.
2510  */
2511 static struct ifnet *
2512 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2513 {
2514         struct bridge_rtnode *brt;
2515
2516         BRIDGE_LOCK_ASSERT(sc);
2517
2518         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2519                 return (NULL);
2520
2521         return (brt->brt_ifp);
2522 }
2523
2524 /*
2525  * bridge_rttrim:
2526  *
2527  *      Trim the routine table so that we have a number
2528  *      of routing entries less than or equal to the
2529  *      maximum number.
2530  */
2531 static void
2532 bridge_rttrim(struct bridge_softc *sc)
2533 {
2534         struct bridge_rtnode *brt, *nbrt;
2535
2536         BRIDGE_LOCK_ASSERT(sc);
2537
2538         /* Make sure we actually need to do this. */
2539         if (sc->sc_brtcnt <= sc->sc_brtmax)
2540                 return;
2541
2542         /* Force an aging cycle; this might trim enough addresses. */
2543         bridge_rtage(sc);
2544         if (sc->sc_brtcnt <= sc->sc_brtmax)
2545                 return;
2546
2547         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2548                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2549                         bridge_rtnode_destroy(sc, brt);
2550                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2551                                 return;
2552                 }
2553         }
2554 }
2555
2556 /*
2557  * bridge_timer:
2558  *
2559  *      Aging timer for the bridge.
2560  */
2561 static void
2562 bridge_timer(void *arg)
2563 {
2564         struct bridge_softc *sc = arg;
2565
2566         BRIDGE_LOCK_ASSERT(sc);
2567
2568         bridge_rtage(sc);
2569
2570         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2571                 callout_reset(&sc->sc_brcallout,
2572                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2573 }
2574
2575 /*
2576  * bridge_rtage:
2577  *
2578  *      Perform an aging cycle.
2579  */
2580 static void
2581 bridge_rtage(struct bridge_softc *sc)
2582 {
2583         struct bridge_rtnode *brt, *nbrt;
2584
2585         BRIDGE_LOCK_ASSERT(sc);
2586
2587         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2588                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2589                         if (time_uptime >= brt->brt_expire)
2590                                 bridge_rtnode_destroy(sc, brt);
2591                 }
2592         }
2593 }
2594
2595 /*
2596  * bridge_rtflush:
2597  *
2598  *      Remove all dynamic addresses from the bridge.
2599  */
2600 static void
2601 bridge_rtflush(struct bridge_softc *sc, int full)
2602 {
2603         struct bridge_rtnode *brt, *nbrt;
2604
2605         BRIDGE_LOCK_ASSERT(sc);
2606
2607         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2608                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2609                         bridge_rtnode_destroy(sc, brt);
2610         }
2611 }
2612
2613 /*
2614  * bridge_rtdaddr:
2615  *
2616  *      Remove an address from the table.
2617  */
2618 static int
2619 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2620 {
2621         struct bridge_rtnode *brt;
2622         int found = 0;
2623
2624         BRIDGE_LOCK_ASSERT(sc);
2625
2626         /*
2627          * If vlan is zero then we want to delete for all vlans so the lookup
2628          * may return more than one.
2629          */
2630         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2631                 bridge_rtnode_destroy(sc, brt);
2632                 found = 1;
2633         }
2634
2635         return (found ? 0 : ENOENT);
2636 }
2637
2638 /*
2639  * bridge_rtdelete:
2640  *
2641  *      Delete routes to a speicifc member interface.
2642  */
2643 static void
2644 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2645 {
2646         struct bridge_rtnode *brt, *nbrt;
2647
2648         BRIDGE_LOCK_ASSERT(sc);
2649
2650         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2651                 if (brt->brt_ifp == ifp && (full ||
2652                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2653                         bridge_rtnode_destroy(sc, brt);
2654         }
2655 }
2656
2657 /*
2658  * bridge_rtable_init:
2659  *
2660  *      Initialize the route table for this bridge.
2661  */
2662 static int
2663 bridge_rtable_init(struct bridge_softc *sc)
2664 {
2665         int i;
2666
2667         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2668             M_DEVBUF, M_NOWAIT);
2669         if (sc->sc_rthash == NULL)
2670                 return (ENOMEM);
2671
2672         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2673                 LIST_INIT(&sc->sc_rthash[i]);
2674
2675         sc->sc_rthash_key = arc4random();
2676
2677         LIST_INIT(&sc->sc_rtlist);
2678
2679         return (0);
2680 }
2681
2682 /*
2683  * bridge_rtable_fini:
2684  *
2685  *      Deconstruct the route table for this bridge.
2686  */
2687 static void
2688 bridge_rtable_fini(struct bridge_softc *sc)
2689 {
2690
2691         KASSERT(sc->sc_brtcnt == 0,
2692             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2693         free(sc->sc_rthash, M_DEVBUF);
2694 }
2695
2696 /*
2697  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2698  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2699  */
2700 #define mix(a, b, c)                                                    \
2701 do {                                                                    \
2702         a -= b; a -= c; a ^= (c >> 13);                                 \
2703         b -= c; b -= a; b ^= (a << 8);                                  \
2704         c -= a; c -= b; c ^= (b >> 13);                                 \
2705         a -= b; a -= c; a ^= (c >> 12);                                 \
2706         b -= c; b -= a; b ^= (a << 16);                                 \
2707         c -= a; c -= b; c ^= (b >> 5);                                  \
2708         a -= b; a -= c; a ^= (c >> 3);                                  \
2709         b -= c; b -= a; b ^= (a << 10);                                 \
2710         c -= a; c -= b; c ^= (b >> 15);                                 \
2711 } while (/*CONSTCOND*/0)
2712
2713 static __inline uint32_t
2714 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2715 {
2716         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2717
2718         b += addr[5] << 8;
2719         b += addr[4];
2720         a += addr[3] << 24;
2721         a += addr[2] << 16;
2722         a += addr[1] << 8;
2723         a += addr[0];
2724
2725         mix(a, b, c);
2726
2727         return (c & BRIDGE_RTHASH_MASK);
2728 }
2729
2730 #undef mix
2731
2732 static int
2733 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2734 {
2735         int i, d;
2736
2737         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2738                 d = ((int)a[i]) - ((int)b[i]);
2739         }
2740
2741         return (d);
2742 }
2743
2744 /*
2745  * bridge_rtnode_lookup:
2746  *
2747  *      Look up a bridge route node for the specified destination. Compare the
2748  *      vlan id or if zero then just return the first match.
2749  */
2750 static struct bridge_rtnode *
2751 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2752 {
2753         struct bridge_rtnode *brt;
2754         uint32_t hash;
2755         int dir;
2756
2757         BRIDGE_LOCK_ASSERT(sc);
2758
2759         hash = bridge_rthash(sc, addr);
2760         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2761                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2762                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2763                         return (brt);
2764                 if (dir > 0)
2765                         return (NULL);
2766         }
2767
2768         return (NULL);
2769 }
2770
2771 /*
2772  * bridge_rtnode_insert:
2773  *
2774  *      Insert the specified bridge node into the route table.  We
2775  *      assume the entry is not already in the table.
2776  */
2777 static int
2778 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2779 {
2780         struct bridge_rtnode *lbrt;
2781         uint32_t hash;
2782         int dir;
2783
2784         BRIDGE_LOCK_ASSERT(sc);
2785
2786         hash = bridge_rthash(sc, brt->brt_addr);
2787
2788         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2789         if (lbrt == NULL) {
2790                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2791                 goto out;
2792         }
2793
2794         do {
2795                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2796                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2797                         return (EEXIST);
2798                 if (dir > 0) {
2799                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2800                         goto out;
2801                 }
2802                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2803                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2804                         goto out;
2805                 }
2806                 lbrt = LIST_NEXT(lbrt, brt_hash);
2807         } while (lbrt != NULL);
2808
2809 #ifdef DIAGNOSTIC
2810         panic("bridge_rtnode_insert: impossible");
2811 #endif
2812
2813 out:
2814         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2815         sc->sc_brtcnt++;
2816
2817         return (0);
2818 }
2819
2820 /*
2821  * bridge_rtnode_destroy:
2822  *
2823  *      Destroy a bridge rtnode.
2824  */
2825 static void
2826 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2827 {
2828         BRIDGE_LOCK_ASSERT(sc);
2829
2830         LIST_REMOVE(brt, brt_hash);
2831
2832         LIST_REMOVE(brt, brt_list);
2833         sc->sc_brtcnt--;
2834         brt->brt_dst->bif_addrcnt--;
2835         uma_zfree(bridge_rtnode_zone, brt);
2836 }
2837
2838 /*
2839  * bridge_rtable_expire:
2840  *
2841  *      Set the expiry time for all routes on an interface.
2842  */
2843 static void
2844 bridge_rtable_expire(struct ifnet *ifp, int age)
2845 {
2846         struct bridge_softc *sc = ifp->if_bridge;
2847         struct bridge_rtnode *brt;
2848
2849         BRIDGE_LOCK(sc);
2850
2851         /*
2852          * If the age is zero then flush, otherwise set all the expiry times to
2853          * age for the interface
2854          */
2855         if (age == 0)
2856                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2857         else {
2858                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2859                         /* Cap the expiry time to 'age' */
2860                         if (brt->brt_ifp == ifp &&
2861                             brt->brt_expire > time_uptime + age &&
2862                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2863                                 brt->brt_expire = time_uptime + age;
2864                 }
2865         }
2866         BRIDGE_UNLOCK(sc);
2867 }
2868
2869 /*
2870  * bridge_state_change:
2871  *
2872  *      Callback from the bridgestp code when a port changes states.
2873  */
2874 static void
2875 bridge_state_change(struct ifnet *ifp, int state)
2876 {
2877         struct bridge_softc *sc = ifp->if_bridge;
2878         static const char *stpstates[] = {
2879                 "disabled",
2880                 "listening",
2881                 "learning",
2882                 "forwarding",
2883                 "blocking",
2884                 "discarding"
2885         };
2886
2887         if (log_stp)
2888                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2889                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2890 }
2891
2892 /*
2893  * Send bridge packets through pfil if they are one of the types pfil can deal
2894  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2895  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2896  * that interface.
2897  */
2898 static int
2899 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2900 {
2901         int snap, error, i, hlen;
2902         struct ether_header *eh1, eh2;
2903         struct ip_fw_args args;
2904         struct ip *ip;
2905         struct llc llc1;
2906         u_int16_t ether_type;
2907
2908         snap = 0;
2909         error = -1;     /* Default error if not error == 0 */
2910
2911 #if 0
2912         /* we may return with the IP fields swapped, ensure its not shared */
2913         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2914 #endif
2915
2916         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2917                 return (0); /* filtering is disabled */
2918
2919         i = min((*mp)->m_pkthdr.len, max_protohdr);
2920         if ((*mp)->m_len < i) {
2921             *mp = m_pullup(*mp, i);
2922             if (*mp == NULL) {
2923                 printf("%s: m_pullup failed\n", __func__);
2924                 return (-1);
2925             }
2926         }
2927
2928         eh1 = mtod(*mp, struct ether_header *);
2929         ether_type = ntohs(eh1->ether_type);
2930
2931         /*
2932          * Check for SNAP/LLC.
2933          */
2934         if (ether_type < ETHERMTU) {
2935                 struct llc *llc2 = (struct llc *)(eh1 + 1);
2936
2937                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2938                     llc2->llc_dsap == LLC_SNAP_LSAP &&
2939                     llc2->llc_ssap == LLC_SNAP_LSAP &&
2940                     llc2->llc_control == LLC_UI) {
2941                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
2942                         snap = 1;
2943                 }
2944         }
2945
2946         /*
2947          * If we're trying to filter bridge traffic, don't look at anything
2948          * other than IP and ARP traffic.  If the filter doesn't understand
2949          * IPv6, don't allow IPv6 through the bridge either.  This is lame
2950          * since if we really wanted, say, an AppleTalk filter, we are hosed,
2951          * but of course we don't have an AppleTalk filter to begin with.
2952          * (Note that since pfil doesn't understand ARP it will pass *ALL*
2953          * ARP traffic.)
2954          */
2955         switch (ether_type) {
2956                 case ETHERTYPE_ARP:
2957                 case ETHERTYPE_REVARP:
2958                         if (pfil_ipfw_arp == 0)
2959                                 return (0); /* Automatically pass */
2960                         break;
2961
2962                 case ETHERTYPE_IP:
2963 #ifdef INET6
2964                 case ETHERTYPE_IPV6:
2965 #endif /* INET6 */
2966                         break;
2967                 default:
2968                         /*
2969                          * Check to see if the user wants to pass non-ip
2970                          * packets, these will not be checked by pfil(9) and
2971                          * passed unconditionally so the default is to drop.
2972                          */
2973                         if (pfil_onlyip)
2974                                 goto bad;
2975         }
2976
2977         /* Strip off the Ethernet header and keep a copy. */
2978         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
2979         m_adj(*mp, ETHER_HDR_LEN);
2980
2981         /* Strip off snap header, if present */
2982         if (snap) {
2983                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
2984                 m_adj(*mp, sizeof(struct llc));
2985         }
2986
2987         /*
2988          * Check the IP header for alignment and errors
2989          */
2990         if (dir == PFIL_IN) {
2991                 switch (ether_type) {
2992                         case ETHERTYPE_IP:
2993                                 error = bridge_ip_checkbasic(mp);
2994                                 break;
2995 #ifdef INET6
2996                         case ETHERTYPE_IPV6:
2997                                 error = bridge_ip6_checkbasic(mp);
2998                                 break;
2999 #endif /* INET6 */
3000                         default:
3001                                 error = 0;
3002                 }
3003                 if (error)
3004                         goto bad;
3005         }
3006
3007         if (IPFW_LOADED && pfil_ipfw != 0 && dir == PFIL_OUT && ifp != NULL) {
3008                 error = -1;
3009                 args.rule = ip_dn_claim_rule(*mp);
3010                 if (args.rule != NULL && fw_one_pass)
3011                         goto ipfwpass; /* packet already partially processed */
3012
3013                 args.m = *mp;
3014                 args.oif = ifp;
3015                 args.next_hop = NULL;
3016                 args.eh = &eh2;
3017                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3018                 i = ip_fw_chk_ptr(&args);
3019                 *mp = args.m;
3020
3021                 if (*mp == NULL)
3022                         return (error);
3023
3024                 if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
3025
3026                         /* put the Ethernet header back on */
3027                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3028                         if (*mp == NULL)
3029                                 return (error);
3030                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3031
3032                         /*
3033                          * Pass the pkt to dummynet, which consumes it. The
3034                          * packet will return to us via bridge_dummynet().
3035                          */
3036                         args.oif = ifp;
3037                         ip_dn_io_ptr(mp, DN_TO_IFB_FWD, &args);
3038                         return (error);
3039                 }
3040
3041                 if (i != IP_FW_PASS) /* drop */
3042                         goto bad;
3043         }
3044
3045 ipfwpass:
3046         error = 0;
3047
3048         /*
3049          * Run the packet through pfil
3050          */
3051         switch (ether_type) {
3052         case ETHERTYPE_IP:
3053                 /*
3054                  * before calling the firewall, swap fields the same as
3055                  * IP does. here we assume the header is contiguous
3056                  */
3057                 ip = mtod(*mp, struct ip *);
3058
3059                 ip->ip_len = ntohs(ip->ip_len);
3060                 ip->ip_off = ntohs(ip->ip_off);
3061
3062                 /*
3063                  * Run pfil on the member interface and the bridge, both can
3064                  * be skipped by clearing pfil_member or pfil_bridge.
3065                  *
3066                  * Keep the order:
3067                  *   in_if -> bridge_if -> out_if
3068                  */
3069                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3070                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3071                                         dir, NULL);
3072
3073                 if (*mp == NULL || error != 0) /* filter may consume */
3074                         break;
3075
3076                 if (pfil_member && ifp != NULL)
3077                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp,
3078                                         dir, NULL);
3079
3080                 if (*mp == NULL || error != 0) /* filter may consume */
3081                         break;
3082
3083                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3084                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3085                                         dir, NULL);
3086
3087                 if (*mp == NULL || error != 0) /* filter may consume */
3088                         break;
3089
3090                 /* check if we need to fragment the packet */
3091                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3092                         i = (*mp)->m_pkthdr.len;
3093                         if (i > ifp->if_mtu) {
3094                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3095                                             &llc1);
3096                                 return (error);
3097                         }
3098                 }
3099
3100                 /* Recalculate the ip checksum and restore byte ordering */
3101                 ip = mtod(*mp, struct ip *);
3102                 hlen = ip->ip_hl << 2;
3103                 if (hlen < sizeof(struct ip))
3104                         goto bad;
3105                 if (hlen > (*mp)->m_len) {
3106                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3107                                 goto bad;
3108                         ip = mtod(*mp, struct ip *);
3109                         if (ip == NULL)
3110                                 goto bad;
3111                 }
3112                 ip->ip_len = htons(ip->ip_len);
3113                 ip->ip_off = htons(ip->ip_off);
3114                 ip->ip_sum = 0;
3115                 if (hlen == sizeof(struct ip))
3116                         ip->ip_sum = in_cksum_hdr(ip);
3117                 else
3118                         ip->ip_sum = in_cksum(*mp, hlen);
3119
3120                 break;
3121 #ifdef INET6
3122         case ETHERTYPE_IPV6:
3123                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3124                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3125                                         dir, NULL);
3126
3127                 if (*mp == NULL || error != 0) /* filter may consume */
3128                         break;
3129
3130                 if (pfil_member && ifp != NULL)
3131                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3132                                         dir, NULL);
3133
3134                 if (*mp == NULL || error != 0) /* filter may consume */
3135                         break;
3136
3137                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3138                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3139                                         dir, NULL);
3140                 break;
3141 #endif
3142         default:
3143                 error = 0;
3144                 break;
3145         }
3146
3147         if (*mp == NULL)
3148                 return (error);
3149         if (error != 0)
3150                 goto bad;
3151
3152         error = -1;
3153
3154         /*
3155          * Finally, put everything back the way it was and return
3156          */
3157         if (snap) {
3158                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3159                 if (*mp == NULL)
3160                         return (error);
3161                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3162         }
3163
3164         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3165         if (*mp == NULL)
3166                 return (error);
3167         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3168
3169         return (0);
3170
3171 bad:
3172         m_freem(*mp);
3173         *mp = NULL;
3174         return (error);
3175 }
3176
3177 /*
3178  * Perform basic checks on header size since
3179  * pfil assumes ip_input has already processed
3180  * it for it.  Cut-and-pasted from ip_input.c.
3181  * Given how simple the IPv6 version is,
3182  * does the IPv4 version really need to be
3183  * this complicated?
3184  *
3185  * XXX Should we update ipstat here, or not?
3186  * XXX Right now we update ipstat but not
3187  * XXX csum_counter.
3188  */
3189 static int
3190 bridge_ip_checkbasic(struct mbuf **mp)
3191 {
3192         struct mbuf *m = *mp;
3193         struct ip *ip;
3194         int len, hlen;
3195         u_short sum;
3196
3197         if (*mp == NULL)
3198                 return (-1);
3199
3200         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3201                 if ((m = m_copyup(m, sizeof(struct ip),
3202                         (max_linkhdr + 3) & ~3)) == NULL) {
3203                         /* XXXJRT new stat, please */
3204                         ipstat.ips_toosmall++;
3205                         goto bad;
3206                 }
3207         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3208                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3209                         ipstat.ips_toosmall++;
3210                         goto bad;
3211                 }
3212         }
3213         ip = mtod(m, struct ip *);
3214         if (ip == NULL) goto bad;
3215
3216         if (ip->ip_v != IPVERSION) {
3217                 ipstat.ips_badvers++;
3218                 goto bad;
3219         }
3220         hlen = ip->ip_hl << 2;
3221         if (hlen < sizeof(struct ip)) { /* minimum header length */
3222                 ipstat.ips_badhlen++;
3223                 goto bad;
3224         }
3225         if (hlen > m->m_len) {
3226                 if ((m = m_pullup(m, hlen)) == 0) {
3227                         ipstat.ips_badhlen++;
3228                         goto bad;
3229                 }
3230                 ip = mtod(m, struct ip *);
3231                 if (ip == NULL) goto bad;
3232         }
3233
3234         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3235                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3236         } else {
3237                 if (hlen == sizeof(struct ip)) {
3238                         sum = in_cksum_hdr(ip);
3239                 } else {
3240                         sum = in_cksum(m, hlen);
3241                 }
3242         }
3243         if (sum) {
3244                 ipstat.ips_badsum++;
3245                 goto bad;
3246         }
3247
3248         /* Retrieve the packet length. */
3249         len = ntohs(ip->ip_len);
3250
3251         /*
3252          * Check for additional length bogosity
3253          */
3254         if (len < hlen) {
3255                 ipstat.ips_badlen++;
3256                 goto bad;
3257         }
3258
3259         /*
3260          * Check that the amount of data in the buffers
3261          * is as at least much as the IP header would have us expect.
3262          * Drop packet if shorter than we expect.
3263          */
3264         if (m->m_pkthdr.len < len) {
3265                 ipstat.ips_tooshort++;
3266                 goto bad;
3267         }
3268
3269         /* Checks out, proceed */
3270         *mp = m;
3271         return (0);
3272
3273 bad:
3274         *mp = m;
3275         return (-1);
3276 }
3277
3278 #ifdef INET6
3279 /*
3280  * Same as above, but for IPv6.
3281  * Cut-and-pasted from ip6_input.c.
3282  * XXX Should we update ip6stat, or not?
3283  */
3284 static int
3285 bridge_ip6_checkbasic(struct mbuf **mp)
3286 {
3287         struct mbuf *m = *mp;
3288         struct ip6_hdr *ip6;
3289
3290         /*
3291          * If the IPv6 header is not aligned, slurp it up into a new
3292          * mbuf with space for link headers, in the event we forward
3293          * it.  Otherwise, if it is aligned, make sure the entire base
3294          * IPv6 header is in the first mbuf of the chain.
3295          */
3296         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3297                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3298                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3299                             (max_linkhdr + 3) & ~3)) == NULL) {
3300                         /* XXXJRT new stat, please */
3301                         ip6stat.ip6s_toosmall++;
3302                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3303                         goto bad;
3304                 }
3305         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3306                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3307                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3308                         ip6stat.ip6s_toosmall++;
3309                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3310                         goto bad;
3311                 }
3312         }
3313
3314         ip6 = mtod(m, struct ip6_hdr *);
3315
3316         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3317                 ip6stat.ip6s_badvers++;
3318                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3319                 goto bad;
3320         }
3321
3322         /* Checks out, proceed */
3323         *mp = m;
3324         return (0);
3325
3326 bad:
3327         *mp = m;
3328         return (-1);
3329 }
3330 #endif /* INET6 */
3331
3332 /*
3333  * bridge_fragment:
3334  *
3335  *      Return a fragmented mbuf chain.
3336  */
3337 static int
3338 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3339     int snap, struct llc *llc)
3340 {
3341         struct mbuf *m0;
3342         struct ip *ip;
3343         int error = -1;
3344
3345         if (m->m_len < sizeof(struct ip) &&
3346             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3347                 goto out;
3348         ip = mtod(m, struct ip *);
3349
3350         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3351                     CSUM_DELAY_IP);
3352         if (error)
3353                 goto out;
3354
3355         /* walk the chain and re-add the Ethernet header */
3356         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3357                 if (error == 0) {
3358                         if (snap) {
3359                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3360                                 if (m0 == NULL) {
3361                                         error = ENOBUFS;
3362                                         continue;
3363                                 }
3364                                 bcopy(llc, mtod(m0, caddr_t),
3365                                     sizeof(struct llc));
3366                         }
3367                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3368                         if (m0 == NULL) {
3369                                 error = ENOBUFS;
3370                                 continue;
3371                         }
3372                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3373                 } else
3374                         m_freem(m);
3375         }
3376
3377         if (error == 0)
3378                 ipstat.ips_fragmented++;
3379
3380         return (error);
3381
3382 out:
3383         if (m != NULL)
3384                 m_freem(m);
3385         return (error);
3386 }