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 #include <sys/rwlock.h>
 104
 105 #include <net/bpf.h>
 106 #include <net/if.h>
 107 #include <net/if_clone.h>
 108 #include <net/if_dl.h>
 109 #include <net/if_types.h>
 110 #include <net/if_var.h>
 111 #include <net/pfil.h>
 112
 113 #include <netinet/in.h> /* for struct arpcom */
 114 #include <netinet/in_systm.h>
 115 #include <netinet/in_var.h>
 116 #include <netinet/ip.h>
 117 #include <netinet/ip_var.h>
 118 #ifdef INET6
 119 #include <netinet/ip6.h>
 120 #include <netinet6/ip6_var.h>
 121 #endif
 122 #if defined(INET) || defined(INET6)
 123 #ifdef DEV_CARP
 124 #include <netinet/ip_carp.h>
 125 #endif
 126 #endif
 127 #include <machine/in_cksum.h>
 128 #include <netinet/if_ether.h> /* for struct arpcom */
 129 #include <net/bridgestp.h>
 130 #include <net/if_bridgevar.h>
 131 #include <net/if_llc.h>
 132 #include <net/if_vlan_var.h>
 133
 134 #include <net/route.h>
 135 #include <netinet/ip_fw.h>
 136 #include <netinet/ip_dummynet.h>
 137
 138 /*
 139  * Size of the route hash table.  Must be a power of two.
 140  */
 141 #ifndef BRIDGE_RTHASH_SIZE
 142 #define BRIDGE_RTHASH_SIZE              1024
 143 #endif
 144
 145 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 146
 147 /*
 148  * Maximum number of addresses to cache.
 149  */
 150 #ifndef BRIDGE_RTABLE_MAX
 151 #define BRIDGE_RTABLE_MAX               100
 152 #endif
 153
 154 /*
 155  * Timeout (in seconds) for entries learned dynamically.
 156  */
 157 #ifndef BRIDGE_RTABLE_TIMEOUT
 158 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 159 #endif
 160
 161 /*
 162  * Number of seconds between walks of the route list.
 163  */
 164 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 165 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 166 #endif
 167
 168 /*
 169  * List of capabilities to possibly mask on the member interface.
 170  */
 171 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 172
 173 /*
 174  * Bridge interface list entry.
 175  */
 176 struct bridge_iflist {
 177         LIST_ENTRY(bridge_iflist) bif_next;
 178         struct ifnet            *bif_ifp;       /* member if */
 179         struct bstp_port        bif_stp;        /* STP state */
 180         uint32_t                bif_flags;      /* member if flags */
 181         int                     bif_savedcaps;  /* saved capabilities */
 182         uint32_t                bif_addrmax;    /* max # of addresses */
 183         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 184         uint32_t                bif_addrexceeded;/* # of address violations */
 185 };
 186
 187 /*
 188  * Bridge route node.
 189  */
 190 struct bridge_rtnode {
 191         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 192         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 193         struct bridge_iflist    *brt_dst;       /* destination if */
 194         unsigned long           brt_expire;     /* expiration time */
 195         uint8_t                 brt_flags;      /* address flags */
 196         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 197         uint16_t                brt_vlan;       /* vlan id */
 198 };
 199 #define brt_ifp                 brt_dst->bif_ifp
 200
 201 /*
 202  * Software state for each bridge.
 203  */
 204 struct bridge_softc {
 205         struct ifnet            *sc_ifp;        /* make this an interface */
 206         LIST_ENTRY(bridge_softc) sc_list;
 207         struct mtx              sc_mtx;
 208         struct cv               sc_cv;
 209         uint32_t                sc_brtmax;      /* max # of addresses */
 210         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 211         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 212         struct callout          sc_brcallout;   /* bridge callout */
 213         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 214         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 215         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 216         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 217         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 218         uint32_t                sc_rthash_key;  /* key for hash */
 219         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 220         struct bstp_state       sc_stp;         /* STP state */
 221         uint32_t                sc_brtexceeded; /* # of cache drops */
 222         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 223         u_char                  sc_defaddr[6];  /* Default MAC address */
 224 };
 225
 226 static struct mtx       bridge_list_mtx;
 227 eventhandler_tag        bridge_detach_cookie = NULL;
 228
 229 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 230
 231 uma_zone_t bridge_rtnode_zone;
 232
 233 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 234 static void     bridge_clone_destroy(struct ifnet *);
 235
 236 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 237 static void     bridge_mutecaps(struct bridge_softc *);
 238 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 239                     int);
 240 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 241 static void     bridge_init(void *);
 242 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 243 static void     bridge_stop(struct ifnet *, int);
 244 static void     bridge_start(struct ifnet *);
 245 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 246 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 247                     struct rtentry *);
 248 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 249                     struct mbuf *);
 250 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 251
 252 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 253                     struct mbuf *m);
 254
 255 static void     bridge_timer(void *);
 256
 257 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 258                     struct mbuf *, int);
 259 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 260
 261 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 262                     uint16_t, struct bridge_iflist *, int, uint8_t);
 263 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 264                     uint16_t);
 265 static void     bridge_rttrim(struct bridge_softc *);
 266 static void     bridge_rtage(struct bridge_softc *);
 267 static void     bridge_rtflush(struct bridge_softc *, int);
 268 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 269                     uint16_t);
 270
 271 static int      bridge_rtable_init(struct bridge_softc *);
 272 static void     bridge_rtable_fini(struct bridge_softc *);
 273
 274 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 275 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 276                     const uint8_t *, uint16_t);
 277 static int      bridge_rtnode_insert(struct bridge_softc *,
 278                     struct bridge_rtnode *);
 279 static void     bridge_rtnode_destroy(struct bridge_softc *,
 280                     struct bridge_rtnode *);
 281 static void     bridge_rtable_expire(struct ifnet *, int);
 282 static void     bridge_state_change(struct ifnet *, int);
 283
 284 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 285                     const char *name);
 286 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 287                     struct ifnet *ifp);
 288 static void     bridge_delete_member(struct bridge_softc *,
 289                     struct bridge_iflist *, int);
 290 static void     bridge_delete_span(struct bridge_softc *,
 291                     struct bridge_iflist *);
 292
 293 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 294 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 295 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 296 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 297 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 298 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 324 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 325                     int);
 326 static int      bridge_ip_checkbasic(struct mbuf **mp);
 327 #ifdef INET6
 328 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 329 #endif /* INET6 */
 330 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 331                     struct ether_header *, int, struct llc *);
 332
 333 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 334 #define VLANTAGOF(_m)   \
 335     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 336
 337 static struct bstp_cb_ops bridge_ops = {
 338         .bcb_state = bridge_state_change,
 339         .bcb_rtage = bridge_rtable_expire
 340 };
 341
 342 SYSCTL_DECL(_net_link);
 343 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 344
 345 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 346 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 347 static int pfil_member = 1; /* run pfil hooks on the member interface */
 348 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 349 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 350 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 351                                    locally destined packets */
 352 static int log_stp   = 0;   /* log STP state changes */
 353 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
 354 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 355     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 356 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 357     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 358 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 359     &pfil_bridge, 0, "Packet filter on the bridge interface");
 360 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 361     &pfil_member, 0, "Packet filter on the member interface");
 362 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 363     &pfil_local_phys, 0,
 364     "Packet filter on the physical interface for locally destined packets");
 365 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 366     &log_stp, 0, "Log STP state changes");
 367 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
 368     &bridge_inherit_mac, 0,
 369     "Inherit MAC address from the first bridge member");
 370
 371 struct bridge_control {
 372         int     (*bc_func)(struct bridge_softc *, void *);
 373         int     bc_argsize;
 374         int     bc_flags;
 375 };
 376
 377 #define BC_F_COPYIN             0x01    /* copy arguments in */
 378 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 379 #define BC_F_SUSER              0x04    /* do super-user check */
 380
 381 const struct bridge_control bridge_control_table[] = {
 382         { bridge_ioctl_add,             sizeof(struct ifbreq),
 383           BC_F_COPYIN|BC_F_SUSER },
 384         { bridge_ioctl_del,             sizeof(struct ifbreq),
 385           BC_F_COPYIN|BC_F_SUSER },
 386
 387         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 388           BC_F_COPYIN|BC_F_COPYOUT },
 389         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 390           BC_F_COPYIN|BC_F_SUSER },
 391
 392         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 393           BC_F_COPYIN|BC_F_SUSER },
 394         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 395           BC_F_COPYOUT },
 396
 397         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 398           BC_F_COPYIN|BC_F_COPYOUT },
 399         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 400           BC_F_COPYIN|BC_F_COPYOUT },
 401
 402         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 403           BC_F_COPYIN|BC_F_SUSER },
 404
 405         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 406           BC_F_COPYIN|BC_F_SUSER },
 407         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 408           BC_F_COPYOUT },
 409
 410         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 411           BC_F_COPYIN|BC_F_SUSER },
 412
 413         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 414           BC_F_COPYIN|BC_F_SUSER },
 415
 416         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 417           BC_F_COPYOUT },
 418         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 419           BC_F_COPYIN|BC_F_SUSER },
 420
 421         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 422           BC_F_COPYOUT },
 423         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 424           BC_F_COPYIN|BC_F_SUSER },
 425
 426         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 427           BC_F_COPYOUT },
 428         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 429           BC_F_COPYIN|BC_F_SUSER },
 430
 431         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 432           BC_F_COPYOUT },
 433         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 434           BC_F_COPYIN|BC_F_SUSER },
 435
 436         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 437           BC_F_COPYIN|BC_F_SUSER },
 438
 439         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 440           BC_F_COPYIN|BC_F_SUSER },
 441
 442         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 443           BC_F_COPYIN|BC_F_SUSER },
 444         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 445           BC_F_COPYIN|BC_F_SUSER },
 446
 447         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 448           BC_F_COPYOUT },
 449
 450         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 451           BC_F_COPYOUT },
 452
 453         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 454           BC_F_COPYIN|BC_F_COPYOUT },
 455
 456         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 457           BC_F_COPYIN|BC_F_SUSER },
 458
 459         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 460           BC_F_COPYIN|BC_F_SUSER },
 461
 462         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 463           BC_F_COPYIN|BC_F_SUSER },
 464
 465 };
 466 const int bridge_control_table_size =
 467     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 468
 469 LIST_HEAD(, bridge_softc) bridge_list;
 470
 471 IFC_SIMPLE_DECLARE(bridge, 0);
 472
 473 static int
 474 bridge_modevent(module_t mod, int type, void *data)
 475 {
 476
 477         switch (type) {
 478         case MOD_LOAD:
 479                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 480                 if_clone_attach(&bridge_cloner);
 481                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 482                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 483                     UMA_ALIGN_PTR, 0);
 484                 LIST_INIT(&bridge_list);
 485                 bridge_input_p = bridge_input;
 486                 bridge_output_p = bridge_output;
 487                 bridge_dn_p = bridge_dummynet;
 488                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 489                     ifnet_departure_event, bridge_ifdetach, NULL,
 490                     EVENTHANDLER_PRI_ANY);
 491                 break;
 492         case MOD_UNLOAD:
 493                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 494                     bridge_detach_cookie);
 495                 if_clone_detach(&bridge_cloner);
 496                 uma_zdestroy(bridge_rtnode_zone);
 497                 bridge_input_p = NULL;
 498                 bridge_output_p = NULL;
 499                 bridge_dn_p = NULL;
 500                 mtx_destroy(&bridge_list_mtx);
 501                 break;
 502         default:
 503                 return (EOPNOTSUPP);
 504         }
 505         return (0);
 506 }
 507
 508 static moduledata_t bridge_mod = {
 509         "if_bridge",
 510         bridge_modevent,
 511         0
 512 };
 513
 514 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 515 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 516
 517 /*
 518  * handler for net.link.bridge.pfil_ipfw
 519  */
 520 static int
 521 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 522 {
 523         int enable = pfil_ipfw;
 524         int error;
 525
 526         error = sysctl_handle_int(oidp, &enable, 0, req);
 527         enable = (enable) ? 1 : 0;
 528
 529         if (enable != pfil_ipfw) {
 530                 pfil_ipfw = enable;
 531
 532                 /*
 533                  * Disable pfil so that ipfw doesnt run twice, if the user
 534                  * really wants both then they can re-enable pfil_bridge and/or
 535                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 536                  * layer2 type.
 537                  */
 538                 if (pfil_ipfw) {
 539                         pfil_onlyip = 0;
 540                         pfil_bridge = 0;
 541                         pfil_member = 0;
 542                 }
 543         }
 544
 545         return (error);
 546 }
 547 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 548             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 549
 550 /*
 551  * bridge_clone_create:
 552  *
 553  *      Create a new bridge instance.
 554  */
 555 static int
 556 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 557 {
 558         struct bridge_softc *sc, *sc2;
 559         struct ifnet *bifp, *ifp;
 560         int retry;
 561
 562         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 563         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 564         if (ifp == NULL) {
 565                 free(sc, M_DEVBUF);
 566                 return (ENOSPC);
 567         }
 568
 569         BRIDGE_LOCK_INIT(sc);
 570         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 571         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 572
 573         /* Initialize our routing table. */
 574         bridge_rtable_init(sc);
 575
 576         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 577
 578         LIST_INIT(&sc->sc_iflist);
 579         LIST_INIT(&sc->sc_spanlist);
 580
 581         ifp->if_softc = sc;
 582         if_initname(ifp, ifc->ifc_name, unit);
 583         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 584         ifp->if_ioctl = bridge_ioctl;
 585         ifp->if_start = bridge_start;
 586         ifp->if_init = bridge_init;
 587         ifp->if_type = IFT_BRIDGE;
 588         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 589         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 590         IFQ_SET_READY(&ifp->if_snd);
 591
 592         /*
 593          * Generate a random ethernet address with a locally administered
 594          * address.
 595          *
 596          * Since we are using random ethernet addresses for the bridge, it is
 597          * possible that we might have address collisions, so make sure that
 598          * this hardware address isn't already in use on another bridge.
 599          */
 600         for (retry = 1; retry != 0;) {
 601                 arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 602                 sc->sc_defaddr[0] &= ~1;        /* clear multicast bit */
 603                 sc->sc_defaddr[0] |= 2;         /* set the LAA bit */
 604                 retry = 0;
 605                 mtx_lock(&bridge_list_mtx);
 606                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 607                         bifp = sc2->sc_ifp;
 608                         if (memcmp(sc->sc_defaddr,
 609                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 610                                 retry = 1;
 611                 }
 612                 mtx_unlock(&bridge_list_mtx);
 613         }
 614
 615         bstp_attach(&sc->sc_stp, &bridge_ops);
 616         ether_ifattach(ifp, sc->sc_defaddr);
 617         /* Now undo some of the damage... */
 618         ifp->if_baudrate = 0;
 619         ifp->if_type = IFT_BRIDGE;
 620
 621         mtx_lock(&bridge_list_mtx);
 622         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 623         mtx_unlock(&bridge_list_mtx);
 624
 625         return (0);
 626 }
 627
 628 /*
 629  * bridge_clone_destroy:
 630  *
 631  *      Destroy a bridge instance.
 632  */
 633 static void
 634 bridge_clone_destroy(struct ifnet *ifp)
 635 {
 636         struct bridge_softc *sc = ifp->if_softc;
 637         struct bridge_iflist *bif;
 638
 639         BRIDGE_LOCK(sc);
 640
 641         bridge_stop(ifp, 1);
 642         ifp->if_flags &= ~IFF_UP;
 643
 644         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 645                 bridge_delete_member(sc, bif, 0);
 646
 647         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 648                 bridge_delete_span(sc, bif);
 649         }
 650
 651         BRIDGE_UNLOCK(sc);
 652
 653         callout_drain(&sc->sc_brcallout);
 654
 655         mtx_lock(&bridge_list_mtx);
 656         LIST_REMOVE(sc, sc_list);
 657         mtx_unlock(&bridge_list_mtx);
 658
 659         bstp_detach(&sc->sc_stp);
 660         ether_ifdetach(ifp);
 661         if_free_type(ifp, IFT_ETHER);
 662
 663         /* Tear down the routing table. */
 664         bridge_rtable_fini(sc);
 665
 666         BRIDGE_LOCK_DESTROY(sc);
 667         free(sc, M_DEVBUF);
 668 }
 669
 670 /*
 671  * bridge_ioctl:
 672  *
 673  *      Handle a control request from the operator.
 674  */
 675 static int
 676 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 677 {
 678         struct bridge_softc *sc = ifp->if_softc;
 679         struct thread *td = curthread;
 680         union {
 681                 struct ifbreq ifbreq;
 682                 struct ifbifconf ifbifconf;
 683                 struct ifbareq ifbareq;
 684                 struct ifbaconf ifbaconf;
 685                 struct ifbrparam ifbrparam;
 686                 struct ifbropreq ifbropreq;
 687         } args;
 688         struct ifdrv *ifd = (struct ifdrv *) data;
 689         const struct bridge_control *bc;
 690         int error = 0;
 691
 692         switch (cmd) {
 693
 694         case SIOCADDMULTI:
 695         case SIOCDELMULTI:
 696                 break;
 697
 698         case SIOCGDRVSPEC:
 699         case SIOCSDRVSPEC:
 700                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 701                         error = EINVAL;
 702                         break;
 703                 }
 704                 bc = &bridge_control_table[ifd->ifd_cmd];
 705
 706                 if (cmd == SIOCGDRVSPEC &&
 707                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 708                         error = EINVAL;
 709                         break;
 710                 }
 711                 else if (cmd == SIOCSDRVSPEC &&
 712                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 713                         error = EINVAL;
 714                         break;
 715                 }
 716
 717                 if (bc->bc_flags & BC_F_SUSER) {
 718                         error = priv_check(td, PRIV_NET_BRIDGE);
 719                         if (error)
 720                                 break;
 721                 }
 722
 723                 if (ifd->ifd_len != bc->bc_argsize ||
 724                     ifd->ifd_len > sizeof(args)) {
 725                         error = EINVAL;
 726                         break;
 727                 }
 728
 729                 bzero(&args, sizeof(args));
 730                 if (bc->bc_flags & BC_F_COPYIN) {
 731                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 732                         if (error)
 733                                 break;
 734                 }
 735
 736                 BRIDGE_LOCK(sc);
 737                 error = (*bc->bc_func)(sc, &args);
 738                 BRIDGE_UNLOCK(sc);
 739                 if (error)
 740                         break;
 741
 742                 if (bc->bc_flags & BC_F_COPYOUT)
 743                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 744
 745                 break;
 746
 747         case SIOCSIFFLAGS:
 748                 if (!(ifp->if_flags & IFF_UP) &&
 749                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 750                         /*
 751                          * If interface is marked down and it is running,
 752                          * then stop and disable it.
 753                          */
 754                         BRIDGE_LOCK(sc);
 755                         bridge_stop(ifp, 1);
 756                         BRIDGE_UNLOCK(sc);
 757                 } else if ((ifp->if_flags & IFF_UP) &&
 758                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 759                         /*
 760                          * If interface is marked up and it is stopped, then
 761                          * start it.
 762                          */
 763                         (*ifp->if_init)(sc);
 764                 }
 765                 break;
 766
 767         case SIOCSIFMTU:
 768                 /* Do not allow the MTU to be changed on the bridge */
 769                 error = EINVAL;
 770                 break;
 771
 772         default:
 773                 /*
 774                  * drop the lock as ether_ioctl() will call bridge_start() and
 775                  * cause the lock to be recursed.
 776                  */
 777                 error = ether_ioctl(ifp, cmd, data);
 778                 break;
 779         }
 780
 781         return (error);
 782 }
 783
 784 /*
 785  * bridge_mutecaps:
 786  *
 787  *      Clear or restore unwanted capabilities on the member interface
 788  */
 789 static void
 790 bridge_mutecaps(struct bridge_softc *sc)
 791 {
 792         struct bridge_iflist *bif;
 793         int enabled, mask;
 794
 795         /* Initial bitmask of capabilities to test */
 796         mask = BRIDGE_IFCAPS_MASK;
 797
 798         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 799                 /* Every member must support it or its disabled */
 800                 mask &= bif->bif_savedcaps;
 801         }
 802
 803         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 804                 enabled = bif->bif_ifp->if_capenable;
 805                 /* strip off mask bits and enable them again if allowed */
 806                 enabled &= ~BRIDGE_IFCAPS_MASK;
 807                 enabled |= mask;
 808                 /*
 809                  * Receive offload can only be enabled if all members also
 810                  * support send offload.
 811                  */
 812                 if ((enabled & IFCAP_TSO) == 0)
 813                         enabled &= ~IFCAP_LRO;
 814
 815                 bridge_set_ifcap(sc, bif, enabled);
 816         }
 817
 818 }
 819
 820 static void
 821 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 822 {
 823         struct ifnet *ifp = bif->bif_ifp;
 824         struct ifreq ifr;
 825         int error;
 826
 827         bzero(&ifr, sizeof(ifr));
 828         ifr.ifr_reqcap = set;
 829
 830         if (ifp->if_capenable != set) {
 831                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 832                 if (error)
 833                         if_printf(sc->sc_ifp,
 834                             "error setting interface capabilities on %s\n",
 835                             ifp->if_xname);
 836         }
 837 }
 838
 839 /*
 840  * bridge_lookup_member:
 841  *
 842  *      Lookup a bridge member interface.
 843  */
 844 static struct bridge_iflist *
 845 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 846 {
 847         struct bridge_iflist *bif;
 848         struct ifnet *ifp;
 849
 850         BRIDGE_LOCK_ASSERT(sc);
 851
 852         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 853                 ifp = bif->bif_ifp;
 854                 if (strcmp(ifp->if_xname, name) == 0)
 855                         return (bif);
 856         }
 857
 858         return (NULL);
 859 }
 860
 861 /*
 862  * bridge_lookup_member_if:
 863  *
 864  *      Lookup a bridge member interface by ifnet*.
 865  */
 866 static struct bridge_iflist *
 867 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 868 {
 869         struct bridge_iflist *bif;
 870
 871         BRIDGE_LOCK_ASSERT(sc);
 872
 873         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 874                 if (bif->bif_ifp == member_ifp)
 875                         return (bif);
 876         }
 877
 878         return (NULL);
 879 }
 880
 881 /*
 882  * bridge_delete_member:
 883  *
 884  *      Delete the specified member interface.
 885  */
 886 static void
 887 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 888     int gone)
 889 {
 890         struct ifnet *ifs = bif->bif_ifp;
 891         struct ifnet *fif = NULL;
 892
 893         BRIDGE_LOCK_ASSERT(sc);
 894
 895         if (bif->bif_flags & IFBIF_STP)
 896                 bstp_disable(&bif->bif_stp);
 897
 898         ifs->if_bridge = NULL;
 899         BRIDGE_XLOCK(sc);
 900         LIST_REMOVE(bif, bif_next);
 901         BRIDGE_XDROP(sc);
 902
 903         /*
 904          * If removing the interface that gave the bridge its mac address, set
 905          * the mac address of the bridge to the address of the next member, or
 906          * to its default address if no members are left.
 907          */
 908         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
 909                 if (LIST_EMPTY(&sc->sc_iflist)) {
 910                         bcopy(sc->sc_defaddr,
 911                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 912                         sc->sc_ifaddr = NULL;
 913                 } else {
 914                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
 915                         bcopy(IF_LLADDR(fif),
 916                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 917                         sc->sc_ifaddr = fif;
 918                 }
 919         }
 920
 921         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 922         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 923         KASSERT(bif->bif_addrcnt == 0,
 924             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 925
 926         BRIDGE_UNLOCK(sc);
 927         if (!gone) {
 928                 switch (ifs->if_type) {
 929                 case IFT_ETHER:
 930                 case IFT_L2VLAN:
 931                         /*
 932                          * Take the interface out of promiscuous mode.
 933                          */
 934                         (void) ifpromisc(ifs, 0);
 935                         break;
 936
 937                 case IFT_GIF:
 938                         break;
 939
 940                 default:
 941 #ifdef DIAGNOSTIC
 942                         panic("bridge_delete_member: impossible");
 943 #endif
 944                         break;
 945                 }
 946                 /* reneable any interface capabilities */
 947                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
 948         }
 949         bstp_destroy(&bif->bif_stp);    /* prepare to free */
 950         BRIDGE_LOCK(sc);
 951         free(bif, M_DEVBUF);
 952 }
 953
 954 /*
 955  * bridge_delete_span:
 956  *
 957  *      Delete the specified span interface.
 958  */
 959 static void
 960 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
 961 {
 962         BRIDGE_LOCK_ASSERT(sc);
 963
 964         KASSERT(bif->bif_ifp->if_bridge == NULL,
 965             ("%s: not a span interface", __func__));
 966
 967         LIST_REMOVE(bif, bif_next);
 968         free(bif, M_DEVBUF);
 969 }
 970
 971 static int
 972 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
 973 {
 974         struct ifbreq *req = arg;
 975         struct bridge_iflist *bif = NULL;
 976         struct ifnet *ifs;
 977         int error = 0;
 978
 979         ifs = ifunit(req->ifbr_ifsname);
 980         if (ifs == NULL)
 981                 return (ENOENT);
 982         if (ifs->if_ioctl == NULL)      /* must be supported */
 983                 return (EINVAL);
 984
 985         /* If it's in the span list, it can't be a member. */
 986         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
 987                 if (ifs == bif->bif_ifp)
 988                         return (EBUSY);
 989
 990         /* Allow the first Ethernet member to define the MTU */
 991         if (ifs->if_type != IFT_GIF) {
 992                 if (LIST_EMPTY(&sc->sc_iflist))
 993                         sc->sc_ifp->if_mtu = ifs->if_mtu;
 994                 else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
 995                         if_printf(sc->sc_ifp, "invalid MTU for %s\n",
 996                             ifs->if_xname);
 997                         return (EINVAL);
 998                 }
 999         }
1000
1001         if (ifs->if_bridge == sc)
1002                 return (EEXIST);
1003
1004         if (ifs->if_bridge != NULL)
1005                 return (EBUSY);
1006
1007         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1008         if (bif == NULL)
1009                 return (ENOMEM);
1010
1011         bif->bif_ifp = ifs;
1012         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1013         bif->bif_savedcaps = ifs->if_capenable;
1014
1015         switch (ifs->if_type) {
1016         case IFT_ETHER:
1017         case IFT_L2VLAN:
1018         case IFT_GIF:
1019                 /* permitted interface types */
1020                 break;
1021         default:
1022                 error = EINVAL;
1023                 goto out;
1024         }
1025
1026         /*
1027          * Assign the interface's MAC address to the bridge if it's the first
1028          * member and the MAC address of the bridge has not been changed from
1029          * the default randomly generated one.
1030          */
1031         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1032             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1033                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1034                 sc->sc_ifaddr = ifs;
1035         }
1036
1037         ifs->if_bridge = sc;
1038         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1039         /*
1040          * XXX: XLOCK HERE!?!
1041          *
1042          * NOTE: insert_***HEAD*** should be safe for the traversals.
1043          */
1044         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1045
1046         /* Set interface capabilities to the intersection set of all members */
1047         bridge_mutecaps(sc);
1048
1049         switch (ifs->if_type) {
1050         case IFT_ETHER:
1051         case IFT_L2VLAN:
1052                 /*
1053                  * Place the interface into promiscuous mode.
1054                  */
1055                 BRIDGE_UNLOCK(sc);
1056                 error = ifpromisc(ifs, 1);
1057                 BRIDGE_LOCK(sc);
1058                 break;
1059         }
1060         if (error)
1061                 bridge_delete_member(sc, bif, 0);
1062 out:
1063         if (error) {
1064                 if (bif != NULL)
1065                         free(bif, M_DEVBUF);
1066         }
1067         return (error);
1068 }
1069
1070 static int
1071 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1072 {
1073         struct ifbreq *req = arg;
1074         struct bridge_iflist *bif;
1075
1076         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1077         if (bif == NULL)
1078                 return (ENOENT);
1079
1080         bridge_delete_member(sc, bif, 0);
1081
1082         return (0);
1083 }
1084
1085 static int
1086 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1087 {
1088         struct ifbreq *req = arg;
1089         struct bridge_iflist *bif;
1090         struct bstp_port *bp;
1091
1092         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1093         if (bif == NULL)
1094                 return (ENOENT);
1095
1096         bp = &bif->bif_stp;
1097         req->ifbr_ifsflags = bif->bif_flags;
1098         req->ifbr_state = bp->bp_state;
1099         req->ifbr_priority = bp->bp_priority;
1100         req->ifbr_path_cost = bp->bp_path_cost;
1101         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1102         req->ifbr_proto = bp->bp_protover;
1103         req->ifbr_role = bp->bp_role;
1104         req->ifbr_stpflags = bp->bp_flags;
1105         req->ifbr_addrcnt = bif->bif_addrcnt;
1106         req->ifbr_addrmax = bif->bif_addrmax;
1107         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1108
1109         /* Copy STP state options as flags */
1110         if (bp->bp_operedge)
1111                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1112         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1113                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1114         if (bp->bp_ptp_link)
1115                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1116         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1117                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1118         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1119                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1120         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1121                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1122         return (0);
1123 }
1124
1125 static int
1126 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1127 {
1128         struct ifbreq *req = arg;
1129         struct bridge_iflist *bif;
1130         struct bstp_port *bp;
1131         int error;
1132
1133         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1134         if (bif == NULL)
1135                 return (ENOENT);
1136         bp = &bif->bif_stp;
1137
1138         if (req->ifbr_ifsflags & IFBIF_SPAN)
1139                 /* SPAN is readonly */
1140                 return (EINVAL);
1141
1142         if (req->ifbr_ifsflags & IFBIF_STP) {
1143                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1144                         error = bstp_enable(&bif->bif_stp);
1145                         if (error)
1146                                 return (error);
1147                 }
1148         } else {
1149                 if ((bif->bif_flags & IFBIF_STP) != 0)
1150                         bstp_disable(&bif->bif_stp);
1151         }
1152
1153         /* Pass on STP flags */
1154         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1155         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1156         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1157         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1158
1159         /* Save the bits relating to the bridge */
1160         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1161
1162         return (0);
1163 }
1164
1165 static int
1166 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1167 {
1168         struct ifbrparam *param = arg;
1169
1170         sc->sc_brtmax = param->ifbrp_csize;
1171         bridge_rttrim(sc);
1172
1173         return (0);
1174 }
1175
1176 static int
1177 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1178 {
1179         struct ifbrparam *param = arg;
1180
1181         param->ifbrp_csize = sc->sc_brtmax;
1182
1183         return (0);
1184 }
1185
1186 static int
1187 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1188 {
1189         struct ifbifconf *bifc = arg;
1190         struct bridge_iflist *bif;
1191         struct ifbreq breq;
1192         char *buf, *outbuf;
1193         int count, buflen, len, error = 0;
1194
1195         count = 0;
1196         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1197                 count++;
1198         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1199                 count++;
1200
1201         buflen = sizeof(breq) * count;
1202         if (bifc->ifbic_len == 0) {
1203                 bifc->ifbic_len = buflen;
1204                 return (0);
1205         }
1206         BRIDGE_UNLOCK(sc);
1207         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1208         BRIDGE_LOCK(sc);
1209
1210         count = 0;
1211         buf = outbuf;
1212         len = min(bifc->ifbic_len, buflen);
1213         bzero(&breq, sizeof(breq));
1214         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1215                 if (len < sizeof(breq))
1216                         break;
1217
1218                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1219                     sizeof(breq.ifbr_ifsname));
1220                 /* Fill in the ifbreq structure */
1221                 error = bridge_ioctl_gifflags(sc, &breq);
1222                 if (error)
1223                         break;
1224                 memcpy(buf, &breq, sizeof(breq));
1225                 count++;
1226                 buf += sizeof(breq);
1227                 len -= sizeof(breq);
1228         }
1229         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1230                 if (len < sizeof(breq))
1231                         break;
1232
1233                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1234                     sizeof(breq.ifbr_ifsname));
1235                 breq.ifbr_ifsflags = bif->bif_flags;
1236                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1237                 memcpy(buf, &breq, sizeof(breq));
1238                 count++;
1239                 buf += sizeof(breq);
1240                 len -= sizeof(breq);
1241         }
1242
1243         BRIDGE_UNLOCK(sc);
1244         bifc->ifbic_len = sizeof(breq) * count;
1245         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1246         BRIDGE_LOCK(sc);
1247         free(outbuf, M_TEMP);
1248         return (error);
1249 }
1250
1251 static int
1252 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1253 {
1254         struct ifbaconf *bac = arg;
1255         struct bridge_rtnode *brt;
1256         struct ifbareq bareq;
1257         char *buf, *outbuf;
1258         int count, buflen, len, error = 0;
1259
1260         if (bac->ifbac_len == 0)
1261                 return (0);
1262
1263         count = 0;
1264         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1265                 count++;
1266         buflen = sizeof(bareq) * count;
1267
1268         BRIDGE_UNLOCK(sc);
1269         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1270         BRIDGE_LOCK(sc);
1271
1272         count = 0;
1273         buf = outbuf;
1274         len = min(bac->ifbac_len, buflen);
1275         bzero(&bareq, sizeof(bareq));
1276         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1277                 if (len < sizeof(bareq))
1278                         goto out;
1279                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1280                     sizeof(bareq.ifba_ifsname));
1281                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1282                 bareq.ifba_vlan = brt->brt_vlan;
1283                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1284                                 time_uptime < brt->brt_expire)
1285                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1286                 else
1287                         bareq.ifba_expire = 0;
1288                 bareq.ifba_flags = brt->brt_flags;
1289
1290                 memcpy(buf, &bareq, sizeof(bareq));
1291                 count++;
1292                 buf += sizeof(bareq);
1293                 len -= sizeof(bareq);
1294         }
1295 out:
1296         BRIDGE_UNLOCK(sc);
1297         bac->ifbac_len = sizeof(bareq) * count;
1298         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1299         BRIDGE_LOCK(sc);
1300         free(outbuf, M_TEMP);
1301         return (error);
1302 }
1303
1304 static int
1305 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1306 {
1307         struct ifbareq *req = arg;
1308         struct bridge_iflist *bif;
1309         int error;
1310
1311         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1312         if (bif == NULL)
1313                 return (ENOENT);
1314
1315         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1316             req->ifba_flags);
1317
1318         return (error);
1319 }
1320
1321 static int
1322 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1323 {
1324         struct ifbrparam *param = arg;
1325
1326         sc->sc_brttimeout = param->ifbrp_ctime;
1327         return (0);
1328 }
1329
1330 static int
1331 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1332 {
1333         struct ifbrparam *param = arg;
1334
1335         param->ifbrp_ctime = sc->sc_brttimeout;
1336         return (0);
1337 }
1338
1339 static int
1340 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1341 {
1342         struct ifbareq *req = arg;
1343
1344         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1345 }
1346
1347 static int
1348 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1349 {
1350         struct ifbreq *req = arg;
1351
1352         bridge_rtflush(sc, req->ifbr_ifsflags);
1353         return (0);
1354 }
1355
1356 static int
1357 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1358 {
1359         struct ifbrparam *param = arg;
1360         struct bstp_state *bs = &sc->sc_stp;
1361
1362         param->ifbrp_prio = bs->bs_bridge_priority;
1363         return (0);
1364 }
1365
1366 static int
1367 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1368 {
1369         struct ifbrparam *param = arg;
1370
1371         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1372 }
1373
1374 static int
1375 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1376 {
1377         struct ifbrparam *param = arg;
1378         struct bstp_state *bs = &sc->sc_stp;
1379
1380         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1381         return (0);
1382 }
1383
1384 static int
1385 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1386 {
1387         struct ifbrparam *param = arg;
1388
1389         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1390 }
1391
1392 static int
1393 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1394 {
1395         struct ifbrparam *param = arg;
1396         struct bstp_state *bs = &sc->sc_stp;
1397
1398         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1399         return (0);
1400 }
1401
1402 static int
1403 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1404 {
1405         struct ifbrparam *param = arg;
1406
1407         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1408 }
1409
1410 static int
1411 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1412 {
1413         struct ifbrparam *param = arg;
1414         struct bstp_state *bs = &sc->sc_stp;
1415
1416         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1417         return (0);
1418 }
1419
1420 static int
1421 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1422 {
1423         struct ifbrparam *param = arg;
1424
1425         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1426 }
1427
1428 static int
1429 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1430 {
1431         struct ifbreq *req = arg;
1432         struct bridge_iflist *bif;
1433
1434         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1435         if (bif == NULL)
1436                 return (ENOENT);
1437
1438         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1439 }
1440
1441 static int
1442 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1443 {
1444         struct ifbreq *req = arg;
1445         struct bridge_iflist *bif;
1446
1447         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1448         if (bif == NULL)
1449                 return (ENOENT);
1450
1451         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1452 }
1453
1454 static int
1455 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1456 {
1457         struct ifbreq *req = arg;
1458         struct bridge_iflist *bif;
1459
1460         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1461         if (bif == NULL)
1462                 return (ENOENT);
1463
1464         bif->bif_addrmax = req->ifbr_addrmax;
1465         return (0);
1466 }
1467
1468 static int
1469 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1470 {
1471         struct ifbreq *req = arg;
1472         struct bridge_iflist *bif = NULL;
1473         struct ifnet *ifs;
1474
1475         ifs = ifunit(req->ifbr_ifsname);
1476         if (ifs == NULL)
1477                 return (ENOENT);
1478
1479         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1480                 if (ifs == bif->bif_ifp)
1481                         return (EBUSY);
1482
1483         if (ifs->if_bridge != NULL)
1484                 return (EBUSY);
1485
1486         switch (ifs->if_type) {
1487                 case IFT_ETHER:
1488                 case IFT_GIF:
1489                 case IFT_L2VLAN:
1490                         break;
1491                 default:
1492                         return (EINVAL);
1493         }
1494
1495         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1496         if (bif == NULL)
1497                 return (ENOMEM);
1498
1499         bif->bif_ifp = ifs;
1500         bif->bif_flags = IFBIF_SPAN;
1501
1502         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1503
1504         return (0);
1505 }
1506
1507 static int
1508 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1509 {
1510         struct ifbreq *req = arg;
1511         struct bridge_iflist *bif;
1512         struct ifnet *ifs;
1513
1514         ifs = ifunit(req->ifbr_ifsname);
1515         if (ifs == NULL)
1516                 return (ENOENT);
1517
1518         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1519                 if (ifs == bif->bif_ifp)
1520                         break;
1521
1522         if (bif == NULL)
1523                 return (ENOENT);
1524
1525         bridge_delete_span(sc, bif);
1526
1527         return (0);
1528 }
1529
1530 static int
1531 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1532 {
1533         struct ifbropreq *req = arg;
1534         struct bstp_state *bs = &sc->sc_stp;
1535         struct bstp_port *root_port;
1536
1537         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1538         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1539         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1540
1541         root_port = bs->bs_root_port;
1542         if (root_port == NULL)
1543                 req->ifbop_root_port = 0;
1544         else
1545                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1546
1547         req->ifbop_holdcount = bs->bs_txholdcount;
1548         req->ifbop_priority = bs->bs_bridge_priority;
1549         req->ifbop_protocol = bs->bs_protover;
1550         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1551         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1552         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1553         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1554         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1555         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1556
1557         return (0);
1558 }
1559
1560 static int
1561 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1562 {
1563         struct ifbrparam *param = arg;
1564
1565         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1566         return (0);
1567 }
1568
1569 static int
1570 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1571 {
1572         struct ifbpstpconf *bifstp = arg;
1573         struct bridge_iflist *bif;
1574         struct bstp_port *bp;
1575         struct ifbpstpreq bpreq;
1576         char *buf, *outbuf;
1577         int count, buflen, len, error = 0;
1578
1579         count = 0;
1580         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1581                 if ((bif->bif_flags & IFBIF_STP) != 0)
1582                         count++;
1583         }
1584
1585         buflen = sizeof(bpreq) * count;
1586         if (bifstp->ifbpstp_len == 0) {
1587                 bifstp->ifbpstp_len = buflen;
1588                 return (0);
1589         }
1590
1591         BRIDGE_UNLOCK(sc);
1592         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1593         BRIDGE_LOCK(sc);
1594
1595         count = 0;
1596         buf = outbuf;
1597         len = min(bifstp->ifbpstp_len, buflen);
1598         bzero(&bpreq, sizeof(bpreq));
1599         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1600                 if (len < sizeof(bpreq))
1601                         break;
1602
1603                 if ((bif->bif_flags & IFBIF_STP) == 0)
1604                         continue;
1605
1606                 bp = &bif->bif_stp;
1607                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1608                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1609                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1610                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1611                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1612                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1613
1614                 memcpy(buf, &bpreq, sizeof(bpreq));
1615                 count++;
1616                 buf += sizeof(bpreq);
1617                 len -= sizeof(bpreq);
1618         }
1619
1620         BRIDGE_UNLOCK(sc);
1621         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1622         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1623         BRIDGE_LOCK(sc);
1624         free(outbuf, M_TEMP);
1625         return (error);
1626 }
1627
1628 static int
1629 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1630 {
1631         struct ifbrparam *param = arg;
1632
1633         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1634 }
1635
1636 static int
1637 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1638 {
1639         struct ifbrparam *param = arg;
1640
1641         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1642 }
1643
1644 /*
1645  * bridge_ifdetach:
1646  *
1647  *      Detach an interface from a bridge.  Called when a member
1648  *      interface is detaching.
1649  */
1650 static void
1651 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1652 {
1653         struct bridge_softc *sc = ifp->if_bridge;
1654         struct bridge_iflist *bif;
1655
1656         /* Check if the interface is a bridge member */
1657         if (sc != NULL) {
1658                 BRIDGE_LOCK(sc);
1659
1660                 bif = bridge_lookup_member_if(sc, ifp);
1661                 if (bif != NULL)
1662                         bridge_delete_member(sc, bif, 1);
1663
1664                 BRIDGE_UNLOCK(sc);
1665                 return;
1666         }
1667
1668         /* Check if the interface is a span port */
1669         mtx_lock(&bridge_list_mtx);
1670         LIST_FOREACH(sc, &bridge_list, sc_list) {
1671                 BRIDGE_LOCK(sc);
1672                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1673                         if (ifp == bif->bif_ifp) {
1674                                 bridge_delete_span(sc, bif);
1675                                 break;
1676                         }
1677
1678                 BRIDGE_UNLOCK(sc);
1679         }
1680         mtx_unlock(&bridge_list_mtx);
1681 }
1682
1683 /*
1684  * bridge_init:
1685  *
1686  *      Initialize a bridge interface.
1687  */
1688 static void
1689 bridge_init(void *xsc)
1690 {
1691         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1692         struct ifnet *ifp = sc->sc_ifp;
1693
1694         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1695                 return;
1696
1697         BRIDGE_LOCK(sc);
1698         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1699             bridge_timer, sc);
1700
1701         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1702         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1703
1704         BRIDGE_UNLOCK(sc);
1705 }
1706
1707 /*
1708  * bridge_stop:
1709  *
1710  *      Stop the bridge interface.
1711  */
1712 static void
1713 bridge_stop(struct ifnet *ifp, int disable)
1714 {
1715         struct bridge_softc *sc = ifp->if_softc;
1716
1717         BRIDGE_LOCK_ASSERT(sc);
1718
1719         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1720                 return;
1721
1722         callout_stop(&sc->sc_brcallout);
1723         bstp_stop(&sc->sc_stp);
1724
1725         bridge_rtflush(sc, IFBF_FLUSHDYN);
1726
1727         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1728 }
1729
1730 /*
1731  * bridge_enqueue:
1732  *
1733  *      Enqueue a packet on a bridge member interface.
1734  *
1735  */
1736 static void
1737 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1738 {
1739         int len, err = 0;
1740         short mflags;
1741         struct mbuf *m0;
1742
1743         len = m->m_pkthdr.len;
1744         mflags = m->m_flags;
1745
1746         /* We may be sending a fragment so traverse the mbuf */
1747         for (; m; m = m0) {
1748                 m0 = m->m_nextpkt;
1749                 m->m_nextpkt = NULL;
1750
1751                 /*
1752                  * If underlying interface can not do VLAN tag insertion itself
1753                  * then attach a packet tag that holds it.
1754                  */
1755                 if ((m->m_flags & M_VLANTAG) &&
1756                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1757                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1758                         if (m == NULL) {
1759                                 if_printf(dst_ifp,
1760                                     "unable to prepend VLAN header\n");
1761                                 dst_ifp->if_oerrors++;
1762                                 continue;
1763                         }
1764                         m->m_flags &= ~M_VLANTAG;
1765                 }
1766
1767                 if (err == 0)
1768                         dst_ifp->if_transmit(dst_ifp, m);
1769         }
1770
1771         if (err == 0) {
1772                 sc->sc_ifp->if_opackets++;
1773                 sc->sc_ifp->if_obytes += len;
1774                 if (mflags & M_MCAST)
1775                         sc->sc_ifp->if_omcasts++;
1776         }
1777 }
1778
1779 /*
1780  * bridge_dummynet:
1781  *
1782  *      Receive a queued packet from dummynet and pass it on to the output
1783  *      interface.
1784  *
1785  *      The mbuf has the Ethernet header already attached.
1786  */
1787 static void
1788 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1789 {
1790         struct bridge_softc *sc;
1791
1792         sc = ifp->if_bridge;
1793
1794         /*
1795          * The packet didnt originate from a member interface. This should only
1796          * ever happen if a member interface is removed while packets are
1797          * queued for it.
1798          */
1799         if (sc == NULL) {
1800                 m_freem(m);
1801                 return;
1802         }
1803
1804         if (PFIL_HOOKED(&inet_pfil_hook)
1805 #ifdef INET6
1806             || PFIL_HOOKED(&inet6_pfil_hook)
1807 #endif
1808             ) {
1809                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1810                         return;
1811                 if (m == NULL)
1812                         return;
1813         }
1814
1815         bridge_enqueue(sc, ifp, m);
1816 }
1817
1818 /*
1819  * bridge_output:
1820  *
1821  *      Send output from a bridge member interface.  This
1822  *      performs the bridging function for locally originated
1823  *      packets.
1824  *
1825  *      The mbuf has the Ethernet header already attached.  We must
1826  *      enqueue or free the mbuf before returning.
1827  */
1828 static int
1829 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1830     struct rtentry *rt)
1831 {
1832         struct ether_header *eh;
1833         struct ifnet *dst_if;
1834         struct bridge_softc *sc;
1835         uint16_t vlan;
1836
1837         if (m->m_len < ETHER_HDR_LEN) {
1838                 m = m_pullup(m, ETHER_HDR_LEN);
1839                 if (m == NULL)
1840                         return (0);
1841         }
1842
1843         eh = mtod(m, struct ether_header *);
1844         sc = ifp->if_bridge;
1845         vlan = VLANTAGOF(m);
1846
1847         BRIDGE_LOCK(sc);
1848
1849         /*
1850          * If bridge is down, but the original output interface is up,
1851          * go ahead and send out that interface.  Otherwise, the packet
1852          * is dropped below.
1853          */
1854         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1855                 dst_if = ifp;
1856                 goto sendunicast;
1857         }
1858
1859         /*
1860          * If the packet is a multicast, or we don't know a better way to
1861          * get there, send to all interfaces.
1862          */
1863         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1864                 dst_if = NULL;
1865         else
1866                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1867         if (dst_if == NULL) {
1868                 struct bridge_iflist *bif;
1869                 struct mbuf *mc;
1870                 int error = 0, used = 0;
1871
1872                 bridge_span(sc, m);
1873
1874                 BRIDGE_LOCK2REF(sc, error);
1875                 if (error) {
1876                         m_freem(m);
1877                         return (0);
1878                 }
1879
1880                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1881                         dst_if = bif->bif_ifp;
1882
1883                         if (dst_if->if_type == IFT_GIF)
1884                                 continue;
1885                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1886                                 continue;
1887
1888                         /*
1889                          * If this is not the original output interface,
1890                          * and the interface is participating in spanning
1891                          * tree, make sure the port is in a state that
1892                          * allows forwarding.
1893                          */
1894                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1895                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1896                                 continue;
1897
1898                         if (LIST_NEXT(bif, bif_next) == NULL) {
1899                                 used = 1;
1900                                 mc = m;
1901                         } else {
1902                                 mc = m_copypacket(m, M_DONTWAIT);
1903                                 if (mc == NULL) {
1904                                         sc->sc_ifp->if_oerrors++;
1905                                         continue;
1906                                 }
1907                         }
1908
1909                         bridge_enqueue(sc, dst_if, mc);
1910                 }
1911                 if (used == 0)
1912                         m_freem(m);
1913                 BRIDGE_UNREF(sc);
1914                 return (0);
1915         }
1916
1917 sendunicast:
1918         /*
1919          * XXX Spanning tree consideration here?
1920          */
1921
1922         bridge_span(sc, m);
1923         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1924                 m_freem(m);
1925                 BRIDGE_UNLOCK(sc);
1926                 return (0);
1927         }
1928
1929         BRIDGE_UNLOCK(sc);
1930         bridge_enqueue(sc, dst_if, m);
1931         return (0);
1932 }
1933
1934 /*
1935  * bridge_start:
1936  *
1937  *      Start output on a bridge.
1938  *
1939  */
1940 static void
1941 bridge_start(struct ifnet *ifp)
1942 {
1943         struct bridge_softc *sc;
1944         struct mbuf *m;
1945         struct ether_header *eh;
1946         struct ifnet *dst_if;
1947
1948         sc = ifp->if_softc;
1949
1950         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1951         for (;;) {
1952                 IFQ_DEQUEUE(&ifp->if_snd, m);
1953                 if (m == 0)
1954                         break;
1955                 ETHER_BPF_MTAP(ifp, m);
1956
1957                 eh = mtod(m, struct ether_header *);
1958                 dst_if = NULL;
1959
1960                 BRIDGE_LOCK(sc);
1961                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1962                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
1963                 }
1964
1965                 if (dst_if == NULL)
1966                         bridge_broadcast(sc, ifp, m, 0);
1967                 else {
1968                         BRIDGE_UNLOCK(sc);
1969                         bridge_enqueue(sc, dst_if, m);
1970                 }
1971         }
1972         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
1973 }
1974
1975 /*
1976  * bridge_forward:
1977  *
1978  *      The forwarding function of the bridge.
1979  *
1980  *      NOTE: Releases the lock on return.
1981  */
1982 static void
1983 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
1984     struct mbuf *m)
1985 {
1986         struct bridge_iflist *dbif;
1987         struct ifnet *src_if, *dst_if, *ifp;
1988         struct ether_header *eh;
1989         uint16_t vlan;
1990         uint8_t *dst;
1991         int error;
1992
1993         src_if = m->m_pkthdr.rcvif;
1994         ifp = sc->sc_ifp;
1995
1996         ifp->if_ipackets++;
1997         ifp->if_ibytes += m->m_pkthdr.len;
1998         vlan = VLANTAGOF(m);
1999
2000         if ((sbif->bif_flags & IFBIF_STP) &&
2001             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2002                 goto drop;
2003
2004         eh = mtod(m, struct ether_header *);
2005         dst = eh->ether_dhost;
2006
2007         /* If the interface is learning, record the address. */
2008         if (sbif->bif_flags & IFBIF_LEARNING) {
2009                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2010                     sbif, 0, IFBAF_DYNAMIC);
2011                 /*
2012                  * If the interface has addresses limits then deny any source
2013                  * that is not in the cache.
2014                  */
2015                 if (error && sbif->bif_addrmax)
2016                         goto drop;
2017         }
2018
2019         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2020             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2021                 goto drop;
2022
2023         /*
2024          * At this point, the port either doesn't participate
2025          * in spanning tree or it is in the forwarding state.
2026          */
2027
2028         /*
2029          * If the packet is unicast, destined for someone on
2030          * "this" side of the bridge, drop it.
2031          */
2032         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2033                 dst_if = bridge_rtlookup(sc, dst, vlan);
2034                 if (src_if == dst_if)
2035                         goto drop;
2036         } else {
2037                 /*
2038                  * Check if its a reserved multicast address, any address
2039                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2040                  * bridge.
2041                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2042                  */
2043                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2044                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2045                     dst[4] == 0x00 && dst[5] <= 0x0f)
2046                         goto drop;
2047
2048                 /* ...forward it to all interfaces. */
2049                 ifp->if_imcasts++;
2050                 dst_if = NULL;
2051         }
2052
2053         /*
2054          * If we have a destination interface which is a member of our bridge,
2055          * OR this is a unicast packet, push it through the bpf(4) machinery.
2056          * For broadcast or multicast packets, don't bother because it will
2057          * be reinjected into ether_input. We do this before we pass the packets
2058          * through the pfil(9) framework, as it is possible that pfil(9) will
2059          * drop the packet, or possibly modify it, making it difficult to debug
2060          * firewall issues on the bridge.
2061          */
2062         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2063                 ETHER_BPF_MTAP(ifp, m);
2064
2065         /* run the packet filter */
2066         if (PFIL_HOOKED(&inet_pfil_hook)
2067 #ifdef INET6
2068             || PFIL_HOOKED(&inet6_pfil_hook)
2069 #endif
2070             ) {
2071                 BRIDGE_UNLOCK(sc);
2072                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2073                         return;
2074                 if (m == NULL)
2075                         return;
2076                 BRIDGE_LOCK(sc);
2077         }
2078
2079         if (dst_if == NULL) {
2080                 bridge_broadcast(sc, src_if, m, 1);
2081                 return;
2082         }
2083
2084         /*
2085          * At this point, we're dealing with a unicast frame
2086          * going to a different interface.
2087          */
2088         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2089                 goto drop;
2090
2091         dbif = bridge_lookup_member_if(sc, dst_if);
2092         if (dbif == NULL)
2093                 /* Not a member of the bridge (anymore?) */
2094                 goto drop;
2095
2096         /* Private segments can not talk to each other */
2097         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2098                 goto drop;
2099
2100         if ((dbif->bif_flags & IFBIF_STP) &&
2101             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2102                 goto drop;
2103
2104         BRIDGE_UNLOCK(sc);
2105
2106         if (PFIL_HOOKED(&inet_pfil_hook)
2107 #ifdef INET6
2108             || PFIL_HOOKED(&inet6_pfil_hook)
2109 #endif
2110             ) {
2111                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2112                         return;
2113                 if (m == NULL)
2114                         return;
2115         }
2116
2117         bridge_enqueue(sc, dst_if, m);
2118         return;
2119
2120 drop:
2121         BRIDGE_UNLOCK(sc);
2122         m_freem(m);
2123 }
2124
2125 /*
2126  * bridge_input:
2127  *
2128  *      Receive input from a member interface.  Queue the packet for
2129  *      bridging if it is not for us.
2130  */
2131 static struct mbuf *
2132 bridge_input(struct ifnet *ifp, struct mbuf *m)
2133 {
2134         struct bridge_softc *sc = ifp->if_bridge;
2135         struct bridge_iflist *bif, *bif2;
2136         struct ifnet *bifp;
2137         struct ether_header *eh;
2138         struct mbuf *mc, *mc2;
2139         uint16_t vlan;
2140         int error;
2141
2142         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2143                 return (m);
2144
2145         bifp = sc->sc_ifp;
2146         vlan = VLANTAGOF(m);
2147
2148         /*
2149          * Implement support for bridge monitoring. If this flag has been
2150          * set on this interface, discard the packet once we push it through
2151          * the bpf(4) machinery, but before we do, increment the byte and
2152          * packet counters associated with this interface.
2153          */
2154         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2155                 m->m_pkthdr.rcvif  = bifp;
2156                 ETHER_BPF_MTAP(bifp, m);
2157                 bifp->if_ipackets++;
2158                 bifp->if_ibytes += m->m_pkthdr.len;
2159                 m_freem(m);
2160                 return (NULL);
2161         }
2162         BRIDGE_LOCK(sc);
2163         bif = bridge_lookup_member_if(sc, ifp);
2164         if (bif == NULL) {
2165                 BRIDGE_UNLOCK(sc);
2166                 return (m);
2167         }
2168
2169         eh = mtod(m, struct ether_header *);
2170
2171         bridge_span(sc, m);
2172
2173         if (m->m_flags & (M_BCAST|M_MCAST)) {
2174                 /* Tap off 802.1D packets; they do not get forwarded. */
2175                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2176                     ETHER_ADDR_LEN) == 0) {
2177                         m = bstp_input(&bif->bif_stp, ifp, m);
2178                         if (m == NULL) {
2179                                 BRIDGE_UNLOCK(sc);
2180                                 return (NULL);
2181                         }
2182                 }
2183
2184                 if ((bif->bif_flags & IFBIF_STP) &&
2185                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2186                         BRIDGE_UNLOCK(sc);
2187                         return (m);
2188                 }
2189
2190                 /*
2191                  * Make a deep copy of the packet and enqueue the copy
2192                  * for bridge processing; return the original packet for
2193                  * local processing.
2194                  */
2195                 mc = m_dup(m, M_DONTWAIT);
2196                 if (mc == NULL) {
2197                         BRIDGE_UNLOCK(sc);
2198                         return (m);
2199                 }
2200
2201                 /* Perform the bridge forwarding function with the copy. */
2202                 bridge_forward(sc, bif, mc);
2203
2204                 /*
2205                  * Reinject the mbuf as arriving on the bridge so we have a
2206                  * chance at claiming multicast packets. We can not loop back
2207                  * here from ether_input as a bridge is never a member of a
2208                  * bridge.
2209                  */
2210                 KASSERT(bifp->if_bridge == NULL,
2211                     ("loop created in bridge_input"));
2212                 mc2 = m_dup(m, M_DONTWAIT);
2213                 if (mc2 != NULL) {
2214                         /* Keep the layer3 header aligned */
2215                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2216                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2217                 }
2218                 if (mc2 != NULL) {
2219                         mc2->m_pkthdr.rcvif = bifp;
2220                         (*bifp->if_input)(bifp, mc2);
2221                 }
2222
2223                 /* Return the original packet for local processing. */
2224                 return (m);
2225         }
2226
2227         if ((bif->bif_flags & IFBIF_STP) &&
2228             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2229                 BRIDGE_UNLOCK(sc);
2230                 return (m);
2231         }
2232
2233 #if (defined(INET) || defined(INET6)) && defined(DEV_CARP)
2234 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2235         || ((iface)->if_carp \
2236             && carp_forus((iface)->if_carp, eh->ether_dhost))
2237 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2238         || ((iface)->if_carp \
2239             && carp_forus((iface)->if_carp, eh->ether_shost))
2240 #else
2241 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2242 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2243 #endif
2244
2245 #ifdef INET6
2246 #   define OR_PFIL_HOOKED_INET6 \
2247         || PFIL_HOOKED(&inet6_pfil_hook)
2248 #else
2249 #   define OR_PFIL_HOOKED_INET6
2250 #endif
2251
2252 #define GRAB_OUR_PACKETS(iface) \
2253         if ((iface)->if_type == IFT_GIF) \
2254                 continue; \
2255         /* It is destined for us. */ \
2256         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2257             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2258             ) {                                                         \
2259                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2260                         ETHER_BPF_MTAP(iface, m);                       \
2261                         iface->if_ipackets++;                           \
2262                         /* Filter on the physical interface. */         \
2263                         if (pfil_local_phys &&                          \
2264                             (PFIL_HOOKED(&inet_pfil_hook)               \
2265                              OR_PFIL_HOOKED_INET6)) {                   \
2266                                 if (bridge_pfil(&m, NULL, ifp,          \
2267                                     PFIL_IN) != 0 || m == NULL) {       \
2268                                         BRIDGE_UNLOCK(sc);              \
2269                                         return (NULL);                  \
2270                                 }                                       \
2271                         }                                               \
2272                 }                                                       \
2273                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2274                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2275                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2276                         if (error && bif->bif_addrmax) {                \
2277                                 BRIDGE_UNLOCK(sc);                      \
2278                                 m_freem(m);                             \
2279                                 return (NULL);                          \
2280                         }                                               \
2281                 }                                                       \
2282                 m->m_pkthdr.rcvif = iface;                              \
2283                 BRIDGE_UNLOCK(sc);                                      \
2284                 return (m);                                             \
2285         }                                                               \
2286                                                                         \
2287         /* We just received a packet that we sent out. */               \
2288         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2289             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2290             ) {                                                         \
2291                 BRIDGE_UNLOCK(sc);                                      \
2292                 m_freem(m);                                             \
2293                 return (NULL);                                          \
2294         }
2295
2296         /*
2297          * Unicast.  Make sure it's not for the bridge.
2298          */
2299         do { GRAB_OUR_PACKETS(bifp) } while (0);
2300
2301         /*
2302          * Give a chance for ifp at first priority. This will help when the
2303          * packet comes through the interface like VLAN's with the same MACs
2304          * on several interfaces from the same bridge. This also will save
2305          * some CPU cycles in case the destination interface and the input
2306          * interface (eq ifp) are the same.
2307          */
2308         do { GRAB_OUR_PACKETS(ifp) } while (0);
2309
2310         /* Now check the all bridge members. */
2311         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2312                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2313         }
2314
2315 #undef OR_CARP_CHECK_WE_ARE_DST
2316 #undef OR_CARP_CHECK_WE_ARE_SRC
2317 #undef OR_PFIL_HOOKED_INET6
2318 #undef GRAB_OUR_PACKETS
2319
2320         /* Perform the bridge forwarding function. */
2321         bridge_forward(sc, bif, m);
2322
2323         return (NULL);
2324 }
2325
2326 /*
2327  * bridge_broadcast:
2328  *
2329  *      Send a frame to all interfaces that are members of
2330  *      the bridge, except for the one on which the packet
2331  *      arrived.
2332  *
2333  *      NOTE: Releases the lock on return.
2334  */
2335 static void
2336 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2337     struct mbuf *m, int runfilt)
2338 {
2339         struct bridge_iflist *dbif, *sbif;
2340         struct mbuf *mc;
2341         struct ifnet *dst_if;
2342         int error = 0, used = 0, i;
2343
2344         sbif = bridge_lookup_member_if(sc, src_if);
2345
2346         BRIDGE_LOCK2REF(sc, error);
2347         if (error) {
2348                 m_freem(m);
2349                 return;
2350         }
2351
2352         /* Filter on the bridge interface before broadcasting */
2353         if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2354 #ifdef INET6
2355             || PFIL_HOOKED(&inet6_pfil_hook)
2356 #endif
2357             )) {
2358                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2359                         goto out;
2360                 if (m == NULL)
2361                         goto out;
2362         }
2363
2364         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2365                 dst_if = dbif->bif_ifp;
2366                 if (dst_if == src_if)
2367                         continue;
2368
2369                 /* Private segments can not talk to each other */
2370                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2371                         continue;
2372
2373                 if ((dbif->bif_flags & IFBIF_STP) &&
2374                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2375                         continue;
2376
2377                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2378                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2379                         continue;
2380
2381                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2382                         continue;
2383
2384                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2385                         mc = m;
2386                         used = 1;
2387                 } else {
2388                         mc = m_dup(m, M_DONTWAIT);
2389                         if (mc == NULL) {
2390                                 sc->sc_ifp->if_oerrors++;
2391                                 continue;
2392                         }
2393                 }
2394
2395                 /*
2396                  * Filter on the output interface. Pass a NULL bridge interface
2397                  * pointer so we do not redundantly filter on the bridge for
2398                  * each interface we broadcast on.
2399                  */
2400                 if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2401 #ifdef INET6
2402                     || PFIL_HOOKED(&inet6_pfil_hook)
2403 #endif
2404                     )) {
2405                         if (used == 0) {
2406                                 /* Keep the layer3 header aligned */
2407                                 i = min(mc->m_pkthdr.len, max_protohdr);
2408                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2409                                 if (mc == NULL) {
2410                                         sc->sc_ifp->if_oerrors++;
2411                                         continue;
2412                                 }
2413                         }
2414                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2415                                 continue;
2416                         if (mc == NULL)
2417                                 continue;
2418                 }
2419
2420                 bridge_enqueue(sc, dst_if, mc);
2421         }
2422         if (used == 0)
2423                 m_freem(m);
2424
2425 out:
2426         BRIDGE_UNREF(sc);
2427 }
2428
2429 /*
2430  * bridge_span:
2431  *
2432  *      Duplicate a packet out one or more interfaces that are in span mode,
2433  *      the original mbuf is unmodified.
2434  */
2435 static void
2436 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2437 {
2438         struct bridge_iflist *bif;
2439         struct ifnet *dst_if;
2440         struct mbuf *mc;
2441
2442         if (LIST_EMPTY(&sc->sc_spanlist))
2443                 return;
2444
2445         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2446                 dst_if = bif->bif_ifp;
2447
2448                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2449                         continue;
2450
2451                 mc = m_copypacket(m, M_DONTWAIT);
2452                 if (mc == NULL) {
2453                         sc->sc_ifp->if_oerrors++;
2454                         continue;
2455                 }
2456
2457                 bridge_enqueue(sc, dst_if, mc);
2458         }
2459 }
2460
2461 /*
2462  * bridge_rtupdate:
2463  *
2464  *      Add a bridge routing entry.
2465  */
2466 static int
2467 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2468     struct bridge_iflist *bif, int setflags, uint8_t flags)
2469 {
2470         struct bridge_rtnode *brt;
2471         int error;
2472
2473         BRIDGE_LOCK_ASSERT(sc);
2474
2475         /* Check the source address is valid and not multicast. */
2476         if (ETHER_IS_MULTICAST(dst) ||
2477             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2478              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2479                 return (EINVAL);
2480
2481         /* 802.1p frames map to vlan 1 */
2482         if (vlan == 0)
2483                 vlan = 1;
2484
2485         /*
2486          * A route for this destination might already exist.  If so,
2487          * update it, otherwise create a new one.
2488          */
2489         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2490                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2491                         sc->sc_brtexceeded++;
2492                         return (ENOSPC);
2493                 }
2494                 /* Check per interface address limits (if enabled) */
2495                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2496                         bif->bif_addrexceeded++;
2497                         return (ENOSPC);
2498                 }
2499
2500                 /*
2501                  * Allocate a new bridge forwarding node, and
2502                  * initialize the expiration time and Ethernet
2503                  * address.
2504                  */
2505                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2506                 if (brt == NULL)
2507                         return (ENOMEM);
2508
2509                 if (bif->bif_flags & IFBIF_STICKY)
2510                         brt->brt_flags = IFBAF_STICKY;
2511                 else
2512                         brt->brt_flags = IFBAF_DYNAMIC;
2513
2514                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2515                 brt->brt_vlan = vlan;
2516
2517                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2518                         uma_zfree(bridge_rtnode_zone, brt);
2519                         return (error);
2520                 }
2521                 brt->brt_dst = bif;
2522                 bif->bif_addrcnt++;
2523         }
2524
2525         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2526             brt->brt_dst != bif) {
2527                 brt->brt_dst->bif_addrcnt--;
2528                 brt->brt_dst = bif;
2529                 brt->brt_dst->bif_addrcnt++;
2530         }
2531
2532         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2533                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2534         if (setflags)
2535                 brt->brt_flags = flags;
2536
2537         return (0);
2538 }
2539
2540 /*
2541  * bridge_rtlookup:
2542  *
2543  *      Lookup the destination interface for an address.
2544  */
2545 static struct ifnet *
2546 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2547 {
2548         struct bridge_rtnode *brt;
2549
2550         BRIDGE_LOCK_ASSERT(sc);
2551
2552         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2553                 return (NULL);
2554
2555         return (brt->brt_ifp);
2556 }
2557
2558 /*
2559  * bridge_rttrim:
2560  *
2561  *      Trim the routine table so that we have a number
2562  *      of routing entries less than or equal to the
2563  *      maximum number.
2564  */
2565 static void
2566 bridge_rttrim(struct bridge_softc *sc)
2567 {
2568         struct bridge_rtnode *brt, *nbrt;
2569
2570         BRIDGE_LOCK_ASSERT(sc);
2571
2572         /* Make sure we actually need to do this. */
2573         if (sc->sc_brtcnt <= sc->sc_brtmax)
2574                 return;
2575
2576         /* Force an aging cycle; this might trim enough addresses. */
2577         bridge_rtage(sc);
2578         if (sc->sc_brtcnt <= sc->sc_brtmax)
2579                 return;
2580
2581         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2582                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2583                         bridge_rtnode_destroy(sc, brt);
2584                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2585                                 return;
2586                 }
2587         }
2588 }
2589
2590 /*
2591  * bridge_timer:
2592  *
2593  *      Aging timer for the bridge.
2594  */
2595 static void
2596 bridge_timer(void *arg)
2597 {
2598         struct bridge_softc *sc = arg;
2599
2600         BRIDGE_LOCK_ASSERT(sc);
2601
2602         bridge_rtage(sc);
2603
2604         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2605                 callout_reset(&sc->sc_brcallout,
2606                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2607 }
2608
2609 /*
2610  * bridge_rtage:
2611  *
2612  *      Perform an aging cycle.
2613  */
2614 static void
2615 bridge_rtage(struct bridge_softc *sc)
2616 {
2617         struct bridge_rtnode *brt, *nbrt;
2618
2619         BRIDGE_LOCK_ASSERT(sc);
2620
2621         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2622                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2623                         if (time_uptime >= brt->brt_expire)
2624                                 bridge_rtnode_destroy(sc, brt);
2625                 }
2626         }
2627 }
2628
2629 /*
2630  * bridge_rtflush:
2631  *
2632  *      Remove all dynamic addresses from the bridge.
2633  */
2634 static void
2635 bridge_rtflush(struct bridge_softc *sc, int full)
2636 {
2637         struct bridge_rtnode *brt, *nbrt;
2638
2639         BRIDGE_LOCK_ASSERT(sc);
2640
2641         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2642                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2643                         bridge_rtnode_destroy(sc, brt);
2644         }
2645 }
2646
2647 /*
2648  * bridge_rtdaddr:
2649  *
2650  *      Remove an address from the table.
2651  */
2652 static int
2653 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2654 {
2655         struct bridge_rtnode *brt;
2656         int found = 0;
2657
2658         BRIDGE_LOCK_ASSERT(sc);
2659
2660         /*
2661          * If vlan is zero then we want to delete for all vlans so the lookup
2662          * may return more than one.
2663          */
2664         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2665                 bridge_rtnode_destroy(sc, brt);
2666                 found = 1;
2667         }
2668
2669         return (found ? 0 : ENOENT);
2670 }
2671
2672 /*
2673  * bridge_rtdelete:
2674  *
2675  *      Delete routes to a speicifc member interface.
2676  */
2677 static void
2678 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2679 {
2680         struct bridge_rtnode *brt, *nbrt;
2681
2682         BRIDGE_LOCK_ASSERT(sc);
2683
2684         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2685                 if (brt->brt_ifp == ifp && (full ||
2686                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2687                         bridge_rtnode_destroy(sc, brt);
2688         }
2689 }
2690
2691 /*
2692  * bridge_rtable_init:
2693  *
2694  *      Initialize the route table for this bridge.
2695  */
2696 static int
2697 bridge_rtable_init(struct bridge_softc *sc)
2698 {
2699         int i;
2700
2701         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2702             M_DEVBUF, M_NOWAIT);
2703         if (sc->sc_rthash == NULL)
2704                 return (ENOMEM);
2705
2706         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2707                 LIST_INIT(&sc->sc_rthash[i]);
2708
2709         sc->sc_rthash_key = arc4random();
2710
2711         LIST_INIT(&sc->sc_rtlist);
2712
2713         return (0);
2714 }
2715
2716 /*
2717  * bridge_rtable_fini:
2718  *
2719  *      Deconstruct the route table for this bridge.
2720  */
2721 static void
2722 bridge_rtable_fini(struct bridge_softc *sc)
2723 {
2724
2725         KASSERT(sc->sc_brtcnt == 0,
2726             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2727         free(sc->sc_rthash, M_DEVBUF);
2728 }
2729
2730 /*
2731  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2732  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2733  */
2734 #define mix(a, b, c)                                                    \
2735 do {                                                                    \
2736         a -= b; a -= c; a ^= (c >> 13);                                 \
2737         b -= c; b -= a; b ^= (a << 8);                                  \
2738         c -= a; c -= b; c ^= (b >> 13);                                 \
2739         a -= b; a -= c; a ^= (c >> 12);                                 \
2740         b -= c; b -= a; b ^= (a << 16);                                 \
2741         c -= a; c -= b; c ^= (b >> 5);                                  \
2742         a -= b; a -= c; a ^= (c >> 3);                                  \
2743         b -= c; b -= a; b ^= (a << 10);                                 \
2744         c -= a; c -= b; c ^= (b >> 15);                                 \
2745 } while (/*CONSTCOND*/0)
2746
2747 static __inline uint32_t
2748 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2749 {
2750         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2751
2752         b += addr[5] << 8;
2753         b += addr[4];
2754         a += addr[3] << 24;
2755         a += addr[2] << 16;
2756         a += addr[1] << 8;
2757         a += addr[0];
2758
2759         mix(a, b, c);
2760
2761         return (c & BRIDGE_RTHASH_MASK);
2762 }
2763
2764 #undef mix
2765
2766 static int
2767 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2768 {
2769         int i, d;
2770
2771         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2772                 d = ((int)a[i]) - ((int)b[i]);
2773         }
2774
2775         return (d);
2776 }
2777
2778 /*
2779  * bridge_rtnode_lookup:
2780  *
2781  *      Look up a bridge route node for the specified destination. Compare the
2782  *      vlan id or if zero then just return the first match.
2783  */
2784 static struct bridge_rtnode *
2785 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2786 {
2787         struct bridge_rtnode *brt;
2788         uint32_t hash;
2789         int dir;
2790
2791         BRIDGE_LOCK_ASSERT(sc);
2792
2793         hash = bridge_rthash(sc, addr);
2794         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2795                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2796                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2797                         return (brt);
2798                 if (dir > 0)
2799                         return (NULL);
2800         }
2801
2802         return (NULL);
2803 }
2804
2805 /*
2806  * bridge_rtnode_insert:
2807  *
2808  *      Insert the specified bridge node into the route table.  We
2809  *      assume the entry is not already in the table.
2810  */
2811 static int
2812 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2813 {
2814         struct bridge_rtnode *lbrt;
2815         uint32_t hash;
2816         int dir;
2817
2818         BRIDGE_LOCK_ASSERT(sc);
2819
2820         hash = bridge_rthash(sc, brt->brt_addr);
2821
2822         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2823         if (lbrt == NULL) {
2824                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2825                 goto out;
2826         }
2827
2828         do {
2829                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2830                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2831                         return (EEXIST);
2832                 if (dir > 0) {
2833                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2834                         goto out;
2835                 }
2836                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2837                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2838                         goto out;
2839                 }
2840                 lbrt = LIST_NEXT(lbrt, brt_hash);
2841         } while (lbrt != NULL);
2842
2843 #ifdef DIAGNOSTIC
2844         panic("bridge_rtnode_insert: impossible");
2845 #endif
2846
2847 out:
2848         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2849         sc->sc_brtcnt++;
2850
2851         return (0);
2852 }
2853
2854 /*
2855  * bridge_rtnode_destroy:
2856  *
2857  *      Destroy a bridge rtnode.
2858  */
2859 static void
2860 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2861 {
2862         BRIDGE_LOCK_ASSERT(sc);
2863
2864         LIST_REMOVE(brt, brt_hash);
2865
2866         LIST_REMOVE(brt, brt_list);
2867         sc->sc_brtcnt--;
2868         brt->brt_dst->bif_addrcnt--;
2869         uma_zfree(bridge_rtnode_zone, brt);
2870 }
2871
2872 /*
2873  * bridge_rtable_expire:
2874  *
2875  *      Set the expiry time for all routes on an interface.
2876  */
2877 static void
2878 bridge_rtable_expire(struct ifnet *ifp, int age)
2879 {
2880         struct bridge_softc *sc = ifp->if_bridge;
2881         struct bridge_rtnode *brt;
2882
2883         BRIDGE_LOCK(sc);
2884
2885         /*
2886          * If the age is zero then flush, otherwise set all the expiry times to
2887          * age for the interface
2888          */
2889         if (age == 0)
2890                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2891         else {
2892                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2893                         /* Cap the expiry time to 'age' */
2894                         if (brt->brt_ifp == ifp &&
2895                             brt->brt_expire > time_uptime + age &&
2896                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2897                                 brt->brt_expire = time_uptime + age;
2898                 }
2899         }
2900         BRIDGE_UNLOCK(sc);
2901 }
2902
2903 /*
2904  * bridge_state_change:
2905  *
2906  *      Callback from the bridgestp code when a port changes states.
2907  */
2908 static void
2909 bridge_state_change(struct ifnet *ifp, int state)
2910 {
2911         struct bridge_softc *sc = ifp->if_bridge;
2912         static const char *stpstates[] = {
2913                 "disabled",
2914                 "listening",
2915                 "learning",
2916                 "forwarding",
2917                 "blocking",
2918                 "discarding"
2919         };
2920
2921         if (log_stp)
2922                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2923                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2924 }
2925
2926 /*
2927  * Send bridge packets through pfil if they are one of the types pfil can deal
2928  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2929  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2930  * that interface.
2931  */
2932 static int
2933 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2934 {
2935         int snap, error, i, hlen;
2936         struct ether_header *eh1, eh2;
2937         struct ip_fw_args args;
2938         struct ip *ip;
2939         struct llc llc1;
2940         u_int16_t ether_type;
2941
2942         snap = 0;
2943         error = -1;     /* Default error if not error == 0 */
2944
2945 #if 0
2946         /* we may return with the IP fields swapped, ensure its not shared */
2947         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2948 #endif
2949
2950         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2951                 return (0); /* filtering is disabled */
2952
2953         i = min((*mp)->m_pkthdr.len, max_protohdr);
2954         if ((*mp)->m_len < i) {
2955             *mp = m_pullup(*mp, i);
2956             if (*mp == NULL) {
2957                 printf("%s: m_pullup failed\n", __func__);
2958                 return (-1);
2959             }
2960         }
2961
2962         eh1 = mtod(*mp, struct ether_header *);
2963         ether_type = ntohs(eh1->ether_type);
2964
2965         /*
2966          * Check for SNAP/LLC.
2967          */
2968         if (ether_type < ETHERMTU) {
2969                 struct llc *llc2 = (struct llc *)(eh1 + 1);
2970
2971                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2972                     llc2->llc_dsap == LLC_SNAP_LSAP &&
2973                     llc2->llc_ssap == LLC_SNAP_LSAP &&
2974                     llc2->llc_control == LLC_UI) {
2975                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
2976                         snap = 1;
2977                 }
2978         }
2979
2980         /*
2981          * If we're trying to filter bridge traffic, don't look at anything
2982          * other than IP and ARP traffic.  If the filter doesn't understand
2983          * IPv6, don't allow IPv6 through the bridge either.  This is lame
2984          * since if we really wanted, say, an AppleTalk filter, we are hosed,
2985          * but of course we don't have an AppleTalk filter to begin with.
2986          * (Note that since pfil doesn't understand ARP it will pass *ALL*
2987          * ARP traffic.)
2988          */
2989         switch (ether_type) {
2990                 case ETHERTYPE_ARP:
2991                 case ETHERTYPE_REVARP:
2992                         if (pfil_ipfw_arp == 0)
2993                                 return (0); /* Automatically pass */
2994                         break;
2995
2996                 case ETHERTYPE_IP:
2997 #ifdef INET6
2998                 case ETHERTYPE_IPV6:
2999 #endif /* INET6 */
3000                         break;
3001                 default:
3002                         /*
3003                          * Check to see if the user wants to pass non-ip
3004                          * packets, these will not be checked by pfil(9) and
3005                          * passed unconditionally so the default is to drop.
3006                          */
3007                         if (pfil_onlyip)
3008                                 goto bad;
3009         }
3010
3011         /* Strip off the Ethernet header and keep a copy. */
3012         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3013         m_adj(*mp, ETHER_HDR_LEN);
3014
3015         /* Strip off snap header, if present */
3016         if (snap) {
3017                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3018                 m_adj(*mp, sizeof(struct llc));
3019         }
3020
3021         /*
3022          * Check the IP header for alignment and errors
3023          */
3024         if (dir == PFIL_IN) {
3025                 switch (ether_type) {
3026                         case ETHERTYPE_IP:
3027                                 error = bridge_ip_checkbasic(mp);
3028                                 break;
3029 #ifdef INET6
3030                         case ETHERTYPE_IPV6:
3031                                 error = bridge_ip6_checkbasic(mp);
3032                                 break;
3033 #endif /* INET6 */
3034                         default:
3035                                 error = 0;
3036                 }
3037                 if (error)
3038                         goto bad;
3039         }
3040
3041         if (ip_fw_chk_ptr && pfil_ipfw != 0 && dir == PFIL_OUT && ifp != NULL) {
3042                 struct dn_pkt_tag *dn_tag;
3043
3044                 error = -1;
3045                 dn_tag = ip_dn_claim_tag(*mp);
3046                 if (dn_tag != NULL) {
3047                         if (dn_tag->rule != NULL && V_fw_one_pass)
3048                                 /* packet already partially processed */
3049                                 goto ipfwpass;
3050                         args.rule = dn_tag->rule; /* matching rule to restart */
3051                         args.rule_id = dn_tag->rule_id;
3052                         args.chain_id = dn_tag->chain_id;
3053                 } else
3054                         args.rule = NULL;
3055
3056                 args.m = *mp;
3057                 args.oif = ifp;
3058                 args.next_hop = NULL;
3059                 args.eh = &eh2;
3060                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3061                 i = ip_fw_chk_ptr(&args);
3062                 *mp = args.m;
3063
3064                 if (*mp == NULL)
3065                         return (error);
3066
3067                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3068
3069                         /* put the Ethernet header back on */
3070                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3071                         if (*mp == NULL)
3072                                 return (error);
3073                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3074
3075                         /*
3076                          * Pass the pkt to dummynet, which consumes it. The
3077                          * packet will return to us via bridge_dummynet().
3078                          */
3079                         args.oif = ifp;
3080                         ip_dn_io_ptr(mp, DN_TO_IFB_FWD, &args);
3081                         return (error);
3082                 }
3083
3084                 if (i != IP_FW_PASS) /* drop */
3085                         goto bad;
3086         }
3087
3088 ipfwpass:
3089         error = 0;
3090
3091         /*
3092          * Run the packet through pfil
3093          */
3094         switch (ether_type) {
3095         case ETHERTYPE_IP:
3096                 /*
3097                  * before calling the firewall, swap fields the same as
3098                  * IP does. here we assume the header is contiguous
3099                  */
3100                 ip = mtod(*mp, struct ip *);
3101
3102                 ip->ip_len = ntohs(ip->ip_len);
3103                 ip->ip_off = ntohs(ip->ip_off);
3104
3105                 /*
3106                  * Run pfil on the member interface and the bridge, both can
3107                  * be skipped by clearing pfil_member or pfil_bridge.
3108                  *
3109                  * Keep the order:
3110                  *   in_if -> bridge_if -> out_if
3111                  */
3112                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3113                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3114                                         dir, NULL);
3115
3116                 if (*mp == NULL || error != 0) /* filter may consume */
3117                         break;
3118
3119                 if (pfil_member && ifp != NULL)
3120                         error = pfil_run_hooks(&inet_pfil_hook, mp, ifp,
3121                                         dir, NULL);
3122
3123                 if (*mp == NULL || error != 0) /* filter may consume */
3124                         break;
3125
3126                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3127                         error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3128                                         dir, NULL);
3129
3130                 if (*mp == NULL || error != 0) /* filter may consume */
3131                         break;
3132
3133                 /* check if we need to fragment the packet */
3134                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3135                         i = (*mp)->m_pkthdr.len;
3136                         if (i > ifp->if_mtu) {
3137                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3138                                             &llc1);
3139                                 return (error);
3140                         }
3141                 }
3142
3143                 /* Recalculate the ip checksum and restore byte ordering */
3144                 ip = mtod(*mp, struct ip *);
3145                 hlen = ip->ip_hl << 2;
3146                 if (hlen < sizeof(struct ip))
3147                         goto bad;
3148                 if (hlen > (*mp)->m_len) {
3149                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3150                                 goto bad;
3151                         ip = mtod(*mp, struct ip *);
3152                         if (ip == NULL)
3153                                 goto bad;
3154                 }
3155                 ip->ip_len = htons(ip->ip_len);
3156                 ip->ip_off = htons(ip->ip_off);
3157                 ip->ip_sum = 0;
3158                 if (hlen == sizeof(struct ip))
3159                         ip->ip_sum = in_cksum_hdr(ip);
3160                 else
3161                         ip->ip_sum = in_cksum(*mp, hlen);
3162
3163                 break;
3164 #ifdef INET6
3165         case ETHERTYPE_IPV6:
3166                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3167                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3168                                         dir, NULL);
3169
3170                 if (*mp == NULL || error != 0) /* filter may consume */
3171                         break;
3172
3173                 if (pfil_member && ifp != NULL)
3174                         error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3175                                         dir, NULL);
3176
3177                 if (*mp == NULL || error != 0) /* filter may consume */
3178                         break;
3179
3180                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3181                         error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3182                                         dir, NULL);
3183                 break;
3184 #endif
3185         default:
3186                 error = 0;
3187                 break;
3188         }
3189
3190         if (*mp == NULL)
3191                 return (error);
3192         if (error != 0)
3193                 goto bad;
3194
3195         error = -1;
3196
3197         /*
3198          * Finally, put everything back the way it was and return
3199          */
3200         if (snap) {
3201                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3202                 if (*mp == NULL)
3203                         return (error);
3204                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3205         }
3206
3207         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3208         if (*mp == NULL)
3209                 return (error);
3210         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3211
3212         return (0);
3213
3214 bad:
3215         m_freem(*mp);
3216         *mp = NULL;
3217         return (error);
3218 }
3219
3220 /*
3221  * Perform basic checks on header size since
3222  * pfil assumes ip_input has already processed
3223  * it for it.  Cut-and-pasted from ip_input.c.
3224  * Given how simple the IPv6 version is,
3225  * does the IPv4 version really need to be
3226  * this complicated?
3227  *
3228  * XXX Should we update ipstat here, or not?
3229  * XXX Right now we update ipstat but not
3230  * XXX csum_counter.
3231  */
3232 static int
3233 bridge_ip_checkbasic(struct mbuf **mp)
3234 {
3235         struct mbuf *m = *mp;
3236         struct ip *ip;
3237         int len, hlen;
3238         u_short sum;
3239
3240         if (*mp == NULL)
3241                 return (-1);
3242
3243         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3244                 if ((m = m_copyup(m, sizeof(struct ip),
3245                         (max_linkhdr + 3) & ~3)) == NULL) {
3246                         /* XXXJRT new stat, please */
3247                         KMOD_IPSTAT_INC(ips_toosmall);
3248                         goto bad;
3249                 }
3250         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3251                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3252                         KMOD_IPSTAT_INC(ips_toosmall);
3253                         goto bad;
3254                 }
3255         }
3256         ip = mtod(m, struct ip *);
3257         if (ip == NULL) goto bad;
3258
3259         if (ip->ip_v != IPVERSION) {
3260                 KMOD_IPSTAT_INC(ips_badvers);
3261                 goto bad;
3262         }
3263         hlen = ip->ip_hl << 2;
3264         if (hlen < sizeof(struct ip)) { /* minimum header length */
3265                 KMOD_IPSTAT_INC(ips_badhlen);
3266                 goto bad;
3267         }
3268         if (hlen > m->m_len) {
3269                 if ((m = m_pullup(m, hlen)) == 0) {
3270                         KMOD_IPSTAT_INC(ips_badhlen);
3271                         goto bad;
3272                 }
3273                 ip = mtod(m, struct ip *);
3274                 if (ip == NULL) goto bad;
3275         }
3276
3277         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3278                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3279         } else {
3280                 if (hlen == sizeof(struct ip)) {
3281                         sum = in_cksum_hdr(ip);
3282                 } else {
3283                         sum = in_cksum(m, hlen);
3284                 }
3285         }
3286         if (sum) {
3287                 KMOD_IPSTAT_INC(ips_badsum);
3288                 goto bad;
3289         }
3290
3291         /* Retrieve the packet length. */
3292         len = ntohs(ip->ip_len);
3293
3294         /*
3295          * Check for additional length bogosity
3296          */
3297         if (len < hlen) {
3298                 KMOD_IPSTAT_INC(ips_badlen);
3299                 goto bad;
3300         }
3301
3302         /*
3303          * Check that the amount of data in the buffers
3304          * is as at least much as the IP header would have us expect.
3305          * Drop packet if shorter than we expect.
3306          */
3307         if (m->m_pkthdr.len < len) {
3308                 KMOD_IPSTAT_INC(ips_tooshort);
3309                 goto bad;
3310         }
3311
3312         /* Checks out, proceed */
3313         *mp = m;
3314         return (0);
3315
3316 bad:
3317         *mp = m;
3318         return (-1);
3319 }
3320
3321 #ifdef INET6
3322 /*
3323  * Same as above, but for IPv6.
3324  * Cut-and-pasted from ip6_input.c.
3325  * XXX Should we update ip6stat, or not?
3326  */
3327 static int
3328 bridge_ip6_checkbasic(struct mbuf **mp)
3329 {
3330         struct mbuf *m = *mp;
3331         struct ip6_hdr *ip6;
3332
3333         /*
3334          * If the IPv6 header is not aligned, slurp it up into a new
3335          * mbuf with space for link headers, in the event we forward
3336          * it.  Otherwise, if it is aligned, make sure the entire base
3337          * IPv6 header is in the first mbuf of the chain.
3338          */
3339         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3340                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3341                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3342                             (max_linkhdr + 3) & ~3)) == NULL) {
3343                         /* XXXJRT new stat, please */
3344                         V_ip6stat.ip6s_toosmall++;
3345                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3346                         goto bad;
3347                 }
3348         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3349                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3350                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3351                         V_ip6stat.ip6s_toosmall++;
3352                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3353                         goto bad;
3354                 }
3355         }
3356
3357         ip6 = mtod(m, struct ip6_hdr *);
3358
3359         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3360                 V_ip6stat.ip6s_badvers++;
3361                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3362                 goto bad;
3363         }
3364
3365         /* Checks out, proceed */
3366         *mp = m;
3367         return (0);
3368
3369 bad:
3370         *mp = m;
3371         return (-1);
3372 }
3373 #endif /* INET6 */
3374
3375 /*
3376  * bridge_fragment:
3377  *
3378  *      Return a fragmented mbuf chain.
3379  */
3380 static int
3381 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3382     int snap, struct llc *llc)
3383 {
3384         struct mbuf *m0;
3385         struct ip *ip;
3386         int error = -1;
3387
3388         if (m->m_len < sizeof(struct ip) &&
3389             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3390                 goto out;
3391         ip = mtod(m, struct ip *);
3392
3393         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3394                     CSUM_DELAY_IP);
3395         if (error)
3396                 goto out;
3397
3398         /* walk the chain and re-add the Ethernet header */
3399         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3400                 if (error == 0) {
3401                         if (snap) {
3402                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3403                                 if (m0 == NULL) {
3404                                         error = ENOBUFS;
3405                                         continue;
3406                                 }
3407                                 bcopy(llc, mtod(m0, caddr_t),
3408                                     sizeof(struct llc));
3409                         }
3410                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3411                         if (m0 == NULL) {
3412                                 error = ENOBUFS;
3413                                 continue;
3414                         }
3415                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3416                 } else
3417                         m_freem(m);
3418         }
3419
3420         if (error == 0)
3421                 KMOD_IPSTAT_INC(ips_fragmented);
3422
3423         return (error);
3424
3425 out:
3426         if (m != NULL)
3427                 m_freem(m);
3428         return (error);
3429 }