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