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
  83 #include <sys/param.h>
  84 #include <sys/mbuf.h>
  85 #include <sys/malloc.h>
  86 #include <sys/protosw.h>
  87 #include <sys/systm.h>
  88 #include <sys/jail.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 #include <net/vnet.h>
 113
 114 #include <netinet/in.h> /* for struct arpcom */
 115 #include <netinet/in_systm.h>
 116 #include <netinet/in_var.h>
 117 #include <netinet/ip.h>
 118 #include <netinet/ip_var.h>
 119 #ifdef INET6
 120 #include <netinet/ip6.h>
 121 #include <netinet6/ip6_var.h>
 122 #endif
 123 #if defined(INET) || defined(INET6)
 124 #include <netinet/ip_carp.h>
 125 #endif
 126 #include <machine/in_cksum.h>
 127 #include <netinet/if_ether.h> /* for struct arpcom */
 128 #include <net/bridgestp.h>
 129 #include <net/if_bridgevar.h>
 130 #include <net/if_llc.h>
 131 #include <net/if_vlan_var.h>
 132
 133 #include <net/route.h>
 134 #include <netinet/ip_fw.h>
 135 #include <netinet/ipfw/ip_fw_private.h>
 136
 137 /*
 138  * Size of the route hash table.  Must be a power of two.
 139  */
 140 #ifndef BRIDGE_RTHASH_SIZE
 141 #define BRIDGE_RTHASH_SIZE              1024
 142 #endif
 143
 144 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 145
 146 /*
 147  * Maximum number of addresses to cache.
 148  */
 149 #ifndef BRIDGE_RTABLE_MAX
 150 #define BRIDGE_RTABLE_MAX               100
 151 #endif
 152
 153 /*
 154  * Timeout (in seconds) for entries learned dynamically.
 155  */
 156 #ifndef BRIDGE_RTABLE_TIMEOUT
 157 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 158 #endif
 159
 160 /*
 161  * Number of seconds between walks of the route list.
 162  */
 163 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 164 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 165 #endif
 166
 167 /*
 168  * List of capabilities to possibly mask on the member interface.
 169  */
 170 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 171
 172 /*
 173  * List of capabilities to strip
 174  */
 175 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 176
 177 /*
 178  * Bridge interface list entry.
 179  */
 180 struct bridge_iflist {
 181         LIST_ENTRY(bridge_iflist) bif_next;
 182         struct ifnet            *bif_ifp;       /* member if */
 183         struct bstp_port        bif_stp;        /* STP state */
 184         uint32_t                bif_flags;      /* member if flags */
 185         int                     bif_savedcaps;  /* saved capabilities */
 186         uint32_t                bif_addrmax;    /* max # of addresses */
 187         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 188         uint32_t                bif_addrexceeded;/* # of address violations */
 189 };
 190
 191 /*
 192  * Bridge route node.
 193  */
 194 struct bridge_rtnode {
 195         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 196         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 197         struct bridge_iflist    *brt_dst;       /* destination if */
 198         unsigned long           brt_expire;     /* expiration time */
 199         uint8_t                 brt_flags;      /* address flags */
 200         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 201         uint16_t                brt_vlan;       /* vlan id */
 202 };
 203 #define brt_ifp                 brt_dst->bif_ifp
 204
 205 /*
 206  * Software state for each bridge.
 207  */
 208 struct bridge_softc {
 209         struct ifnet            *sc_ifp;        /* make this an interface */
 210         LIST_ENTRY(bridge_softc) sc_list;
 211         struct mtx              sc_mtx;
 212         struct cv               sc_cv;
 213         uint32_t                sc_brtmax;      /* max # of addresses */
 214         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 215         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 216         struct callout          sc_brcallout;   /* bridge callout */
 217         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 218         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 219         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 220         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 221         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 222         uint32_t                sc_rthash_key;  /* key for hash */
 223         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 224         struct bstp_state       sc_stp;         /* STP state */
 225         uint32_t                sc_brtexceeded; /* # of cache drops */
 226         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 227         u_char                  sc_defaddr[6];  /* Default MAC address */
 228 };
 229
 230 static struct mtx       bridge_list_mtx;
 231 eventhandler_tag        bridge_detach_cookie = NULL;
 232
 233 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 234
 235 uma_zone_t bridge_rtnode_zone;
 236
 237 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 238 static void     bridge_clone_destroy(struct ifnet *);
 239
 240 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 241 static void     bridge_mutecaps(struct bridge_softc *);
 242 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 243                     int);
 244 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 245 static void     bridge_init(void *);
 246 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 247 static void     bridge_stop(struct ifnet *, int);
 248 static void     bridge_start(struct ifnet *);
 249 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 250 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 251                     struct rtentry *);
 252 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 253                     struct mbuf *);
 254 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 255
 256 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 257                     struct mbuf *m);
 258
 259 static void     bridge_timer(void *);
 260
 261 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 262                     struct mbuf *, int);
 263 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 264
 265 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 266                     uint16_t, struct bridge_iflist *, int, uint8_t);
 267 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 268                     uint16_t);
 269 static void     bridge_rttrim(struct bridge_softc *);
 270 static void     bridge_rtage(struct bridge_softc *);
 271 static void     bridge_rtflush(struct bridge_softc *, int);
 272 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 273                     uint16_t);
 274
 275 static int      bridge_rtable_init(struct bridge_softc *);
 276 static void     bridge_rtable_fini(struct bridge_softc *);
 277
 278 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 279 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 280                     const uint8_t *, uint16_t);
 281 static int      bridge_rtnode_insert(struct bridge_softc *,
 282                     struct bridge_rtnode *);
 283 static void     bridge_rtnode_destroy(struct bridge_softc *,
 284                     struct bridge_rtnode *);
 285 static void     bridge_rtable_expire(struct ifnet *, int);
 286 static void     bridge_state_change(struct ifnet *, int);
 287
 288 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 289                     const char *name);
 290 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 291                     struct ifnet *ifp);
 292 static void     bridge_delete_member(struct bridge_softc *,
 293                     struct bridge_iflist *, int);
 294 static void     bridge_delete_span(struct bridge_softc *,
 295                     struct bridge_iflist *);
 296
 297 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 298 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 328 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 329                     int);
 330 static int      bridge_ip_checkbasic(struct mbuf **mp);
 331 #ifdef INET6
 332 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 333 #endif /* INET6 */
 334 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 335                     struct ether_header *, int, struct llc *);
 336
 337 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 338 #define VLANTAGOF(_m)   \
 339     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 340
 341 static struct bstp_cb_ops bridge_ops = {
 342         .bcb_state = bridge_state_change,
 343         .bcb_rtage = bridge_rtable_expire
 344 };
 345
 346 SYSCTL_DECL(_net_link);
 347 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 348
 349 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 350 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 351 static int pfil_member = 1; /* run pfil hooks on the member interface */
 352 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 353 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 354 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 355                                    locally destined packets */
 356 static int log_stp   = 0;   /* log STP state changes */
 357 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
 358 TUNABLE_INT("net.link.bridge.pfil_onlyip", &pfil_onlyip);
 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 TUNABLE_INT("net.link.bridge.ipfw_arp", &pfil_ipfw_arp);
 362 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 363     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 364 TUNABLE_INT("net.link.bridge.pfil_bridge", &pfil_bridge);
 365 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 366     &pfil_bridge, 0, "Packet filter on the bridge interface");
 367 TUNABLE_INT("net.link.bridge.pfil_member", &pfil_member);
 368 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 369     &pfil_member, 0, "Packet filter on the member interface");
 370 TUNABLE_INT("net.link.bridge.pfil_local_phys", &pfil_local_phys);
 371 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 372     &pfil_local_phys, 0,
 373     "Packet filter on the physical interface for locally destined packets");
 374 TUNABLE_INT("net.link.bridge.log_stp", &log_stp);
 375 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 376     &log_stp, 0, "Log STP state changes");
 377 TUNABLE_INT("net.link.bridge.inherit_mac", &bridge_inherit_mac);
 378 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
 379     &bridge_inherit_mac, 0,
 380     "Inherit MAC address from the first bridge member");
 381
 382 struct bridge_control {
 383         int     (*bc_func)(struct bridge_softc *, void *);
 384         int     bc_argsize;
 385         int     bc_flags;
 386 };
 387
 388 #define BC_F_COPYIN             0x01    /* copy arguments in */
 389 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 390 #define BC_F_SUSER              0x04    /* do super-user check */
 391
 392 const struct bridge_control bridge_control_table[] = {
 393         { bridge_ioctl_add,             sizeof(struct ifbreq),
 394           BC_F_COPYIN|BC_F_SUSER },
 395         { bridge_ioctl_del,             sizeof(struct ifbreq),
 396           BC_F_COPYIN|BC_F_SUSER },
 397
 398         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 399           BC_F_COPYIN|BC_F_COPYOUT },
 400         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 401           BC_F_COPYIN|BC_F_SUSER },
 402
 403         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 404           BC_F_COPYIN|BC_F_SUSER },
 405         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 406           BC_F_COPYOUT },
 407
 408         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 409           BC_F_COPYIN|BC_F_COPYOUT },
 410         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 411           BC_F_COPYIN|BC_F_COPYOUT },
 412
 413         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 414           BC_F_COPYIN|BC_F_SUSER },
 415
 416         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 417           BC_F_COPYIN|BC_F_SUSER },
 418         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 419           BC_F_COPYOUT },
 420
 421         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 422           BC_F_COPYIN|BC_F_SUSER },
 423
 424         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 425           BC_F_COPYIN|BC_F_SUSER },
 426
 427         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 428           BC_F_COPYOUT },
 429         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 430           BC_F_COPYIN|BC_F_SUSER },
 431
 432         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 433           BC_F_COPYOUT },
 434         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 435           BC_F_COPYIN|BC_F_SUSER },
 436
 437         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 438           BC_F_COPYOUT },
 439         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 440           BC_F_COPYIN|BC_F_SUSER },
 441
 442         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 443           BC_F_COPYOUT },
 444         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 445           BC_F_COPYIN|BC_F_SUSER },
 446
 447         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 448           BC_F_COPYIN|BC_F_SUSER },
 449
 450         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 451           BC_F_COPYIN|BC_F_SUSER },
 452
 453         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 454           BC_F_COPYIN|BC_F_SUSER },
 455         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 456           BC_F_COPYIN|BC_F_SUSER },
 457
 458         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 459           BC_F_COPYOUT },
 460
 461         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 462           BC_F_COPYOUT },
 463
 464         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 465           BC_F_COPYIN|BC_F_COPYOUT },
 466
 467         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 468           BC_F_COPYIN|BC_F_SUSER },
 469
 470         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 471           BC_F_COPYIN|BC_F_SUSER },
 472
 473         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 474           BC_F_COPYIN|BC_F_SUSER },
 475
 476 };
 477 const int bridge_control_table_size =
 478     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 479
 480 LIST_HEAD(, bridge_softc) bridge_list;
 481
 482 IFC_SIMPLE_DECLARE(bridge, 0);
 483
 484 static int
 485 bridge_modevent(module_t mod, int type, void *data)
 486 {
 487
 488         switch (type) {
 489         case MOD_LOAD:
 490                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 491                 if_clone_attach(&bridge_cloner);
 492                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 493                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 494                     UMA_ALIGN_PTR, 0);
 495                 LIST_INIT(&bridge_list);
 496                 bridge_input_p = bridge_input;
 497                 bridge_output_p = bridge_output;
 498                 bridge_dn_p = bridge_dummynet;
 499                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 500                     ifnet_departure_event, bridge_ifdetach, NULL,
 501                     EVENTHANDLER_PRI_ANY);
 502                 break;
 503         case MOD_UNLOAD:
 504                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 505                     bridge_detach_cookie);
 506                 if_clone_detach(&bridge_cloner);
 507                 uma_zdestroy(bridge_rtnode_zone);
 508                 bridge_input_p = NULL;
 509                 bridge_output_p = NULL;
 510                 bridge_dn_p = NULL;
 511                 mtx_destroy(&bridge_list_mtx);
 512                 break;
 513         default:
 514                 return (EOPNOTSUPP);
 515         }
 516         return (0);
 517 }
 518
 519 static moduledata_t bridge_mod = {
 520         "if_bridge",
 521         bridge_modevent,
 522         0
 523 };
 524
 525 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 526 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 527
 528 /*
 529  * handler for net.link.bridge.pfil_ipfw
 530  */
 531 static int
 532 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 533 {
 534         int enable = pfil_ipfw;
 535         int error;
 536
 537         error = sysctl_handle_int(oidp, &enable, 0, req);
 538         enable = (enable) ? 1 : 0;
 539
 540         if (enable != pfil_ipfw) {
 541                 pfil_ipfw = enable;
 542
 543                 /*
 544                  * Disable pfil so that ipfw doesnt run twice, if the user
 545                  * really wants both then they can re-enable pfil_bridge and/or
 546                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 547                  * layer2 type.
 548                  */
 549                 if (pfil_ipfw) {
 550                         pfil_onlyip = 0;
 551                         pfil_bridge = 0;
 552                         pfil_member = 0;
 553                 }
 554         }
 555
 556         return (error);
 557 }
 558 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 559             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 560
 561 /*
 562  * bridge_clone_create:
 563  *
 564  *      Create a new bridge instance.
 565  */
 566 static int
 567 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 568 {
 569         struct bridge_softc *sc, *sc2;
 570         struct ifnet *bifp, *ifp;
 571         int fb, retry;
 572         unsigned long hostid;
 573
 574         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 575         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 576         if (ifp == NULL) {
 577                 free(sc, M_DEVBUF);
 578                 return (ENOSPC);
 579         }
 580
 581         BRIDGE_LOCK_INIT(sc);
 582         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 583         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 584
 585         /* Initialize our routing table. */
 586         bridge_rtable_init(sc);
 587
 588         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 589
 590         LIST_INIT(&sc->sc_iflist);
 591         LIST_INIT(&sc->sc_spanlist);
 592
 593         ifp->if_softc = sc;
 594         if_initname(ifp, ifc->ifc_name, unit);
 595         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 596         ifp->if_ioctl = bridge_ioctl;
 597         ifp->if_start = bridge_start;
 598         ifp->if_init = bridge_init;
 599         ifp->if_type = IFT_BRIDGE;
 600         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 601         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 602         IFQ_SET_READY(&ifp->if_snd);
 603
 604         /*
 605          * Generate an ethernet address with a locally administered address.
 606          *
 607          * Since we are using random ethernet addresses for the bridge, it is
 608          * possible that we might have address collisions, so make sure that
 609          * this hardware address isn't already in use on another bridge.
 610          * The first try uses the hostid and falls back to arc4rand().
 611          */
 612         fb = 0;
 613         getcredhostid(curthread->td_ucred, &hostid);
 614         for (retry = 1; retry != 0;) {
 615                 if (fb || hostid == 0) {
 616                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 617                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 618                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 619                 } else {
 620                         sc->sc_defaddr[0] = 0x2;
 621                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 622                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 623                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 624                         sc->sc_defaddr[4] =  hostid        & 0xff;
 625                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 626                 }
 627
 628                 fb = 1;
 629                 retry = 0;
 630                 mtx_lock(&bridge_list_mtx);
 631                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 632                         bifp = sc2->sc_ifp;
 633                         if (memcmp(sc->sc_defaddr,
 634                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 635                                 retry = 1;
 636                 }
 637                 mtx_unlock(&bridge_list_mtx);
 638         }
 639
 640         bstp_attach(&sc->sc_stp, &bridge_ops);
 641         ether_ifattach(ifp, sc->sc_defaddr);
 642         /* Now undo some of the damage... */
 643         ifp->if_baudrate = 0;
 644         ifp->if_type = IFT_BRIDGE;
 645
 646         mtx_lock(&bridge_list_mtx);
 647         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 648         mtx_unlock(&bridge_list_mtx);
 649
 650         return (0);
 651 }
 652
 653 /*
 654  * bridge_clone_destroy:
 655  *
 656  *      Destroy a bridge instance.
 657  */
 658 static void
 659 bridge_clone_destroy(struct ifnet *ifp)
 660 {
 661         struct bridge_softc *sc = ifp->if_softc;
 662         struct bridge_iflist *bif;
 663
 664         BRIDGE_LOCK(sc);
 665
 666         bridge_stop(ifp, 1);
 667         ifp->if_flags &= ~IFF_UP;
 668
 669         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 670                 bridge_delete_member(sc, bif, 0);
 671
 672         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 673                 bridge_delete_span(sc, bif);
 674         }
 675
 676         BRIDGE_UNLOCK(sc);
 677
 678         callout_drain(&sc->sc_brcallout);
 679
 680         mtx_lock(&bridge_list_mtx);
 681         LIST_REMOVE(sc, sc_list);
 682         mtx_unlock(&bridge_list_mtx);
 683
 684         bstp_detach(&sc->sc_stp);
 685         ether_ifdetach(ifp);
 686         if_free(ifp);
 687
 688         /* Tear down the routing table. */
 689         bridge_rtable_fini(sc);
 690
 691         BRIDGE_LOCK_DESTROY(sc);
 692         free(sc, M_DEVBUF);
 693 }
 694
 695 /*
 696  * bridge_ioctl:
 697  *
 698  *      Handle a control request from the operator.
 699  */
 700 static int
 701 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 702 {
 703         struct bridge_softc *sc = ifp->if_softc;
 704         struct ifreq *ifr = (struct ifreq *)data;
 705         struct bridge_iflist *bif;
 706         struct thread *td = curthread;
 707         union {
 708                 struct ifbreq ifbreq;
 709                 struct ifbifconf ifbifconf;
 710                 struct ifbareq ifbareq;
 711                 struct ifbaconf ifbaconf;
 712                 struct ifbrparam ifbrparam;
 713                 struct ifbropreq ifbropreq;
 714         } args;
 715         struct ifdrv *ifd = (struct ifdrv *) data;
 716         const struct bridge_control *bc;
 717         int error = 0;
 718
 719         switch (cmd) {
 720
 721         case SIOCADDMULTI:
 722         case SIOCDELMULTI:
 723                 break;
 724
 725         case SIOCGDRVSPEC:
 726         case SIOCSDRVSPEC:
 727                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 728                         error = EINVAL;
 729                         break;
 730                 }
 731                 bc = &bridge_control_table[ifd->ifd_cmd];
 732
 733                 if (cmd == SIOCGDRVSPEC &&
 734                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 735                         error = EINVAL;
 736                         break;
 737                 }
 738                 else if (cmd == SIOCSDRVSPEC &&
 739                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 740                         error = EINVAL;
 741                         break;
 742                 }
 743
 744                 if (bc->bc_flags & BC_F_SUSER) {
 745                         error = priv_check(td, PRIV_NET_BRIDGE);
 746                         if (error)
 747                                 break;
 748                 }
 749
 750                 if (ifd->ifd_len != bc->bc_argsize ||
 751                     ifd->ifd_len > sizeof(args)) {
 752                         error = EINVAL;
 753                         break;
 754                 }
 755
 756                 bzero(&args, sizeof(args));
 757                 if (bc->bc_flags & BC_F_COPYIN) {
 758                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 759                         if (error)
 760                                 break;
 761                 }
 762
 763                 BRIDGE_LOCK(sc);
 764                 error = (*bc->bc_func)(sc, &args);
 765                 BRIDGE_UNLOCK(sc);
 766                 if (error)
 767                         break;
 768
 769                 if (bc->bc_flags & BC_F_COPYOUT)
 770                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 771
 772                 break;
 773
 774         case SIOCSIFFLAGS:
 775                 if (!(ifp->if_flags & IFF_UP) &&
 776                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 777                         /*
 778                          * If interface is marked down and it is running,
 779                          * then stop and disable it.
 780                          */
 781                         BRIDGE_LOCK(sc);
 782                         bridge_stop(ifp, 1);
 783                         BRIDGE_UNLOCK(sc);
 784                 } else if ((ifp->if_flags & IFF_UP) &&
 785                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 786                         /*
 787                          * If interface is marked up and it is stopped, then
 788                          * start it.
 789                          */
 790                         (*ifp->if_init)(sc);
 791                 }
 792                 break;
 793
 794         case SIOCSIFMTU:
 795                 if (ifr->ifr_mtu < 576) {
 796                         error = EINVAL;
 797                         break;
 798                 }
 799                 if (LIST_EMPTY(&sc->sc_iflist)) {
 800                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 801                         break;
 802                 }
 803                 BRIDGE_LOCK(sc);
 804                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 805                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 806                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
 807                                     " != %d\n", sc->sc_ifp->if_xname,
 808                                     bif->bif_ifp->if_mtu,
 809                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 810                                 error = EINVAL;
 811                                 break;
 812                         }
 813                 }
 814                 if (!error)
 815                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 816                 BRIDGE_UNLOCK(sc);
 817                 break;
 818         default:
 819                 /*
 820                  * drop the lock as ether_ioctl() will call bridge_start() and
 821                  * cause the lock to be recursed.
 822                  */
 823                 error = ether_ioctl(ifp, cmd, data);
 824                 break;
 825         }
 826
 827         return (error);
 828 }
 829
 830 /*
 831  * bridge_mutecaps:
 832  *
 833  *      Clear or restore unwanted capabilities on the member interface
 834  */
 835 static void
 836 bridge_mutecaps(struct bridge_softc *sc)
 837 {
 838         struct bridge_iflist *bif;
 839         int enabled, mask;
 840
 841         /* Initial bitmask of capabilities to test */
 842         mask = BRIDGE_IFCAPS_MASK;
 843
 844         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 845                 /* Every member must support it or its disabled */
 846                 mask &= bif->bif_savedcaps;
 847         }
 848
 849         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 850                 enabled = bif->bif_ifp->if_capenable;
 851                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 852                 /* strip off mask bits and enable them again if allowed */
 853                 enabled &= ~BRIDGE_IFCAPS_MASK;
 854                 enabled |= mask;
 855                 bridge_set_ifcap(sc, bif, enabled);
 856         }
 857
 858 }
 859
 860 static void
 861 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 862 {
 863         struct ifnet *ifp = bif->bif_ifp;
 864         struct ifreq ifr;
 865         int error;
 866
 867         bzero(&ifr, sizeof(ifr));
 868         ifr.ifr_reqcap = set;
 869
 870         if (ifp->if_capenable != set) {
 871                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 872                 if (error)
 873                         if_printf(sc->sc_ifp,
 874                             "error setting interface capabilities on %s\n",
 875                             ifp->if_xname);
 876         }
 877 }
 878
 879 /*
 880  * bridge_lookup_member:
 881  *
 882  *      Lookup a bridge member interface.
 883  */
 884 static struct bridge_iflist *
 885 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 886 {
 887         struct bridge_iflist *bif;
 888         struct ifnet *ifp;
 889
 890         BRIDGE_LOCK_ASSERT(sc);
 891
 892         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 893                 ifp = bif->bif_ifp;
 894                 if (strcmp(ifp->if_xname, name) == 0)
 895                         return (bif);
 896         }
 897
 898         return (NULL);
 899 }
 900
 901 /*
 902  * bridge_lookup_member_if:
 903  *
 904  *      Lookup a bridge member interface by ifnet*.
 905  */
 906 static struct bridge_iflist *
 907 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 908 {
 909         struct bridge_iflist *bif;
 910
 911         BRIDGE_LOCK_ASSERT(sc);
 912
 913         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 914                 if (bif->bif_ifp == member_ifp)
 915                         return (bif);
 916         }
 917
 918         return (NULL);
 919 }
 920
 921 /*
 922  * bridge_delete_member:
 923  *
 924  *      Delete the specified member interface.
 925  */
 926 static void
 927 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 928     int gone)
 929 {
 930         struct ifnet *ifs = bif->bif_ifp;
 931         struct ifnet *fif = NULL;
 932
 933         BRIDGE_LOCK_ASSERT(sc);
 934
 935         if (bif->bif_flags & IFBIF_STP)
 936                 bstp_disable(&bif->bif_stp);
 937
 938         ifs->if_bridge = NULL;
 939         BRIDGE_XLOCK(sc);
 940         LIST_REMOVE(bif, bif_next);
 941         BRIDGE_XDROP(sc);
 942
 943         /*
 944          * If removing the interface that gave the bridge its mac address, set
 945          * the mac address of the bridge to the address of the next member, or
 946          * to its default address if no members are left.
 947          */
 948         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
 949                 if (LIST_EMPTY(&sc->sc_iflist)) {
 950                         bcopy(sc->sc_defaddr,
 951                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 952                         sc->sc_ifaddr = NULL;
 953                 } else {
 954                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
 955                         bcopy(IF_LLADDR(fif),
 956                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 957                         sc->sc_ifaddr = fif;
 958                 }
 959                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
 960         }
 961
 962         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 963         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 964         KASSERT(bif->bif_addrcnt == 0,
 965             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 966
 967         BRIDGE_UNLOCK(sc);
 968         if (!gone) {
 969                 switch (ifs->if_type) {
 970                 case IFT_ETHER:
 971                 case IFT_L2VLAN:
 972                         /*
 973                          * Take the interface out of promiscuous mode.
 974                          */
 975                         (void) ifpromisc(ifs, 0);
 976                         break;
 977
 978                 case IFT_GIF:
 979                         break;
 980
 981                 default:
 982 #ifdef DIAGNOSTIC
 983                         panic("bridge_delete_member: impossible");
 984 #endif
 985                         break;
 986                 }
 987                 /* reneable any interface capabilities */
 988                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
 989         }
 990         bstp_destroy(&bif->bif_stp);    /* prepare to free */
 991         BRIDGE_LOCK(sc);
 992         free(bif, M_DEVBUF);
 993 }
 994
 995 /*
 996  * bridge_delete_span:
 997  *
 998  *      Delete the specified span interface.
 999  */
1000 static void
1001 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1002 {
1003         BRIDGE_LOCK_ASSERT(sc);
1004
1005         KASSERT(bif->bif_ifp->if_bridge == NULL,
1006             ("%s: not a span interface", __func__));
1007
1008         LIST_REMOVE(bif, bif_next);
1009         free(bif, M_DEVBUF);
1010 }
1011
1012 static int
1013 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1014 {
1015         struct ifbreq *req = arg;
1016         struct bridge_iflist *bif = NULL;
1017         struct ifnet *ifs;
1018         int error = 0;
1019
1020         ifs = ifunit(req->ifbr_ifsname);
1021         if (ifs == NULL)
1022                 return (ENOENT);
1023         if (ifs->if_ioctl == NULL)      /* must be supported */
1024                 return (EINVAL);
1025
1026         /* If it's in the span list, it can't be a member. */
1027         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1028                 if (ifs == bif->bif_ifp)
1029                         return (EBUSY);
1030
1031         if (ifs->if_bridge == sc)
1032                 return (EEXIST);
1033
1034         if (ifs->if_bridge != NULL)
1035                 return (EBUSY);
1036
1037         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1038         if (bif == NULL)
1039                 return (ENOMEM);
1040
1041         bif->bif_ifp = ifs;
1042         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1043         bif->bif_savedcaps = ifs->if_capenable;
1044
1045         switch (ifs->if_type) {
1046         case IFT_ETHER:
1047         case IFT_L2VLAN:
1048         case IFT_GIF:
1049                 /* permitted interface types */
1050                 break;
1051         default:
1052                 error = EINVAL;
1053                 goto out;
1054         }
1055
1056         /* Allow the first Ethernet member to define the MTU */
1057         if (LIST_EMPTY(&sc->sc_iflist))
1058                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1059         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1060                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1061                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1062                 error = EINVAL;
1063                 goto out;
1064         }
1065
1066         /*
1067          * Assign the interface's MAC address to the bridge if it's the first
1068          * member and the MAC address of the bridge has not been changed from
1069          * the default randomly generated one.
1070          */
1071         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1072             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1073                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1074                 sc->sc_ifaddr = ifs;
1075                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1076         }
1077
1078         ifs->if_bridge = sc;
1079         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1080         /*
1081          * XXX: XLOCK HERE!?!
1082          *
1083          * NOTE: insert_***HEAD*** should be safe for the traversals.
1084          */
1085         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1086
1087         /* Set interface capabilities to the intersection set of all members */
1088         bridge_mutecaps(sc);
1089
1090         switch (ifs->if_type) {
1091         case IFT_ETHER:
1092         case IFT_L2VLAN:
1093                 /*
1094                  * Place the interface into promiscuous mode.
1095                  */
1096                 BRIDGE_UNLOCK(sc);
1097                 error = ifpromisc(ifs, 1);
1098                 BRIDGE_LOCK(sc);
1099                 break;
1100         }
1101         if (error)
1102                 bridge_delete_member(sc, bif, 0);
1103 out:
1104         if (error) {
1105                 if (bif != NULL)
1106                         free(bif, M_DEVBUF);
1107         }
1108         return (error);
1109 }
1110
1111 static int
1112 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1113 {
1114         struct ifbreq *req = arg;
1115         struct bridge_iflist *bif;
1116
1117         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1118         if (bif == NULL)
1119                 return (ENOENT);
1120
1121         bridge_delete_member(sc, bif, 0);
1122
1123         return (0);
1124 }
1125
1126 static int
1127 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1128 {
1129         struct ifbreq *req = arg;
1130         struct bridge_iflist *bif;
1131         struct bstp_port *bp;
1132
1133         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1134         if (bif == NULL)
1135                 return (ENOENT);
1136
1137         bp = &bif->bif_stp;
1138         req->ifbr_ifsflags = bif->bif_flags;
1139         req->ifbr_state = bp->bp_state;
1140         req->ifbr_priority = bp->bp_priority;
1141         req->ifbr_path_cost = bp->bp_path_cost;
1142         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1143         req->ifbr_proto = bp->bp_protover;
1144         req->ifbr_role = bp->bp_role;
1145         req->ifbr_stpflags = bp->bp_flags;
1146         req->ifbr_addrcnt = bif->bif_addrcnt;
1147         req->ifbr_addrmax = bif->bif_addrmax;
1148         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1149
1150         /* Copy STP state options as flags */
1151         if (bp->bp_operedge)
1152                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1153         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1154                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1155         if (bp->bp_ptp_link)
1156                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1157         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1158                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1159         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1160                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1161         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1162                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1163         return (0);
1164 }
1165
1166 static int
1167 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1168 {
1169         struct ifbreq *req = arg;
1170         struct bridge_iflist *bif;
1171         struct bstp_port *bp;
1172         int error;
1173
1174         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1175         if (bif == NULL)
1176                 return (ENOENT);
1177         bp = &bif->bif_stp;
1178
1179         if (req->ifbr_ifsflags & IFBIF_SPAN)
1180                 /* SPAN is readonly */
1181                 return (EINVAL);
1182
1183         if (req->ifbr_ifsflags & IFBIF_STP) {
1184                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1185                         error = bstp_enable(&bif->bif_stp);
1186                         if (error)
1187                                 return (error);
1188                 }
1189         } else {
1190                 if ((bif->bif_flags & IFBIF_STP) != 0)
1191                         bstp_disable(&bif->bif_stp);
1192         }
1193
1194         /* Pass on STP flags */
1195         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1196         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1197         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1198         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1199
1200         /* Save the bits relating to the bridge */
1201         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1202
1203         return (0);
1204 }
1205
1206 static int
1207 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1208 {
1209         struct ifbrparam *param = arg;
1210
1211         sc->sc_brtmax = param->ifbrp_csize;
1212         bridge_rttrim(sc);
1213
1214         return (0);
1215 }
1216
1217 static int
1218 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1219 {
1220         struct ifbrparam *param = arg;
1221
1222         param->ifbrp_csize = sc->sc_brtmax;
1223
1224         return (0);
1225 }
1226
1227 static int
1228 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1229 {
1230         struct ifbifconf *bifc = arg;
1231         struct bridge_iflist *bif;
1232         struct ifbreq breq;
1233         char *buf, *outbuf;
1234         int count, buflen, len, error = 0;
1235
1236         count = 0;
1237         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1238                 count++;
1239         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1240                 count++;
1241
1242         buflen = sizeof(breq) * count;
1243         if (bifc->ifbic_len == 0) {
1244                 bifc->ifbic_len = buflen;
1245                 return (0);
1246         }
1247         BRIDGE_UNLOCK(sc);
1248         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1249         BRIDGE_LOCK(sc);
1250
1251         count = 0;
1252         buf = outbuf;
1253         len = min(bifc->ifbic_len, buflen);
1254         bzero(&breq, sizeof(breq));
1255         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1256                 if (len < sizeof(breq))
1257                         break;
1258
1259                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1260                     sizeof(breq.ifbr_ifsname));
1261                 /* Fill in the ifbreq structure */
1262                 error = bridge_ioctl_gifflags(sc, &breq);
1263                 if (error)
1264                         break;
1265                 memcpy(buf, &breq, sizeof(breq));
1266                 count++;
1267                 buf += sizeof(breq);
1268                 len -= sizeof(breq);
1269         }
1270         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1271                 if (len < sizeof(breq))
1272                         break;
1273
1274                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1275                     sizeof(breq.ifbr_ifsname));
1276                 breq.ifbr_ifsflags = bif->bif_flags;
1277                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1278                 memcpy(buf, &breq, sizeof(breq));
1279                 count++;
1280                 buf += sizeof(breq);
1281                 len -= sizeof(breq);
1282         }
1283
1284         BRIDGE_UNLOCK(sc);
1285         bifc->ifbic_len = sizeof(breq) * count;
1286         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1287         BRIDGE_LOCK(sc);
1288         free(outbuf, M_TEMP);
1289         return (error);
1290 }
1291
1292 static int
1293 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1294 {
1295         struct ifbaconf *bac = arg;
1296         struct bridge_rtnode *brt;
1297         struct ifbareq bareq;
1298         char *buf, *outbuf;
1299         int count, buflen, len, error = 0;
1300
1301         if (bac->ifbac_len == 0)
1302                 return (0);
1303
1304         count = 0;
1305         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1306                 count++;
1307         buflen = sizeof(bareq) * count;
1308
1309         BRIDGE_UNLOCK(sc);
1310         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1311         BRIDGE_LOCK(sc);
1312
1313         count = 0;
1314         buf = outbuf;
1315         len = min(bac->ifbac_len, buflen);
1316         bzero(&bareq, sizeof(bareq));
1317         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1318                 if (len < sizeof(bareq))
1319                         goto out;
1320                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1321                     sizeof(bareq.ifba_ifsname));
1322                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1323                 bareq.ifba_vlan = brt->brt_vlan;
1324                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1325                                 time_uptime < brt->brt_expire)
1326                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1327                 else
1328                         bareq.ifba_expire = 0;
1329                 bareq.ifba_flags = brt->brt_flags;
1330
1331                 memcpy(buf, &bareq, sizeof(bareq));
1332                 count++;
1333                 buf += sizeof(bareq);
1334                 len -= sizeof(bareq);
1335         }
1336 out:
1337         BRIDGE_UNLOCK(sc);
1338         bac->ifbac_len = sizeof(bareq) * count;
1339         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1340         BRIDGE_LOCK(sc);
1341         free(outbuf, M_TEMP);
1342         return (error);
1343 }
1344
1345 static int
1346 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1347 {
1348         struct ifbareq *req = arg;
1349         struct bridge_iflist *bif;
1350         int error;
1351
1352         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1353         if (bif == NULL)
1354                 return (ENOENT);
1355
1356         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1357             req->ifba_flags);
1358
1359         return (error);
1360 }
1361
1362 static int
1363 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1364 {
1365         struct ifbrparam *param = arg;
1366
1367         sc->sc_brttimeout = param->ifbrp_ctime;
1368         return (0);
1369 }
1370
1371 static int
1372 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1373 {
1374         struct ifbrparam *param = arg;
1375
1376         param->ifbrp_ctime = sc->sc_brttimeout;
1377         return (0);
1378 }
1379
1380 static int
1381 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1382 {
1383         struct ifbareq *req = arg;
1384
1385         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1386 }
1387
1388 static int
1389 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1390 {
1391         struct ifbreq *req = arg;
1392
1393         bridge_rtflush(sc, req->ifbr_ifsflags);
1394         return (0);
1395 }
1396
1397 static int
1398 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1399 {
1400         struct ifbrparam *param = arg;
1401         struct bstp_state *bs = &sc->sc_stp;
1402
1403         param->ifbrp_prio = bs->bs_bridge_priority;
1404         return (0);
1405 }
1406
1407 static int
1408 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1409 {
1410         struct ifbrparam *param = arg;
1411
1412         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1413 }
1414
1415 static int
1416 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1417 {
1418         struct ifbrparam *param = arg;
1419         struct bstp_state *bs = &sc->sc_stp;
1420
1421         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1422         return (0);
1423 }
1424
1425 static int
1426 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1427 {
1428         struct ifbrparam *param = arg;
1429
1430         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1431 }
1432
1433 static int
1434 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1435 {
1436         struct ifbrparam *param = arg;
1437         struct bstp_state *bs = &sc->sc_stp;
1438
1439         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1440         return (0);
1441 }
1442
1443 static int
1444 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1445 {
1446         struct ifbrparam *param = arg;
1447
1448         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1449 }
1450
1451 static int
1452 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1453 {
1454         struct ifbrparam *param = arg;
1455         struct bstp_state *bs = &sc->sc_stp;
1456
1457         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1458         return (0);
1459 }
1460
1461 static int
1462 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1463 {
1464         struct ifbrparam *param = arg;
1465
1466         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1467 }
1468
1469 static int
1470 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1471 {
1472         struct ifbreq *req = arg;
1473         struct bridge_iflist *bif;
1474
1475         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1476         if (bif == NULL)
1477                 return (ENOENT);
1478
1479         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1480 }
1481
1482 static int
1483 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1484 {
1485         struct ifbreq *req = arg;
1486         struct bridge_iflist *bif;
1487
1488         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1489         if (bif == NULL)
1490                 return (ENOENT);
1491
1492         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1493 }
1494
1495 static int
1496 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1497 {
1498         struct ifbreq *req = arg;
1499         struct bridge_iflist *bif;
1500
1501         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1502         if (bif == NULL)
1503                 return (ENOENT);
1504
1505         bif->bif_addrmax = req->ifbr_addrmax;
1506         return (0);
1507 }
1508
1509 static int
1510 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1511 {
1512         struct ifbreq *req = arg;
1513         struct bridge_iflist *bif = NULL;
1514         struct ifnet *ifs;
1515
1516         ifs = ifunit(req->ifbr_ifsname);
1517         if (ifs == NULL)
1518                 return (ENOENT);
1519
1520         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1521                 if (ifs == bif->bif_ifp)
1522                         return (EBUSY);
1523
1524         if (ifs->if_bridge != NULL)
1525                 return (EBUSY);
1526
1527         switch (ifs->if_type) {
1528                 case IFT_ETHER:
1529                 case IFT_GIF:
1530                 case IFT_L2VLAN:
1531                         break;
1532                 default:
1533                         return (EINVAL);
1534         }
1535
1536         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1537         if (bif == NULL)
1538                 return (ENOMEM);
1539
1540         bif->bif_ifp = ifs;
1541         bif->bif_flags = IFBIF_SPAN;
1542
1543         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1544
1545         return (0);
1546 }
1547
1548 static int
1549 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1550 {
1551         struct ifbreq *req = arg;
1552         struct bridge_iflist *bif;
1553         struct ifnet *ifs;
1554
1555         ifs = ifunit(req->ifbr_ifsname);
1556         if (ifs == NULL)
1557                 return (ENOENT);
1558
1559         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1560                 if (ifs == bif->bif_ifp)
1561                         break;
1562
1563         if (bif == NULL)
1564                 return (ENOENT);
1565
1566         bridge_delete_span(sc, bif);
1567
1568         return (0);
1569 }
1570
1571 static int
1572 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1573 {
1574         struct ifbropreq *req = arg;
1575         struct bstp_state *bs = &sc->sc_stp;
1576         struct bstp_port *root_port;
1577
1578         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1579         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1580         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1581
1582         root_port = bs->bs_root_port;
1583         if (root_port == NULL)
1584                 req->ifbop_root_port = 0;
1585         else
1586                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1587
1588         req->ifbop_holdcount = bs->bs_txholdcount;
1589         req->ifbop_priority = bs->bs_bridge_priority;
1590         req->ifbop_protocol = bs->bs_protover;
1591         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1592         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1593         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1594         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1595         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1596         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1597
1598         return (0);
1599 }
1600
1601 static int
1602 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1603 {
1604         struct ifbrparam *param = arg;
1605
1606         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1607         return (0);
1608 }
1609
1610 static int
1611 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1612 {
1613         struct ifbpstpconf *bifstp = arg;
1614         struct bridge_iflist *bif;
1615         struct bstp_port *bp;
1616         struct ifbpstpreq bpreq;
1617         char *buf, *outbuf;
1618         int count, buflen, len, error = 0;
1619
1620         count = 0;
1621         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1622                 if ((bif->bif_flags & IFBIF_STP) != 0)
1623                         count++;
1624         }
1625
1626         buflen = sizeof(bpreq) * count;
1627         if (bifstp->ifbpstp_len == 0) {
1628                 bifstp->ifbpstp_len = buflen;
1629                 return (0);
1630         }
1631
1632         BRIDGE_UNLOCK(sc);
1633         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1634         BRIDGE_LOCK(sc);
1635
1636         count = 0;
1637         buf = outbuf;
1638         len = min(bifstp->ifbpstp_len, buflen);
1639         bzero(&bpreq, sizeof(bpreq));
1640         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1641                 if (len < sizeof(bpreq))
1642                         break;
1643
1644                 if ((bif->bif_flags & IFBIF_STP) == 0)
1645                         continue;
1646
1647                 bp = &bif->bif_stp;
1648                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1649                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1650                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1651                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1652                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1653                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1654
1655                 memcpy(buf, &bpreq, sizeof(bpreq));
1656                 count++;
1657                 buf += sizeof(bpreq);
1658                 len -= sizeof(bpreq);
1659         }
1660
1661         BRIDGE_UNLOCK(sc);
1662         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1663         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1664         BRIDGE_LOCK(sc);
1665         free(outbuf, M_TEMP);
1666         return (error);
1667 }
1668
1669 static int
1670 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1671 {
1672         struct ifbrparam *param = arg;
1673
1674         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1675 }
1676
1677 static int
1678 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1679 {
1680         struct ifbrparam *param = arg;
1681
1682         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1683 }
1684
1685 /*
1686  * bridge_ifdetach:
1687  *
1688  *      Detach an interface from a bridge.  Called when a member
1689  *      interface is detaching.
1690  */
1691 static void
1692 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1693 {
1694         struct bridge_softc *sc = ifp->if_bridge;
1695         struct bridge_iflist *bif;
1696
1697         /* Check if the interface is a bridge member */
1698         if (sc != NULL) {
1699                 BRIDGE_LOCK(sc);
1700
1701                 bif = bridge_lookup_member_if(sc, ifp);
1702                 if (bif != NULL)
1703                         bridge_delete_member(sc, bif, 1);
1704
1705                 BRIDGE_UNLOCK(sc);
1706                 return;
1707         }
1708
1709         /* Check if the interface is a span port */
1710         mtx_lock(&bridge_list_mtx);
1711         LIST_FOREACH(sc, &bridge_list, sc_list) {
1712                 BRIDGE_LOCK(sc);
1713                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1714                         if (ifp == bif->bif_ifp) {
1715                                 bridge_delete_span(sc, bif);
1716                                 break;
1717                         }
1718
1719                 BRIDGE_UNLOCK(sc);
1720         }
1721         mtx_unlock(&bridge_list_mtx);
1722 }
1723
1724 /*
1725  * bridge_init:
1726  *
1727  *      Initialize a bridge interface.
1728  */
1729 static void
1730 bridge_init(void *xsc)
1731 {
1732         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1733         struct ifnet *ifp = sc->sc_ifp;
1734
1735         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1736                 return;
1737
1738         BRIDGE_LOCK(sc);
1739         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1740             bridge_timer, sc);
1741
1742         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1743         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1744
1745         BRIDGE_UNLOCK(sc);
1746 }
1747
1748 /*
1749  * bridge_stop:
1750  *
1751  *      Stop the bridge interface.
1752  */
1753 static void
1754 bridge_stop(struct ifnet *ifp, int disable)
1755 {
1756         struct bridge_softc *sc = ifp->if_softc;
1757
1758         BRIDGE_LOCK_ASSERT(sc);
1759
1760         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1761                 return;
1762
1763         callout_stop(&sc->sc_brcallout);
1764         bstp_stop(&sc->sc_stp);
1765
1766         bridge_rtflush(sc, IFBF_FLUSHDYN);
1767
1768         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1769 }
1770
1771 /*
1772  * bridge_enqueue:
1773  *
1774  *      Enqueue a packet on a bridge member interface.
1775  *
1776  */
1777 static void
1778 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1779 {
1780         int len, err = 0;
1781         short mflags;
1782         struct mbuf *m0;
1783
1784         len = m->m_pkthdr.len;
1785         mflags = m->m_flags;
1786
1787         /* We may be sending a fragment so traverse the mbuf */
1788         for (; m; m = m0) {
1789                 m0 = m->m_nextpkt;
1790                 m->m_nextpkt = NULL;
1791
1792                 /*
1793                  * If underlying interface can not do VLAN tag insertion itself
1794                  * then attach a packet tag that holds it.
1795                  */
1796                 if ((m->m_flags & M_VLANTAG) &&
1797                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1798                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1799                         if (m == NULL) {
1800                                 if_printf(dst_ifp,
1801                                     "unable to prepend VLAN header\n");
1802                                 dst_ifp->if_oerrors++;
1803                                 continue;
1804                         }
1805                         m->m_flags &= ~M_VLANTAG;
1806                 }
1807
1808                 if (err == 0)
1809                         dst_ifp->if_transmit(dst_ifp, m);
1810         }
1811
1812         if (err == 0) {
1813                 sc->sc_ifp->if_opackets++;
1814                 sc->sc_ifp->if_obytes += len;
1815                 if (mflags & M_MCAST)
1816                         sc->sc_ifp->if_omcasts++;
1817         }
1818 }
1819
1820 /*
1821  * bridge_dummynet:
1822  *
1823  *      Receive a queued packet from dummynet and pass it on to the output
1824  *      interface.
1825  *
1826  *      The mbuf has the Ethernet header already attached.
1827  */
1828 static void
1829 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1830 {
1831         struct bridge_softc *sc;
1832
1833         sc = ifp->if_bridge;
1834
1835         /*
1836          * The packet didnt originate from a member interface. This should only
1837          * ever happen if a member interface is removed while packets are
1838          * queued for it.
1839          */
1840         if (sc == NULL) {
1841                 m_freem(m);
1842                 return;
1843         }
1844
1845         if (PFIL_HOOKED(&V_inet_pfil_hook)
1846 #ifdef INET6
1847             || PFIL_HOOKED(&V_inet6_pfil_hook)
1848 #endif
1849             ) {
1850                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1851                         return;
1852                 if (m == NULL)
1853                         return;
1854         }
1855
1856         bridge_enqueue(sc, ifp, m);
1857 }
1858
1859 /*
1860  * bridge_output:
1861  *
1862  *      Send output from a bridge member interface.  This
1863  *      performs the bridging function for locally originated
1864  *      packets.
1865  *
1866  *      The mbuf has the Ethernet header already attached.  We must
1867  *      enqueue or free the mbuf before returning.
1868  */
1869 static int
1870 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1871     struct rtentry *rt)
1872 {
1873         struct ether_header *eh;
1874         struct ifnet *dst_if;
1875         struct bridge_softc *sc;
1876         uint16_t vlan;
1877
1878         if (m->m_len < ETHER_HDR_LEN) {
1879                 m = m_pullup(m, ETHER_HDR_LEN);
1880                 if (m == NULL)
1881                         return (0);
1882         }
1883
1884         eh = mtod(m, struct ether_header *);
1885         sc = ifp->if_bridge;
1886         vlan = VLANTAGOF(m);
1887
1888         BRIDGE_LOCK(sc);
1889
1890         /*
1891          * If bridge is down, but the original output interface is up,
1892          * go ahead and send out that interface.  Otherwise, the packet
1893          * is dropped below.
1894          */
1895         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1896                 dst_if = ifp;
1897                 goto sendunicast;
1898         }
1899
1900         /*
1901          * If the packet is a multicast, or we don't know a better way to
1902          * get there, send to all interfaces.
1903          */
1904         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1905                 dst_if = NULL;
1906         else
1907                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1908         if (dst_if == NULL) {
1909                 struct bridge_iflist *bif;
1910                 struct mbuf *mc;
1911                 int error = 0, used = 0;
1912
1913                 bridge_span(sc, m);
1914
1915                 BRIDGE_LOCK2REF(sc, error);
1916                 if (error) {
1917                         m_freem(m);
1918                         return (0);
1919                 }
1920
1921                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1922                         dst_if = bif->bif_ifp;
1923
1924                         if (dst_if->if_type == IFT_GIF)
1925                                 continue;
1926                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1927                                 continue;
1928
1929                         /*
1930                          * If this is not the original output interface,
1931                          * and the interface is participating in spanning
1932                          * tree, make sure the port is in a state that
1933                          * allows forwarding.
1934                          */
1935                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1936                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1937                                 continue;
1938
1939                         if (LIST_NEXT(bif, bif_next) == NULL) {
1940                                 used = 1;
1941                                 mc = m;
1942                         } else {
1943                                 mc = m_copypacket(m, M_DONTWAIT);
1944                                 if (mc == NULL) {
1945                                         sc->sc_ifp->if_oerrors++;
1946                                         continue;
1947                                 }
1948                         }
1949
1950                         bridge_enqueue(sc, dst_if, mc);
1951                 }
1952                 if (used == 0)
1953                         m_freem(m);
1954                 BRIDGE_UNREF(sc);
1955                 return (0);
1956         }
1957
1958 sendunicast:
1959         /*
1960          * XXX Spanning tree consideration here?
1961          */
1962
1963         bridge_span(sc, m);
1964         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1965                 m_freem(m);
1966                 BRIDGE_UNLOCK(sc);
1967                 return (0);
1968         }
1969
1970         BRIDGE_UNLOCK(sc);
1971         bridge_enqueue(sc, dst_if, m);
1972         return (0);
1973 }
1974
1975 /*
1976  * bridge_start:
1977  *
1978  *      Start output on a bridge.
1979  *
1980  */
1981 static void
1982 bridge_start(struct ifnet *ifp)
1983 {
1984         struct bridge_softc *sc;
1985         struct mbuf *m;
1986         struct ether_header *eh;
1987         struct ifnet *dst_if;
1988
1989         sc = ifp->if_softc;
1990
1991         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1992         for (;;) {
1993                 IFQ_DEQUEUE(&ifp->if_snd, m);
1994                 if (m == 0)
1995                         break;
1996                 ETHER_BPF_MTAP(ifp, m);
1997
1998                 eh = mtod(m, struct ether_header *);
1999                 dst_if = NULL;
2000
2001                 BRIDGE_LOCK(sc);
2002                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2003                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
2004                 }
2005
2006                 if (dst_if == NULL)
2007                         bridge_broadcast(sc, ifp, m, 0);
2008                 else {
2009                         BRIDGE_UNLOCK(sc);
2010                         bridge_enqueue(sc, dst_if, m);
2011                 }
2012         }
2013         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2014 }
2015
2016 /*
2017  * bridge_forward:
2018  *
2019  *      The forwarding function of the bridge.
2020  *
2021  *      NOTE: Releases the lock on return.
2022  */
2023 static void
2024 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2025     struct mbuf *m)
2026 {
2027         struct bridge_iflist *dbif;
2028         struct ifnet *src_if, *dst_if, *ifp;
2029         struct ether_header *eh;
2030         uint16_t vlan;
2031         uint8_t *dst;
2032         int error;
2033
2034         src_if = m->m_pkthdr.rcvif;
2035         ifp = sc->sc_ifp;
2036
2037         ifp->if_ipackets++;
2038         ifp->if_ibytes += m->m_pkthdr.len;
2039         vlan = VLANTAGOF(m);
2040
2041         if ((sbif->bif_flags & IFBIF_STP) &&
2042             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2043                 goto drop;
2044
2045         eh = mtod(m, struct ether_header *);
2046         dst = eh->ether_dhost;
2047
2048         /* If the interface is learning, record the address. */
2049         if (sbif->bif_flags & IFBIF_LEARNING) {
2050                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2051                     sbif, 0, IFBAF_DYNAMIC);
2052                 /*
2053                  * If the interface has addresses limits then deny any source
2054                  * that is not in the cache.
2055                  */
2056                 if (error && sbif->bif_addrmax)
2057                         goto drop;
2058         }
2059
2060         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2061             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2062                 goto drop;
2063
2064         /*
2065          * At this point, the port either doesn't participate
2066          * in spanning tree or it is in the forwarding state.
2067          */
2068
2069         /*
2070          * If the packet is unicast, destined for someone on
2071          * "this" side of the bridge, drop it.
2072          */
2073         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2074                 dst_if = bridge_rtlookup(sc, dst, vlan);
2075                 if (src_if == dst_if)
2076                         goto drop;
2077         } else {
2078                 /*
2079                  * Check if its a reserved multicast address, any address
2080                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2081                  * bridge.
2082                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2083                  */
2084                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2085                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2086                     dst[4] == 0x00 && dst[5] <= 0x0f)
2087                         goto drop;
2088
2089                 /* ...forward it to all interfaces. */
2090                 ifp->if_imcasts++;
2091                 dst_if = NULL;
2092         }
2093
2094         /*
2095          * If we have a destination interface which is a member of our bridge,
2096          * OR this is a unicast packet, push it through the bpf(4) machinery.
2097          * For broadcast or multicast packets, don't bother because it will
2098          * be reinjected into ether_input. We do this before we pass the packets
2099          * through the pfil(9) framework, as it is possible that pfil(9) will
2100          * drop the packet, or possibly modify it, making it difficult to debug
2101          * firewall issues on the bridge.
2102          */
2103         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2104                 ETHER_BPF_MTAP(ifp, m);
2105
2106         /* run the packet filter */
2107         if (PFIL_HOOKED(&V_inet_pfil_hook)
2108 #ifdef INET6
2109             || PFIL_HOOKED(&V_inet6_pfil_hook)
2110 #endif
2111             ) {
2112                 BRIDGE_UNLOCK(sc);
2113                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2114                         return;
2115                 if (m == NULL)
2116                         return;
2117                 BRIDGE_LOCK(sc);
2118         }
2119
2120         if (dst_if == NULL) {
2121                 bridge_broadcast(sc, src_if, m, 1);
2122                 return;
2123         }
2124
2125         /*
2126          * At this point, we're dealing with a unicast frame
2127          * going to a different interface.
2128          */
2129         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2130                 goto drop;
2131
2132         dbif = bridge_lookup_member_if(sc, dst_if);
2133         if (dbif == NULL)
2134                 /* Not a member of the bridge (anymore?) */
2135                 goto drop;
2136
2137         /* Private segments can not talk to each other */
2138         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2139                 goto drop;
2140
2141         if ((dbif->bif_flags & IFBIF_STP) &&
2142             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2143                 goto drop;
2144
2145         BRIDGE_UNLOCK(sc);
2146
2147         if (PFIL_HOOKED(&V_inet_pfil_hook)
2148 #ifdef INET6
2149             || PFIL_HOOKED(&V_inet6_pfil_hook)
2150 #endif
2151             ) {
2152                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2153                         return;
2154                 if (m == NULL)
2155                         return;
2156         }
2157
2158         bridge_enqueue(sc, dst_if, m);
2159         return;
2160
2161 drop:
2162         BRIDGE_UNLOCK(sc);
2163         m_freem(m);
2164 }
2165
2166 /*
2167  * bridge_input:
2168  *
2169  *      Receive input from a member interface.  Queue the packet for
2170  *      bridging if it is not for us.
2171  */
2172 static struct mbuf *
2173 bridge_input(struct ifnet *ifp, struct mbuf *m)
2174 {
2175         struct bridge_softc *sc = ifp->if_bridge;
2176         struct bridge_iflist *bif, *bif2;
2177         struct ifnet *bifp;
2178         struct ether_header *eh;
2179         struct mbuf *mc, *mc2;
2180         uint16_t vlan;
2181         int error;
2182
2183         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2184                 return (m);
2185
2186         bifp = sc->sc_ifp;
2187         vlan = VLANTAGOF(m);
2188
2189         /*
2190          * Implement support for bridge monitoring. If this flag has been
2191          * set on this interface, discard the packet once we push it through
2192          * the bpf(4) machinery, but before we do, increment the byte and
2193          * packet counters associated with this interface.
2194          */
2195         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2196                 m->m_pkthdr.rcvif  = bifp;
2197                 ETHER_BPF_MTAP(bifp, m);
2198                 bifp->if_ipackets++;
2199                 bifp->if_ibytes += m->m_pkthdr.len;
2200                 m_freem(m);
2201                 return (NULL);
2202         }
2203         BRIDGE_LOCK(sc);
2204         bif = bridge_lookup_member_if(sc, ifp);
2205         if (bif == NULL) {
2206                 BRIDGE_UNLOCK(sc);
2207                 return (m);
2208         }
2209
2210         eh = mtod(m, struct ether_header *);
2211
2212         bridge_span(sc, m);
2213
2214         if (m->m_flags & (M_BCAST|M_MCAST)) {
2215                 /* Tap off 802.1D packets; they do not get forwarded. */
2216                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2217                     ETHER_ADDR_LEN) == 0) {
2218                         m = bstp_input(&bif->bif_stp, ifp, m);
2219                         if (m == NULL) {
2220                                 BRIDGE_UNLOCK(sc);
2221                                 return (NULL);
2222                         }
2223                 }
2224
2225                 if ((bif->bif_flags & IFBIF_STP) &&
2226                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2227                         BRIDGE_UNLOCK(sc);
2228                         return (m);
2229                 }
2230
2231                 /*
2232                  * Make a deep copy of the packet and enqueue the copy
2233                  * for bridge processing; return the original packet for
2234                  * local processing.
2235                  */
2236                 mc = m_dup(m, M_DONTWAIT);
2237                 if (mc == NULL) {
2238                         BRIDGE_UNLOCK(sc);
2239                         return (m);
2240                 }
2241
2242                 /* Perform the bridge forwarding function with the copy. */
2243                 bridge_forward(sc, bif, mc);
2244
2245                 /*
2246                  * Reinject the mbuf as arriving on the bridge so we have a
2247                  * chance at claiming multicast packets. We can not loop back
2248                  * here from ether_input as a bridge is never a member of a
2249                  * bridge.
2250                  */
2251                 KASSERT(bifp->if_bridge == NULL,
2252                     ("loop created in bridge_input"));
2253                 mc2 = m_dup(m, M_DONTWAIT);
2254                 if (mc2 != NULL) {
2255                         /* Keep the layer3 header aligned */
2256                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2257                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2258                 }
2259                 if (mc2 != NULL) {
2260                         mc2->m_pkthdr.rcvif = bifp;
2261                         (*bifp->if_input)(bifp, mc2);
2262                 }
2263
2264                 /* Return the original packet for local processing. */
2265                 return (m);
2266         }
2267
2268         if ((bif->bif_flags & IFBIF_STP) &&
2269             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2270                 BRIDGE_UNLOCK(sc);
2271                 return (m);
2272         }
2273
2274 #if (defined(INET) || defined(INET6))
2275 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2276         || ((iface)->if_carp \
2277             && (*carp_forus_p)((iface), eh->ether_dhost))
2278 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2279         || ((iface)->if_carp \
2280             && (*carp_forus_p)((iface), eh->ether_shost))
2281 #else
2282 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2283 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2284 #endif
2285
2286 #ifdef INET6
2287 #   define OR_PFIL_HOOKED_INET6 \
2288         || PFIL_HOOKED(&V_inet6_pfil_hook)
2289 #else
2290 #   define OR_PFIL_HOOKED_INET6
2291 #endif
2292
2293 #define GRAB_OUR_PACKETS(iface) \
2294         if ((iface)->if_type == IFT_GIF) \
2295                 continue; \
2296         /* It is destined for us. */ \
2297         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2298             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2299             ) {                                                         \
2300                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2301                         ETHER_BPF_MTAP(iface, m);                       \
2302                         iface->if_ipackets++;                           \
2303                         /* Filter on the physical interface. */         \
2304                         if (pfil_local_phys &&                          \
2305                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2306                              OR_PFIL_HOOKED_INET6)) {                   \
2307                                 if (bridge_pfil(&m, NULL, ifp,          \
2308                                     PFIL_IN) != 0 || m == NULL) {       \
2309                                         BRIDGE_UNLOCK(sc);              \
2310                                         return (NULL);                  \
2311                                 }                                       \
2312                         }                                               \
2313                 }                                                       \
2314                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2315                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2316                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2317                         if (error && bif->bif_addrmax) {                \
2318                                 BRIDGE_UNLOCK(sc);                      \
2319                                 m_freem(m);                             \
2320                                 return (NULL);                          \
2321                         }                                               \
2322                 }                                                       \
2323                 m->m_pkthdr.rcvif = iface;                              \
2324                 BRIDGE_UNLOCK(sc);                                      \
2325                 return (m);                                             \
2326         }                                                               \
2327                                                                         \
2328         /* We just received a packet that we sent out. */               \
2329         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2330             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2331             ) {                                                         \
2332                 BRIDGE_UNLOCK(sc);                                      \
2333                 m_freem(m);                                             \
2334                 return (NULL);                                          \
2335         }
2336
2337         /*
2338          * Unicast.  Make sure it's not for the bridge.
2339          */
2340         do { GRAB_OUR_PACKETS(bifp) } while (0);
2341
2342         /*
2343          * Give a chance for ifp at first priority. This will help when the
2344          * packet comes through the interface like VLAN's with the same MACs
2345          * on several interfaces from the same bridge. This also will save
2346          * some CPU cycles in case the destination interface and the input
2347          * interface (eq ifp) are the same.
2348          */
2349         do { GRAB_OUR_PACKETS(ifp) } while (0);
2350
2351         /* Now check the all bridge members. */
2352         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2353                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2354         }
2355
2356 #undef OR_CARP_CHECK_WE_ARE_DST
2357 #undef OR_CARP_CHECK_WE_ARE_SRC
2358 #undef OR_PFIL_HOOKED_INET6
2359 #undef GRAB_OUR_PACKETS
2360
2361         /* Perform the bridge forwarding function. */
2362         bridge_forward(sc, bif, m);
2363
2364         return (NULL);
2365 }
2366
2367 /*
2368  * bridge_broadcast:
2369  *
2370  *      Send a frame to all interfaces that are members of
2371  *      the bridge, except for the one on which the packet
2372  *      arrived.
2373  *
2374  *      NOTE: Releases the lock on return.
2375  */
2376 static void
2377 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2378     struct mbuf *m, int runfilt)
2379 {
2380         struct bridge_iflist *dbif, *sbif;
2381         struct mbuf *mc;
2382         struct ifnet *dst_if;
2383         int error = 0, used = 0, i;
2384
2385         sbif = bridge_lookup_member_if(sc, src_if);
2386
2387         BRIDGE_LOCK2REF(sc, error);
2388         if (error) {
2389                 m_freem(m);
2390                 return;
2391         }
2392
2393         /* Filter on the bridge interface before broadcasting */
2394         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2395 #ifdef INET6
2396             || PFIL_HOOKED(&V_inet6_pfil_hook)
2397 #endif
2398             )) {
2399                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2400                         goto out;
2401                 if (m == NULL)
2402                         goto out;
2403         }
2404
2405         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2406                 dst_if = dbif->bif_ifp;
2407                 if (dst_if == src_if)
2408                         continue;
2409
2410                 /* Private segments can not talk to each other */
2411                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2412                         continue;
2413
2414                 if ((dbif->bif_flags & IFBIF_STP) &&
2415                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2416                         continue;
2417
2418                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2419                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2420                         continue;
2421
2422                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2423                         continue;
2424
2425                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2426                         mc = m;
2427                         used = 1;
2428                 } else {
2429                         mc = m_dup(m, M_DONTWAIT);
2430                         if (mc == NULL) {
2431                                 sc->sc_ifp->if_oerrors++;
2432                                 continue;
2433                         }
2434                 }
2435
2436                 /*
2437                  * Filter on the output interface. Pass a NULL bridge interface
2438                  * pointer so we do not redundantly filter on the bridge for
2439                  * each interface we broadcast on.
2440                  */
2441                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2442 #ifdef INET6
2443                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2444 #endif
2445                     )) {
2446                         if (used == 0) {
2447                                 /* Keep the layer3 header aligned */
2448                                 i = min(mc->m_pkthdr.len, max_protohdr);
2449                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2450                                 if (mc == NULL) {
2451                                         sc->sc_ifp->if_oerrors++;
2452                                         continue;
2453                                 }
2454                         }
2455                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2456                                 continue;
2457                         if (mc == NULL)
2458                                 continue;
2459                 }
2460
2461                 bridge_enqueue(sc, dst_if, mc);
2462         }
2463         if (used == 0)
2464                 m_freem(m);
2465
2466 out:
2467         BRIDGE_UNREF(sc);
2468 }
2469
2470 /*
2471  * bridge_span:
2472  *
2473  *      Duplicate a packet out one or more interfaces that are in span mode,
2474  *      the original mbuf is unmodified.
2475  */
2476 static void
2477 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2478 {
2479         struct bridge_iflist *bif;
2480         struct ifnet *dst_if;
2481         struct mbuf *mc;
2482
2483         if (LIST_EMPTY(&sc->sc_spanlist))
2484                 return;
2485
2486         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2487                 dst_if = bif->bif_ifp;
2488
2489                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2490                         continue;
2491
2492                 mc = m_copypacket(m, M_DONTWAIT);
2493                 if (mc == NULL) {
2494                         sc->sc_ifp->if_oerrors++;
2495                         continue;
2496                 }
2497
2498                 bridge_enqueue(sc, dst_if, mc);
2499         }
2500 }
2501
2502 /*
2503  * bridge_rtupdate:
2504  *
2505  *      Add a bridge routing entry.
2506  */
2507 static int
2508 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2509     struct bridge_iflist *bif, int setflags, uint8_t flags)
2510 {
2511         struct bridge_rtnode *brt;
2512         int error;
2513
2514         BRIDGE_LOCK_ASSERT(sc);
2515
2516         /* Check the source address is valid and not multicast. */
2517         if (ETHER_IS_MULTICAST(dst) ||
2518             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2519              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2520                 return (EINVAL);
2521
2522         /* 802.1p frames map to vlan 1 */
2523         if (vlan == 0)
2524                 vlan = 1;
2525
2526         /*
2527          * A route for this destination might already exist.  If so,
2528          * update it, otherwise create a new one.
2529          */
2530         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2531                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2532                         sc->sc_brtexceeded++;
2533                         return (ENOSPC);
2534                 }
2535                 /* Check per interface address limits (if enabled) */
2536                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2537                         bif->bif_addrexceeded++;
2538                         return (ENOSPC);
2539                 }
2540
2541                 /*
2542                  * Allocate a new bridge forwarding node, and
2543                  * initialize the expiration time and Ethernet
2544                  * address.
2545                  */
2546                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2547                 if (brt == NULL)
2548                         return (ENOMEM);
2549
2550                 if (bif->bif_flags & IFBIF_STICKY)
2551                         brt->brt_flags = IFBAF_STICKY;
2552                 else
2553                         brt->brt_flags = IFBAF_DYNAMIC;
2554
2555                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2556                 brt->brt_vlan = vlan;
2557
2558                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2559                         uma_zfree(bridge_rtnode_zone, brt);
2560                         return (error);
2561                 }
2562                 brt->brt_dst = bif;
2563                 bif->bif_addrcnt++;
2564         }
2565
2566         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2567             brt->brt_dst != bif) {
2568                 brt->brt_dst->bif_addrcnt--;
2569                 brt->brt_dst = bif;
2570                 brt->brt_dst->bif_addrcnt++;
2571         }
2572
2573         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2574                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2575         if (setflags)
2576                 brt->brt_flags = flags;
2577
2578         return (0);
2579 }
2580
2581 /*
2582  * bridge_rtlookup:
2583  *
2584  *      Lookup the destination interface for an address.
2585  */
2586 static struct ifnet *
2587 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2588 {
2589         struct bridge_rtnode *brt;
2590
2591         BRIDGE_LOCK_ASSERT(sc);
2592
2593         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2594                 return (NULL);
2595
2596         return (brt->brt_ifp);
2597 }
2598
2599 /*
2600  * bridge_rttrim:
2601  *
2602  *      Trim the routine table so that we have a number
2603  *      of routing entries less than or equal to the
2604  *      maximum number.
2605  */
2606 static void
2607 bridge_rttrim(struct bridge_softc *sc)
2608 {
2609         struct bridge_rtnode *brt, *nbrt;
2610
2611         BRIDGE_LOCK_ASSERT(sc);
2612
2613         /* Make sure we actually need to do this. */
2614         if (sc->sc_brtcnt <= sc->sc_brtmax)
2615                 return;
2616
2617         /* Force an aging cycle; this might trim enough addresses. */
2618         bridge_rtage(sc);
2619         if (sc->sc_brtcnt <= sc->sc_brtmax)
2620                 return;
2621
2622         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2623                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2624                         bridge_rtnode_destroy(sc, brt);
2625                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2626                                 return;
2627                 }
2628         }
2629 }
2630
2631 /*
2632  * bridge_timer:
2633  *
2634  *      Aging timer for the bridge.
2635  */
2636 static void
2637 bridge_timer(void *arg)
2638 {
2639         struct bridge_softc *sc = arg;
2640
2641         BRIDGE_LOCK_ASSERT(sc);
2642
2643         bridge_rtage(sc);
2644
2645         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2646                 callout_reset(&sc->sc_brcallout,
2647                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2648 }
2649
2650 /*
2651  * bridge_rtage:
2652  *
2653  *      Perform an aging cycle.
2654  */
2655 static void
2656 bridge_rtage(struct bridge_softc *sc)
2657 {
2658         struct bridge_rtnode *brt, *nbrt;
2659
2660         BRIDGE_LOCK_ASSERT(sc);
2661
2662         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2663                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2664                         if (time_uptime >= brt->brt_expire)
2665                                 bridge_rtnode_destroy(sc, brt);
2666                 }
2667         }
2668 }
2669
2670 /*
2671  * bridge_rtflush:
2672  *
2673  *      Remove all dynamic addresses from the bridge.
2674  */
2675 static void
2676 bridge_rtflush(struct bridge_softc *sc, int full)
2677 {
2678         struct bridge_rtnode *brt, *nbrt;
2679
2680         BRIDGE_LOCK_ASSERT(sc);
2681
2682         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2683                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2684                         bridge_rtnode_destroy(sc, brt);
2685         }
2686 }
2687
2688 /*
2689  * bridge_rtdaddr:
2690  *
2691  *      Remove an address from the table.
2692  */
2693 static int
2694 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2695 {
2696         struct bridge_rtnode *brt;
2697         int found = 0;
2698
2699         BRIDGE_LOCK_ASSERT(sc);
2700
2701         /*
2702          * If vlan is zero then we want to delete for all vlans so the lookup
2703          * may return more than one.
2704          */
2705         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2706                 bridge_rtnode_destroy(sc, brt);
2707                 found = 1;
2708         }
2709
2710         return (found ? 0 : ENOENT);
2711 }
2712
2713 /*
2714  * bridge_rtdelete:
2715  *
2716  *      Delete routes to a speicifc member interface.
2717  */
2718 static void
2719 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2720 {
2721         struct bridge_rtnode *brt, *nbrt;
2722
2723         BRIDGE_LOCK_ASSERT(sc);
2724
2725         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2726                 if (brt->brt_ifp == ifp && (full ||
2727                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2728                         bridge_rtnode_destroy(sc, brt);
2729         }
2730 }
2731
2732 /*
2733  * bridge_rtable_init:
2734  *
2735  *      Initialize the route table for this bridge.
2736  */
2737 static int
2738 bridge_rtable_init(struct bridge_softc *sc)
2739 {
2740         int i;
2741
2742         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2743             M_DEVBUF, M_NOWAIT);
2744         if (sc->sc_rthash == NULL)
2745                 return (ENOMEM);
2746
2747         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2748                 LIST_INIT(&sc->sc_rthash[i]);
2749
2750         sc->sc_rthash_key = arc4random();
2751
2752         LIST_INIT(&sc->sc_rtlist);
2753
2754         return (0);
2755 }
2756
2757 /*
2758  * bridge_rtable_fini:
2759  *
2760  *      Deconstruct the route table for this bridge.
2761  */
2762 static void
2763 bridge_rtable_fini(struct bridge_softc *sc)
2764 {
2765
2766         KASSERT(sc->sc_brtcnt == 0,
2767             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2768         free(sc->sc_rthash, M_DEVBUF);
2769 }
2770
2771 /*
2772  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2773  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2774  */
2775 #define mix(a, b, c)                                                    \
2776 do {                                                                    \
2777         a -= b; a -= c; a ^= (c >> 13);                                 \
2778         b -= c; b -= a; b ^= (a << 8);                                  \
2779         c -= a; c -= b; c ^= (b >> 13);                                 \
2780         a -= b; a -= c; a ^= (c >> 12);                                 \
2781         b -= c; b -= a; b ^= (a << 16);                                 \
2782         c -= a; c -= b; c ^= (b >> 5);                                  \
2783         a -= b; a -= c; a ^= (c >> 3);                                  \
2784         b -= c; b -= a; b ^= (a << 10);                                 \
2785         c -= a; c -= b; c ^= (b >> 15);                                 \
2786 } while (/*CONSTCOND*/0)
2787
2788 static __inline uint32_t
2789 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2790 {
2791         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2792
2793         b += addr[5] << 8;
2794         b += addr[4];
2795         a += addr[3] << 24;
2796         a += addr[2] << 16;
2797         a += addr[1] << 8;
2798         a += addr[0];
2799
2800         mix(a, b, c);
2801
2802         return (c & BRIDGE_RTHASH_MASK);
2803 }
2804
2805 #undef mix
2806
2807 static int
2808 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2809 {
2810         int i, d;
2811
2812         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2813                 d = ((int)a[i]) - ((int)b[i]);
2814         }
2815
2816         return (d);
2817 }
2818
2819 /*
2820  * bridge_rtnode_lookup:
2821  *
2822  *      Look up a bridge route node for the specified destination. Compare the
2823  *      vlan id or if zero then just return the first match.
2824  */
2825 static struct bridge_rtnode *
2826 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2827 {
2828         struct bridge_rtnode *brt;
2829         uint32_t hash;
2830         int dir;
2831
2832         BRIDGE_LOCK_ASSERT(sc);
2833
2834         hash = bridge_rthash(sc, addr);
2835         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2836                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2837                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2838                         return (brt);
2839                 if (dir > 0)
2840                         return (NULL);
2841         }
2842
2843         return (NULL);
2844 }
2845
2846 /*
2847  * bridge_rtnode_insert:
2848  *
2849  *      Insert the specified bridge node into the route table.  We
2850  *      assume the entry is not already in the table.
2851  */
2852 static int
2853 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2854 {
2855         struct bridge_rtnode *lbrt;
2856         uint32_t hash;
2857         int dir;
2858
2859         BRIDGE_LOCK_ASSERT(sc);
2860
2861         hash = bridge_rthash(sc, brt->brt_addr);
2862
2863         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2864         if (lbrt == NULL) {
2865                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2866                 goto out;
2867         }
2868
2869         do {
2870                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2871                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2872                         return (EEXIST);
2873                 if (dir > 0) {
2874                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2875                         goto out;
2876                 }
2877                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2878                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2879                         goto out;
2880                 }
2881                 lbrt = LIST_NEXT(lbrt, brt_hash);
2882         } while (lbrt != NULL);
2883
2884 #ifdef DIAGNOSTIC
2885         panic("bridge_rtnode_insert: impossible");
2886 #endif
2887
2888 out:
2889         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2890         sc->sc_brtcnt++;
2891
2892         return (0);
2893 }
2894
2895 /*
2896  * bridge_rtnode_destroy:
2897  *
2898  *      Destroy a bridge rtnode.
2899  */
2900 static void
2901 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2902 {
2903         BRIDGE_LOCK_ASSERT(sc);
2904
2905         LIST_REMOVE(brt, brt_hash);
2906
2907         LIST_REMOVE(brt, brt_list);
2908         sc->sc_brtcnt--;
2909         brt->brt_dst->bif_addrcnt--;
2910         uma_zfree(bridge_rtnode_zone, brt);
2911 }
2912
2913 /*
2914  * bridge_rtable_expire:
2915  *
2916  *      Set the expiry time for all routes on an interface.
2917  */
2918 static void
2919 bridge_rtable_expire(struct ifnet *ifp, int age)
2920 {
2921         struct bridge_softc *sc = ifp->if_bridge;
2922         struct bridge_rtnode *brt;
2923
2924         BRIDGE_LOCK(sc);
2925
2926         /*
2927          * If the age is zero then flush, otherwise set all the expiry times to
2928          * age for the interface
2929          */
2930         if (age == 0)
2931                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2932         else {
2933                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2934                         /* Cap the expiry time to 'age' */
2935                         if (brt->brt_ifp == ifp &&
2936                             brt->brt_expire > time_uptime + age &&
2937                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2938                                 brt->brt_expire = time_uptime + age;
2939                 }
2940         }
2941         BRIDGE_UNLOCK(sc);
2942 }
2943
2944 /*
2945  * bridge_state_change:
2946  *
2947  *      Callback from the bridgestp code when a port changes states.
2948  */
2949 static void
2950 bridge_state_change(struct ifnet *ifp, int state)
2951 {
2952         struct bridge_softc *sc = ifp->if_bridge;
2953         static const char *stpstates[] = {
2954                 "disabled",
2955                 "listening",
2956                 "learning",
2957                 "forwarding",
2958                 "blocking",
2959                 "discarding"
2960         };
2961
2962         if (log_stp)
2963                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2964                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2965 }
2966
2967 /*
2968  * Send bridge packets through pfil if they are one of the types pfil can deal
2969  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2970  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2971  * that interface.
2972  */
2973 static int
2974 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2975 {
2976         int snap, error, i, hlen;
2977         struct ether_header *eh1, eh2;
2978         struct ip_fw_args args;
2979         struct ip *ip;
2980         struct llc llc1;
2981         u_int16_t ether_type;
2982
2983         snap = 0;
2984         error = -1;     /* Default error if not error == 0 */
2985
2986 #if 0
2987         /* we may return with the IP fields swapped, ensure its not shared */
2988         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2989 #endif
2990
2991         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2992                 return (0); /* filtering is disabled */
2993
2994         i = min((*mp)->m_pkthdr.len, max_protohdr);
2995         if ((*mp)->m_len < i) {
2996             *mp = m_pullup(*mp, i);
2997             if (*mp == NULL) {
2998                 printf("%s: m_pullup failed\n", __func__);
2999                 return (-1);
3000             }
3001         }
3002
3003         eh1 = mtod(*mp, struct ether_header *);
3004         ether_type = ntohs(eh1->ether_type);
3005
3006         /*
3007          * Check for SNAP/LLC.
3008          */
3009         if (ether_type < ETHERMTU) {
3010                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3011
3012                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3013                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3014                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3015                     llc2->llc_control == LLC_UI) {
3016                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3017                         snap = 1;
3018                 }
3019         }
3020
3021         /*
3022          * If we're trying to filter bridge traffic, don't look at anything
3023          * other than IP and ARP traffic.  If the filter doesn't understand
3024          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3025          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3026          * but of course we don't have an AppleTalk filter to begin with.
3027          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3028          * ARP traffic.)
3029          */
3030         switch (ether_type) {
3031                 case ETHERTYPE_ARP:
3032                 case ETHERTYPE_REVARP:
3033                         if (pfil_ipfw_arp == 0)
3034                                 return (0); /* Automatically pass */
3035                         break;
3036
3037                 case ETHERTYPE_IP:
3038 #ifdef INET6
3039                 case ETHERTYPE_IPV6:
3040 #endif /* INET6 */
3041                         break;
3042                 default:
3043                         /*
3044                          * Check to see if the user wants to pass non-ip
3045                          * packets, these will not be checked by pfil(9) and
3046                          * passed unconditionally so the default is to drop.
3047                          */
3048                         if (pfil_onlyip)
3049                                 goto bad;
3050         }
3051
3052         /* Strip off the Ethernet header and keep a copy. */
3053         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3054         m_adj(*mp, ETHER_HDR_LEN);
3055
3056         /* Strip off snap header, if present */
3057         if (snap) {
3058                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3059                 m_adj(*mp, sizeof(struct llc));
3060         }
3061
3062         /*
3063          * Check the IP header for alignment and errors
3064          */
3065         if (dir == PFIL_IN) {
3066                 switch (ether_type) {
3067                         case ETHERTYPE_IP:
3068                                 error = bridge_ip_checkbasic(mp);
3069                                 break;
3070 #ifdef INET6
3071                         case ETHERTYPE_IPV6:
3072                                 error = bridge_ip6_checkbasic(mp);
3073                                 break;
3074 #endif /* INET6 */
3075                         default:
3076                                 error = 0;
3077                 }
3078                 if (error)
3079                         goto bad;
3080         }
3081
3082         /* XXX this section is also in if_ethersubr.c */
3083         // XXX PFIL_OUT or DIR_OUT ?
3084         if (V_ip_fw_chk_ptr && pfil_ipfw != 0 &&
3085                         dir == PFIL_OUT && ifp != NULL) {
3086                 struct m_tag *mtag;
3087
3088                 error = -1;
3089                 /* fetch the start point from existing tags, if any */
3090                 mtag = m_tag_locate(*mp, MTAG_IPFW_RULE, 0, NULL);
3091                 if (mtag == NULL) {
3092                         args.rule.slot = 0;
3093                 } else {
3094                         struct ipfw_rule_ref *r;
3095
3096                         /* XXX can we free the tag after use ? */
3097                         mtag->m_tag_id = PACKET_TAG_NONE;
3098                         r = (struct ipfw_rule_ref *)(mtag + 1);
3099                         /* packet already partially processed ? */
3100                         if (r->info & IPFW_ONEPASS)
3101                                 goto ipfwpass;
3102                         args.rule = *r;
3103                 }
3104
3105                 args.m = *mp;
3106                 args.oif = ifp;
3107                 args.next_hop = NULL;
3108                 args.next_hop6 = NULL;
3109                 args.eh = &eh2;
3110                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3111                 i = V_ip_fw_chk_ptr(&args);
3112                 *mp = args.m;
3113
3114                 if (*mp == NULL)
3115                         return (error);
3116
3117                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3118
3119                         /* put the Ethernet header back on */
3120                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3121                         if (*mp == NULL)
3122                                 return (error);
3123                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3124
3125                         /*
3126                          * Pass the pkt to dummynet, which consumes it. The
3127                          * packet will return to us via bridge_dummynet().
3128                          */
3129                         args.oif = ifp;
3130                         ip_dn_io_ptr(mp, DIR_FWD | PROTO_IFB, &args);
3131                         return (error);
3132                 }
3133
3134                 if (i != IP_FW_PASS) /* drop */
3135                         goto bad;
3136         }
3137
3138 ipfwpass:
3139         error = 0;
3140
3141         /*
3142          * Run the packet through pfil
3143          */
3144         switch (ether_type) {
3145         case ETHERTYPE_IP:
3146                 /*
3147                  * before calling the firewall, swap fields the same as
3148                  * IP does. here we assume the header is contiguous
3149                  */
3150                 ip = mtod(*mp, struct ip *);
3151
3152                 ip->ip_len = ntohs(ip->ip_len);
3153                 ip->ip_off = ntohs(ip->ip_off);
3154
3155                 /*
3156                  * Run pfil on the member interface and the bridge, both can
3157                  * be skipped by clearing pfil_member or pfil_bridge.
3158                  *
3159                  * Keep the order:
3160                  *   in_if -> bridge_if -> out_if
3161                  */
3162                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3163                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3164                                         dir, NULL);
3165
3166                 if (*mp == NULL || error != 0) /* filter may consume */
3167                         break;
3168
3169                 if (pfil_member && ifp != NULL)
3170                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3171                                         dir, NULL);
3172
3173                 if (*mp == NULL || error != 0) /* filter may consume */
3174                         break;
3175
3176                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3177                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3178                                         dir, NULL);
3179
3180                 if (*mp == NULL || error != 0) /* filter may consume */
3181                         break;
3182
3183                 /* check if we need to fragment the packet */
3184                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3185                         i = (*mp)->m_pkthdr.len;
3186                         if (i > ifp->if_mtu) {
3187                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3188                                             &llc1);
3189                                 return (error);
3190                         }
3191                 }
3192
3193                 /* Recalculate the ip checksum and restore byte ordering */
3194                 ip = mtod(*mp, struct ip *);
3195                 hlen = ip->ip_hl << 2;
3196                 if (hlen < sizeof(struct ip))
3197                         goto bad;
3198                 if (hlen > (*mp)->m_len) {
3199                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3200                                 goto bad;
3201                         ip = mtod(*mp, struct ip *);
3202                         if (ip == NULL)
3203                                 goto bad;
3204                 }
3205                 ip->ip_len = htons(ip->ip_len);
3206                 ip->ip_off = htons(ip->ip_off);
3207                 ip->ip_sum = 0;
3208                 if (hlen == sizeof(struct ip))
3209                         ip->ip_sum = in_cksum_hdr(ip);
3210                 else
3211                         ip->ip_sum = in_cksum(*mp, hlen);
3212
3213                 break;
3214 #ifdef INET6
3215         case ETHERTYPE_IPV6:
3216                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3217                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3218                                         dir, NULL);
3219
3220                 if (*mp == NULL || error != 0) /* filter may consume */
3221                         break;
3222
3223                 if (pfil_member && ifp != NULL)
3224                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3225                                         dir, NULL);
3226
3227                 if (*mp == NULL || error != 0) /* filter may consume */
3228                         break;
3229
3230                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3231                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3232                                         dir, NULL);
3233                 break;
3234 #endif
3235         default:
3236                 error = 0;
3237                 break;
3238         }
3239
3240         if (*mp == NULL)
3241                 return (error);
3242         if (error != 0)
3243                 goto bad;
3244
3245         error = -1;
3246
3247         /*
3248          * Finally, put everything back the way it was and return
3249          */
3250         if (snap) {
3251                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3252                 if (*mp == NULL)
3253                         return (error);
3254                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3255         }
3256
3257         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3258         if (*mp == NULL)
3259                 return (error);
3260         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3261
3262         return (0);
3263
3264 bad:
3265         m_freem(*mp);
3266         *mp = NULL;
3267         return (error);
3268 }
3269
3270 /*
3271  * Perform basic checks on header size since
3272  * pfil assumes ip_input has already processed
3273  * it for it.  Cut-and-pasted from ip_input.c.
3274  * Given how simple the IPv6 version is,
3275  * does the IPv4 version really need to be
3276  * this complicated?
3277  *
3278  * XXX Should we update ipstat here, or not?
3279  * XXX Right now we update ipstat but not
3280  * XXX csum_counter.
3281  */
3282 static int
3283 bridge_ip_checkbasic(struct mbuf **mp)
3284 {
3285         struct mbuf *m = *mp;
3286         struct ip *ip;
3287         int len, hlen;
3288         u_short sum;
3289
3290         if (*mp == NULL)
3291                 return (-1);
3292
3293         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3294                 if ((m = m_copyup(m, sizeof(struct ip),
3295                         (max_linkhdr + 3) & ~3)) == NULL) {
3296                         /* XXXJRT new stat, please */
3297                         KMOD_IPSTAT_INC(ips_toosmall);
3298                         goto bad;
3299                 }
3300         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3301                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3302                         KMOD_IPSTAT_INC(ips_toosmall);
3303                         goto bad;
3304                 }
3305         }
3306         ip = mtod(m, struct ip *);
3307         if (ip == NULL) goto bad;
3308
3309         if (ip->ip_v != IPVERSION) {
3310                 KMOD_IPSTAT_INC(ips_badvers);
3311                 goto bad;
3312         }
3313         hlen = ip->ip_hl << 2;
3314         if (hlen < sizeof(struct ip)) { /* minimum header length */
3315                 KMOD_IPSTAT_INC(ips_badhlen);
3316                 goto bad;
3317         }
3318         if (hlen > m->m_len) {
3319                 if ((m = m_pullup(m, hlen)) == 0) {
3320                         KMOD_IPSTAT_INC(ips_badhlen);
3321                         goto bad;
3322                 }
3323                 ip = mtod(m, struct ip *);
3324                 if (ip == NULL) goto bad;
3325         }
3326
3327         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3328                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3329         } else {
3330                 if (hlen == sizeof(struct ip)) {
3331                         sum = in_cksum_hdr(ip);
3332                 } else {
3333                         sum = in_cksum(m, hlen);
3334                 }
3335         }
3336         if (sum) {
3337                 KMOD_IPSTAT_INC(ips_badsum);
3338                 goto bad;
3339         }
3340
3341         /* Retrieve the packet length. */
3342         len = ntohs(ip->ip_len);
3343
3344         /*
3345          * Check for additional length bogosity
3346          */
3347         if (len < hlen) {
3348                 KMOD_IPSTAT_INC(ips_badlen);
3349                 goto bad;
3350         }
3351
3352         /*
3353          * Check that the amount of data in the buffers
3354          * is as at least much as the IP header would have us expect.
3355          * Drop packet if shorter than we expect.
3356          */
3357         if (m->m_pkthdr.len < len) {
3358                 KMOD_IPSTAT_INC(ips_tooshort);
3359                 goto bad;
3360         }
3361
3362         /* Checks out, proceed */
3363         *mp = m;
3364         return (0);
3365
3366 bad:
3367         *mp = m;
3368         return (-1);
3369 }
3370
3371 #ifdef INET6
3372 /*
3373  * Same as above, but for IPv6.
3374  * Cut-and-pasted from ip6_input.c.
3375  * XXX Should we update ip6stat, or not?
3376  */
3377 static int
3378 bridge_ip6_checkbasic(struct mbuf **mp)
3379 {
3380         struct mbuf *m = *mp;
3381         struct ip6_hdr *ip6;
3382
3383         /*
3384          * If the IPv6 header is not aligned, slurp it up into a new
3385          * mbuf with space for link headers, in the event we forward
3386          * it.  Otherwise, if it is aligned, make sure the entire base
3387          * IPv6 header is in the first mbuf of the chain.
3388          */
3389         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3390                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3391                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3392                             (max_linkhdr + 3) & ~3)) == NULL) {
3393                         /* XXXJRT new stat, please */
3394                         V_ip6stat.ip6s_toosmall++;
3395                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3396                         goto bad;
3397                 }
3398         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3399                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3400                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3401                         V_ip6stat.ip6s_toosmall++;
3402                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3403                         goto bad;
3404                 }
3405         }
3406
3407         ip6 = mtod(m, struct ip6_hdr *);
3408
3409         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3410                 V_ip6stat.ip6s_badvers++;
3411                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3412                 goto bad;
3413         }
3414
3415         /* Checks out, proceed */
3416         *mp = m;
3417         return (0);
3418
3419 bad:
3420         *mp = m;
3421         return (-1);
3422 }
3423 #endif /* INET6 */
3424
3425 /*
3426  * bridge_fragment:
3427  *
3428  *      Return a fragmented mbuf chain.
3429  */
3430 static int
3431 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3432     int snap, struct llc *llc)
3433 {
3434         struct mbuf *m0;
3435         struct ip *ip;
3436         int error = -1;
3437
3438         if (m->m_len < sizeof(struct ip) &&
3439             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3440                 goto out;
3441         ip = mtod(m, struct ip *);
3442
3443         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3444                     CSUM_DELAY_IP);
3445         if (error)
3446                 goto out;
3447
3448         /* walk the chain and re-add the Ethernet header */
3449         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3450                 if (error == 0) {
3451                         if (snap) {
3452                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3453                                 if (m0 == NULL) {
3454                                         error = ENOBUFS;
3455                                         continue;
3456                                 }
3457                                 bcopy(llc, mtod(m0, caddr_t),
3458                                     sizeof(struct llc));
3459                         }
3460                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3461                         if (m0 == NULL) {
3462                                 error = ENOBUFS;
3463                                 continue;
3464                         }
3465                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3466                 } else
3467                         m_freem(m);
3468         }
3469
3470         if (error == 0)
3471                 KMOD_IPSTAT_INC(ips_fragmented);
3472
3473         return (error);
3474
3475 out:
3476         if (m != NULL)
3477                 m_freem(m);
3478         return (error);
3479 }