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 #include <netinet6/in6_ifattach.h>
 123 #endif
 124 #if defined(INET) || defined(INET6)
 125 #include <netinet/ip_carp.h>
 126 #endif
 127 #include <machine/in_cksum.h>
 128 #include <netinet/if_ether.h> /* for struct arpcom */
 129 #include <net/bridgestp.h>
 130 #include <net/if_bridgevar.h>
 131 #include <net/if_llc.h>
 132 #include <net/if_vlan_var.h>
 133
 134 #include <net/route.h>
 135 #include <netinet/ip_fw.h>
 136 #include <netpfil/ipfw/ip_fw_private.h>
 137
 138 /*
 139  * Size of the route hash table.  Must be a power of two.
 140  */
 141 #ifndef BRIDGE_RTHASH_SIZE
 142 #define BRIDGE_RTHASH_SIZE              1024
 143 #endif
 144
 145 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 146
 147 /*
 148  * Default maximum number of addresses to cache.
 149  */
 150 #ifndef BRIDGE_RTABLE_MAX
 151 #define BRIDGE_RTABLE_MAX               2000
 152 #endif
 153
 154 /*
 155  * Timeout (in seconds) for entries learned dynamically.
 156  */
 157 #ifndef BRIDGE_RTABLE_TIMEOUT
 158 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 159 #endif
 160
 161 /*
 162  * Number of seconds between walks of the route list.
 163  */
 164 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 165 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 166 #endif
 167
 168 /*
 169  * List of capabilities to possibly mask on the member interface.
 170  */
 171 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 172
 173 /*
 174  * List of capabilities to strip
 175  */
 176 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 177
 178 /*
 179  * Bridge interface list entry.
 180  */
 181 struct bridge_iflist {
 182         LIST_ENTRY(bridge_iflist) bif_next;
 183         struct ifnet            *bif_ifp;       /* member if */
 184         struct bstp_port        bif_stp;        /* STP state */
 185         uint32_t                bif_flags;      /* member if flags */
 186         int                     bif_savedcaps;  /* saved capabilities */
 187         uint32_t                bif_addrmax;    /* max # of addresses */
 188         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 189         uint32_t                bif_addrexceeded;/* # of address violations */
 190 };
 191
 192 /*
 193  * Bridge route node.
 194  */
 195 struct bridge_rtnode {
 196         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 197         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 198         struct bridge_iflist    *brt_dst;       /* destination if */
 199         unsigned long           brt_expire;     /* expiration time */
 200         uint8_t                 brt_flags;      /* address flags */
 201         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 202         uint16_t                brt_vlan;       /* vlan id */
 203 };
 204 #define brt_ifp                 brt_dst->bif_ifp
 205
 206 /*
 207  * Software state for each bridge.
 208  */
 209 struct bridge_softc {
 210         struct ifnet            *sc_ifp;        /* make this an interface */
 211         LIST_ENTRY(bridge_softc) sc_list;
 212         struct mtx              sc_mtx;
 213         struct cv               sc_cv;
 214         uint32_t                sc_brtmax;      /* max # of addresses */
 215         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 216         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 217         struct callout          sc_brcallout;   /* bridge callout */
 218         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 219         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 220         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 221         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 222         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 223         uint32_t                sc_rthash_key;  /* key for hash */
 224         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 225         struct bstp_state       sc_stp;         /* STP state */
 226         uint32_t                sc_brtexceeded; /* # of cache drops */
 227         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 228         u_char                  sc_defaddr[6];  /* Default MAC address */
 229 };
 230
 231 static struct mtx       bridge_list_mtx;
 232 eventhandler_tag        bridge_detach_cookie = NULL;
 233
 234 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 235
 236 uma_zone_t bridge_rtnode_zone;
 237
 238 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 239 static void     bridge_clone_destroy(struct ifnet *);
 240
 241 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 242 static void     bridge_mutecaps(struct bridge_softc *);
 243 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 244                     int);
 245 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 246 static void     bridge_init(void *);
 247 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 248 static void     bridge_stop(struct ifnet *, int);
 249 static void     bridge_start(struct ifnet *);
 250 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 251 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 252                     struct rtentry *);
 253 static void     bridge_enqueue(struct bridge_softc *, struct ifnet *,
 254                     struct mbuf *);
 255 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 256
 257 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 258                     struct mbuf *m);
 259
 260 static void     bridge_timer(void *);
 261
 262 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 263                     struct mbuf *, int);
 264 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 265
 266 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 267                     uint16_t, struct bridge_iflist *, int, uint8_t);
 268 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 269                     uint16_t);
 270 static void     bridge_rttrim(struct bridge_softc *);
 271 static void     bridge_rtage(struct bridge_softc *);
 272 static void     bridge_rtflush(struct bridge_softc *, int);
 273 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 274                     uint16_t);
 275
 276 static int      bridge_rtable_init(struct bridge_softc *);
 277 static void     bridge_rtable_fini(struct bridge_softc *);
 278
 279 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 280 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 281                     const uint8_t *, uint16_t);
 282 static int      bridge_rtnode_insert(struct bridge_softc *,
 283                     struct bridge_rtnode *);
 284 static void     bridge_rtnode_destroy(struct bridge_softc *,
 285                     struct bridge_rtnode *);
 286 static void     bridge_rtable_expire(struct ifnet *, int);
 287 static void     bridge_state_change(struct ifnet *, int);
 288
 289 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 290                     const char *name);
 291 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 292                     struct ifnet *ifp);
 293 static void     bridge_delete_member(struct bridge_softc *,
 294                     struct bridge_iflist *, int);
 295 static void     bridge_delete_span(struct bridge_softc *,
 296                     struct bridge_iflist *);
 297
 298 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 328 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 329 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 330                     int);
 331 static int      bridge_ip_checkbasic(struct mbuf **mp);
 332 #ifdef INET6
 333 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 334 #endif /* INET6 */
 335 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 336                     struct ether_header *, int, struct llc *);
 337 static void     bridge_linkstate(struct ifnet *ifp);
 338 static void     bridge_linkcheck(struct bridge_softc *sc);
 339
 340 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
 341
 342 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 343 #define VLANTAGOF(_m)   \
 344     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 345
 346 static struct bstp_cb_ops bridge_ops = {
 347         .bcb_state = bridge_state_change,
 348         .bcb_rtage = bridge_rtable_expire
 349 };
 350
 351 SYSCTL_DECL(_net_link);
 352 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 353
 354 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
 355 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
 356 static int pfil_member = 1; /* run pfil hooks on the member interface */
 357 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
 358 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
 359 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
 360                                    locally destined packets */
 361 static int log_stp   = 0;   /* log STP state changes */
 362 static int bridge_inherit_mac = 0;   /* share MAC with first bridge member */
 363 TUNABLE_INT("net.link.bridge.pfil_onlyip", &pfil_onlyip);
 364 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
 365     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
 366 TUNABLE_INT("net.link.bridge.ipfw_arp", &pfil_ipfw_arp);
 367 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
 368     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
 369 TUNABLE_INT("net.link.bridge.pfil_bridge", &pfil_bridge);
 370 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
 371     &pfil_bridge, 0, "Packet filter on the bridge interface");
 372 TUNABLE_INT("net.link.bridge.pfil_member", &pfil_member);
 373 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
 374     &pfil_member, 0, "Packet filter on the member interface");
 375 TUNABLE_INT("net.link.bridge.pfil_local_phys", &pfil_local_phys);
 376 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
 377     &pfil_local_phys, 0,
 378     "Packet filter on the physical interface for locally destined packets");
 379 TUNABLE_INT("net.link.bridge.log_stp", &log_stp);
 380 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
 381     &log_stp, 0, "Log STP state changes");
 382 TUNABLE_INT("net.link.bridge.inherit_mac", &bridge_inherit_mac);
 383 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac, CTLFLAG_RW,
 384     &bridge_inherit_mac, 0,
 385     "Inherit MAC address from the first bridge member");
 386
 387 struct bridge_control {
 388         int     (*bc_func)(struct bridge_softc *, void *);
 389         int     bc_argsize;
 390         int     bc_flags;
 391 };
 392
 393 #define BC_F_COPYIN             0x01    /* copy arguments in */
 394 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 395 #define BC_F_SUSER              0x04    /* do super-user check */
 396
 397 const struct bridge_control bridge_control_table[] = {
 398         { bridge_ioctl_add,             sizeof(struct ifbreq),
 399           BC_F_COPYIN|BC_F_SUSER },
 400         { bridge_ioctl_del,             sizeof(struct ifbreq),
 401           BC_F_COPYIN|BC_F_SUSER },
 402
 403         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 404           BC_F_COPYIN|BC_F_COPYOUT },
 405         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 406           BC_F_COPYIN|BC_F_SUSER },
 407
 408         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 409           BC_F_COPYIN|BC_F_SUSER },
 410         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 411           BC_F_COPYOUT },
 412
 413         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 414           BC_F_COPYIN|BC_F_COPYOUT },
 415         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 416           BC_F_COPYIN|BC_F_COPYOUT },
 417
 418         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 419           BC_F_COPYIN|BC_F_SUSER },
 420
 421         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 422           BC_F_COPYIN|BC_F_SUSER },
 423         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 424           BC_F_COPYOUT },
 425
 426         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 427           BC_F_COPYIN|BC_F_SUSER },
 428
 429         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 430           BC_F_COPYIN|BC_F_SUSER },
 431
 432         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 433           BC_F_COPYOUT },
 434         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 435           BC_F_COPYIN|BC_F_SUSER },
 436
 437         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 438           BC_F_COPYOUT },
 439         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 440           BC_F_COPYIN|BC_F_SUSER },
 441
 442         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 443           BC_F_COPYOUT },
 444         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 445           BC_F_COPYIN|BC_F_SUSER },
 446
 447         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 448           BC_F_COPYOUT },
 449         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 450           BC_F_COPYIN|BC_F_SUSER },
 451
 452         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 453           BC_F_COPYIN|BC_F_SUSER },
 454
 455         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 456           BC_F_COPYIN|BC_F_SUSER },
 457
 458         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 459           BC_F_COPYIN|BC_F_SUSER },
 460         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 461           BC_F_COPYIN|BC_F_SUSER },
 462
 463         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 464           BC_F_COPYOUT },
 465
 466         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 467           BC_F_COPYOUT },
 468
 469         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 470           BC_F_COPYIN|BC_F_COPYOUT },
 471
 472         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 473           BC_F_COPYIN|BC_F_SUSER },
 474
 475         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 476           BC_F_COPYIN|BC_F_SUSER },
 477
 478         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 479           BC_F_COPYIN|BC_F_SUSER },
 480
 481 };
 482 const int bridge_control_table_size =
 483     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
 484
 485 LIST_HEAD(, bridge_softc) bridge_list;
 486
 487 IFC_SIMPLE_DECLARE(bridge, 0);
 488
 489 static int
 490 bridge_modevent(module_t mod, int type, void *data)
 491 {
 492
 493         switch (type) {
 494         case MOD_LOAD:
 495                 mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
 496                 if_clone_attach(&bridge_cloner);
 497                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 498                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 499                     UMA_ALIGN_PTR, 0);
 500                 LIST_INIT(&bridge_list);
 501                 bridge_input_p = bridge_input;
 502                 bridge_output_p = bridge_output;
 503                 bridge_dn_p = bridge_dummynet;
 504                 bridge_linkstate_p = bridge_linkstate;
 505                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 506                     ifnet_departure_event, bridge_ifdetach, NULL,
 507                     EVENTHANDLER_PRI_ANY);
 508                 break;
 509         case MOD_UNLOAD:
 510                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 511                     bridge_detach_cookie);
 512                 if_clone_detach(&bridge_cloner);
 513                 uma_zdestroy(bridge_rtnode_zone);
 514                 bridge_input_p = NULL;
 515                 bridge_output_p = NULL;
 516                 bridge_dn_p = NULL;
 517                 bridge_linkstate_p = NULL;
 518                 mtx_destroy(&bridge_list_mtx);
 519                 break;
 520         default:
 521                 return (EOPNOTSUPP);
 522         }
 523         return (0);
 524 }
 525
 526 static moduledata_t bridge_mod = {
 527         "if_bridge",
 528         bridge_modevent,
 529         0
 530 };
 531
 532 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 533 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 534
 535 /*
 536  * handler for net.link.bridge.pfil_ipfw
 537  */
 538 static int
 539 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 540 {
 541         int enable = pfil_ipfw;
 542         int error;
 543
 544         error = sysctl_handle_int(oidp, &enable, 0, req);
 545         enable = (enable) ? 1 : 0;
 546
 547         if (enable != pfil_ipfw) {
 548                 pfil_ipfw = enable;
 549
 550                 /*
 551                  * Disable pfil so that ipfw doesnt run twice, if the user
 552                  * really wants both then they can re-enable pfil_bridge and/or
 553                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 554                  * layer2 type.
 555                  */
 556                 if (pfil_ipfw) {
 557                         pfil_onlyip = 0;
 558                         pfil_bridge = 0;
 559                         pfil_member = 0;
 560                 }
 561         }
 562
 563         return (error);
 564 }
 565 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
 566             &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
 567
 568 /*
 569  * bridge_clone_create:
 570  *
 571  *      Create a new bridge instance.
 572  */
 573 static int
 574 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 575 {
 576         struct bridge_softc *sc, *sc2;
 577         struct ifnet *bifp, *ifp;
 578         int fb, retry;
 579         unsigned long hostid;
 580
 581         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 582         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 583         if (ifp == NULL) {
 584                 free(sc, M_DEVBUF);
 585                 return (ENOSPC);
 586         }
 587
 588         BRIDGE_LOCK_INIT(sc);
 589         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 590         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 591
 592         /* Initialize our routing table. */
 593         bridge_rtable_init(sc);
 594
 595         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 596
 597         LIST_INIT(&sc->sc_iflist);
 598         LIST_INIT(&sc->sc_spanlist);
 599
 600         ifp->if_softc = sc;
 601         if_initname(ifp, ifc->ifc_name, unit);
 602         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 603         ifp->if_ioctl = bridge_ioctl;
 604         ifp->if_start = bridge_start;
 605         ifp->if_init = bridge_init;
 606         ifp->if_type = IFT_BRIDGE;
 607         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 608         ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
 609         IFQ_SET_READY(&ifp->if_snd);
 610
 611         /*
 612          * Generate an ethernet address with a locally administered address.
 613          *
 614          * Since we are using random ethernet addresses for the bridge, it is
 615          * possible that we might have address collisions, so make sure that
 616          * this hardware address isn't already in use on another bridge.
 617          * The first try uses the hostid and falls back to arc4rand().
 618          */
 619         fb = 0;
 620         getcredhostid(curthread->td_ucred, &hostid);
 621         for (retry = 1; retry != 0;) {
 622                 if (fb || hostid == 0) {
 623                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 624                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 625                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 626                 } else {
 627                         sc->sc_defaddr[0] = 0x2;
 628                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 629                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 630                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 631                         sc->sc_defaddr[4] =  hostid        & 0xff;
 632                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 633                 }
 634
 635                 fb = 1;
 636                 retry = 0;
 637                 mtx_lock(&bridge_list_mtx);
 638                 LIST_FOREACH(sc2, &bridge_list, sc_list) {
 639                         bifp = sc2->sc_ifp;
 640                         if (memcmp(sc->sc_defaddr,
 641                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
 642                                 retry = 1;
 643                 }
 644                 mtx_unlock(&bridge_list_mtx);
 645         }
 646
 647         bstp_attach(&sc->sc_stp, &bridge_ops);
 648         ether_ifattach(ifp, sc->sc_defaddr);
 649         /* Now undo some of the damage... */
 650         ifp->if_baudrate = 0;
 651         ifp->if_type = IFT_BRIDGE;
 652
 653         mtx_lock(&bridge_list_mtx);
 654         LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
 655         mtx_unlock(&bridge_list_mtx);
 656
 657         return (0);
 658 }
 659
 660 /*
 661  * bridge_clone_destroy:
 662  *
 663  *      Destroy a bridge instance.
 664  */
 665 static void
 666 bridge_clone_destroy(struct ifnet *ifp)
 667 {
 668         struct bridge_softc *sc = ifp->if_softc;
 669         struct bridge_iflist *bif;
 670
 671         BRIDGE_LOCK(sc);
 672
 673         bridge_stop(ifp, 1);
 674         ifp->if_flags &= ~IFF_UP;
 675
 676         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 677                 bridge_delete_member(sc, bif, 0);
 678
 679         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 680                 bridge_delete_span(sc, bif);
 681         }
 682
 683         BRIDGE_UNLOCK(sc);
 684
 685         callout_drain(&sc->sc_brcallout);
 686
 687         mtx_lock(&bridge_list_mtx);
 688         LIST_REMOVE(sc, sc_list);
 689         mtx_unlock(&bridge_list_mtx);
 690
 691         bstp_detach(&sc->sc_stp);
 692         ether_ifdetach(ifp);
 693         if_free_type(ifp, IFT_ETHER);
 694
 695         /* Tear down the routing table. */
 696         bridge_rtable_fini(sc);
 697
 698         BRIDGE_LOCK_DESTROY(sc);
 699         free(sc, M_DEVBUF);
 700 }
 701
 702 /*
 703  * bridge_ioctl:
 704  *
 705  *      Handle a control request from the operator.
 706  */
 707 static int
 708 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 709 {
 710         struct bridge_softc *sc = ifp->if_softc;
 711         struct ifreq *ifr = (struct ifreq *)data;
 712         struct bridge_iflist *bif;
 713         struct thread *td = curthread;
 714         union {
 715                 struct ifbreq ifbreq;
 716                 struct ifbifconf ifbifconf;
 717                 struct ifbareq ifbareq;
 718                 struct ifbaconf ifbaconf;
 719                 struct ifbrparam ifbrparam;
 720                 struct ifbropreq ifbropreq;
 721         } args;
 722         struct ifdrv *ifd = (struct ifdrv *) data;
 723         const struct bridge_control *bc;
 724         int error = 0;
 725
 726         switch (cmd) {
 727
 728         case SIOCADDMULTI:
 729         case SIOCDELMULTI:
 730                 break;
 731
 732         case SIOCGDRVSPEC:
 733         case SIOCSDRVSPEC:
 734                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 735                         error = EINVAL;
 736                         break;
 737                 }
 738                 bc = &bridge_control_table[ifd->ifd_cmd];
 739
 740                 if (cmd == SIOCGDRVSPEC &&
 741                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 742                         error = EINVAL;
 743                         break;
 744                 }
 745                 else if (cmd == SIOCSDRVSPEC &&
 746                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 747                         error = EINVAL;
 748                         break;
 749                 }
 750
 751                 if (bc->bc_flags & BC_F_SUSER) {
 752                         error = priv_check(td, PRIV_NET_BRIDGE);
 753                         if (error)
 754                                 break;
 755                 }
 756
 757                 if (ifd->ifd_len != bc->bc_argsize ||
 758                     ifd->ifd_len > sizeof(args)) {
 759                         error = EINVAL;
 760                         break;
 761                 }
 762
 763                 bzero(&args, sizeof(args));
 764                 if (bc->bc_flags & BC_F_COPYIN) {
 765                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 766                         if (error)
 767                                 break;
 768                 }
 769
 770                 BRIDGE_LOCK(sc);
 771                 error = (*bc->bc_func)(sc, &args);
 772                 BRIDGE_UNLOCK(sc);
 773                 if (error)
 774                         break;
 775
 776                 if (bc->bc_flags & BC_F_COPYOUT)
 777                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 778
 779                 break;
 780
 781         case SIOCSIFFLAGS:
 782                 if (!(ifp->if_flags & IFF_UP) &&
 783                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 784                         /*
 785                          * If interface is marked down and it is running,
 786                          * then stop and disable it.
 787                          */
 788                         BRIDGE_LOCK(sc);
 789                         bridge_stop(ifp, 1);
 790                         BRIDGE_UNLOCK(sc);
 791                 } else if ((ifp->if_flags & IFF_UP) &&
 792                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 793                         /*
 794                          * If interface is marked up and it is stopped, then
 795                          * start it.
 796                          */
 797                         (*ifp->if_init)(sc);
 798                 }
 799                 break;
 800
 801         case SIOCSIFMTU:
 802                 if (ifr->ifr_mtu < 576) {
 803                         error = EINVAL;
 804                         break;
 805                 }
 806                 if (LIST_EMPTY(&sc->sc_iflist)) {
 807                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 808                         break;
 809                 }
 810                 BRIDGE_LOCK(sc);
 811                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 812                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 813                                 log(LOG_NOTICE, "%s: invalid MTU: %lu(%s)"
 814                                     " != %d\n", sc->sc_ifp->if_xname,
 815                                     bif->bif_ifp->if_mtu,
 816                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 817                                 error = EINVAL;
 818                                 break;
 819                         }
 820                 }
 821                 if (!error)
 822                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 823                 BRIDGE_UNLOCK(sc);
 824                 break;
 825         default:
 826                 /*
 827                  * drop the lock as ether_ioctl() will call bridge_start() and
 828                  * cause the lock to be recursed.
 829                  */
 830                 error = ether_ioctl(ifp, cmd, data);
 831                 break;
 832         }
 833
 834         return (error);
 835 }
 836
 837 /*
 838  * bridge_mutecaps:
 839  *
 840  *      Clear or restore unwanted capabilities on the member interface
 841  */
 842 static void
 843 bridge_mutecaps(struct bridge_softc *sc)
 844 {
 845         struct bridge_iflist *bif;
 846         int enabled, mask;
 847
 848         /* Initial bitmask of capabilities to test */
 849         mask = BRIDGE_IFCAPS_MASK;
 850
 851         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 852                 /* Every member must support it or its disabled */
 853                 mask &= bif->bif_savedcaps;
 854         }
 855
 856         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 857                 enabled = bif->bif_ifp->if_capenable;
 858                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 859                 /* strip off mask bits and enable them again if allowed */
 860                 enabled &= ~BRIDGE_IFCAPS_MASK;
 861                 enabled |= mask;
 862                 bridge_set_ifcap(sc, bif, enabled);
 863         }
 864
 865 }
 866
 867 static void
 868 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 869 {
 870         struct ifnet *ifp = bif->bif_ifp;
 871         struct ifreq ifr;
 872         int error;
 873
 874         bzero(&ifr, sizeof(ifr));
 875         ifr.ifr_reqcap = set;
 876
 877         if (ifp->if_capenable != set) {
 878                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 879                 if (error)
 880                         if_printf(sc->sc_ifp,
 881                             "error setting interface capabilities on %s\n",
 882                             ifp->if_xname);
 883         }
 884 }
 885
 886 /*
 887  * bridge_lookup_member:
 888  *
 889  *      Lookup a bridge member interface.
 890  */
 891 static struct bridge_iflist *
 892 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 893 {
 894         struct bridge_iflist *bif;
 895         struct ifnet *ifp;
 896
 897         BRIDGE_LOCK_ASSERT(sc);
 898
 899         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 900                 ifp = bif->bif_ifp;
 901                 if (strcmp(ifp->if_xname, name) == 0)
 902                         return (bif);
 903         }
 904
 905         return (NULL);
 906 }
 907
 908 /*
 909  * bridge_lookup_member_if:
 910  *
 911  *      Lookup a bridge member interface by ifnet*.
 912  */
 913 static struct bridge_iflist *
 914 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 915 {
 916         struct bridge_iflist *bif;
 917
 918         BRIDGE_LOCK_ASSERT(sc);
 919
 920         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 921                 if (bif->bif_ifp == member_ifp)
 922                         return (bif);
 923         }
 924
 925         return (NULL);
 926 }
 927
 928 /*
 929  * bridge_delete_member:
 930  *
 931  *      Delete the specified member interface.
 932  */
 933 static void
 934 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 935     int gone)
 936 {
 937         struct ifnet *ifs = bif->bif_ifp;
 938         struct ifnet *fif = NULL;
 939
 940         BRIDGE_LOCK_ASSERT(sc);
 941
 942         if (bif->bif_flags & IFBIF_STP)
 943                 bstp_disable(&bif->bif_stp);
 944
 945         ifs->if_bridge = NULL;
 946         BRIDGE_XLOCK(sc);
 947         LIST_REMOVE(bif, bif_next);
 948         BRIDGE_XDROP(sc);
 949
 950         /*
 951          * If removing the interface that gave the bridge its mac address, set
 952          * the mac address of the bridge to the address of the next member, or
 953          * to its default address if no members are left.
 954          */
 955         if (bridge_inherit_mac && sc->sc_ifaddr == ifs) {
 956                 if (LIST_EMPTY(&sc->sc_iflist)) {
 957                         bcopy(sc->sc_defaddr,
 958                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 959                         sc->sc_ifaddr = NULL;
 960                 } else {
 961                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
 962                         bcopy(IF_LLADDR(fif),
 963                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
 964                         sc->sc_ifaddr = fif;
 965                 }
 966                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
 967         }
 968
 969         bridge_linkcheck(sc);
 970         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
 971         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
 972         KASSERT(bif->bif_addrcnt == 0,
 973             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
 974
 975         BRIDGE_UNLOCK(sc);
 976         if (!gone) {
 977                 switch (ifs->if_type) {
 978                 case IFT_ETHER:
 979                 case IFT_L2VLAN:
 980                         /*
 981                          * Take the interface out of promiscuous mode.
 982                          */
 983                         (void) ifpromisc(ifs, 0);
 984                         break;
 985
 986                 case IFT_GIF:
 987                         break;
 988
 989                 default:
 990 #ifdef DIAGNOSTIC
 991                         panic("bridge_delete_member: impossible");
 992 #endif
 993                         break;
 994                 }
 995                 /* reneable any interface capabilities */
 996                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
 997         }
 998         bstp_destroy(&bif->bif_stp);    /* prepare to free */
 999         BRIDGE_LOCK(sc);
1000         free(bif, M_DEVBUF);
1001 }
1002
1003 /*
1004  * bridge_delete_span:
1005  *
1006  *      Delete the specified span interface.
1007  */
1008 static void
1009 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1010 {
1011         BRIDGE_LOCK_ASSERT(sc);
1012
1013         KASSERT(bif->bif_ifp->if_bridge == NULL,
1014             ("%s: not a span interface", __func__));
1015
1016         LIST_REMOVE(bif, bif_next);
1017         free(bif, M_DEVBUF);
1018 }
1019
1020 static int
1021 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1022 {
1023         struct ifbreq *req = arg;
1024         struct bridge_iflist *bif = NULL;
1025         struct ifnet *ifs;
1026         int error = 0;
1027
1028         ifs = ifunit(req->ifbr_ifsname);
1029         if (ifs == NULL)
1030                 return (ENOENT);
1031         if (ifs->if_ioctl == NULL)      /* must be supported */
1032                 return (EINVAL);
1033
1034         /* If it's in the span list, it can't be a member. */
1035         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1036                 if (ifs == bif->bif_ifp)
1037                         return (EBUSY);
1038
1039         if (ifs->if_bridge == sc)
1040                 return (EEXIST);
1041
1042         if (ifs->if_bridge != NULL)
1043                 return (EBUSY);
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                 return (EINVAL);
1053         }
1054
1055 #ifdef INET6
1056         /*
1057          * Two valid inet6 addresses with link-local scope must not be
1058          * on the parent interface and the member interfaces at the
1059          * same time.  This restriction is needed to prevent violation
1060          * of link-local scope zone.  Attempts to add a member
1061          * interface which has inet6 addresses when the parent has
1062          * inet6 triggers removal of all inet6 addresses on the member
1063          * interface.
1064          */
1065
1066         /* Check if the parent interface has a link-local scope addr. */
1067         if (in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1068                 /*
1069                  * If any, remove all inet6 addresses from the member
1070                  * interfaces.
1071                  */
1072                 BRIDGE_XLOCK(sc);
1073                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1074                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1075                                 BRIDGE_UNLOCK(sc);
1076                                 in6_ifdetach(bif->bif_ifp);
1077                                 BRIDGE_LOCK(sc);
1078                                 if_printf(sc->sc_ifp,
1079                                     "IPv6 addresses on %s have been removed "
1080                                     "before adding it as a member to prevent "
1081                                     "IPv6 address scope violation.\n",
1082                                     bif->bif_ifp->if_xname);
1083                         }
1084                 }
1085                 BRIDGE_XDROP(sc);
1086                 if (in6ifa_llaonifp(ifs)) {
1087                         BRIDGE_UNLOCK(sc);
1088                         in6_ifdetach(ifs);
1089                         BRIDGE_LOCK(sc);
1090                         if_printf(sc->sc_ifp,
1091                             "IPv6 addresses on %s have been removed "
1092                             "before adding it as a member to prevent "
1093                             "IPv6 address scope violation.\n",
1094                             ifs->if_xname);
1095                 }
1096         } else {
1097                 struct in6_ifaddr *ia6_m, *ia6_s;
1098                 /*
1099                  * If not, check whether one of the existing member
1100                  * interfaces have inet6 address.  If any, remove
1101                  * inet6 addresses on the interface to be added.
1102                  */
1103                 ia6_m = NULL;
1104                 BRIDGE_XLOCK(sc);
1105                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1106                         ia6_m = in6ifa_llaonifp(bif->bif_ifp);
1107                         if (ia6_m != NULL)
1108                                 break;
1109                 }
1110                 BRIDGE_XDROP(sc);
1111                 ia6_s = in6ifa_llaonifp(ifs);
1112
1113                 if (ia6_m != NULL && ia6_s != NULL) {
1114                         BRIDGE_UNLOCK(sc);
1115                         in6_ifdetach(ifs);
1116                         BRIDGE_LOCK(sc);
1117                         if_printf(sc->sc_ifp, "IPv6 addresses on %s have "
1118                                   "been removed before adding it as a member "
1119                                   "to prevent IPv6 address scope violation.\n",
1120                                   ifs->if_xname);
1121                 }
1122         }
1123 #endif
1124         /* Allow the first Ethernet member to define the MTU */
1125         if (LIST_EMPTY(&sc->sc_iflist))
1126                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1127         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1128                 if_printf(sc->sc_ifp, "invalid MTU: %lu(%s) != %lu\n",
1129                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1130                 return (EINVAL);
1131         }
1132
1133         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1134         if (bif == NULL)
1135                 return (ENOMEM);
1136
1137         bif->bif_ifp = ifs;
1138         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1139         bif->bif_savedcaps = ifs->if_capenable;
1140
1141         /*
1142          * Assign the interface's MAC address to the bridge if it's the first
1143          * member and the MAC address of the bridge has not been changed from
1144          * the default randomly generated one.
1145          */
1146         if (bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1147             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1148                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1149                 sc->sc_ifaddr = ifs;
1150                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1151         }
1152
1153         ifs->if_bridge = sc;
1154         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1155         /*
1156          * XXX: XLOCK HERE!?!
1157          *
1158          * NOTE: insert_***HEAD*** should be safe for the traversals.
1159          */
1160         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1161
1162         /* Set interface capabilities to the intersection set of all members */
1163         bridge_mutecaps(sc);
1164         bridge_linkcheck(sc);
1165
1166         /* Place the interface into promiscuous mode */
1167         switch (ifs->if_type) {
1168                 case IFT_ETHER:
1169                 case IFT_L2VLAN:
1170                         BRIDGE_UNLOCK(sc);
1171                         error = ifpromisc(ifs, 1);
1172                         BRIDGE_LOCK(sc);
1173                         break;
1174         }
1175
1176         if (error) {
1177                 bridge_delete_member(sc, bif, 0);
1178                 free(bif, M_DEVBUF);
1179         }
1180         return (error);
1181 }
1182
1183 static int
1184 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1185 {
1186         struct ifbreq *req = arg;
1187         struct bridge_iflist *bif;
1188
1189         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1190         if (bif == NULL)
1191                 return (ENOENT);
1192
1193         bridge_delete_member(sc, bif, 0);
1194
1195         return (0);
1196 }
1197
1198 static int
1199 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1200 {
1201         struct ifbreq *req = arg;
1202         struct bridge_iflist *bif;
1203         struct bstp_port *bp;
1204
1205         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1206         if (bif == NULL)
1207                 return (ENOENT);
1208
1209         bp = &bif->bif_stp;
1210         req->ifbr_ifsflags = bif->bif_flags;
1211         req->ifbr_state = bp->bp_state;
1212         req->ifbr_priority = bp->bp_priority;
1213         req->ifbr_path_cost = bp->bp_path_cost;
1214         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1215         req->ifbr_proto = bp->bp_protover;
1216         req->ifbr_role = bp->bp_role;
1217         req->ifbr_stpflags = bp->bp_flags;
1218         req->ifbr_addrcnt = bif->bif_addrcnt;
1219         req->ifbr_addrmax = bif->bif_addrmax;
1220         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1221
1222         /* Copy STP state options as flags */
1223         if (bp->bp_operedge)
1224                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1225         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1226                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1227         if (bp->bp_ptp_link)
1228                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1229         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1230                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1231         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1232                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1233         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1234                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1235         return (0);
1236 }
1237
1238 static int
1239 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1240 {
1241         struct ifbreq *req = arg;
1242         struct bridge_iflist *bif;
1243         struct bstp_port *bp;
1244         int error;
1245
1246         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1247         if (bif == NULL)
1248                 return (ENOENT);
1249         bp = &bif->bif_stp;
1250
1251         if (req->ifbr_ifsflags & IFBIF_SPAN)
1252                 /* SPAN is readonly */
1253                 return (EINVAL);
1254
1255         if (req->ifbr_ifsflags & IFBIF_STP) {
1256                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1257                         error = bstp_enable(&bif->bif_stp);
1258                         if (error)
1259                                 return (error);
1260                 }
1261         } else {
1262                 if ((bif->bif_flags & IFBIF_STP) != 0)
1263                         bstp_disable(&bif->bif_stp);
1264         }
1265
1266         /* Pass on STP flags */
1267         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1268         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1269         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1270         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1271
1272         /* Save the bits relating to the bridge */
1273         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1274
1275         return (0);
1276 }
1277
1278 static int
1279 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1280 {
1281         struct ifbrparam *param = arg;
1282
1283         sc->sc_brtmax = param->ifbrp_csize;
1284         bridge_rttrim(sc);
1285
1286         return (0);
1287 }
1288
1289 static int
1290 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1291 {
1292         struct ifbrparam *param = arg;
1293
1294         param->ifbrp_csize = sc->sc_brtmax;
1295
1296         return (0);
1297 }
1298
1299 static int
1300 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1301 {
1302         struct ifbifconf *bifc = arg;
1303         struct bridge_iflist *bif;
1304         struct ifbreq breq;
1305         char *buf, *outbuf;
1306         int count, buflen, len, error = 0;
1307
1308         count = 0;
1309         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1310                 count++;
1311         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1312                 count++;
1313
1314         buflen = sizeof(breq) * count;
1315         if (bifc->ifbic_len == 0) {
1316                 bifc->ifbic_len = buflen;
1317                 return (0);
1318         }
1319         BRIDGE_UNLOCK(sc);
1320         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1321         BRIDGE_LOCK(sc);
1322
1323         count = 0;
1324         buf = outbuf;
1325         len = min(bifc->ifbic_len, buflen);
1326         bzero(&breq, sizeof(breq));
1327         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1328                 if (len < sizeof(breq))
1329                         break;
1330
1331                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1332                     sizeof(breq.ifbr_ifsname));
1333                 /* Fill in the ifbreq structure */
1334                 error = bridge_ioctl_gifflags(sc, &breq);
1335                 if (error)
1336                         break;
1337                 memcpy(buf, &breq, sizeof(breq));
1338                 count++;
1339                 buf += sizeof(breq);
1340                 len -= sizeof(breq);
1341         }
1342         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1343                 if (len < sizeof(breq))
1344                         break;
1345
1346                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1347                     sizeof(breq.ifbr_ifsname));
1348                 breq.ifbr_ifsflags = bif->bif_flags;
1349                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1350                 memcpy(buf, &breq, sizeof(breq));
1351                 count++;
1352                 buf += sizeof(breq);
1353                 len -= sizeof(breq);
1354         }
1355
1356         BRIDGE_UNLOCK(sc);
1357         bifc->ifbic_len = sizeof(breq) * count;
1358         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1359         BRIDGE_LOCK(sc);
1360         free(outbuf, M_TEMP);
1361         return (error);
1362 }
1363
1364 static int
1365 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1366 {
1367         struct ifbaconf *bac = arg;
1368         struct bridge_rtnode *brt;
1369         struct ifbareq bareq;
1370         char *buf, *outbuf;
1371         int count, buflen, len, error = 0;
1372
1373         if (bac->ifbac_len == 0)
1374                 return (0);
1375
1376         count = 0;
1377         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1378                 count++;
1379         buflen = sizeof(bareq) * count;
1380
1381         BRIDGE_UNLOCK(sc);
1382         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1383         BRIDGE_LOCK(sc);
1384
1385         count = 0;
1386         buf = outbuf;
1387         len = min(bac->ifbac_len, buflen);
1388         bzero(&bareq, sizeof(bareq));
1389         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1390                 if (len < sizeof(bareq))
1391                         goto out;
1392                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1393                     sizeof(bareq.ifba_ifsname));
1394                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1395                 bareq.ifba_vlan = brt->brt_vlan;
1396                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1397                                 time_uptime < brt->brt_expire)
1398                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1399                 else
1400                         bareq.ifba_expire = 0;
1401                 bareq.ifba_flags = brt->brt_flags;
1402
1403                 memcpy(buf, &bareq, sizeof(bareq));
1404                 count++;
1405                 buf += sizeof(bareq);
1406                 len -= sizeof(bareq);
1407         }
1408 out:
1409         BRIDGE_UNLOCK(sc);
1410         bac->ifbac_len = sizeof(bareq) * count;
1411         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1412         BRIDGE_LOCK(sc);
1413         free(outbuf, M_TEMP);
1414         return (error);
1415 }
1416
1417 static int
1418 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1419 {
1420         struct ifbareq *req = arg;
1421         struct bridge_iflist *bif;
1422         int error;
1423
1424         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1425         if (bif == NULL)
1426                 return (ENOENT);
1427
1428         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1429             req->ifba_flags);
1430
1431         return (error);
1432 }
1433
1434 static int
1435 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1436 {
1437         struct ifbrparam *param = arg;
1438
1439         sc->sc_brttimeout = param->ifbrp_ctime;
1440         return (0);
1441 }
1442
1443 static int
1444 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1445 {
1446         struct ifbrparam *param = arg;
1447
1448         param->ifbrp_ctime = sc->sc_brttimeout;
1449         return (0);
1450 }
1451
1452 static int
1453 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1454 {
1455         struct ifbareq *req = arg;
1456
1457         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1458 }
1459
1460 static int
1461 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1462 {
1463         struct ifbreq *req = arg;
1464
1465         bridge_rtflush(sc, req->ifbr_ifsflags);
1466         return (0);
1467 }
1468
1469 static int
1470 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1471 {
1472         struct ifbrparam *param = arg;
1473         struct bstp_state *bs = &sc->sc_stp;
1474
1475         param->ifbrp_prio = bs->bs_bridge_priority;
1476         return (0);
1477 }
1478
1479 static int
1480 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1481 {
1482         struct ifbrparam *param = arg;
1483
1484         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1485 }
1486
1487 static int
1488 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1489 {
1490         struct ifbrparam *param = arg;
1491         struct bstp_state *bs = &sc->sc_stp;
1492
1493         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1494         return (0);
1495 }
1496
1497 static int
1498 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1499 {
1500         struct ifbrparam *param = arg;
1501
1502         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1503 }
1504
1505 static int
1506 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1507 {
1508         struct ifbrparam *param = arg;
1509         struct bstp_state *bs = &sc->sc_stp;
1510
1511         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1512         return (0);
1513 }
1514
1515 static int
1516 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1517 {
1518         struct ifbrparam *param = arg;
1519
1520         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1521 }
1522
1523 static int
1524 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1525 {
1526         struct ifbrparam *param = arg;
1527         struct bstp_state *bs = &sc->sc_stp;
1528
1529         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1530         return (0);
1531 }
1532
1533 static int
1534 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1535 {
1536         struct ifbrparam *param = arg;
1537
1538         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1539 }
1540
1541 static int
1542 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1543 {
1544         struct ifbreq *req = arg;
1545         struct bridge_iflist *bif;
1546
1547         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1548         if (bif == NULL)
1549                 return (ENOENT);
1550
1551         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1552 }
1553
1554 static int
1555 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1556 {
1557         struct ifbreq *req = arg;
1558         struct bridge_iflist *bif;
1559
1560         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1561         if (bif == NULL)
1562                 return (ENOENT);
1563
1564         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1565 }
1566
1567 static int
1568 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1569 {
1570         struct ifbreq *req = arg;
1571         struct bridge_iflist *bif;
1572
1573         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1574         if (bif == NULL)
1575                 return (ENOENT);
1576
1577         bif->bif_addrmax = req->ifbr_addrmax;
1578         return (0);
1579 }
1580
1581 static int
1582 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1583 {
1584         struct ifbreq *req = arg;
1585         struct bridge_iflist *bif = NULL;
1586         struct ifnet *ifs;
1587
1588         ifs = ifunit(req->ifbr_ifsname);
1589         if (ifs == NULL)
1590                 return (ENOENT);
1591
1592         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1593                 if (ifs == bif->bif_ifp)
1594                         return (EBUSY);
1595
1596         if (ifs->if_bridge != NULL)
1597                 return (EBUSY);
1598
1599         switch (ifs->if_type) {
1600                 case IFT_ETHER:
1601                 case IFT_GIF:
1602                 case IFT_L2VLAN:
1603                         break;
1604                 default:
1605                         return (EINVAL);
1606         }
1607
1608         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1609         if (bif == NULL)
1610                 return (ENOMEM);
1611
1612         bif->bif_ifp = ifs;
1613         bif->bif_flags = IFBIF_SPAN;
1614
1615         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1616
1617         return (0);
1618 }
1619
1620 static int
1621 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1622 {
1623         struct ifbreq *req = arg;
1624         struct bridge_iflist *bif;
1625         struct ifnet *ifs;
1626
1627         ifs = ifunit(req->ifbr_ifsname);
1628         if (ifs == NULL)
1629                 return (ENOENT);
1630
1631         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1632                 if (ifs == bif->bif_ifp)
1633                         break;
1634
1635         if (bif == NULL)
1636                 return (ENOENT);
1637
1638         bridge_delete_span(sc, bif);
1639
1640         return (0);
1641 }
1642
1643 static int
1644 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1645 {
1646         struct ifbropreq *req = arg;
1647         struct bstp_state *bs = &sc->sc_stp;
1648         struct bstp_port *root_port;
1649
1650         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1651         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1652         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1653
1654         root_port = bs->bs_root_port;
1655         if (root_port == NULL)
1656                 req->ifbop_root_port = 0;
1657         else
1658                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1659
1660         req->ifbop_holdcount = bs->bs_txholdcount;
1661         req->ifbop_priority = bs->bs_bridge_priority;
1662         req->ifbop_protocol = bs->bs_protover;
1663         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1664         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1665         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1666         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1667         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1668         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1669
1670         return (0);
1671 }
1672
1673 static int
1674 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1675 {
1676         struct ifbrparam *param = arg;
1677
1678         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1679         return (0);
1680 }
1681
1682 static int
1683 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1684 {
1685         struct ifbpstpconf *bifstp = arg;
1686         struct bridge_iflist *bif;
1687         struct bstp_port *bp;
1688         struct ifbpstpreq bpreq;
1689         char *buf, *outbuf;
1690         int count, buflen, len, error = 0;
1691
1692         count = 0;
1693         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1694                 if ((bif->bif_flags & IFBIF_STP) != 0)
1695                         count++;
1696         }
1697
1698         buflen = sizeof(bpreq) * count;
1699         if (bifstp->ifbpstp_len == 0) {
1700                 bifstp->ifbpstp_len = buflen;
1701                 return (0);
1702         }
1703
1704         BRIDGE_UNLOCK(sc);
1705         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1706         BRIDGE_LOCK(sc);
1707
1708         count = 0;
1709         buf = outbuf;
1710         len = min(bifstp->ifbpstp_len, buflen);
1711         bzero(&bpreq, sizeof(bpreq));
1712         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1713                 if (len < sizeof(bpreq))
1714                         break;
1715
1716                 if ((bif->bif_flags & IFBIF_STP) == 0)
1717                         continue;
1718
1719                 bp = &bif->bif_stp;
1720                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1721                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1722                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1723                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1724                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1725                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1726
1727                 memcpy(buf, &bpreq, sizeof(bpreq));
1728                 count++;
1729                 buf += sizeof(bpreq);
1730                 len -= sizeof(bpreq);
1731         }
1732
1733         BRIDGE_UNLOCK(sc);
1734         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1735         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1736         BRIDGE_LOCK(sc);
1737         free(outbuf, M_TEMP);
1738         return (error);
1739 }
1740
1741 static int
1742 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1743 {
1744         struct ifbrparam *param = arg;
1745
1746         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1747 }
1748
1749 static int
1750 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1751 {
1752         struct ifbrparam *param = arg;
1753
1754         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1755 }
1756
1757 /*
1758  * bridge_ifdetach:
1759  *
1760  *      Detach an interface from a bridge.  Called when a member
1761  *      interface is detaching.
1762  */
1763 static void
1764 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1765 {
1766         struct bridge_softc *sc = ifp->if_bridge;
1767         struct bridge_iflist *bif;
1768
1769         if (ifp->if_flags & IFF_RENAMING)
1770                 return;
1771
1772         /* Check if the interface is a bridge member */
1773         if (sc != NULL) {
1774                 BRIDGE_LOCK(sc);
1775
1776                 bif = bridge_lookup_member_if(sc, ifp);
1777                 if (bif != NULL)
1778                         bridge_delete_member(sc, bif, 1);
1779
1780                 BRIDGE_UNLOCK(sc);
1781                 return;
1782         }
1783
1784         /* Check if the interface is a span port */
1785         mtx_lock(&bridge_list_mtx);
1786         LIST_FOREACH(sc, &bridge_list, sc_list) {
1787                 BRIDGE_LOCK(sc);
1788                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1789                         if (ifp == bif->bif_ifp) {
1790                                 bridge_delete_span(sc, bif);
1791                                 break;
1792                         }
1793
1794                 BRIDGE_UNLOCK(sc);
1795         }
1796         mtx_unlock(&bridge_list_mtx);
1797 }
1798
1799 /*
1800  * bridge_init:
1801  *
1802  *      Initialize a bridge interface.
1803  */
1804 static void
1805 bridge_init(void *xsc)
1806 {
1807         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1808         struct ifnet *ifp = sc->sc_ifp;
1809
1810         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1811                 return;
1812
1813         BRIDGE_LOCK(sc);
1814         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1815             bridge_timer, sc);
1816
1817         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1818         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1819
1820         BRIDGE_UNLOCK(sc);
1821 }
1822
1823 /*
1824  * bridge_stop:
1825  *
1826  *      Stop the bridge interface.
1827  */
1828 static void
1829 bridge_stop(struct ifnet *ifp, int disable)
1830 {
1831         struct bridge_softc *sc = ifp->if_softc;
1832
1833         BRIDGE_LOCK_ASSERT(sc);
1834
1835         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1836                 return;
1837
1838         callout_stop(&sc->sc_brcallout);
1839         bstp_stop(&sc->sc_stp);
1840
1841         bridge_rtflush(sc, IFBF_FLUSHDYN);
1842
1843         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1844 }
1845
1846 /*
1847  * bridge_enqueue:
1848  *
1849  *      Enqueue a packet on a bridge member interface.
1850  *
1851  */
1852 static void
1853 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1854 {
1855         int len, err = 0;
1856         short mflags;
1857         struct mbuf *m0;
1858
1859         len = m->m_pkthdr.len;
1860         mflags = m->m_flags;
1861
1862         /* We may be sending a fragment so traverse the mbuf */
1863         for (; m; m = m0) {
1864                 m0 = m->m_nextpkt;
1865                 m->m_nextpkt = NULL;
1866
1867                 /*
1868                  * If underlying interface can not do VLAN tag insertion itself
1869                  * then attach a packet tag that holds it.
1870                  */
1871                 if ((m->m_flags & M_VLANTAG) &&
1872                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1873                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1874                         if (m == NULL) {
1875                                 if_printf(dst_ifp,
1876                                     "unable to prepend VLAN header\n");
1877                                 dst_ifp->if_oerrors++;
1878                                 continue;
1879                         }
1880                         m->m_flags &= ~M_VLANTAG;
1881                 }
1882
1883                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1884                         m_freem(m0);
1885                         break;
1886                 }
1887         }
1888
1889         if (err == 0) {
1890                 sc->sc_ifp->if_opackets++;
1891                 sc->sc_ifp->if_obytes += len;
1892                 if (mflags & M_MCAST)
1893                         sc->sc_ifp->if_omcasts++;
1894         }
1895 }
1896
1897 /*
1898  * bridge_dummynet:
1899  *
1900  *      Receive a queued packet from dummynet and pass it on to the output
1901  *      interface.
1902  *
1903  *      The mbuf has the Ethernet header already attached.
1904  */
1905 static void
1906 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1907 {
1908         struct bridge_softc *sc;
1909
1910         sc = ifp->if_bridge;
1911
1912         /*
1913          * The packet didnt originate from a member interface. This should only
1914          * ever happen if a member interface is removed while packets are
1915          * queued for it.
1916          */
1917         if (sc == NULL) {
1918                 m_freem(m);
1919                 return;
1920         }
1921
1922         if (PFIL_HOOKED(&V_inet_pfil_hook)
1923 #ifdef INET6
1924             || PFIL_HOOKED(&V_inet6_pfil_hook)
1925 #endif
1926             ) {
1927                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1928                         return;
1929                 if (m == NULL)
1930                         return;
1931         }
1932
1933         bridge_enqueue(sc, ifp, m);
1934 }
1935
1936 /*
1937  * bridge_output:
1938  *
1939  *      Send output from a bridge member interface.  This
1940  *      performs the bridging function for locally originated
1941  *      packets.
1942  *
1943  *      The mbuf has the Ethernet header already attached.  We must
1944  *      enqueue or free the mbuf before returning.
1945  */
1946 static int
1947 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1948     struct rtentry *rt)
1949 {
1950         struct ether_header *eh;
1951         struct ifnet *dst_if;
1952         struct bridge_softc *sc;
1953         uint16_t vlan;
1954
1955         if (m->m_len < ETHER_HDR_LEN) {
1956                 m = m_pullup(m, ETHER_HDR_LEN);
1957                 if (m == NULL)
1958                         return (0);
1959         }
1960
1961         eh = mtod(m, struct ether_header *);
1962         sc = ifp->if_bridge;
1963         vlan = VLANTAGOF(m);
1964
1965         BRIDGE_LOCK(sc);
1966
1967         /*
1968          * If bridge is down, but the original output interface is up,
1969          * go ahead and send out that interface.  Otherwise, the packet
1970          * is dropped below.
1971          */
1972         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1973                 dst_if = ifp;
1974                 goto sendunicast;
1975         }
1976
1977         /*
1978          * If the packet is a multicast, or we don't know a better way to
1979          * get there, send to all interfaces.
1980          */
1981         if (ETHER_IS_MULTICAST(eh->ether_dhost))
1982                 dst_if = NULL;
1983         else
1984                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1985         if (dst_if == NULL) {
1986                 struct bridge_iflist *bif;
1987                 struct mbuf *mc;
1988                 int error = 0, used = 0;
1989
1990                 bridge_span(sc, m);
1991
1992                 BRIDGE_LOCK2REF(sc, error);
1993                 if (error) {
1994                         m_freem(m);
1995                         return (0);
1996                 }
1997
1998                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1999                         dst_if = bif->bif_ifp;
2000
2001                         if (dst_if->if_type == IFT_GIF)
2002                                 continue;
2003                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2004                                 continue;
2005
2006                         /*
2007                          * If this is not the original output interface,
2008                          * and the interface is participating in spanning
2009                          * tree, make sure the port is in a state that
2010                          * allows forwarding.
2011                          */
2012                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2013                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2014                                 continue;
2015
2016                         if (LIST_NEXT(bif, bif_next) == NULL) {
2017                                 used = 1;
2018                                 mc = m;
2019                         } else {
2020                                 mc = m_copypacket(m, M_DONTWAIT);
2021                                 if (mc == NULL) {
2022                                         sc->sc_ifp->if_oerrors++;
2023                                         continue;
2024                                 }
2025                         }
2026
2027                         bridge_enqueue(sc, dst_if, mc);
2028                 }
2029                 if (used == 0)
2030                         m_freem(m);
2031                 BRIDGE_UNREF(sc);
2032                 return (0);
2033         }
2034
2035 sendunicast:
2036         /*
2037          * XXX Spanning tree consideration here?
2038          */
2039
2040         bridge_span(sc, m);
2041         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2042                 m_freem(m);
2043                 BRIDGE_UNLOCK(sc);
2044                 return (0);
2045         }
2046
2047         BRIDGE_UNLOCK(sc);
2048         bridge_enqueue(sc, dst_if, m);
2049         return (0);
2050 }
2051
2052 /*
2053  * bridge_start:
2054  *
2055  *      Start output on a bridge.
2056  *
2057  */
2058 static void
2059 bridge_start(struct ifnet *ifp)
2060 {
2061         struct bridge_softc *sc;
2062         struct mbuf *m;
2063         struct ether_header *eh;
2064         struct ifnet *dst_if;
2065
2066         sc = ifp->if_softc;
2067
2068         ifp->if_drv_flags |= IFF_DRV_OACTIVE;
2069         for (;;) {
2070                 IFQ_DEQUEUE(&ifp->if_snd, m);
2071                 if (m == 0)
2072                         break;
2073                 ETHER_BPF_MTAP(ifp, m);
2074
2075                 eh = mtod(m, struct ether_header *);
2076                 dst_if = NULL;
2077
2078                 BRIDGE_LOCK(sc);
2079                 if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2080                         dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
2081                 }
2082
2083                 if (dst_if == NULL)
2084                         bridge_broadcast(sc, ifp, m, 0);
2085                 else {
2086                         BRIDGE_UNLOCK(sc);
2087                         bridge_enqueue(sc, dst_if, m);
2088                 }
2089         }
2090         ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
2091 }
2092
2093 /*
2094  * bridge_forward:
2095  *
2096  *      The forwarding function of the bridge.
2097  *
2098  *      NOTE: Releases the lock on return.
2099  */
2100 static void
2101 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2102     struct mbuf *m)
2103 {
2104         struct bridge_iflist *dbif;
2105         struct ifnet *src_if, *dst_if, *ifp;
2106         struct ether_header *eh;
2107         uint16_t vlan;
2108         uint8_t *dst;
2109         int error;
2110
2111         src_if = m->m_pkthdr.rcvif;
2112         ifp = sc->sc_ifp;
2113
2114         ifp->if_ipackets++;
2115         ifp->if_ibytes += m->m_pkthdr.len;
2116         vlan = VLANTAGOF(m);
2117
2118         if ((sbif->bif_flags & IFBIF_STP) &&
2119             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2120                 goto drop;
2121
2122         eh = mtod(m, struct ether_header *);
2123         dst = eh->ether_dhost;
2124
2125         /* If the interface is learning, record the address. */
2126         if (sbif->bif_flags & IFBIF_LEARNING) {
2127                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2128                     sbif, 0, IFBAF_DYNAMIC);
2129                 /*
2130                  * If the interface has addresses limits then deny any source
2131                  * that is not in the cache.
2132                  */
2133                 if (error && sbif->bif_addrmax)
2134                         goto drop;
2135         }
2136
2137         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2138             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2139                 goto drop;
2140
2141         /*
2142          * At this point, the port either doesn't participate
2143          * in spanning tree or it is in the forwarding state.
2144          */
2145
2146         /*
2147          * If the packet is unicast, destined for someone on
2148          * "this" side of the bridge, drop it.
2149          */
2150         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2151                 dst_if = bridge_rtlookup(sc, dst, vlan);
2152                 if (src_if == dst_if)
2153                         goto drop;
2154         } else {
2155                 /*
2156                  * Check if its a reserved multicast address, any address
2157                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2158                  * bridge.
2159                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2160                  */
2161                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2162                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2163                     dst[4] == 0x00 && dst[5] <= 0x0f)
2164                         goto drop;
2165
2166                 /* ...forward it to all interfaces. */
2167                 ifp->if_imcasts++;
2168                 dst_if = NULL;
2169         }
2170
2171         /*
2172          * If we have a destination interface which is a member of our bridge,
2173          * OR this is a unicast packet, push it through the bpf(4) machinery.
2174          * For broadcast or multicast packets, don't bother because it will
2175          * be reinjected into ether_input. We do this before we pass the packets
2176          * through the pfil(9) framework, as it is possible that pfil(9) will
2177          * drop the packet, or possibly modify it, making it difficult to debug
2178          * firewall issues on the bridge.
2179          */
2180         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2181                 ETHER_BPF_MTAP(ifp, m);
2182
2183         /* run the packet filter */
2184         if (PFIL_HOOKED(&V_inet_pfil_hook)
2185 #ifdef INET6
2186             || PFIL_HOOKED(&V_inet6_pfil_hook)
2187 #endif
2188             ) {
2189                 BRIDGE_UNLOCK(sc);
2190                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2191                         return;
2192                 if (m == NULL)
2193                         return;
2194                 BRIDGE_LOCK(sc);
2195         }
2196
2197         if (dst_if == NULL) {
2198                 bridge_broadcast(sc, src_if, m, 1);
2199                 return;
2200         }
2201
2202         /*
2203          * At this point, we're dealing with a unicast frame
2204          * going to a different interface.
2205          */
2206         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2207                 goto drop;
2208
2209         dbif = bridge_lookup_member_if(sc, dst_if);
2210         if (dbif == NULL)
2211                 /* Not a member of the bridge (anymore?) */
2212                 goto drop;
2213
2214         /* Private segments can not talk to each other */
2215         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2216                 goto drop;
2217
2218         if ((dbif->bif_flags & IFBIF_STP) &&
2219             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2220                 goto drop;
2221
2222         BRIDGE_UNLOCK(sc);
2223
2224         if (PFIL_HOOKED(&V_inet_pfil_hook)
2225 #ifdef INET6
2226             || PFIL_HOOKED(&V_inet6_pfil_hook)
2227 #endif
2228             ) {
2229                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2230                         return;
2231                 if (m == NULL)
2232                         return;
2233         }
2234
2235         bridge_enqueue(sc, dst_if, m);
2236         return;
2237
2238 drop:
2239         BRIDGE_UNLOCK(sc);
2240         m_freem(m);
2241 }
2242
2243 /*
2244  * bridge_input:
2245  *
2246  *      Receive input from a member interface.  Queue the packet for
2247  *      bridging if it is not for us.
2248  */
2249 static struct mbuf *
2250 bridge_input(struct ifnet *ifp, struct mbuf *m)
2251 {
2252         struct bridge_softc *sc = ifp->if_bridge;
2253         struct bridge_iflist *bif, *bif2;
2254         struct ifnet *bifp;
2255         struct ether_header *eh;
2256         struct mbuf *mc, *mc2;
2257         uint16_t vlan;
2258         int error;
2259
2260         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2261                 return (m);
2262
2263         bifp = sc->sc_ifp;
2264         vlan = VLANTAGOF(m);
2265
2266         /*
2267          * Implement support for bridge monitoring. If this flag has been
2268          * set on this interface, discard the packet once we push it through
2269          * the bpf(4) machinery, but before we do, increment the byte and
2270          * packet counters associated with this interface.
2271          */
2272         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2273                 m->m_pkthdr.rcvif  = bifp;
2274                 ETHER_BPF_MTAP(bifp, m);
2275                 bifp->if_ipackets++;
2276                 bifp->if_ibytes += m->m_pkthdr.len;
2277                 m_freem(m);
2278                 return (NULL);
2279         }
2280         BRIDGE_LOCK(sc);
2281         bif = bridge_lookup_member_if(sc, ifp);
2282         if (bif == NULL) {
2283                 BRIDGE_UNLOCK(sc);
2284                 return (m);
2285         }
2286
2287         eh = mtod(m, struct ether_header *);
2288
2289         bridge_span(sc, m);
2290
2291         if (m->m_flags & (M_BCAST|M_MCAST)) {
2292                 /* Tap off 802.1D packets; they do not get forwarded. */
2293                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2294                     ETHER_ADDR_LEN) == 0) {
2295                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2296                         BRIDGE_UNLOCK(sc);
2297                         return (NULL);
2298                 }
2299
2300                 if ((bif->bif_flags & IFBIF_STP) &&
2301                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2302                         BRIDGE_UNLOCK(sc);
2303                         return (m);
2304                 }
2305
2306                 /*
2307                  * Make a deep copy of the packet and enqueue the copy
2308                  * for bridge processing; return the original packet for
2309                  * local processing.
2310                  */
2311                 mc = m_dup(m, M_DONTWAIT);
2312                 if (mc == NULL) {
2313                         BRIDGE_UNLOCK(sc);
2314                         return (m);
2315                 }
2316
2317                 /* Perform the bridge forwarding function with the copy. */
2318                 bridge_forward(sc, bif, mc);
2319
2320                 /*
2321                  * Reinject the mbuf as arriving on the bridge so we have a
2322                  * chance at claiming multicast packets. We can not loop back
2323                  * here from ether_input as a bridge is never a member of a
2324                  * bridge.
2325                  */
2326                 KASSERT(bifp->if_bridge == NULL,
2327                     ("loop created in bridge_input"));
2328                 mc2 = m_dup(m, M_DONTWAIT);
2329                 if (mc2 != NULL) {
2330                         /* Keep the layer3 header aligned */
2331                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2332                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2333                 }
2334                 if (mc2 != NULL) {
2335                         mc2->m_pkthdr.rcvif = bifp;
2336                         (*bifp->if_input)(bifp, mc2);
2337                 }
2338
2339                 /* Return the original packet for local processing. */
2340                 return (m);
2341         }
2342
2343         if ((bif->bif_flags & IFBIF_STP) &&
2344             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2345                 BRIDGE_UNLOCK(sc);
2346                 return (m);
2347         }
2348
2349 #if (defined(INET) || defined(INET6))
2350 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2351         || ((iface)->if_carp \
2352             && (*carp_forus_p)((iface), eh->ether_dhost))
2353 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2354         || ((iface)->if_carp \
2355             && (*carp_forus_p)((iface), eh->ether_shost))
2356 #else
2357 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2358 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2359 #endif
2360
2361 #ifdef INET6
2362 #   define OR_PFIL_HOOKED_INET6 \
2363         || PFIL_HOOKED(&V_inet6_pfil_hook)
2364 #else
2365 #   define OR_PFIL_HOOKED_INET6
2366 #endif
2367
2368 #define GRAB_OUR_PACKETS(iface) \
2369         if ((iface)->if_type == IFT_GIF) \
2370                 continue; \
2371         /* It is destined for us. */ \
2372         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2373             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2374             ) {                                                         \
2375                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2376                         ETHER_BPF_MTAP(iface, m);                       \
2377                         iface->if_ipackets++;                           \
2378                         /* Filter on the physical interface. */         \
2379                         if (pfil_local_phys &&                          \
2380                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2381                              OR_PFIL_HOOKED_INET6)) {                   \
2382                                 if (bridge_pfil(&m, NULL, ifp,          \
2383                                     PFIL_IN) != 0 || m == NULL) {       \
2384                                         BRIDGE_UNLOCK(sc);              \
2385                                         return (NULL);                  \
2386                                 }                                       \
2387                                 eh = mtod(m, struct ether_header *);    \
2388                         }                                               \
2389                 }                                                       \
2390                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2391                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2392                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2393                         if (error && bif->bif_addrmax) {                \
2394                                 BRIDGE_UNLOCK(sc);                      \
2395                                 m_freem(m);                             \
2396                                 return (NULL);                          \
2397                         }                                               \
2398                 }                                                       \
2399                 m->m_pkthdr.rcvif = iface;                              \
2400                 BRIDGE_UNLOCK(sc);                                      \
2401                 return (m);                                             \
2402         }                                                               \
2403                                                                         \
2404         /* We just received a packet that we sent out. */               \
2405         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2406             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2407             ) {                                                         \
2408                 BRIDGE_UNLOCK(sc);                                      \
2409                 m_freem(m);                                             \
2410                 return (NULL);                                          \
2411         }
2412
2413         /*
2414          * Unicast.  Make sure it's not for the bridge.
2415          */
2416         do { GRAB_OUR_PACKETS(bifp) } while (0);
2417
2418         /*
2419          * Give a chance for ifp at first priority. This will help when the
2420          * packet comes through the interface like VLAN's with the same MACs
2421          * on several interfaces from the same bridge. This also will save
2422          * some CPU cycles in case the destination interface and the input
2423          * interface (eq ifp) are the same.
2424          */
2425         do { GRAB_OUR_PACKETS(ifp) } while (0);
2426
2427         /* Now check the all bridge members. */
2428         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2429                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2430         }
2431
2432 #undef OR_CARP_CHECK_WE_ARE_DST
2433 #undef OR_CARP_CHECK_WE_ARE_SRC
2434 #undef OR_PFIL_HOOKED_INET6
2435 #undef GRAB_OUR_PACKETS
2436
2437         /* Perform the bridge forwarding function. */
2438         bridge_forward(sc, bif, m);
2439
2440         return (NULL);
2441 }
2442
2443 /*
2444  * bridge_broadcast:
2445  *
2446  *      Send a frame to all interfaces that are members of
2447  *      the bridge, except for the one on which the packet
2448  *      arrived.
2449  *
2450  *      NOTE: Releases the lock on return.
2451  */
2452 static void
2453 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2454     struct mbuf *m, int runfilt)
2455 {
2456         struct bridge_iflist *dbif, *sbif;
2457         struct mbuf *mc;
2458         struct ifnet *dst_if;
2459         int error = 0, used = 0, i;
2460
2461         sbif = bridge_lookup_member_if(sc, src_if);
2462
2463         BRIDGE_LOCK2REF(sc, error);
2464         if (error) {
2465                 m_freem(m);
2466                 return;
2467         }
2468
2469         /* Filter on the bridge interface before broadcasting */
2470         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2471 #ifdef INET6
2472             || PFIL_HOOKED(&V_inet6_pfil_hook)
2473 #endif
2474             )) {
2475                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2476                         goto out;
2477                 if (m == NULL)
2478                         goto out;
2479         }
2480
2481         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2482                 dst_if = dbif->bif_ifp;
2483                 if (dst_if == src_if)
2484                         continue;
2485
2486                 /* Private segments can not talk to each other */
2487                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2488                         continue;
2489
2490                 if ((dbif->bif_flags & IFBIF_STP) &&
2491                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2492                         continue;
2493
2494                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2495                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2496                         continue;
2497
2498                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2499                         continue;
2500
2501                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2502                         mc = m;
2503                         used = 1;
2504                 } else {
2505                         mc = m_dup(m, M_DONTWAIT);
2506                         if (mc == NULL) {
2507                                 sc->sc_ifp->if_oerrors++;
2508                                 continue;
2509                         }
2510                 }
2511
2512                 /*
2513                  * Filter on the output interface. Pass a NULL bridge interface
2514                  * pointer so we do not redundantly filter on the bridge for
2515                  * each interface we broadcast on.
2516                  */
2517                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2518 #ifdef INET6
2519                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2520 #endif
2521                     )) {
2522                         if (used == 0) {
2523                                 /* Keep the layer3 header aligned */
2524                                 i = min(mc->m_pkthdr.len, max_protohdr);
2525                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2526                                 if (mc == NULL) {
2527                                         sc->sc_ifp->if_oerrors++;
2528                                         continue;
2529                                 }
2530                         }
2531                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2532                                 continue;
2533                         if (mc == NULL)
2534                                 continue;
2535                 }
2536
2537                 bridge_enqueue(sc, dst_if, mc);
2538         }
2539         if (used == 0)
2540                 m_freem(m);
2541
2542 out:
2543         BRIDGE_UNREF(sc);
2544 }
2545
2546 /*
2547  * bridge_span:
2548  *
2549  *      Duplicate a packet out one or more interfaces that are in span mode,
2550  *      the original mbuf is unmodified.
2551  */
2552 static void
2553 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2554 {
2555         struct bridge_iflist *bif;
2556         struct ifnet *dst_if;
2557         struct mbuf *mc;
2558
2559         if (LIST_EMPTY(&sc->sc_spanlist))
2560                 return;
2561
2562         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2563                 dst_if = bif->bif_ifp;
2564
2565                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2566                         continue;
2567
2568                 mc = m_copypacket(m, M_DONTWAIT);
2569                 if (mc == NULL) {
2570                         sc->sc_ifp->if_oerrors++;
2571                         continue;
2572                 }
2573
2574                 bridge_enqueue(sc, dst_if, mc);
2575         }
2576 }
2577
2578 /*
2579  * bridge_rtupdate:
2580  *
2581  *      Add a bridge routing entry.
2582  */
2583 static int
2584 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2585     struct bridge_iflist *bif, int setflags, uint8_t flags)
2586 {
2587         struct bridge_rtnode *brt;
2588         int error;
2589
2590         BRIDGE_LOCK_ASSERT(sc);
2591
2592         /* Check the source address is valid and not multicast. */
2593         if (ETHER_IS_MULTICAST(dst) ||
2594             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2595              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2596                 return (EINVAL);
2597
2598         /* 802.1p frames map to vlan 1 */
2599         if (vlan == 0)
2600                 vlan = 1;
2601
2602         /*
2603          * A route for this destination might already exist.  If so,
2604          * update it, otherwise create a new one.
2605          */
2606         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2607                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2608                         sc->sc_brtexceeded++;
2609                         return (ENOSPC);
2610                 }
2611                 /* Check per interface address limits (if enabled) */
2612                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2613                         bif->bif_addrexceeded++;
2614                         return (ENOSPC);
2615                 }
2616
2617                 /*
2618                  * Allocate a new bridge forwarding node, and
2619                  * initialize the expiration time and Ethernet
2620                  * address.
2621                  */
2622                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2623                 if (brt == NULL)
2624                         return (ENOMEM);
2625
2626                 if (bif->bif_flags & IFBIF_STICKY)
2627                         brt->brt_flags = IFBAF_STICKY;
2628                 else
2629                         brt->brt_flags = IFBAF_DYNAMIC;
2630
2631                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2632                 brt->brt_vlan = vlan;
2633
2634                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2635                         uma_zfree(bridge_rtnode_zone, brt);
2636                         return (error);
2637                 }
2638                 brt->brt_dst = bif;
2639                 bif->bif_addrcnt++;
2640         }
2641
2642         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2643             brt->brt_dst != bif) {
2644                 brt->brt_dst->bif_addrcnt--;
2645                 brt->brt_dst = bif;
2646                 brt->brt_dst->bif_addrcnt++;
2647         }
2648
2649         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2650                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2651         if (setflags)
2652                 brt->brt_flags = flags;
2653
2654         return (0);
2655 }
2656
2657 /*
2658  * bridge_rtlookup:
2659  *
2660  *      Lookup the destination interface for an address.
2661  */
2662 static struct ifnet *
2663 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2664 {
2665         struct bridge_rtnode *brt;
2666
2667         BRIDGE_LOCK_ASSERT(sc);
2668
2669         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2670                 return (NULL);
2671
2672         return (brt->brt_ifp);
2673 }
2674
2675 /*
2676  * bridge_rttrim:
2677  *
2678  *      Trim the routine table so that we have a number
2679  *      of routing entries less than or equal to the
2680  *      maximum number.
2681  */
2682 static void
2683 bridge_rttrim(struct bridge_softc *sc)
2684 {
2685         struct bridge_rtnode *brt, *nbrt;
2686
2687         BRIDGE_LOCK_ASSERT(sc);
2688
2689         /* Make sure we actually need to do this. */
2690         if (sc->sc_brtcnt <= sc->sc_brtmax)
2691                 return;
2692
2693         /* Force an aging cycle; this might trim enough addresses. */
2694         bridge_rtage(sc);
2695         if (sc->sc_brtcnt <= sc->sc_brtmax)
2696                 return;
2697
2698         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2699                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2700                         bridge_rtnode_destroy(sc, brt);
2701                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2702                                 return;
2703                 }
2704         }
2705 }
2706
2707 /*
2708  * bridge_timer:
2709  *
2710  *      Aging timer for the bridge.
2711  */
2712 static void
2713 bridge_timer(void *arg)
2714 {
2715         struct bridge_softc *sc = arg;
2716
2717         BRIDGE_LOCK_ASSERT(sc);
2718
2719         bridge_rtage(sc);
2720
2721         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2722                 callout_reset(&sc->sc_brcallout,
2723                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2724 }
2725
2726 /*
2727  * bridge_rtage:
2728  *
2729  *      Perform an aging cycle.
2730  */
2731 static void
2732 bridge_rtage(struct bridge_softc *sc)
2733 {
2734         struct bridge_rtnode *brt, *nbrt;
2735
2736         BRIDGE_LOCK_ASSERT(sc);
2737
2738         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2739                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2740                         if (time_uptime >= brt->brt_expire)
2741                                 bridge_rtnode_destroy(sc, brt);
2742                 }
2743         }
2744 }
2745
2746 /*
2747  * bridge_rtflush:
2748  *
2749  *      Remove all dynamic addresses from the bridge.
2750  */
2751 static void
2752 bridge_rtflush(struct bridge_softc *sc, int full)
2753 {
2754         struct bridge_rtnode *brt, *nbrt;
2755
2756         BRIDGE_LOCK_ASSERT(sc);
2757
2758         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2759                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2760                         bridge_rtnode_destroy(sc, brt);
2761         }
2762 }
2763
2764 /*
2765  * bridge_rtdaddr:
2766  *
2767  *      Remove an address from the table.
2768  */
2769 static int
2770 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2771 {
2772         struct bridge_rtnode *brt;
2773         int found = 0;
2774
2775         BRIDGE_LOCK_ASSERT(sc);
2776
2777         /*
2778          * If vlan is zero then we want to delete for all vlans so the lookup
2779          * may return more than one.
2780          */
2781         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2782                 bridge_rtnode_destroy(sc, brt);
2783                 found = 1;
2784         }
2785
2786         return (found ? 0 : ENOENT);
2787 }
2788
2789 /*
2790  * bridge_rtdelete:
2791  *
2792  *      Delete routes to a speicifc member interface.
2793  */
2794 static void
2795 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2796 {
2797         struct bridge_rtnode *brt, *nbrt;
2798
2799         BRIDGE_LOCK_ASSERT(sc);
2800
2801         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2802                 if (brt->brt_ifp == ifp && (full ||
2803                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2804                         bridge_rtnode_destroy(sc, brt);
2805         }
2806 }
2807
2808 /*
2809  * bridge_rtable_init:
2810  *
2811  *      Initialize the route table for this bridge.
2812  */
2813 static int
2814 bridge_rtable_init(struct bridge_softc *sc)
2815 {
2816         int i;
2817
2818         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2819             M_DEVBUF, M_NOWAIT);
2820         if (sc->sc_rthash == NULL)
2821                 return (ENOMEM);
2822
2823         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2824                 LIST_INIT(&sc->sc_rthash[i]);
2825
2826         sc->sc_rthash_key = arc4random();
2827
2828         LIST_INIT(&sc->sc_rtlist);
2829
2830         return (0);
2831 }
2832
2833 /*
2834  * bridge_rtable_fini:
2835  *
2836  *      Deconstruct the route table for this bridge.
2837  */
2838 static void
2839 bridge_rtable_fini(struct bridge_softc *sc)
2840 {
2841
2842         KASSERT(sc->sc_brtcnt == 0,
2843             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2844         free(sc->sc_rthash, M_DEVBUF);
2845 }
2846
2847 /*
2848  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2849  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2850  */
2851 #define mix(a, b, c)                                                    \
2852 do {                                                                    \
2853         a -= b; a -= c; a ^= (c >> 13);                                 \
2854         b -= c; b -= a; b ^= (a << 8);                                  \
2855         c -= a; c -= b; c ^= (b >> 13);                                 \
2856         a -= b; a -= c; a ^= (c >> 12);                                 \
2857         b -= c; b -= a; b ^= (a << 16);                                 \
2858         c -= a; c -= b; c ^= (b >> 5);                                  \
2859         a -= b; a -= c; a ^= (c >> 3);                                  \
2860         b -= c; b -= a; b ^= (a << 10);                                 \
2861         c -= a; c -= b; c ^= (b >> 15);                                 \
2862 } while (/*CONSTCOND*/0)
2863
2864 static __inline uint32_t
2865 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2866 {
2867         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2868
2869         b += addr[5] << 8;
2870         b += addr[4];
2871         a += addr[3] << 24;
2872         a += addr[2] << 16;
2873         a += addr[1] << 8;
2874         a += addr[0];
2875
2876         mix(a, b, c);
2877
2878         return (c & BRIDGE_RTHASH_MASK);
2879 }
2880
2881 #undef mix
2882
2883 static int
2884 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2885 {
2886         int i, d;
2887
2888         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2889                 d = ((int)a[i]) - ((int)b[i]);
2890         }
2891
2892         return (d);
2893 }
2894
2895 /*
2896  * bridge_rtnode_lookup:
2897  *
2898  *      Look up a bridge route node for the specified destination. Compare the
2899  *      vlan id or if zero then just return the first match.
2900  */
2901 static struct bridge_rtnode *
2902 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2903 {
2904         struct bridge_rtnode *brt;
2905         uint32_t hash;
2906         int dir;
2907
2908         BRIDGE_LOCK_ASSERT(sc);
2909
2910         hash = bridge_rthash(sc, addr);
2911         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2912                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2913                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2914                         return (brt);
2915                 if (dir > 0)
2916                         return (NULL);
2917         }
2918
2919         return (NULL);
2920 }
2921
2922 /*
2923  * bridge_rtnode_insert:
2924  *
2925  *      Insert the specified bridge node into the route table.  We
2926  *      assume the entry is not already in the table.
2927  */
2928 static int
2929 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2930 {
2931         struct bridge_rtnode *lbrt;
2932         uint32_t hash;
2933         int dir;
2934
2935         BRIDGE_LOCK_ASSERT(sc);
2936
2937         hash = bridge_rthash(sc, brt->brt_addr);
2938
2939         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2940         if (lbrt == NULL) {
2941                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2942                 goto out;
2943         }
2944
2945         do {
2946                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2947                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2948                         return (EEXIST);
2949                 if (dir > 0) {
2950                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2951                         goto out;
2952                 }
2953                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2954                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2955                         goto out;
2956                 }
2957                 lbrt = LIST_NEXT(lbrt, brt_hash);
2958         } while (lbrt != NULL);
2959
2960 #ifdef DIAGNOSTIC
2961         panic("bridge_rtnode_insert: impossible");
2962 #endif
2963
2964 out:
2965         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2966         sc->sc_brtcnt++;
2967
2968         return (0);
2969 }
2970
2971 /*
2972  * bridge_rtnode_destroy:
2973  *
2974  *      Destroy a bridge rtnode.
2975  */
2976 static void
2977 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2978 {
2979         BRIDGE_LOCK_ASSERT(sc);
2980
2981         LIST_REMOVE(brt, brt_hash);
2982
2983         LIST_REMOVE(brt, brt_list);
2984         sc->sc_brtcnt--;
2985         brt->brt_dst->bif_addrcnt--;
2986         uma_zfree(bridge_rtnode_zone, brt);
2987 }
2988
2989 /*
2990  * bridge_rtable_expire:
2991  *
2992  *      Set the expiry time for all routes on an interface.
2993  */
2994 static void
2995 bridge_rtable_expire(struct ifnet *ifp, int age)
2996 {
2997         struct bridge_softc *sc = ifp->if_bridge;
2998         struct bridge_rtnode *brt;
2999
3000         BRIDGE_LOCK(sc);
3001
3002         /*
3003          * If the age is zero then flush, otherwise set all the expiry times to
3004          * age for the interface
3005          */
3006         if (age == 0)
3007                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3008         else {
3009                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3010                         /* Cap the expiry time to 'age' */
3011                         if (brt->brt_ifp == ifp &&
3012                             brt->brt_expire > time_uptime + age &&
3013                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3014                                 brt->brt_expire = time_uptime + age;
3015                 }
3016         }
3017         BRIDGE_UNLOCK(sc);
3018 }
3019
3020 /*
3021  * bridge_state_change:
3022  *
3023  *      Callback from the bridgestp code when a port changes states.
3024  */
3025 static void
3026 bridge_state_change(struct ifnet *ifp, int state)
3027 {
3028         struct bridge_softc *sc = ifp->if_bridge;
3029         static const char *stpstates[] = {
3030                 "disabled",
3031                 "listening",
3032                 "learning",
3033                 "forwarding",
3034                 "blocking",
3035                 "discarding"
3036         };
3037
3038         if (log_stp)
3039                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3040                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3041 }
3042
3043 /*
3044  * Send bridge packets through pfil if they are one of the types pfil can deal
3045  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3046  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3047  * that interface.
3048  */
3049 static int
3050 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3051 {
3052         int snap, error, i, hlen;
3053         struct ether_header *eh1, eh2;
3054         struct ip_fw_args args;
3055         struct ip *ip;
3056         struct llc llc1;
3057         u_int16_t ether_type;
3058
3059         snap = 0;
3060         error = -1;     /* Default error if not error == 0 */
3061
3062 #if 0
3063         /* we may return with the IP fields swapped, ensure its not shared */
3064         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3065 #endif
3066
3067         if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
3068                 return (0); /* filtering is disabled */
3069
3070         i = min((*mp)->m_pkthdr.len, max_protohdr);
3071         if ((*mp)->m_len < i) {
3072             *mp = m_pullup(*mp, i);
3073             if (*mp == NULL) {
3074                 printf("%s: m_pullup failed\n", __func__);
3075                 return (-1);
3076             }
3077         }
3078
3079         eh1 = mtod(*mp, struct ether_header *);
3080         ether_type = ntohs(eh1->ether_type);
3081
3082         /*
3083          * Check for SNAP/LLC.
3084          */
3085         if (ether_type < ETHERMTU) {
3086                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3087
3088                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3089                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3090                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3091                     llc2->llc_control == LLC_UI) {
3092                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3093                         snap = 1;
3094                 }
3095         }
3096
3097         /*
3098          * If we're trying to filter bridge traffic, don't look at anything
3099          * other than IP and ARP traffic.  If the filter doesn't understand
3100          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3101          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3102          * but of course we don't have an AppleTalk filter to begin with.
3103          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3104          * ARP traffic.)
3105          */
3106         switch (ether_type) {
3107                 case ETHERTYPE_ARP:
3108                 case ETHERTYPE_REVARP:
3109                         if (pfil_ipfw_arp == 0)
3110                                 return (0); /* Automatically pass */
3111                         break;
3112
3113                 case ETHERTYPE_IP:
3114 #ifdef INET6
3115                 case ETHERTYPE_IPV6:
3116 #endif /* INET6 */
3117                         break;
3118                 default:
3119                         /*
3120                          * Check to see if the user wants to pass non-ip
3121                          * packets, these will not be checked by pfil(9) and
3122                          * passed unconditionally so the default is to drop.
3123                          */
3124                         if (pfil_onlyip)
3125                                 goto bad;
3126         }
3127
3128         /* Strip off the Ethernet header and keep a copy. */
3129         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3130         m_adj(*mp, ETHER_HDR_LEN);
3131
3132         /* Strip off snap header, if present */
3133         if (snap) {
3134                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3135                 m_adj(*mp, sizeof(struct llc));
3136         }
3137
3138         /*
3139          * Check the IP header for alignment and errors
3140          */
3141         if (dir == PFIL_IN) {
3142                 switch (ether_type) {
3143                         case ETHERTYPE_IP:
3144                                 error = bridge_ip_checkbasic(mp);
3145                                 break;
3146 #ifdef INET6
3147                         case ETHERTYPE_IPV6:
3148                                 error = bridge_ip6_checkbasic(mp);
3149                                 break;
3150 #endif /* INET6 */
3151                         default:
3152                                 error = 0;
3153                 }
3154                 if (error)
3155                         goto bad;
3156         }
3157
3158         /* XXX this section is also in if_ethersubr.c */
3159         // XXX PFIL_OUT or DIR_OUT ?
3160         if (V_ip_fw_chk_ptr && pfil_ipfw != 0 &&
3161                         dir == PFIL_OUT && ifp != NULL) {
3162                 struct m_tag *mtag;
3163
3164                 error = -1;
3165                 /* fetch the start point from existing tags, if any */
3166                 mtag = m_tag_locate(*mp, MTAG_IPFW_RULE, 0, NULL);
3167                 if (mtag == NULL) {
3168                         args.rule.slot = 0;
3169                 } else {
3170                         struct ipfw_rule_ref *r;
3171
3172                         /* XXX can we free the tag after use ? */
3173                         mtag->m_tag_id = PACKET_TAG_NONE;
3174                         r = (struct ipfw_rule_ref *)(mtag + 1);
3175                         /* packet already partially processed ? */
3176                         if (r->info & IPFW_ONEPASS)
3177                                 goto ipfwpass;
3178                         args.rule = *r;
3179                 }
3180
3181                 args.m = *mp;
3182                 args.oif = ifp;
3183                 args.next_hop = NULL;
3184                 args.next_hop6 = NULL;
3185                 args.eh = &eh2;
3186                 args.inp = NULL;        /* used by ipfw uid/gid/jail rules */
3187                 i = V_ip_fw_chk_ptr(&args);
3188                 *mp = args.m;
3189
3190                 if (*mp == NULL)
3191                         return (error);
3192
3193                 if (ip_dn_io_ptr && (i == IP_FW_DUMMYNET)) {
3194
3195                         /* put the Ethernet header back on */
3196                         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3197                         if (*mp == NULL)
3198                                 return (error);
3199                         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3200
3201                         /*
3202                          * Pass the pkt to dummynet, which consumes it. The
3203                          * packet will return to us via bridge_dummynet().
3204                          */
3205                         args.oif = ifp;
3206                         ip_dn_io_ptr(mp, DIR_FWD | PROTO_IFB, &args);
3207                         return (error);
3208                 }
3209
3210                 if (i != IP_FW_PASS) /* drop */
3211                         goto bad;
3212         }
3213
3214 ipfwpass:
3215         error = 0;
3216
3217         /*
3218          * Run the packet through pfil
3219          */
3220         switch (ether_type) {
3221         case ETHERTYPE_IP:
3222                 /*
3223                  * before calling the firewall, swap fields the same as
3224                  * IP does. here we assume the header is contiguous
3225                  */
3226                 ip = mtod(*mp, struct ip *);
3227
3228                 ip->ip_len = ntohs(ip->ip_len);
3229                 ip->ip_off = ntohs(ip->ip_off);
3230
3231                 /*
3232                  * Run pfil on the member interface and the bridge, both can
3233                  * be skipped by clearing pfil_member or pfil_bridge.
3234                  *
3235                  * Keep the order:
3236                  *   in_if -> bridge_if -> out_if
3237                  */
3238                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3239                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3240                                         dir, NULL);
3241
3242                 if (*mp == NULL || error != 0) /* filter may consume */
3243                         break;
3244
3245                 if (pfil_member && ifp != NULL)
3246                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3247                                         dir, NULL);
3248
3249                 if (*mp == NULL || error != 0) /* filter may consume */
3250                         break;
3251
3252                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3253                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3254                                         dir, NULL);
3255
3256                 if (*mp == NULL || error != 0) /* filter may consume */
3257                         break;
3258
3259                 /* check if we need to fragment the packet */
3260                 if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3261                         i = (*mp)->m_pkthdr.len;
3262                         if (i > ifp->if_mtu) {
3263                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3264                                             &llc1);
3265                                 return (error);
3266                         }
3267                 }
3268
3269                 /* Recalculate the ip checksum and restore byte ordering */
3270                 ip = mtod(*mp, struct ip *);
3271                 hlen = ip->ip_hl << 2;
3272                 if (hlen < sizeof(struct ip))
3273                         goto bad;
3274                 if (hlen > (*mp)->m_len) {
3275                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3276                                 goto bad;
3277                         ip = mtod(*mp, struct ip *);
3278                         if (ip == NULL)
3279                                 goto bad;
3280                 }
3281                 ip->ip_len = htons(ip->ip_len);
3282                 ip->ip_off = htons(ip->ip_off);
3283                 ip->ip_sum = 0;
3284                 if (hlen == sizeof(struct ip))
3285                         ip->ip_sum = in_cksum_hdr(ip);
3286                 else
3287                         ip->ip_sum = in_cksum(*mp, hlen);
3288
3289                 break;
3290 #ifdef INET6
3291         case ETHERTYPE_IPV6:
3292                 if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3293                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3294                                         dir, NULL);
3295
3296                 if (*mp == NULL || error != 0) /* filter may consume */
3297                         break;
3298
3299                 if (pfil_member && ifp != NULL)
3300                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3301                                         dir, NULL);
3302
3303                 if (*mp == NULL || error != 0) /* filter may consume */
3304                         break;
3305
3306                 if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3307                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3308                                         dir, NULL);
3309                 break;
3310 #endif
3311         default:
3312                 error = 0;
3313                 break;
3314         }
3315
3316         if (*mp == NULL)
3317                 return (error);
3318         if (error != 0)
3319                 goto bad;
3320
3321         error = -1;
3322
3323         /*
3324          * Finally, put everything back the way it was and return
3325          */
3326         if (snap) {
3327                 M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3328                 if (*mp == NULL)
3329                         return (error);
3330                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3331         }
3332
3333         M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3334         if (*mp == NULL)
3335                 return (error);
3336         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3337
3338         return (0);
3339
3340 bad:
3341         m_freem(*mp);
3342         *mp = NULL;
3343         return (error);
3344 }
3345
3346 /*
3347  * Perform basic checks on header size since
3348  * pfil assumes ip_input has already processed
3349  * it for it.  Cut-and-pasted from ip_input.c.
3350  * Given how simple the IPv6 version is,
3351  * does the IPv4 version really need to be
3352  * this complicated?
3353  *
3354  * XXX Should we update ipstat here, or not?
3355  * XXX Right now we update ipstat but not
3356  * XXX csum_counter.
3357  */
3358 static int
3359 bridge_ip_checkbasic(struct mbuf **mp)
3360 {
3361         struct mbuf *m = *mp;
3362         struct ip *ip;
3363         int len, hlen;
3364         u_short sum;
3365
3366         if (*mp == NULL)
3367                 return (-1);
3368
3369         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3370                 if ((m = m_copyup(m, sizeof(struct ip),
3371                         (max_linkhdr + 3) & ~3)) == NULL) {
3372                         /* XXXJRT new stat, please */
3373                         KMOD_IPSTAT_INC(ips_toosmall);
3374                         goto bad;
3375                 }
3376         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3377                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3378                         KMOD_IPSTAT_INC(ips_toosmall);
3379                         goto bad;
3380                 }
3381         }
3382         ip = mtod(m, struct ip *);
3383         if (ip == NULL) goto bad;
3384
3385         if (ip->ip_v != IPVERSION) {
3386                 KMOD_IPSTAT_INC(ips_badvers);
3387                 goto bad;
3388         }
3389         hlen = ip->ip_hl << 2;
3390         if (hlen < sizeof(struct ip)) { /* minimum header length */
3391                 KMOD_IPSTAT_INC(ips_badhlen);
3392                 goto bad;
3393         }
3394         if (hlen > m->m_len) {
3395                 if ((m = m_pullup(m, hlen)) == 0) {
3396                         KMOD_IPSTAT_INC(ips_badhlen);
3397                         goto bad;
3398                 }
3399                 ip = mtod(m, struct ip *);
3400                 if (ip == NULL) goto bad;
3401         }
3402
3403         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3404                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3405         } else {
3406                 if (hlen == sizeof(struct ip)) {
3407                         sum = in_cksum_hdr(ip);
3408                 } else {
3409                         sum = in_cksum(m, hlen);
3410                 }
3411         }
3412         if (sum) {
3413                 KMOD_IPSTAT_INC(ips_badsum);
3414                 goto bad;
3415         }
3416
3417         /* Retrieve the packet length. */
3418         len = ntohs(ip->ip_len);
3419
3420         /*
3421          * Check for additional length bogosity
3422          */
3423         if (len < hlen) {
3424                 KMOD_IPSTAT_INC(ips_badlen);
3425                 goto bad;
3426         }
3427
3428         /*
3429          * Check that the amount of data in the buffers
3430          * is as at least much as the IP header would have us expect.
3431          * Drop packet if shorter than we expect.
3432          */
3433         if (m->m_pkthdr.len < len) {
3434                 KMOD_IPSTAT_INC(ips_tooshort);
3435                 goto bad;
3436         }
3437
3438         /* Checks out, proceed */
3439         *mp = m;
3440         return (0);
3441
3442 bad:
3443         *mp = m;
3444         return (-1);
3445 }
3446
3447 #ifdef INET6
3448 /*
3449  * Same as above, but for IPv6.
3450  * Cut-and-pasted from ip6_input.c.
3451  * XXX Should we update ip6stat, or not?
3452  */
3453 static int
3454 bridge_ip6_checkbasic(struct mbuf **mp)
3455 {
3456         struct mbuf *m = *mp;
3457         struct ip6_hdr *ip6;
3458
3459         /*
3460          * If the IPv6 header is not aligned, slurp it up into a new
3461          * mbuf with space for link headers, in the event we forward
3462          * it.  Otherwise, if it is aligned, make sure the entire base
3463          * IPv6 header is in the first mbuf of the chain.
3464          */
3465         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3466                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3467                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3468                             (max_linkhdr + 3) & ~3)) == NULL) {
3469                         /* XXXJRT new stat, please */
3470                         IP6STAT_INC(ip6s_toosmall);
3471                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3472                         goto bad;
3473                 }
3474         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3475                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3476                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3477                         IP6STAT_INC(ip6s_toosmall);
3478                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3479                         goto bad;
3480                 }
3481         }
3482
3483         ip6 = mtod(m, struct ip6_hdr *);
3484
3485         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3486                 IP6STAT_INC(ip6s_badvers);
3487                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3488                 goto bad;
3489         }
3490
3491         /* Checks out, proceed */
3492         *mp = m;
3493         return (0);
3494
3495 bad:
3496         *mp = m;
3497         return (-1);
3498 }
3499 #endif /* INET6 */
3500
3501 /*
3502  * bridge_fragment:
3503  *
3504  *      Return a fragmented mbuf chain.
3505  */
3506 static int
3507 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3508     int snap, struct llc *llc)
3509 {
3510         struct mbuf *m0;
3511         struct ip *ip;
3512         int error = -1;
3513
3514         if (m->m_len < sizeof(struct ip) &&
3515             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3516                 goto out;
3517         ip = mtod(m, struct ip *);
3518
3519         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3520                     CSUM_DELAY_IP);
3521         if (error)
3522                 goto out;
3523
3524         /* walk the chain and re-add the Ethernet header */
3525         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3526                 if (error == 0) {
3527                         if (snap) {
3528                                 M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3529                                 if (m0 == NULL) {
3530                                         error = ENOBUFS;
3531                                         continue;
3532                                 }
3533                                 bcopy(llc, mtod(m0, caddr_t),
3534                                     sizeof(struct llc));
3535                         }
3536                         M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3537                         if (m0 == NULL) {
3538                                 error = ENOBUFS;
3539                                 continue;
3540                         }
3541                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3542                 } else
3543                         m_freem(m);
3544         }
3545
3546         if (error == 0)
3547                 KMOD_IPSTAT_INC(ips_fragmented);
3548
3549         return (error);
3550
3551 out:
3552         if (m != NULL)
3553                 m_freem(m);
3554         return (error);
3555 }
3556
3557 static void
3558 bridge_linkstate(struct ifnet *ifp)
3559 {
3560         struct bridge_softc *sc = ifp->if_bridge;
3561         struct bridge_iflist *bif;
3562
3563         BRIDGE_LOCK(sc);
3564         bif = bridge_lookup_member_if(sc, ifp);
3565         if (bif == NULL) {
3566                 BRIDGE_UNLOCK(sc);
3567                 return;
3568         }
3569         bridge_linkcheck(sc);
3570         BRIDGE_UNLOCK(sc);
3571
3572         bstp_linkstate(&bif->bif_stp);
3573 }
3574
3575 static void
3576 bridge_linkcheck(struct bridge_softc *sc)
3577 {
3578         struct bridge_iflist *bif;
3579         int new_link, hasls;
3580
3581         BRIDGE_LOCK_ASSERT(sc);
3582         new_link = LINK_STATE_DOWN;
3583         hasls = 0;
3584         /* Our link is considered up if at least one of our ports is active */
3585         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3586                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3587                         hasls++;
3588                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3589                         new_link = LINK_STATE_UP;
3590                         break;
3591                 }
3592         }
3593         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3594                 /* If no interfaces support link-state then we default to up */
3595                 new_link = LINK_STATE_UP;
3596         }
3597         if_link_state_change(sc->sc_ifp, new_link);
3598 }