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 heterogeneous 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/eventhandler.h>
  85 #include <sys/mbuf.h>
  86 #include <sys/malloc.h>
  87 #include <sys/protosw.h>
  88 #include <sys/systm.h>
  89 #include <sys/jail.h>
  90 #include <sys/time.h>
  91 #include <sys/socket.h> /* for net/if.h */
  92 #include <sys/sockio.h>
  93 #include <sys/ctype.h>  /* string functions */
  94 #include <sys/kernel.h>
  95 #include <sys/random.h>
  96 #include <sys/syslog.h>
  97 #include <sys/sysctl.h>
  98 #include <vm/uma.h>
  99 #include <sys/module.h>
 100 #include <sys/priv.h>
 101 #include <sys/proc.h>
 102 #include <sys/lock.h>
 103 #include <sys/mutex.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>
 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>
 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
 136 /*
 137  * Size of the route hash table.  Must be a power of two.
 138  */
 139 #ifndef BRIDGE_RTHASH_SIZE
 140 #define BRIDGE_RTHASH_SIZE              1024
 141 #endif
 142
 143 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 144
 145 /*
 146  * Default maximum number of addresses to cache.
 147  */
 148 #ifndef BRIDGE_RTABLE_MAX
 149 #define BRIDGE_RTABLE_MAX               2000
 150 #endif
 151
 152 /*
 153  * Timeout (in seconds) for entries learned dynamically.
 154  */
 155 #ifndef BRIDGE_RTABLE_TIMEOUT
 156 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 157 #endif
 158
 159 /*
 160  * Number of seconds between walks of the route list.
 161  */
 162 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 163 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 164 #endif
 165
 166 /*
 167  * List of capabilities to possibly mask on the member interface.
 168  */
 169 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 170                                          IFCAP_TXCSUM_IPV6)
 171
 172 /*
 173  * List of capabilities to strip
 174  */
 175 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 176
 177 /*
 178  * Bridge interface list entry.
 179  */
 180 struct bridge_iflist {
 181         LIST_ENTRY(bridge_iflist) bif_next;
 182         struct ifnet            *bif_ifp;       /* member if */
 183         struct bstp_port        bif_stp;        /* STP state */
 184         uint32_t                bif_flags;      /* member if flags */
 185         int                     bif_savedcaps;  /* saved capabilities */
 186         uint32_t                bif_addrmax;    /* max # of addresses */
 187         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 188         uint32_t                bif_addrexceeded;/* # of address violations */
 189 };
 190
 191 /*
 192  * Bridge route node.
 193  */
 194 struct bridge_rtnode {
 195         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 196         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 197         struct bridge_iflist    *brt_dst;       /* destination if */
 198         unsigned long           brt_expire;     /* expiration time */
 199         uint8_t                 brt_flags;      /* address flags */
 200         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 201         uint16_t                brt_vlan;       /* vlan id */
 202 };
 203 #define brt_ifp                 brt_dst->bif_ifp
 204
 205 /*
 206  * Software state for each bridge.
 207  */
 208 struct bridge_softc {
 209         struct ifnet            *sc_ifp;        /* make this an interface */
 210         LIST_ENTRY(bridge_softc) sc_list;
 211         struct mtx              sc_mtx;
 212         struct cv               sc_cv;
 213         uint32_t                sc_brtmax;      /* max # of addresses */
 214         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 215         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 216         struct callout          sc_brcallout;   /* bridge callout */
 217         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 218         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 219         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 220         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 221         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 222         uint32_t                sc_rthash_key;  /* key for hash */
 223         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 224         struct bstp_state       sc_stp;         /* STP state */
 225         uint32_t                sc_brtexceeded; /* # of cache drops */
 226         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 227         u_char                  sc_defaddr[6];  /* Default MAC address */
 228 };
 229
 230 static VNET_DEFINE(struct mtx, bridge_list_mtx);
 231 #define V_bridge_list_mtx       VNET(bridge_list_mtx)
 232 static eventhandler_tag bridge_detach_cookie;
 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 int      bridge_transmit(struct ifnet *, struct mbuf *);
 250 static void     bridge_qflush(struct ifnet *);
 251 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 252 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 253                     struct rtentry *);
 254 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 255                     struct mbuf *);
 256 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 257
 258 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 259                     struct mbuf *m);
 260
 261 static void     bridge_timer(void *);
 262
 263 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 264                     struct mbuf *, int);
 265 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 266
 267 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 268                     uint16_t, struct bridge_iflist *, int, uint8_t);
 269 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 270                     uint16_t);
 271 static void     bridge_rttrim(struct bridge_softc *);
 272 static void     bridge_rtage(struct bridge_softc *);
 273 static void     bridge_rtflush(struct bridge_softc *, int);
 274 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 275                     uint16_t);
 276
 277 static void     bridge_rtable_init(struct bridge_softc *);
 278 static void     bridge_rtable_fini(struct bridge_softc *);
 279
 280 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 281 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 282                     const uint8_t *, uint16_t);
 283 static int      bridge_rtnode_insert(struct bridge_softc *,
 284                     struct bridge_rtnode *);
 285 static void     bridge_rtnode_destroy(struct bridge_softc *,
 286                     struct bridge_rtnode *);
 287 static void     bridge_rtable_expire(struct ifnet *, int);
 288 static void     bridge_state_change(struct ifnet *, int);
 289
 290 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 291                     const char *name);
 292 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 293                     struct ifnet *ifp);
 294 static void     bridge_delete_member(struct bridge_softc *,
 295                     struct bridge_iflist *, int);
 296 static void     bridge_delete_span(struct bridge_softc *,
 297                     struct bridge_iflist *);
 298
 299 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 328 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 329 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 330 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 331                     int);
 332 static int      bridge_ip_checkbasic(struct mbuf **mp);
 333 #ifdef INET6
 334 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 335 #endif /* INET6 */
 336 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 337                     struct ether_header *, int, struct llc *);
 338 static void     bridge_linkstate(struct ifnet *ifp);
 339 static void     bridge_linkcheck(struct bridge_softc *sc);
 340
 341 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
 342
 343 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 344 #define VLANTAGOF(_m)   \
 345     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 346
 347 static struct bstp_cb_ops bridge_ops = {
 348         .bcb_state = bridge_state_change,
 349         .bcb_rtage = bridge_rtable_expire
 350 };
 351
 352 SYSCTL_DECL(_net_link);
 353 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 354
 355 /* only pass IP[46] packets when pfil is enabled */
 356 static VNET_DEFINE(int, pfil_onlyip) = 1;
 357 #define V_pfil_onlyip   VNET(pfil_onlyip)
 358 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 359     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 360     "Only pass IP packets when pfil is enabled");
 361
 362 /* run pfil hooks on the bridge interface */
 363 static VNET_DEFINE(int, pfil_bridge) = 1;
 364 #define V_pfil_bridge   VNET(pfil_bridge)
 365 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 366     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 367     "Packet filter on the bridge interface");
 368
 369 /* layer2 filter with ipfw */
 370 static VNET_DEFINE(int, pfil_ipfw);
 371 #define V_pfil_ipfw     VNET(pfil_ipfw)
 372
 373 /* layer2 ARP filter with ipfw */
 374 static VNET_DEFINE(int, pfil_ipfw_arp);
 375 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 376 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 377     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 378     "Filter ARP packets through IPFW layer2");
 379
 380 /* run pfil hooks on the member interface */
 381 static VNET_DEFINE(int, pfil_member) = 1;
 382 #define V_pfil_member   VNET(pfil_member)
 383 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 384     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 385     "Packet filter on the member interface");
 386
 387 /* run pfil hooks on the physical interface for locally destined packets */
 388 static VNET_DEFINE(int, pfil_local_phys);
 389 #define V_pfil_local_phys       VNET(pfil_local_phys)
 390 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 391     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 392     "Packet filter on the physical interface for locally destined packets");
 393
 394 /* log STP state changes */
 395 static VNET_DEFINE(int, log_stp);
 396 #define V_log_stp       VNET(log_stp)
 397 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 398     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 399     "Log STP state changes");
 400
 401 /* share MAC with first bridge member */
 402 static VNET_DEFINE(int, bridge_inherit_mac);
 403 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 404 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 405     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 406     "Inherit MAC address from the first bridge member");
 407
 408 static VNET_DEFINE(int, allow_llz_overlap) = 0;
 409 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 410 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 411     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 412     "Allow overlap of link-local scope "
 413     "zones of a bridge interface and the member interfaces");
 414
 415 struct bridge_control {
 416         int     (*bc_func)(struct bridge_softc *, void *);
 417         int     bc_argsize;
 418         int     bc_flags;
 419 };
 420
 421 #define BC_F_COPYIN             0x01    /* copy arguments in */
 422 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 423 #define BC_F_SUSER              0x04    /* do super-user check */
 424
 425 const struct bridge_control bridge_control_table[] = {
 426         { bridge_ioctl_add,             sizeof(struct ifbreq),
 427           BC_F_COPYIN|BC_F_SUSER },
 428         { bridge_ioctl_del,             sizeof(struct ifbreq),
 429           BC_F_COPYIN|BC_F_SUSER },
 430
 431         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 432           BC_F_COPYIN|BC_F_COPYOUT },
 433         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 434           BC_F_COPYIN|BC_F_SUSER },
 435
 436         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 437           BC_F_COPYIN|BC_F_SUSER },
 438         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 439           BC_F_COPYOUT },
 440
 441         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 442           BC_F_COPYIN|BC_F_COPYOUT },
 443         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 444           BC_F_COPYIN|BC_F_COPYOUT },
 445
 446         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 447           BC_F_COPYIN|BC_F_SUSER },
 448
 449         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 450           BC_F_COPYIN|BC_F_SUSER },
 451         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 452           BC_F_COPYOUT },
 453
 454         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 455           BC_F_COPYIN|BC_F_SUSER },
 456
 457         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 458           BC_F_COPYIN|BC_F_SUSER },
 459
 460         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 461           BC_F_COPYOUT },
 462         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 463           BC_F_COPYIN|BC_F_SUSER },
 464
 465         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 466           BC_F_COPYOUT },
 467         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 468           BC_F_COPYIN|BC_F_SUSER },
 469
 470         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 471           BC_F_COPYOUT },
 472         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 473           BC_F_COPYIN|BC_F_SUSER },
 474
 475         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 476           BC_F_COPYOUT },
 477         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 478           BC_F_COPYIN|BC_F_SUSER },
 479
 480         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 481           BC_F_COPYIN|BC_F_SUSER },
 482
 483         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 484           BC_F_COPYIN|BC_F_SUSER },
 485
 486         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 487           BC_F_COPYIN|BC_F_SUSER },
 488         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 489           BC_F_COPYIN|BC_F_SUSER },
 490
 491         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 492           BC_F_COPYOUT },
 493
 494         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 495           BC_F_COPYOUT },
 496
 497         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 498           BC_F_COPYIN|BC_F_COPYOUT },
 499
 500         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 501           BC_F_COPYIN|BC_F_SUSER },
 502
 503         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 504           BC_F_COPYIN|BC_F_SUSER },
 505
 506         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 507           BC_F_COPYIN|BC_F_SUSER },
 508
 509 };
 510 const int bridge_control_table_size = nitems(bridge_control_table);
 511
 512 static VNET_DEFINE(LIST_HEAD(, bridge_softc), bridge_list);
 513 #define V_bridge_list   VNET(bridge_list)
 514 #define BRIDGE_LIST_LOCK_INIT(x)        mtx_init(&V_bridge_list_mtx,    \
 515                                             "if_bridge list", NULL, MTX_DEF)
 516 #define BRIDGE_LIST_LOCK_DESTROY(x)     mtx_destroy(&V_bridge_list_mtx)
 517 #define BRIDGE_LIST_LOCK(x)             mtx_lock(&V_bridge_list_mtx)
 518 #define BRIDGE_LIST_UNLOCK(x)           mtx_unlock(&V_bridge_list_mtx)
 519
 520 static VNET_DEFINE(struct if_clone *, bridge_cloner);
 521 #define V_bridge_cloner VNET(bridge_cloner)
 522
 523 static const char bridge_name[] = "bridge";
 524
 525 static void
 526 vnet_bridge_init(const void *unused __unused)
 527 {
 528
 529         BRIDGE_LIST_LOCK_INIT();
 530         LIST_INIT(&V_bridge_list);
 531         V_bridge_cloner = if_clone_simple(bridge_name,
 532             bridge_clone_create, bridge_clone_destroy, 0);
 533 }
 534 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 535     vnet_bridge_init, NULL);
 536
 537 static void
 538 vnet_bridge_uninit(const void *unused __unused)
 539 {
 540
 541         if_clone_detach(V_bridge_cloner);
 542         V_bridge_cloner = NULL;
 543         BRIDGE_LIST_LOCK_DESTROY();
 544 }
 545 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 546     vnet_bridge_uninit, NULL);
 547
 548 static int
 549 bridge_modevent(module_t mod, int type, void *data)
 550 {
 551
 552         switch (type) {
 553         case MOD_LOAD:
 554                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 555                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 556                     UMA_ALIGN_PTR, 0);
 557                 bridge_input_p = bridge_input;
 558                 bridge_output_p = bridge_output;
 559                 bridge_dn_p = bridge_dummynet;
 560                 bridge_linkstate_p = bridge_linkstate;
 561                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 562                     ifnet_departure_event, bridge_ifdetach, NULL,
 563                     EVENTHANDLER_PRI_ANY);
 564                 break;
 565         case MOD_UNLOAD:
 566                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 567                     bridge_detach_cookie);
 568                 uma_zdestroy(bridge_rtnode_zone);
 569                 bridge_input_p = NULL;
 570                 bridge_output_p = NULL;
 571                 bridge_dn_p = NULL;
 572                 bridge_linkstate_p = NULL;
 573                 break;
 574         default:
 575                 return (EOPNOTSUPP);
 576         }
 577         return (0);
 578 }
 579
 580 static moduledata_t bridge_mod = {
 581         "if_bridge",
 582         bridge_modevent,
 583         0
 584 };
 585
 586 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 587 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 588
 589 /*
 590  * handler for net.link.bridge.ipfw
 591  */
 592 static int
 593 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 594 {
 595         int enable = V_pfil_ipfw;
 596         int error;
 597
 598         error = sysctl_handle_int(oidp, &enable, 0, req);
 599         enable &= 1;
 600
 601         if (enable != V_pfil_ipfw) {
 602                 V_pfil_ipfw = enable;
 603
 604                 /*
 605                  * Disable pfil so that ipfw doesnt run twice, if the user
 606                  * really wants both then they can re-enable pfil_bridge and/or
 607                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 608                  * layer2 type.
 609                  */
 610                 if (V_pfil_ipfw) {
 611                         V_pfil_onlyip = 0;
 612                         V_pfil_bridge = 0;
 613                         V_pfil_member = 0;
 614                 }
 615         }
 616
 617         return (error);
 618 }
 619 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 620     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 621     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 622     "Layer2 filter with IPFW");
 623
 624 /*
 625  * bridge_clone_create:
 626  *
 627  *      Create a new bridge instance.
 628  */
 629 static int
 630 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 631 {
 632         struct bridge_softc *sc, *sc2;
 633         struct ifnet *bifp, *ifp;
 634         int fb, retry;
 635         unsigned long hostid;
 636
 637         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 638         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 639         if (ifp == NULL) {
 640                 free(sc, M_DEVBUF);
 641                 return (ENOSPC);
 642         }
 643
 644         BRIDGE_LOCK_INIT(sc);
 645         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 646         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 647
 648         /* Initialize our routing table. */
 649         bridge_rtable_init(sc);
 650
 651         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 652
 653         LIST_INIT(&sc->sc_iflist);
 654         LIST_INIT(&sc->sc_spanlist);
 655
 656         ifp->if_softc = sc;
 657         if_initname(ifp, bridge_name, unit);
 658         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 659         ifp->if_ioctl = bridge_ioctl;
 660         ifp->if_transmit = bridge_transmit;
 661         ifp->if_qflush = bridge_qflush;
 662         ifp->if_init = bridge_init;
 663         ifp->if_type = IFT_BRIDGE;
 664
 665         /*
 666          * Generate an ethernet address with a locally administered address.
 667          *
 668          * Since we are using random ethernet addresses for the bridge, it is
 669          * possible that we might have address collisions, so make sure that
 670          * this hardware address isn't already in use on another bridge.
 671          * The first try uses the hostid and falls back to arc4rand().
 672          */
 673         fb = 0;
 674         getcredhostid(curthread->td_ucred, &hostid);
 675         do {
 676                 if (fb || hostid == 0) {
 677                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 678                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 679                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 680                 } else {
 681                         sc->sc_defaddr[0] = 0x2;
 682                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 683                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 684                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 685                         sc->sc_defaddr[4] =  hostid        & 0xff;
 686                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 687                 }
 688
 689                 fb = 1;
 690                 retry = 0;
 691                 BRIDGE_LIST_LOCK();
 692                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 693                         bifp = sc2->sc_ifp;
 694                         if (memcmp(sc->sc_defaddr,
 695                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 696                                 retry = 1;
 697                                 break;
 698                         }
 699                 }
 700                 BRIDGE_LIST_UNLOCK();
 701         } while (retry == 1);
 702
 703         bstp_attach(&sc->sc_stp, &bridge_ops);
 704         ether_ifattach(ifp, sc->sc_defaddr);
 705         /* Now undo some of the damage... */
 706         ifp->if_baudrate = 0;
 707         ifp->if_type = IFT_BRIDGE;
 708
 709         BRIDGE_LIST_LOCK();
 710         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 711         BRIDGE_LIST_UNLOCK();
 712
 713         return (0);
 714 }
 715
 716 /*
 717  * bridge_clone_destroy:
 718  *
 719  *      Destroy a bridge instance.
 720  */
 721 static void
 722 bridge_clone_destroy(struct ifnet *ifp)
 723 {
 724         struct bridge_softc *sc = ifp->if_softc;
 725         struct bridge_iflist *bif;
 726
 727         BRIDGE_LOCK(sc);
 728
 729         bridge_stop(ifp, 1);
 730         ifp->if_flags &= ~IFF_UP;
 731
 732         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 733                 bridge_delete_member(sc, bif, 0);
 734
 735         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 736                 bridge_delete_span(sc, bif);
 737         }
 738
 739         BRIDGE_UNLOCK(sc);
 740
 741         callout_drain(&sc->sc_brcallout);
 742
 743         BRIDGE_LIST_LOCK();
 744         LIST_REMOVE(sc, sc_list);
 745         BRIDGE_LIST_UNLOCK();
 746
 747         bstp_detach(&sc->sc_stp);
 748         ether_ifdetach(ifp);
 749         if_free(ifp);
 750
 751         /* Tear down the routing table. */
 752         bridge_rtable_fini(sc);
 753
 754         BRIDGE_LOCK_DESTROY(sc);
 755         free(sc, M_DEVBUF);
 756 }
 757
 758 /*
 759  * bridge_ioctl:
 760  *
 761  *      Handle a control request from the operator.
 762  */
 763 static int
 764 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 765 {
 766         struct bridge_softc *sc = ifp->if_softc;
 767         struct ifreq *ifr = (struct ifreq *)data;
 768         struct bridge_iflist *bif;
 769         struct thread *td = curthread;
 770         union {
 771                 struct ifbreq ifbreq;
 772                 struct ifbifconf ifbifconf;
 773                 struct ifbareq ifbareq;
 774                 struct ifbaconf ifbaconf;
 775                 struct ifbrparam ifbrparam;
 776                 struct ifbropreq ifbropreq;
 777         } args;
 778         struct ifdrv *ifd = (struct ifdrv *) data;
 779         const struct bridge_control *bc;
 780         int error = 0;
 781
 782         switch (cmd) {
 783
 784         case SIOCADDMULTI:
 785         case SIOCDELMULTI:
 786                 break;
 787
 788         case SIOCGDRVSPEC:
 789         case SIOCSDRVSPEC:
 790                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 791                         error = EINVAL;
 792                         break;
 793                 }
 794                 bc = &bridge_control_table[ifd->ifd_cmd];
 795
 796                 if (cmd == SIOCGDRVSPEC &&
 797                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 798                         error = EINVAL;
 799                         break;
 800                 }
 801                 else if (cmd == SIOCSDRVSPEC &&
 802                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 803                         error = EINVAL;
 804                         break;
 805                 }
 806
 807                 if (bc->bc_flags & BC_F_SUSER) {
 808                         error = priv_check(td, PRIV_NET_BRIDGE);
 809                         if (error)
 810                                 break;
 811                 }
 812
 813                 if (ifd->ifd_len != bc->bc_argsize ||
 814                     ifd->ifd_len > sizeof(args)) {
 815                         error = EINVAL;
 816                         break;
 817                 }
 818
 819                 bzero(&args, sizeof(args));
 820                 if (bc->bc_flags & BC_F_COPYIN) {
 821                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 822                         if (error)
 823                                 break;
 824                 }
 825
 826                 BRIDGE_LOCK(sc);
 827                 error = (*bc->bc_func)(sc, &args);
 828                 BRIDGE_UNLOCK(sc);
 829                 if (error)
 830                         break;
 831
 832                 if (bc->bc_flags & BC_F_COPYOUT)
 833                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 834
 835                 break;
 836
 837         case SIOCSIFFLAGS:
 838                 if (!(ifp->if_flags & IFF_UP) &&
 839                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 840                         /*
 841                          * If interface is marked down and it is running,
 842                          * then stop and disable it.
 843                          */
 844                         BRIDGE_LOCK(sc);
 845                         bridge_stop(ifp, 1);
 846                         BRIDGE_UNLOCK(sc);
 847                 } else if ((ifp->if_flags & IFF_UP) &&
 848                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 849                         /*
 850                          * If interface is marked up and it is stopped, then
 851                          * start it.
 852                          */
 853                         (*ifp->if_init)(sc);
 854                 }
 855                 break;
 856
 857         case SIOCSIFMTU:
 858                 if (ifr->ifr_mtu < 576) {
 859                         error = EINVAL;
 860                         break;
 861                 }
 862                 if (LIST_EMPTY(&sc->sc_iflist)) {
 863                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 864                         break;
 865                 }
 866                 BRIDGE_LOCK(sc);
 867                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 868                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 869                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 870                                     " != %d\n", sc->sc_ifp->if_xname,
 871                                     bif->bif_ifp->if_mtu,
 872                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 873                                 error = EINVAL;
 874                                 break;
 875                         }
 876                 }
 877                 if (!error)
 878                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 879                 BRIDGE_UNLOCK(sc);
 880                 break;
 881         default:
 882                 /*
 883                  * drop the lock as ether_ioctl() will call bridge_start() and
 884                  * cause the lock to be recursed.
 885                  */
 886                 error = ether_ioctl(ifp, cmd, data);
 887                 break;
 888         }
 889
 890         return (error);
 891 }
 892
 893 /*
 894  * bridge_mutecaps:
 895  *
 896  *      Clear or restore unwanted capabilities on the member interface
 897  */
 898 static void
 899 bridge_mutecaps(struct bridge_softc *sc)
 900 {
 901         struct bridge_iflist *bif;
 902         int enabled, mask;
 903
 904         /* Initial bitmask of capabilities to test */
 905         mask = BRIDGE_IFCAPS_MASK;
 906
 907         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 908                 /* Every member must support it or its disabled */
 909                 mask &= bif->bif_savedcaps;
 910         }
 911
 912         BRIDGE_XLOCK(sc);
 913         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 914                 enabled = bif->bif_ifp->if_capenable;
 915                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 916                 /* strip off mask bits and enable them again if allowed */
 917                 enabled &= ~BRIDGE_IFCAPS_MASK;
 918                 enabled |= mask;
 919                 BRIDGE_UNLOCK(sc);
 920                 bridge_set_ifcap(sc, bif, enabled);
 921                 BRIDGE_LOCK(sc);
 922         }
 923         BRIDGE_XDROP(sc);
 924
 925 }
 926
 927 static void
 928 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 929 {
 930         struct ifnet *ifp = bif->bif_ifp;
 931         struct ifreq ifr;
 932         int error;
 933
 934         BRIDGE_UNLOCK_ASSERT(sc);
 935
 936         bzero(&ifr, sizeof(ifr));
 937         ifr.ifr_reqcap = set;
 938
 939         if (ifp->if_capenable != set) {
 940                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 941                 if (error)
 942                         if_printf(sc->sc_ifp,
 943                             "error setting capabilities on %s: %d\n",
 944                             ifp->if_xname, error);
 945                 if ((ifp->if_capenable & ~set) != 0)
 946                         if_printf(sc->sc_ifp,
 947                             "can't disable some capabilities on %s: 0x%x\n",
 948                             ifp->if_xname, ifp->if_capenable & ~set);
 949         }
 950 }
 951
 952 /*
 953  * bridge_lookup_member:
 954  *
 955  *      Lookup a bridge member interface.
 956  */
 957 static struct bridge_iflist *
 958 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 959 {
 960         struct bridge_iflist *bif;
 961         struct ifnet *ifp;
 962
 963         BRIDGE_LOCK_ASSERT(sc);
 964
 965         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 966                 ifp = bif->bif_ifp;
 967                 if (strcmp(ifp->if_xname, name) == 0)
 968                         return (bif);
 969         }
 970
 971         return (NULL);
 972 }
 973
 974 /*
 975  * bridge_lookup_member_if:
 976  *
 977  *      Lookup a bridge member interface by ifnet*.
 978  */
 979 static struct bridge_iflist *
 980 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 981 {
 982         struct bridge_iflist *bif;
 983
 984         BRIDGE_LOCK_ASSERT(sc);
 985
 986         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 987                 if (bif->bif_ifp == member_ifp)
 988                         return (bif);
 989         }
 990
 991         return (NULL);
 992 }
 993
 994 /*
 995  * bridge_delete_member:
 996  *
 997  *      Delete the specified member interface.
 998  */
 999 static void
1000 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1001     int gone)
1002 {
1003         struct ifnet *ifs = bif->bif_ifp;
1004         struct ifnet *fif = NULL;
1005
1006         BRIDGE_LOCK_ASSERT(sc);
1007
1008         if (bif->bif_flags & IFBIF_STP)
1009                 bstp_disable(&bif->bif_stp);
1010
1011         ifs->if_bridge = NULL;
1012         BRIDGE_XLOCK(sc);
1013         LIST_REMOVE(bif, bif_next);
1014         BRIDGE_XDROP(sc);
1015
1016         /*
1017          * If removing the interface that gave the bridge its mac address, set
1018          * the mac address of the bridge to the address of the next member, or
1019          * to its default address if no members are left.
1020          */
1021         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1022                 if (LIST_EMPTY(&sc->sc_iflist)) {
1023                         bcopy(sc->sc_defaddr,
1024                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1025                         sc->sc_ifaddr = NULL;
1026                 } else {
1027                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
1028                         bcopy(IF_LLADDR(fif),
1029                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1030                         sc->sc_ifaddr = fif;
1031                 }
1032                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1033         }
1034
1035         bridge_linkcheck(sc);
1036         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1037         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1038         KASSERT(bif->bif_addrcnt == 0,
1039             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1040
1041         BRIDGE_UNLOCK(sc);
1042         if (!gone) {
1043                 switch (ifs->if_type) {
1044                 case IFT_ETHER:
1045                 case IFT_L2VLAN:
1046                         /*
1047                          * Take the interface out of promiscuous mode, but only
1048                          * if it was promiscuous in the first place. It might
1049                          * not be if we're in the bridge_ioctl_add() error path.
1050                          */
1051                         if (ifs->if_flags & IFF_PROMISC)
1052                                 (void) ifpromisc(ifs, 0);
1053                         break;
1054
1055                 case IFT_GIF:
1056                         break;
1057
1058                 default:
1059 #ifdef DIAGNOSTIC
1060                         panic("bridge_delete_member: impossible");
1061 #endif
1062                         break;
1063                 }
1064                 /* reneable any interface capabilities */
1065                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1066         }
1067         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1068         BRIDGE_LOCK(sc);
1069         free(bif, M_DEVBUF);
1070 }
1071
1072 /*
1073  * bridge_delete_span:
1074  *
1075  *      Delete the specified span interface.
1076  */
1077 static void
1078 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1079 {
1080         BRIDGE_LOCK_ASSERT(sc);
1081
1082         KASSERT(bif->bif_ifp->if_bridge == NULL,
1083             ("%s: not a span interface", __func__));
1084
1085         LIST_REMOVE(bif, bif_next);
1086         free(bif, M_DEVBUF);
1087 }
1088
1089 static int
1090 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1091 {
1092         struct ifbreq *req = arg;
1093         struct bridge_iflist *bif = NULL;
1094         struct ifnet *ifs;
1095         int error = 0;
1096
1097         ifs = ifunit(req->ifbr_ifsname);
1098         if (ifs == NULL)
1099                 return (ENOENT);
1100         if (ifs->if_ioctl == NULL)      /* must be supported */
1101                 return (EINVAL);
1102
1103         /* If it's in the span list, it can't be a member. */
1104         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1105                 if (ifs == bif->bif_ifp)
1106                         return (EBUSY);
1107
1108         if (ifs->if_bridge == sc)
1109                 return (EEXIST);
1110
1111         if (ifs->if_bridge != NULL)
1112                 return (EBUSY);
1113
1114         switch (ifs->if_type) {
1115         case IFT_ETHER:
1116         case IFT_L2VLAN:
1117         case IFT_GIF:
1118                 /* permitted interface types */
1119                 break;
1120         default:
1121                 return (EINVAL);
1122         }
1123
1124 #ifdef INET6
1125         /*
1126          * Two valid inet6 addresses with link-local scope must not be
1127          * on the parent interface and the member interfaces at the
1128          * same time.  This restriction is needed to prevent violation
1129          * of link-local scope zone.  Attempts to add a member
1130          * interface which has inet6 addresses when the parent has
1131          * inet6 triggers removal of all inet6 addresses on the member
1132          * interface.
1133          */
1134
1135         /* Check if the parent interface has a link-local scope addr. */
1136         if (V_allow_llz_overlap == 0 &&
1137             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1138                 /*
1139                  * If any, remove all inet6 addresses from the member
1140                  * interfaces.
1141                  */
1142                 BRIDGE_XLOCK(sc);
1143                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1144                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1145                                 BRIDGE_UNLOCK(sc);
1146                                 in6_ifdetach(bif->bif_ifp);
1147                                 BRIDGE_LOCK(sc);
1148                                 if_printf(sc->sc_ifp,
1149                                     "IPv6 addresses on %s have been removed "
1150                                     "before adding it as a member to prevent "
1151                                     "IPv6 address scope violation.\n",
1152                                     bif->bif_ifp->if_xname);
1153                         }
1154                 }
1155                 BRIDGE_XDROP(sc);
1156                 if (in6ifa_llaonifp(ifs)) {
1157                         BRIDGE_UNLOCK(sc);
1158                         in6_ifdetach(ifs);
1159                         BRIDGE_LOCK(sc);
1160                         if_printf(sc->sc_ifp,
1161                             "IPv6 addresses on %s have been removed "
1162                             "before adding it as a member to prevent "
1163                             "IPv6 address scope violation.\n",
1164                             ifs->if_xname);
1165                 }
1166         }
1167 #endif
1168         /* Allow the first Ethernet member to define the MTU */
1169         if (LIST_EMPTY(&sc->sc_iflist))
1170                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1171         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1172                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1173                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1174                 return (EINVAL);
1175         }
1176
1177         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1178         if (bif == NULL)
1179                 return (ENOMEM);
1180
1181         bif->bif_ifp = ifs;
1182         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1183         bif->bif_savedcaps = ifs->if_capenable;
1184
1185         /*
1186          * Assign the interface's MAC address to the bridge if it's the first
1187          * member and the MAC address of the bridge has not been changed from
1188          * the default randomly generated one.
1189          */
1190         if (V_bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1191             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1192                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1193                 sc->sc_ifaddr = ifs;
1194                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1195         }
1196
1197         ifs->if_bridge = sc;
1198         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1199         /*
1200          * XXX: XLOCK HERE!?!
1201          *
1202          * NOTE: insert_***HEAD*** should be safe for the traversals.
1203          */
1204         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1205
1206         /* Set interface capabilities to the intersection set of all members */
1207         bridge_mutecaps(sc);
1208         bridge_linkcheck(sc);
1209
1210         /* Place the interface into promiscuous mode */
1211         switch (ifs->if_type) {
1212                 case IFT_ETHER:
1213                 case IFT_L2VLAN:
1214                         BRIDGE_UNLOCK(sc);
1215                         error = ifpromisc(ifs, 1);
1216                         BRIDGE_LOCK(sc);
1217                         break;
1218         }
1219
1220         if (error)
1221                 bridge_delete_member(sc, bif, 0);
1222         return (error);
1223 }
1224
1225 static int
1226 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1227 {
1228         struct ifbreq *req = arg;
1229         struct bridge_iflist *bif;
1230
1231         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1232         if (bif == NULL)
1233                 return (ENOENT);
1234
1235         bridge_delete_member(sc, bif, 0);
1236
1237         return (0);
1238 }
1239
1240 static int
1241 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1242 {
1243         struct ifbreq *req = arg;
1244         struct bridge_iflist *bif;
1245         struct bstp_port *bp;
1246
1247         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1248         if (bif == NULL)
1249                 return (ENOENT);
1250
1251         bp = &bif->bif_stp;
1252         req->ifbr_ifsflags = bif->bif_flags;
1253         req->ifbr_state = bp->bp_state;
1254         req->ifbr_priority = bp->bp_priority;
1255         req->ifbr_path_cost = bp->bp_path_cost;
1256         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1257         req->ifbr_proto = bp->bp_protover;
1258         req->ifbr_role = bp->bp_role;
1259         req->ifbr_stpflags = bp->bp_flags;
1260         req->ifbr_addrcnt = bif->bif_addrcnt;
1261         req->ifbr_addrmax = bif->bif_addrmax;
1262         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1263
1264         /* Copy STP state options as flags */
1265         if (bp->bp_operedge)
1266                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1267         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1268                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1269         if (bp->bp_ptp_link)
1270                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1271         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1272                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1273         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1274                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1275         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1276                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1277         return (0);
1278 }
1279
1280 static int
1281 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1282 {
1283         struct ifbreq *req = arg;
1284         struct bridge_iflist *bif;
1285         struct bstp_port *bp;
1286         int error;
1287
1288         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1289         if (bif == NULL)
1290                 return (ENOENT);
1291         bp = &bif->bif_stp;
1292
1293         if (req->ifbr_ifsflags & IFBIF_SPAN)
1294                 /* SPAN is readonly */
1295                 return (EINVAL);
1296
1297         if (req->ifbr_ifsflags & IFBIF_STP) {
1298                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1299                         error = bstp_enable(&bif->bif_stp);
1300                         if (error)
1301                                 return (error);
1302                 }
1303         } else {
1304                 if ((bif->bif_flags & IFBIF_STP) != 0)
1305                         bstp_disable(&bif->bif_stp);
1306         }
1307
1308         /* Pass on STP flags */
1309         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1310         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1311         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1312         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1313
1314         /* Save the bits relating to the bridge */
1315         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1316
1317         return (0);
1318 }
1319
1320 static int
1321 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1322 {
1323         struct ifbrparam *param = arg;
1324
1325         sc->sc_brtmax = param->ifbrp_csize;
1326         bridge_rttrim(sc);
1327
1328         return (0);
1329 }
1330
1331 static int
1332 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1333 {
1334         struct ifbrparam *param = arg;
1335
1336         param->ifbrp_csize = sc->sc_brtmax;
1337
1338         return (0);
1339 }
1340
1341 static int
1342 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1343 {
1344         struct ifbifconf *bifc = arg;
1345         struct bridge_iflist *bif;
1346         struct ifbreq breq;
1347         char *buf, *outbuf;
1348         int count, buflen, len, error = 0;
1349
1350         count = 0;
1351         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1352                 count++;
1353         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1354                 count++;
1355
1356         buflen = sizeof(breq) * count;
1357         if (bifc->ifbic_len == 0) {
1358                 bifc->ifbic_len = buflen;
1359                 return (0);
1360         }
1361         BRIDGE_UNLOCK(sc);
1362         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1363         BRIDGE_LOCK(sc);
1364
1365         count = 0;
1366         buf = outbuf;
1367         len = min(bifc->ifbic_len, buflen);
1368         bzero(&breq, sizeof(breq));
1369         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1370                 if (len < sizeof(breq))
1371                         break;
1372
1373                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1374                     sizeof(breq.ifbr_ifsname));
1375                 /* Fill in the ifbreq structure */
1376                 error = bridge_ioctl_gifflags(sc, &breq);
1377                 if (error)
1378                         break;
1379                 memcpy(buf, &breq, sizeof(breq));
1380                 count++;
1381                 buf += sizeof(breq);
1382                 len -= sizeof(breq);
1383         }
1384         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1385                 if (len < sizeof(breq))
1386                         break;
1387
1388                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1389                     sizeof(breq.ifbr_ifsname));
1390                 breq.ifbr_ifsflags = bif->bif_flags;
1391                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1392                 memcpy(buf, &breq, sizeof(breq));
1393                 count++;
1394                 buf += sizeof(breq);
1395                 len -= sizeof(breq);
1396         }
1397
1398         BRIDGE_UNLOCK(sc);
1399         bifc->ifbic_len = sizeof(breq) * count;
1400         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1401         BRIDGE_LOCK(sc);
1402         free(outbuf, M_TEMP);
1403         return (error);
1404 }
1405
1406 static int
1407 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1408 {
1409         struct ifbaconf *bac = arg;
1410         struct bridge_rtnode *brt;
1411         struct ifbareq bareq;
1412         char *buf, *outbuf;
1413         int count, buflen, len, error = 0;
1414
1415         if (bac->ifbac_len == 0)
1416                 return (0);
1417
1418         count = 0;
1419         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1420                 count++;
1421         buflen = sizeof(bareq) * count;
1422
1423         BRIDGE_UNLOCK(sc);
1424         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1425         BRIDGE_LOCK(sc);
1426
1427         count = 0;
1428         buf = outbuf;
1429         len = min(bac->ifbac_len, buflen);
1430         bzero(&bareq, sizeof(bareq));
1431         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1432                 if (len < sizeof(bareq))
1433                         goto out;
1434                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1435                     sizeof(bareq.ifba_ifsname));
1436                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1437                 bareq.ifba_vlan = brt->brt_vlan;
1438                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1439                                 time_uptime < brt->brt_expire)
1440                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1441                 else
1442                         bareq.ifba_expire = 0;
1443                 bareq.ifba_flags = brt->brt_flags;
1444
1445                 memcpy(buf, &bareq, sizeof(bareq));
1446                 count++;
1447                 buf += sizeof(bareq);
1448                 len -= sizeof(bareq);
1449         }
1450 out:
1451         BRIDGE_UNLOCK(sc);
1452         bac->ifbac_len = sizeof(bareq) * count;
1453         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1454         BRIDGE_LOCK(sc);
1455         free(outbuf, M_TEMP);
1456         return (error);
1457 }
1458
1459 static int
1460 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1461 {
1462         struct ifbareq *req = arg;
1463         struct bridge_iflist *bif;
1464         int error;
1465
1466         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1467         if (bif == NULL)
1468                 return (ENOENT);
1469
1470         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1471             req->ifba_flags);
1472
1473         return (error);
1474 }
1475
1476 static int
1477 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1478 {
1479         struct ifbrparam *param = arg;
1480
1481         sc->sc_brttimeout = param->ifbrp_ctime;
1482         return (0);
1483 }
1484
1485 static int
1486 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1487 {
1488         struct ifbrparam *param = arg;
1489
1490         param->ifbrp_ctime = sc->sc_brttimeout;
1491         return (0);
1492 }
1493
1494 static int
1495 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1496 {
1497         struct ifbareq *req = arg;
1498
1499         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1500 }
1501
1502 static int
1503 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1504 {
1505         struct ifbreq *req = arg;
1506
1507         bridge_rtflush(sc, req->ifbr_ifsflags);
1508         return (0);
1509 }
1510
1511 static int
1512 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1513 {
1514         struct ifbrparam *param = arg;
1515         struct bstp_state *bs = &sc->sc_stp;
1516
1517         param->ifbrp_prio = bs->bs_bridge_priority;
1518         return (0);
1519 }
1520
1521 static int
1522 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1523 {
1524         struct ifbrparam *param = arg;
1525
1526         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1527 }
1528
1529 static int
1530 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1531 {
1532         struct ifbrparam *param = arg;
1533         struct bstp_state *bs = &sc->sc_stp;
1534
1535         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1536         return (0);
1537 }
1538
1539 static int
1540 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1541 {
1542         struct ifbrparam *param = arg;
1543
1544         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1545 }
1546
1547 static int
1548 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1549 {
1550         struct ifbrparam *param = arg;
1551         struct bstp_state *bs = &sc->sc_stp;
1552
1553         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1554         return (0);
1555 }
1556
1557 static int
1558 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1559 {
1560         struct ifbrparam *param = arg;
1561
1562         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1563 }
1564
1565 static int
1566 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1567 {
1568         struct ifbrparam *param = arg;
1569         struct bstp_state *bs = &sc->sc_stp;
1570
1571         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1572         return (0);
1573 }
1574
1575 static int
1576 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1577 {
1578         struct ifbrparam *param = arg;
1579
1580         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1581 }
1582
1583 static int
1584 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1585 {
1586         struct ifbreq *req = arg;
1587         struct bridge_iflist *bif;
1588
1589         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1590         if (bif == NULL)
1591                 return (ENOENT);
1592
1593         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1594 }
1595
1596 static int
1597 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1598 {
1599         struct ifbreq *req = arg;
1600         struct bridge_iflist *bif;
1601
1602         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1603         if (bif == NULL)
1604                 return (ENOENT);
1605
1606         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1607 }
1608
1609 static int
1610 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1611 {
1612         struct ifbreq *req = arg;
1613         struct bridge_iflist *bif;
1614
1615         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1616         if (bif == NULL)
1617                 return (ENOENT);
1618
1619         bif->bif_addrmax = req->ifbr_addrmax;
1620         return (0);
1621 }
1622
1623 static int
1624 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1625 {
1626         struct ifbreq *req = arg;
1627         struct bridge_iflist *bif = NULL;
1628         struct ifnet *ifs;
1629
1630         ifs = ifunit(req->ifbr_ifsname);
1631         if (ifs == NULL)
1632                 return (ENOENT);
1633
1634         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1635                 if (ifs == bif->bif_ifp)
1636                         return (EBUSY);
1637
1638         if (ifs->if_bridge != NULL)
1639                 return (EBUSY);
1640
1641         switch (ifs->if_type) {
1642                 case IFT_ETHER:
1643                 case IFT_GIF:
1644                 case IFT_L2VLAN:
1645                         break;
1646                 default:
1647                         return (EINVAL);
1648         }
1649
1650         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1651         if (bif == NULL)
1652                 return (ENOMEM);
1653
1654         bif->bif_ifp = ifs;
1655         bif->bif_flags = IFBIF_SPAN;
1656
1657         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1658
1659         return (0);
1660 }
1661
1662 static int
1663 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1664 {
1665         struct ifbreq *req = arg;
1666         struct bridge_iflist *bif;
1667         struct ifnet *ifs;
1668
1669         ifs = ifunit(req->ifbr_ifsname);
1670         if (ifs == NULL)
1671                 return (ENOENT);
1672
1673         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1674                 if (ifs == bif->bif_ifp)
1675                         break;
1676
1677         if (bif == NULL)
1678                 return (ENOENT);
1679
1680         bridge_delete_span(sc, bif);
1681
1682         return (0);
1683 }
1684
1685 static int
1686 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1687 {
1688         struct ifbropreq *req = arg;
1689         struct bstp_state *bs = &sc->sc_stp;
1690         struct bstp_port *root_port;
1691
1692         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1693         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1694         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1695
1696         root_port = bs->bs_root_port;
1697         if (root_port == NULL)
1698                 req->ifbop_root_port = 0;
1699         else
1700                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1701
1702         req->ifbop_holdcount = bs->bs_txholdcount;
1703         req->ifbop_priority = bs->bs_bridge_priority;
1704         req->ifbop_protocol = bs->bs_protover;
1705         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1706         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1707         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1708         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1709         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1710         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1711
1712         return (0);
1713 }
1714
1715 static int
1716 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1717 {
1718         struct ifbrparam *param = arg;
1719
1720         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1721         return (0);
1722 }
1723
1724 static int
1725 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1726 {
1727         struct ifbpstpconf *bifstp = arg;
1728         struct bridge_iflist *bif;
1729         struct bstp_port *bp;
1730         struct ifbpstpreq bpreq;
1731         char *buf, *outbuf;
1732         int count, buflen, len, error = 0;
1733
1734         count = 0;
1735         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1736                 if ((bif->bif_flags & IFBIF_STP) != 0)
1737                         count++;
1738         }
1739
1740         buflen = sizeof(bpreq) * count;
1741         if (bifstp->ifbpstp_len == 0) {
1742                 bifstp->ifbpstp_len = buflen;
1743                 return (0);
1744         }
1745
1746         BRIDGE_UNLOCK(sc);
1747         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1748         BRIDGE_LOCK(sc);
1749
1750         count = 0;
1751         buf = outbuf;
1752         len = min(bifstp->ifbpstp_len, buflen);
1753         bzero(&bpreq, sizeof(bpreq));
1754         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1755                 if (len < sizeof(bpreq))
1756                         break;
1757
1758                 if ((bif->bif_flags & IFBIF_STP) == 0)
1759                         continue;
1760
1761                 bp = &bif->bif_stp;
1762                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1763                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1764                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1765                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1766                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1767                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1768
1769                 memcpy(buf, &bpreq, sizeof(bpreq));
1770                 count++;
1771                 buf += sizeof(bpreq);
1772                 len -= sizeof(bpreq);
1773         }
1774
1775         BRIDGE_UNLOCK(sc);
1776         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1777         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1778         BRIDGE_LOCK(sc);
1779         free(outbuf, M_TEMP);
1780         return (error);
1781 }
1782
1783 static int
1784 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1785 {
1786         struct ifbrparam *param = arg;
1787
1788         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1789 }
1790
1791 static int
1792 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1793 {
1794         struct ifbrparam *param = arg;
1795
1796         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1797 }
1798
1799 /*
1800  * bridge_ifdetach:
1801  *
1802  *      Detach an interface from a bridge.  Called when a member
1803  *      interface is detaching.
1804  */
1805 static void
1806 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1807 {
1808         struct bridge_softc *sc = ifp->if_bridge;
1809         struct bridge_iflist *bif;
1810
1811         if (ifp->if_flags & IFF_RENAMING)
1812                 return;
1813         if (V_bridge_cloner == NULL) {
1814                 /*
1815                  * This detach handler can be called after
1816                  * vnet_bridge_uninit().  Just return in that case.
1817                  */
1818                 return;
1819         }
1820         /* Check if the interface is a bridge member */
1821         if (sc != NULL) {
1822                 BRIDGE_LOCK(sc);
1823
1824                 bif = bridge_lookup_member_if(sc, ifp);
1825                 if (bif != NULL)
1826                         bridge_delete_member(sc, bif, 1);
1827
1828                 BRIDGE_UNLOCK(sc);
1829                 return;
1830         }
1831
1832         /* Check if the interface is a span port */
1833         BRIDGE_LIST_LOCK();
1834         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1835                 BRIDGE_LOCK(sc);
1836                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1837                         if (ifp == bif->bif_ifp) {
1838                                 bridge_delete_span(sc, bif);
1839                                 break;
1840                         }
1841
1842                 BRIDGE_UNLOCK(sc);
1843         }
1844         BRIDGE_LIST_UNLOCK();
1845 }
1846
1847 /*
1848  * bridge_init:
1849  *
1850  *      Initialize a bridge interface.
1851  */
1852 static void
1853 bridge_init(void *xsc)
1854 {
1855         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1856         struct ifnet *ifp = sc->sc_ifp;
1857
1858         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1859                 return;
1860
1861         BRIDGE_LOCK(sc);
1862         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1863             bridge_timer, sc);
1864
1865         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1866         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1867
1868         BRIDGE_UNLOCK(sc);
1869 }
1870
1871 /*
1872  * bridge_stop:
1873  *
1874  *      Stop the bridge interface.
1875  */
1876 static void
1877 bridge_stop(struct ifnet *ifp, int disable)
1878 {
1879         struct bridge_softc *sc = ifp->if_softc;
1880
1881         BRIDGE_LOCK_ASSERT(sc);
1882
1883         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1884                 return;
1885
1886         callout_stop(&sc->sc_brcallout);
1887         bstp_stop(&sc->sc_stp);
1888
1889         bridge_rtflush(sc, IFBF_FLUSHDYN);
1890
1891         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1892 }
1893
1894 /*
1895  * bridge_enqueue:
1896  *
1897  *      Enqueue a packet on a bridge member interface.
1898  *
1899  */
1900 static int
1901 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1902 {
1903         int len, err = 0;
1904         short mflags;
1905         struct mbuf *m0;
1906
1907         /* We may be sending a fragment so traverse the mbuf */
1908         for (; m; m = m0) {
1909                 m0 = m->m_nextpkt;
1910                 m->m_nextpkt = NULL;
1911                 len = m->m_pkthdr.len;
1912                 mflags = m->m_flags;
1913
1914                 /*
1915                  * If underlying interface can not do VLAN tag insertion itself
1916                  * then attach a packet tag that holds it.
1917                  */
1918                 if ((m->m_flags & M_VLANTAG) &&
1919                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1920                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1921                         if (m == NULL) {
1922                                 if_printf(dst_ifp,
1923                                     "unable to prepend VLAN header\n");
1924                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
1925                                 continue;
1926                         }
1927                         m->m_flags &= ~M_VLANTAG;
1928                 }
1929
1930                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
1931                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1932                         m_freem(m0);
1933                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
1934                         break;
1935                 }
1936
1937                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
1938                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
1939                 if (mflags & M_MCAST)
1940                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
1941         }
1942
1943         return (err);
1944 }
1945
1946 /*
1947  * bridge_dummynet:
1948  *
1949  *      Receive a queued packet from dummynet and pass it on to the output
1950  *      interface.
1951  *
1952  *      The mbuf has the Ethernet header already attached.
1953  */
1954 static void
1955 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1956 {
1957         struct bridge_softc *sc;
1958
1959         sc = ifp->if_bridge;
1960
1961         /*
1962          * The packet didnt originate from a member interface. This should only
1963          * ever happen if a member interface is removed while packets are
1964          * queued for it.
1965          */
1966         if (sc == NULL) {
1967                 m_freem(m);
1968                 return;
1969         }
1970
1971         if (PFIL_HOOKED(&V_inet_pfil_hook)
1972 #ifdef INET6
1973             || PFIL_HOOKED(&V_inet6_pfil_hook)
1974 #endif
1975             ) {
1976                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1977                         return;
1978                 if (m == NULL)
1979                         return;
1980         }
1981
1982         bridge_enqueue(sc, ifp, m);
1983 }
1984
1985 /*
1986  * bridge_output:
1987  *
1988  *      Send output from a bridge member interface.  This
1989  *      performs the bridging function for locally originated
1990  *      packets.
1991  *
1992  *      The mbuf has the Ethernet header already attached.  We must
1993  *      enqueue or free the mbuf before returning.
1994  */
1995 static int
1996 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1997     struct rtentry *rt)
1998 {
1999         struct ether_header *eh;
2000         struct ifnet *dst_if;
2001         struct bridge_softc *sc;
2002         uint16_t vlan;
2003
2004         if (m->m_len < ETHER_HDR_LEN) {
2005                 m = m_pullup(m, ETHER_HDR_LEN);
2006                 if (m == NULL)
2007                         return (0);
2008         }
2009
2010         eh = mtod(m, struct ether_header *);
2011         sc = ifp->if_bridge;
2012         vlan = VLANTAGOF(m);
2013
2014         BRIDGE_LOCK(sc);
2015
2016         /*
2017          * If bridge is down, but the original output interface is up,
2018          * go ahead and send out that interface.  Otherwise, the packet
2019          * is dropped below.
2020          */
2021         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2022                 dst_if = ifp;
2023                 goto sendunicast;
2024         }
2025
2026         /*
2027          * If the packet is a multicast, or we don't know a better way to
2028          * get there, send to all interfaces.
2029          */
2030         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2031                 dst_if = NULL;
2032         else
2033                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2034         if (dst_if == NULL) {
2035                 struct bridge_iflist *bif;
2036                 struct mbuf *mc;
2037                 int error = 0, used = 0;
2038
2039                 bridge_span(sc, m);
2040
2041                 BRIDGE_LOCK2REF(sc, error);
2042                 if (error) {
2043                         m_freem(m);
2044                         return (0);
2045                 }
2046
2047                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2048                         dst_if = bif->bif_ifp;
2049
2050                         if (dst_if->if_type == IFT_GIF)
2051                                 continue;
2052                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2053                                 continue;
2054
2055                         /*
2056                          * If this is not the original output interface,
2057                          * and the interface is participating in spanning
2058                          * tree, make sure the port is in a state that
2059                          * allows forwarding.
2060                          */
2061                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2062                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2063                                 continue;
2064
2065                         if (LIST_NEXT(bif, bif_next) == NULL) {
2066                                 used = 1;
2067                                 mc = m;
2068                         } else {
2069                                 mc = m_copypacket(m, M_NOWAIT);
2070                                 if (mc == NULL) {
2071                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2072                                         continue;
2073                                 }
2074                         }
2075
2076                         bridge_enqueue(sc, dst_if, mc);
2077                 }
2078                 if (used == 0)
2079                         m_freem(m);
2080                 BRIDGE_UNREF(sc);
2081                 return (0);
2082         }
2083
2084 sendunicast:
2085         /*
2086          * XXX Spanning tree consideration here?
2087          */
2088
2089         bridge_span(sc, m);
2090         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2091                 m_freem(m);
2092                 BRIDGE_UNLOCK(sc);
2093                 return (0);
2094         }
2095
2096         BRIDGE_UNLOCK(sc);
2097         bridge_enqueue(sc, dst_if, m);
2098         return (0);
2099 }
2100
2101 /*
2102  * bridge_transmit:
2103  *
2104  *      Do output on a bridge.
2105  *
2106  */
2107 static int
2108 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2109 {
2110         struct bridge_softc *sc;
2111         struct ether_header *eh;
2112         struct ifnet *dst_if;
2113         int error = 0;
2114
2115         sc = ifp->if_softc;
2116
2117         ETHER_BPF_MTAP(ifp, m);
2118
2119         eh = mtod(m, struct ether_header *);
2120
2121         BRIDGE_LOCK(sc);
2122         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2123             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2124                 BRIDGE_UNLOCK(sc);
2125                 error = bridge_enqueue(sc, dst_if, m);
2126         } else
2127                 bridge_broadcast(sc, ifp, m, 0);
2128
2129         return (error);
2130 }
2131
2132 /*
2133  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2134  */
2135 static void
2136 bridge_qflush(struct ifnet *ifp __unused)
2137 {
2138 }
2139
2140 /*
2141  * bridge_forward:
2142  *
2143  *      The forwarding function of the bridge.
2144  *
2145  *      NOTE: Releases the lock on return.
2146  */
2147 static void
2148 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2149     struct mbuf *m)
2150 {
2151         struct bridge_iflist *dbif;
2152         struct ifnet *src_if, *dst_if, *ifp;
2153         struct ether_header *eh;
2154         uint16_t vlan;
2155         uint8_t *dst;
2156         int error;
2157
2158         src_if = m->m_pkthdr.rcvif;
2159         ifp = sc->sc_ifp;
2160
2161         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2162         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2163         vlan = VLANTAGOF(m);
2164
2165         if ((sbif->bif_flags & IFBIF_STP) &&
2166             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2167                 goto drop;
2168
2169         eh = mtod(m, struct ether_header *);
2170         dst = eh->ether_dhost;
2171
2172         /* If the interface is learning, record the address. */
2173         if (sbif->bif_flags & IFBIF_LEARNING) {
2174                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2175                     sbif, 0, IFBAF_DYNAMIC);
2176                 /*
2177                  * If the interface has addresses limits then deny any source
2178                  * that is not in the cache.
2179                  */
2180                 if (error && sbif->bif_addrmax)
2181                         goto drop;
2182         }
2183
2184         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2185             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2186                 goto drop;
2187
2188         /*
2189          * At this point, the port either doesn't participate
2190          * in spanning tree or it is in the forwarding state.
2191          */
2192
2193         /*
2194          * If the packet is unicast, destined for someone on
2195          * "this" side of the bridge, drop it.
2196          */
2197         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2198                 dst_if = bridge_rtlookup(sc, dst, vlan);
2199                 if (src_if == dst_if)
2200                         goto drop;
2201         } else {
2202                 /*
2203                  * Check if its a reserved multicast address, any address
2204                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2205                  * bridge.
2206                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2207                  */
2208                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2209                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2210                     dst[4] == 0x00 && dst[5] <= 0x0f)
2211                         goto drop;
2212
2213                 /* ...forward it to all interfaces. */
2214                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2215                 dst_if = NULL;
2216         }
2217
2218         /*
2219          * If we have a destination interface which is a member of our bridge,
2220          * OR this is a unicast packet, push it through the bpf(4) machinery.
2221          * For broadcast or multicast packets, don't bother because it will
2222          * be reinjected into ether_input. We do this before we pass the packets
2223          * through the pfil(9) framework, as it is possible that pfil(9) will
2224          * drop the packet, or possibly modify it, making it difficult to debug
2225          * firewall issues on the bridge.
2226          */
2227         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2228                 ETHER_BPF_MTAP(ifp, m);
2229
2230         /* run the packet filter */
2231         if (PFIL_HOOKED(&V_inet_pfil_hook)
2232 #ifdef INET6
2233             || PFIL_HOOKED(&V_inet6_pfil_hook)
2234 #endif
2235             ) {
2236                 BRIDGE_UNLOCK(sc);
2237                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2238                         return;
2239                 if (m == NULL)
2240                         return;
2241                 BRIDGE_LOCK(sc);
2242         }
2243
2244         if (dst_if == NULL) {
2245                 bridge_broadcast(sc, src_if, m, 1);
2246                 return;
2247         }
2248
2249         /*
2250          * At this point, we're dealing with a unicast frame
2251          * going to a different interface.
2252          */
2253         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2254                 goto drop;
2255
2256         dbif = bridge_lookup_member_if(sc, dst_if);
2257         if (dbif == NULL)
2258                 /* Not a member of the bridge (anymore?) */
2259                 goto drop;
2260
2261         /* Private segments can not talk to each other */
2262         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2263                 goto drop;
2264
2265         if ((dbif->bif_flags & IFBIF_STP) &&
2266             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2267                 goto drop;
2268
2269         BRIDGE_UNLOCK(sc);
2270
2271         if (PFIL_HOOKED(&V_inet_pfil_hook)
2272 #ifdef INET6
2273             || PFIL_HOOKED(&V_inet6_pfil_hook)
2274 #endif
2275             ) {
2276                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2277                         return;
2278                 if (m == NULL)
2279                         return;
2280         }
2281
2282         bridge_enqueue(sc, dst_if, m);
2283         return;
2284
2285 drop:
2286         BRIDGE_UNLOCK(sc);
2287         m_freem(m);
2288 }
2289
2290 /*
2291  * bridge_input:
2292  *
2293  *      Receive input from a member interface.  Queue the packet for
2294  *      bridging if it is not for us.
2295  */
2296 static struct mbuf *
2297 bridge_input(struct ifnet *ifp, struct mbuf *m)
2298 {
2299         struct bridge_softc *sc = ifp->if_bridge;
2300         struct bridge_iflist *bif, *bif2;
2301         struct ifnet *bifp;
2302         struct ether_header *eh;
2303         struct mbuf *mc, *mc2;
2304         uint16_t vlan;
2305         int error;
2306
2307         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2308                 return (m);
2309
2310         bifp = sc->sc_ifp;
2311         vlan = VLANTAGOF(m);
2312
2313         /*
2314          * Implement support for bridge monitoring. If this flag has been
2315          * set on this interface, discard the packet once we push it through
2316          * the bpf(4) machinery, but before we do, increment the byte and
2317          * packet counters associated with this interface.
2318          */
2319         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2320                 m->m_pkthdr.rcvif  = bifp;
2321                 ETHER_BPF_MTAP(bifp, m);
2322                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2323                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2324                 m_freem(m);
2325                 return (NULL);
2326         }
2327         BRIDGE_LOCK(sc);
2328         bif = bridge_lookup_member_if(sc, ifp);
2329         if (bif == NULL) {
2330                 BRIDGE_UNLOCK(sc);
2331                 return (m);
2332         }
2333
2334         eh = mtod(m, struct ether_header *);
2335
2336         bridge_span(sc, m);
2337
2338         if (m->m_flags & (M_BCAST|M_MCAST)) {
2339                 /* Tap off 802.1D packets; they do not get forwarded. */
2340                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2341                     ETHER_ADDR_LEN) == 0) {
2342                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2343                         BRIDGE_UNLOCK(sc);
2344                         return (NULL);
2345                 }
2346
2347                 if ((bif->bif_flags & IFBIF_STP) &&
2348                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2349                         BRIDGE_UNLOCK(sc);
2350                         return (m);
2351                 }
2352
2353                 /*
2354                  * Make a deep copy of the packet and enqueue the copy
2355                  * for bridge processing; return the original packet for
2356                  * local processing.
2357                  */
2358                 mc = m_dup(m, M_NOWAIT);
2359                 if (mc == NULL) {
2360                         BRIDGE_UNLOCK(sc);
2361                         return (m);
2362                 }
2363
2364                 /* Perform the bridge forwarding function with the copy. */
2365                 bridge_forward(sc, bif, mc);
2366
2367                 /*
2368                  * Reinject the mbuf as arriving on the bridge so we have a
2369                  * chance at claiming multicast packets. We can not loop back
2370                  * here from ether_input as a bridge is never a member of a
2371                  * bridge.
2372                  */
2373                 KASSERT(bifp->if_bridge == NULL,
2374                     ("loop created in bridge_input"));
2375                 mc2 = m_dup(m, M_NOWAIT);
2376                 if (mc2 != NULL) {
2377                         /* Keep the layer3 header aligned */
2378                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2379                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2380                 }
2381                 if (mc2 != NULL) {
2382                         mc2->m_pkthdr.rcvif = bifp;
2383                         (*bifp->if_input)(bifp, mc2);
2384                 }
2385
2386                 /* Return the original packet for local processing. */
2387                 return (m);
2388         }
2389
2390         if ((bif->bif_flags & IFBIF_STP) &&
2391             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2392                 BRIDGE_UNLOCK(sc);
2393                 return (m);
2394         }
2395
2396 #if (defined(INET) || defined(INET6))
2397 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2398         || ((iface)->if_carp \
2399             && (*carp_forus_p)((iface), eh->ether_dhost))
2400 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2401         || ((iface)->if_carp \
2402             && (*carp_forus_p)((iface), eh->ether_shost))
2403 #else
2404 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2405 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2406 #endif
2407
2408 #ifdef INET6
2409 #   define OR_PFIL_HOOKED_INET6 \
2410         || PFIL_HOOKED(&V_inet6_pfil_hook)
2411 #else
2412 #   define OR_PFIL_HOOKED_INET6
2413 #endif
2414
2415 #define GRAB_OUR_PACKETS(iface) \
2416         if ((iface)->if_type == IFT_GIF) \
2417                 continue; \
2418         /* It is destined for us. */ \
2419         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2420             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2421             ) {                                                         \
2422                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2423                         ETHER_BPF_MTAP(iface, m);                       \
2424                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2425                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2426                         /* Filter on the physical interface. */         \
2427                         if (V_pfil_local_phys &&                        \
2428                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2429                              OR_PFIL_HOOKED_INET6)) {                   \
2430                                 if (bridge_pfil(&m, NULL, ifp,          \
2431                                     PFIL_IN) != 0 || m == NULL) {       \
2432                                         BRIDGE_UNLOCK(sc);              \
2433                                         return (NULL);                  \
2434                                 }                                       \
2435                                 eh = mtod(m, struct ether_header *);    \
2436                         }                                               \
2437                 }                                                       \
2438                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2439                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2440                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2441                         if (error && bif->bif_addrmax) {                \
2442                                 BRIDGE_UNLOCK(sc);                      \
2443                                 m_freem(m);                             \
2444                                 return (NULL);                          \
2445                         }                                               \
2446                 }                                                       \
2447                 m->m_pkthdr.rcvif = iface;                              \
2448                 BRIDGE_UNLOCK(sc);                                      \
2449                 return (m);                                             \
2450         }                                                               \
2451                                                                         \
2452         /* We just received a packet that we sent out. */               \
2453         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2454             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2455             ) {                                                         \
2456                 BRIDGE_UNLOCK(sc);                                      \
2457                 m_freem(m);                                             \
2458                 return (NULL);                                          \
2459         }
2460
2461         /*
2462          * Unicast.  Make sure it's not for the bridge.
2463          */
2464         do { GRAB_OUR_PACKETS(bifp) } while (0);
2465
2466         /*
2467          * Give a chance for ifp at first priority. This will help when the
2468          * packet comes through the interface like VLAN's with the same MACs
2469          * on several interfaces from the same bridge. This also will save
2470          * some CPU cycles in case the destination interface and the input
2471          * interface (eq ifp) are the same.
2472          */
2473         do { GRAB_OUR_PACKETS(ifp) } while (0);
2474
2475         /* Now check the all bridge members. */
2476         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2477                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2478         }
2479
2480 #undef OR_CARP_CHECK_WE_ARE_DST
2481 #undef OR_CARP_CHECK_WE_ARE_SRC
2482 #undef OR_PFIL_HOOKED_INET6
2483 #undef GRAB_OUR_PACKETS
2484
2485         /* Perform the bridge forwarding function. */
2486         bridge_forward(sc, bif, m);
2487
2488         return (NULL);
2489 }
2490
2491 /*
2492  * bridge_broadcast:
2493  *
2494  *      Send a frame to all interfaces that are members of
2495  *      the bridge, except for the one on which the packet
2496  *      arrived.
2497  *
2498  *      NOTE: Releases the lock on return.
2499  */
2500 static void
2501 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2502     struct mbuf *m, int runfilt)
2503 {
2504         struct bridge_iflist *dbif, *sbif;
2505         struct mbuf *mc;
2506         struct ifnet *dst_if;
2507         int error = 0, used = 0, i;
2508
2509         sbif = bridge_lookup_member_if(sc, src_if);
2510
2511         BRIDGE_LOCK2REF(sc, error);
2512         if (error) {
2513                 m_freem(m);
2514                 return;
2515         }
2516
2517         /* Filter on the bridge interface before broadcasting */
2518         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2519 #ifdef INET6
2520             || PFIL_HOOKED(&V_inet6_pfil_hook)
2521 #endif
2522             )) {
2523                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2524                         goto out;
2525                 if (m == NULL)
2526                         goto out;
2527         }
2528
2529         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2530                 dst_if = dbif->bif_ifp;
2531                 if (dst_if == src_if)
2532                         continue;
2533
2534                 /* Private segments can not talk to each other */
2535                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2536                         continue;
2537
2538                 if ((dbif->bif_flags & IFBIF_STP) &&
2539                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2540                         continue;
2541
2542                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2543                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2544                         continue;
2545
2546                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2547                         continue;
2548
2549                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2550                         mc = m;
2551                         used = 1;
2552                 } else {
2553                         mc = m_dup(m, M_NOWAIT);
2554                         if (mc == NULL) {
2555                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2556                                 continue;
2557                         }
2558                 }
2559
2560                 /*
2561                  * Filter on the output interface. Pass a NULL bridge interface
2562                  * pointer so we do not redundantly filter on the bridge for
2563                  * each interface we broadcast on.
2564                  */
2565                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2566 #ifdef INET6
2567                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2568 #endif
2569                     )) {
2570                         if (used == 0) {
2571                                 /* Keep the layer3 header aligned */
2572                                 i = min(mc->m_pkthdr.len, max_protohdr);
2573                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2574                                 if (mc == NULL) {
2575                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2576                                         continue;
2577                                 }
2578                         }
2579                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2580                                 continue;
2581                         if (mc == NULL)
2582                                 continue;
2583                 }
2584
2585                 bridge_enqueue(sc, dst_if, mc);
2586         }
2587         if (used == 0)
2588                 m_freem(m);
2589
2590 out:
2591         BRIDGE_UNREF(sc);
2592 }
2593
2594 /*
2595  * bridge_span:
2596  *
2597  *      Duplicate a packet out one or more interfaces that are in span mode,
2598  *      the original mbuf is unmodified.
2599  */
2600 static void
2601 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2602 {
2603         struct bridge_iflist *bif;
2604         struct ifnet *dst_if;
2605         struct mbuf *mc;
2606
2607         if (LIST_EMPTY(&sc->sc_spanlist))
2608                 return;
2609
2610         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2611                 dst_if = bif->bif_ifp;
2612
2613                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2614                         continue;
2615
2616                 mc = m_copypacket(m, M_NOWAIT);
2617                 if (mc == NULL) {
2618                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2619                         continue;
2620                 }
2621
2622                 bridge_enqueue(sc, dst_if, mc);
2623         }
2624 }
2625
2626 /*
2627  * bridge_rtupdate:
2628  *
2629  *      Add a bridge routing entry.
2630  */
2631 static int
2632 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2633     struct bridge_iflist *bif, int setflags, uint8_t flags)
2634 {
2635         struct bridge_rtnode *brt;
2636         int error;
2637
2638         BRIDGE_LOCK_ASSERT(sc);
2639
2640         /* Check the source address is valid and not multicast. */
2641         if (ETHER_IS_MULTICAST(dst) ||
2642             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2643              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2644                 return (EINVAL);
2645
2646         /* 802.1p frames map to vlan 1 */
2647         if (vlan == 0)
2648                 vlan = 1;
2649
2650         /*
2651          * A route for this destination might already exist.  If so,
2652          * update it, otherwise create a new one.
2653          */
2654         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2655                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2656                         sc->sc_brtexceeded++;
2657                         return (ENOSPC);
2658                 }
2659                 /* Check per interface address limits (if enabled) */
2660                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2661                         bif->bif_addrexceeded++;
2662                         return (ENOSPC);
2663                 }
2664
2665                 /*
2666                  * Allocate a new bridge forwarding node, and
2667                  * initialize the expiration time and Ethernet
2668                  * address.
2669                  */
2670                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2671                 if (brt == NULL)
2672                         return (ENOMEM);
2673
2674                 if (bif->bif_flags & IFBIF_STICKY)
2675                         brt->brt_flags = IFBAF_STICKY;
2676                 else
2677                         brt->brt_flags = IFBAF_DYNAMIC;
2678
2679                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2680                 brt->brt_vlan = vlan;
2681
2682                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2683                         uma_zfree(bridge_rtnode_zone, brt);
2684                         return (error);
2685                 }
2686                 brt->brt_dst = bif;
2687                 bif->bif_addrcnt++;
2688         }
2689
2690         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2691             brt->brt_dst != bif) {
2692                 brt->brt_dst->bif_addrcnt--;
2693                 brt->brt_dst = bif;
2694                 brt->brt_dst->bif_addrcnt++;
2695         }
2696
2697         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2698                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2699         if (setflags)
2700                 brt->brt_flags = flags;
2701
2702         return (0);
2703 }
2704
2705 /*
2706  * bridge_rtlookup:
2707  *
2708  *      Lookup the destination interface for an address.
2709  */
2710 static struct ifnet *
2711 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2712 {
2713         struct bridge_rtnode *brt;
2714
2715         BRIDGE_LOCK_ASSERT(sc);
2716
2717         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2718                 return (NULL);
2719
2720         return (brt->brt_ifp);
2721 }
2722
2723 /*
2724  * bridge_rttrim:
2725  *
2726  *      Trim the routine table so that we have a number
2727  *      of routing entries less than or equal to the
2728  *      maximum number.
2729  */
2730 static void
2731 bridge_rttrim(struct bridge_softc *sc)
2732 {
2733         struct bridge_rtnode *brt, *nbrt;
2734
2735         BRIDGE_LOCK_ASSERT(sc);
2736
2737         /* Make sure we actually need to do this. */
2738         if (sc->sc_brtcnt <= sc->sc_brtmax)
2739                 return;
2740
2741         /* Force an aging cycle; this might trim enough addresses. */
2742         bridge_rtage(sc);
2743         if (sc->sc_brtcnt <= sc->sc_brtmax)
2744                 return;
2745
2746         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2747                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2748                         bridge_rtnode_destroy(sc, brt);
2749                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2750                                 return;
2751                 }
2752         }
2753 }
2754
2755 /*
2756  * bridge_timer:
2757  *
2758  *      Aging timer for the bridge.
2759  */
2760 static void
2761 bridge_timer(void *arg)
2762 {
2763         struct bridge_softc *sc = arg;
2764
2765         BRIDGE_LOCK_ASSERT(sc);
2766
2767         bridge_rtage(sc);
2768
2769         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2770                 callout_reset(&sc->sc_brcallout,
2771                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2772 }
2773
2774 /*
2775  * bridge_rtage:
2776  *
2777  *      Perform an aging cycle.
2778  */
2779 static void
2780 bridge_rtage(struct bridge_softc *sc)
2781 {
2782         struct bridge_rtnode *brt, *nbrt;
2783
2784         BRIDGE_LOCK_ASSERT(sc);
2785
2786         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2787                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2788                         if (time_uptime >= brt->brt_expire)
2789                                 bridge_rtnode_destroy(sc, brt);
2790                 }
2791         }
2792 }
2793
2794 /*
2795  * bridge_rtflush:
2796  *
2797  *      Remove all dynamic addresses from the bridge.
2798  */
2799 static void
2800 bridge_rtflush(struct bridge_softc *sc, int full)
2801 {
2802         struct bridge_rtnode *brt, *nbrt;
2803
2804         BRIDGE_LOCK_ASSERT(sc);
2805
2806         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2807                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2808                         bridge_rtnode_destroy(sc, brt);
2809         }
2810 }
2811
2812 /*
2813  * bridge_rtdaddr:
2814  *
2815  *      Remove an address from the table.
2816  */
2817 static int
2818 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2819 {
2820         struct bridge_rtnode *brt;
2821         int found = 0;
2822
2823         BRIDGE_LOCK_ASSERT(sc);
2824
2825         /*
2826          * If vlan is zero then we want to delete for all vlans so the lookup
2827          * may return more than one.
2828          */
2829         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2830                 bridge_rtnode_destroy(sc, brt);
2831                 found = 1;
2832         }
2833
2834         return (found ? 0 : ENOENT);
2835 }
2836
2837 /*
2838  * bridge_rtdelete:
2839  *
2840  *      Delete routes to a speicifc member interface.
2841  */
2842 static void
2843 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2844 {
2845         struct bridge_rtnode *brt, *nbrt;
2846
2847         BRIDGE_LOCK_ASSERT(sc);
2848
2849         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2850                 if (brt->brt_ifp == ifp && (full ||
2851                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2852                         bridge_rtnode_destroy(sc, brt);
2853         }
2854 }
2855
2856 /*
2857  * bridge_rtable_init:
2858  *
2859  *      Initialize the route table for this bridge.
2860  */
2861 static void
2862 bridge_rtable_init(struct bridge_softc *sc)
2863 {
2864         int i;
2865
2866         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2867             M_DEVBUF, M_WAITOK);
2868
2869         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2870                 LIST_INIT(&sc->sc_rthash[i]);
2871
2872         sc->sc_rthash_key = arc4random();
2873         LIST_INIT(&sc->sc_rtlist);
2874 }
2875
2876 /*
2877  * bridge_rtable_fini:
2878  *
2879  *      Deconstruct the route table for this bridge.
2880  */
2881 static void
2882 bridge_rtable_fini(struct bridge_softc *sc)
2883 {
2884
2885         KASSERT(sc->sc_brtcnt == 0,
2886             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2887         free(sc->sc_rthash, M_DEVBUF);
2888 }
2889
2890 /*
2891  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2892  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2893  */
2894 #define mix(a, b, c)                                                    \
2895 do {                                                                    \
2896         a -= b; a -= c; a ^= (c >> 13);                                 \
2897         b -= c; b -= a; b ^= (a << 8);                                  \
2898         c -= a; c -= b; c ^= (b >> 13);                                 \
2899         a -= b; a -= c; a ^= (c >> 12);                                 \
2900         b -= c; b -= a; b ^= (a << 16);                                 \
2901         c -= a; c -= b; c ^= (b >> 5);                                  \
2902         a -= b; a -= c; a ^= (c >> 3);                                  \
2903         b -= c; b -= a; b ^= (a << 10);                                 \
2904         c -= a; c -= b; c ^= (b >> 15);                                 \
2905 } while (/*CONSTCOND*/0)
2906
2907 static __inline uint32_t
2908 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2909 {
2910         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2911
2912         b += addr[5] << 8;
2913         b += addr[4];
2914         a += addr[3] << 24;
2915         a += addr[2] << 16;
2916         a += addr[1] << 8;
2917         a += addr[0];
2918
2919         mix(a, b, c);
2920
2921         return (c & BRIDGE_RTHASH_MASK);
2922 }
2923
2924 #undef mix
2925
2926 static int
2927 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2928 {
2929         int i, d;
2930
2931         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2932                 d = ((int)a[i]) - ((int)b[i]);
2933         }
2934
2935         return (d);
2936 }
2937
2938 /*
2939  * bridge_rtnode_lookup:
2940  *
2941  *      Look up a bridge route node for the specified destination. Compare the
2942  *      vlan id or if zero then just return the first match.
2943  */
2944 static struct bridge_rtnode *
2945 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2946 {
2947         struct bridge_rtnode *brt;
2948         uint32_t hash;
2949         int dir;
2950
2951         BRIDGE_LOCK_ASSERT(sc);
2952
2953         hash = bridge_rthash(sc, addr);
2954         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2955                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2956                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2957                         return (brt);
2958                 if (dir > 0)
2959                         return (NULL);
2960         }
2961
2962         return (NULL);
2963 }
2964
2965 /*
2966  * bridge_rtnode_insert:
2967  *
2968  *      Insert the specified bridge node into the route table.  We
2969  *      assume the entry is not already in the table.
2970  */
2971 static int
2972 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2973 {
2974         struct bridge_rtnode *lbrt;
2975         uint32_t hash;
2976         int dir;
2977
2978         BRIDGE_LOCK_ASSERT(sc);
2979
2980         hash = bridge_rthash(sc, brt->brt_addr);
2981
2982         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2983         if (lbrt == NULL) {
2984                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2985                 goto out;
2986         }
2987
2988         do {
2989                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2990                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2991                         return (EEXIST);
2992                 if (dir > 0) {
2993                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2994                         goto out;
2995                 }
2996                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2997                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2998                         goto out;
2999                 }
3000                 lbrt = LIST_NEXT(lbrt, brt_hash);
3001         } while (lbrt != NULL);
3002
3003 #ifdef DIAGNOSTIC
3004         panic("bridge_rtnode_insert: impossible");
3005 #endif
3006
3007 out:
3008         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3009         sc->sc_brtcnt++;
3010
3011         return (0);
3012 }
3013
3014 /*
3015  * bridge_rtnode_destroy:
3016  *
3017  *      Destroy a bridge rtnode.
3018  */
3019 static void
3020 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3021 {
3022         BRIDGE_LOCK_ASSERT(sc);
3023
3024         LIST_REMOVE(brt, brt_hash);
3025
3026         LIST_REMOVE(brt, brt_list);
3027         sc->sc_brtcnt--;
3028         brt->brt_dst->bif_addrcnt--;
3029         uma_zfree(bridge_rtnode_zone, brt);
3030 }
3031
3032 /*
3033  * bridge_rtable_expire:
3034  *
3035  *      Set the expiry time for all routes on an interface.
3036  */
3037 static void
3038 bridge_rtable_expire(struct ifnet *ifp, int age)
3039 {
3040         struct bridge_softc *sc = ifp->if_bridge;
3041         struct bridge_rtnode *brt;
3042
3043         BRIDGE_LOCK(sc);
3044
3045         /*
3046          * If the age is zero then flush, otherwise set all the expiry times to
3047          * age for the interface
3048          */
3049         if (age == 0)
3050                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3051         else {
3052                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3053                         /* Cap the expiry time to 'age' */
3054                         if (brt->brt_ifp == ifp &&
3055                             brt->brt_expire > time_uptime + age &&
3056                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3057                                 brt->brt_expire = time_uptime + age;
3058                 }
3059         }
3060         BRIDGE_UNLOCK(sc);
3061 }
3062
3063 /*
3064  * bridge_state_change:
3065  *
3066  *      Callback from the bridgestp code when a port changes states.
3067  */
3068 static void
3069 bridge_state_change(struct ifnet *ifp, int state)
3070 {
3071         struct bridge_softc *sc = ifp->if_bridge;
3072         static const char *stpstates[] = {
3073                 "disabled",
3074                 "listening",
3075                 "learning",
3076                 "forwarding",
3077                 "blocking",
3078                 "discarding"
3079         };
3080
3081         CURVNET_SET(ifp->if_vnet);
3082         if (V_log_stp)
3083                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3084                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3085         CURVNET_RESTORE();
3086 }
3087
3088 /*
3089  * Send bridge packets through pfil if they are one of the types pfil can deal
3090  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3091  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3092  * that interface.
3093  */
3094 static int
3095 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3096 {
3097         int snap, error, i, hlen;
3098         struct ether_header *eh1, eh2;
3099         struct ip *ip;
3100         struct llc llc1;
3101         u_int16_t ether_type;
3102
3103         snap = 0;
3104         error = -1;     /* Default error if not error == 0 */
3105
3106 #if 0
3107         /* we may return with the IP fields swapped, ensure its not shared */
3108         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3109 #endif
3110
3111         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3112                 return (0); /* filtering is disabled */
3113
3114         i = min((*mp)->m_pkthdr.len, max_protohdr);
3115         if ((*mp)->m_len < i) {
3116             *mp = m_pullup(*mp, i);
3117             if (*mp == NULL) {
3118                 printf("%s: m_pullup failed\n", __func__);
3119                 return (-1);
3120             }
3121         }
3122
3123         eh1 = mtod(*mp, struct ether_header *);
3124         ether_type = ntohs(eh1->ether_type);
3125
3126         /*
3127          * Check for SNAP/LLC.
3128          */
3129         if (ether_type < ETHERMTU) {
3130                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3131
3132                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3133                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3134                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3135                     llc2->llc_control == LLC_UI) {
3136                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3137                         snap = 1;
3138                 }
3139         }
3140
3141         /*
3142          * If we're trying to filter bridge traffic, don't look at anything
3143          * other than IP and ARP traffic.  If the filter doesn't understand
3144          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3145          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3146          * but of course we don't have an AppleTalk filter to begin with.
3147          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3148          * ARP traffic.)
3149          */
3150         switch (ether_type) {
3151                 case ETHERTYPE_ARP:
3152                 case ETHERTYPE_REVARP:
3153                         if (V_pfil_ipfw_arp == 0)
3154                                 return (0); /* Automatically pass */
3155                         break;
3156
3157                 case ETHERTYPE_IP:
3158 #ifdef INET6
3159                 case ETHERTYPE_IPV6:
3160 #endif /* INET6 */
3161                         break;
3162                 default:
3163                         /*
3164                          * Check to see if the user wants to pass non-ip
3165                          * packets, these will not be checked by pfil(9) and
3166                          * passed unconditionally so the default is to drop.
3167                          */
3168                         if (V_pfil_onlyip)
3169                                 goto bad;
3170         }
3171
3172         /* Run the packet through pfil before stripping link headers */
3173         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3174                         dir == PFIL_OUT && ifp != NULL) {
3175
3176                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, NULL);
3177
3178                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3179                         return (error);
3180         }
3181
3182         /* Strip off the Ethernet header and keep a copy. */
3183         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3184         m_adj(*mp, ETHER_HDR_LEN);
3185
3186         /* Strip off snap header, if present */
3187         if (snap) {
3188                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3189                 m_adj(*mp, sizeof(struct llc));
3190         }
3191
3192         /*
3193          * Check the IP header for alignment and errors
3194          */
3195         if (dir == PFIL_IN) {
3196                 switch (ether_type) {
3197                         case ETHERTYPE_IP:
3198                                 error = bridge_ip_checkbasic(mp);
3199                                 break;
3200 #ifdef INET6
3201                         case ETHERTYPE_IPV6:
3202                                 error = bridge_ip6_checkbasic(mp);
3203                                 break;
3204 #endif /* INET6 */
3205                         default:
3206                                 error = 0;
3207                 }
3208                 if (error)
3209                         goto bad;
3210         }
3211
3212         error = 0;
3213
3214         /*
3215          * Run the packet through pfil
3216          */
3217         switch (ether_type) {
3218         case ETHERTYPE_IP:
3219                 /*
3220                  * Run pfil on the member interface and the bridge, both can
3221                  * be skipped by clearing pfil_member or pfil_bridge.
3222                  *
3223                  * Keep the order:
3224                  *   in_if -> bridge_if -> out_if
3225                  */
3226                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3227                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3228                                         dir, NULL);
3229
3230                 if (*mp == NULL || error != 0) /* filter may consume */
3231                         break;
3232
3233                 if (V_pfil_member && ifp != NULL)
3234                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3235                                         dir, NULL);
3236
3237                 if (*mp == NULL || error != 0) /* filter may consume */
3238                         break;
3239
3240                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3241                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3242                                         dir, NULL);
3243
3244                 if (*mp == NULL || error != 0) /* filter may consume */
3245                         break;
3246
3247                 /* check if we need to fragment the packet */
3248                 /* bridge_fragment generates a mbuf chain of packets */
3249                 /* that already include eth headers */
3250                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3251                         i = (*mp)->m_pkthdr.len;
3252                         if (i > ifp->if_mtu) {
3253                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3254                                             &llc1);
3255                                 return (error);
3256                         }
3257                 }
3258
3259                 /* Recalculate the ip checksum. */
3260                 ip = mtod(*mp, struct ip *);
3261                 hlen = ip->ip_hl << 2;
3262                 if (hlen < sizeof(struct ip))
3263                         goto bad;
3264                 if (hlen > (*mp)->m_len) {
3265                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3266                                 goto bad;
3267                         ip = mtod(*mp, struct ip *);
3268                         if (ip == NULL)
3269                                 goto bad;
3270                 }
3271                 ip->ip_sum = 0;
3272                 if (hlen == sizeof(struct ip))
3273                         ip->ip_sum = in_cksum_hdr(ip);
3274                 else
3275                         ip->ip_sum = in_cksum(*mp, hlen);
3276
3277                 break;
3278 #ifdef INET6
3279         case ETHERTYPE_IPV6:
3280                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3281                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3282                                         dir, NULL);
3283
3284                 if (*mp == NULL || error != 0) /* filter may consume */
3285                         break;
3286
3287                 if (V_pfil_member && ifp != NULL)
3288                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3289                                         dir, NULL);
3290
3291                 if (*mp == NULL || error != 0) /* filter may consume */
3292                         break;
3293
3294                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3295                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3296                                         dir, NULL);
3297                 break;
3298 #endif
3299         default:
3300                 error = 0;
3301                 break;
3302         }
3303
3304         if (*mp == NULL)
3305                 return (error);
3306         if (error != 0)
3307                 goto bad;
3308
3309         error = -1;
3310
3311         /*
3312          * Finally, put everything back the way it was and return
3313          */
3314         if (snap) {
3315                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3316                 if (*mp == NULL)
3317                         return (error);
3318                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3319         }
3320
3321         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3322         if (*mp == NULL)
3323                 return (error);
3324         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3325
3326         return (0);
3327
3328 bad:
3329         m_freem(*mp);
3330         *mp = NULL;
3331         return (error);
3332 }
3333
3334 /*
3335  * Perform basic checks on header size since
3336  * pfil assumes ip_input has already processed
3337  * it for it.  Cut-and-pasted from ip_input.c.
3338  * Given how simple the IPv6 version is,
3339  * does the IPv4 version really need to be
3340  * this complicated?
3341  *
3342  * XXX Should we update ipstat here, or not?
3343  * XXX Right now we update ipstat but not
3344  * XXX csum_counter.
3345  */
3346 static int
3347 bridge_ip_checkbasic(struct mbuf **mp)
3348 {
3349         struct mbuf *m = *mp;
3350         struct ip *ip;
3351         int len, hlen;
3352         u_short sum;
3353
3354         if (*mp == NULL)
3355                 return (-1);
3356
3357         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3358                 if ((m = m_copyup(m, sizeof(struct ip),
3359                         (max_linkhdr + 3) & ~3)) == NULL) {
3360                         /* XXXJRT new stat, please */
3361                         KMOD_IPSTAT_INC(ips_toosmall);
3362                         goto bad;
3363                 }
3364         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3365                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3366                         KMOD_IPSTAT_INC(ips_toosmall);
3367                         goto bad;
3368                 }
3369         }
3370         ip = mtod(m, struct ip *);
3371         if (ip == NULL) goto bad;
3372
3373         if (ip->ip_v != IPVERSION) {
3374                 KMOD_IPSTAT_INC(ips_badvers);
3375                 goto bad;
3376         }
3377         hlen = ip->ip_hl << 2;
3378         if (hlen < sizeof(struct ip)) { /* minimum header length */
3379                 KMOD_IPSTAT_INC(ips_badhlen);
3380                 goto bad;
3381         }
3382         if (hlen > m->m_len) {
3383                 if ((m = m_pullup(m, hlen)) == NULL) {
3384                         KMOD_IPSTAT_INC(ips_badhlen);
3385                         goto bad;
3386                 }
3387                 ip = mtod(m, struct ip *);
3388                 if (ip == NULL) goto bad;
3389         }
3390
3391         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3392                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3393         } else {
3394                 if (hlen == sizeof(struct ip)) {
3395                         sum = in_cksum_hdr(ip);
3396                 } else {
3397                         sum = in_cksum(m, hlen);
3398                 }
3399         }
3400         if (sum) {
3401                 KMOD_IPSTAT_INC(ips_badsum);
3402                 goto bad;
3403         }
3404
3405         /* Retrieve the packet length. */
3406         len = ntohs(ip->ip_len);
3407
3408         /*
3409          * Check for additional length bogosity
3410          */
3411         if (len < hlen) {
3412                 KMOD_IPSTAT_INC(ips_badlen);
3413                 goto bad;
3414         }
3415
3416         /*
3417          * Check that the amount of data in the buffers
3418          * is as at least much as the IP header would have us expect.
3419          * Drop packet if shorter than we expect.
3420          */
3421         if (m->m_pkthdr.len < len) {
3422                 KMOD_IPSTAT_INC(ips_tooshort);
3423                 goto bad;
3424         }
3425
3426         /* Checks out, proceed */
3427         *mp = m;
3428         return (0);
3429
3430 bad:
3431         *mp = m;
3432         return (-1);
3433 }
3434
3435 #ifdef INET6
3436 /*
3437  * Same as above, but for IPv6.
3438  * Cut-and-pasted from ip6_input.c.
3439  * XXX Should we update ip6stat, or not?
3440  */
3441 static int
3442 bridge_ip6_checkbasic(struct mbuf **mp)
3443 {
3444         struct mbuf *m = *mp;
3445         struct ip6_hdr *ip6;
3446
3447         /*
3448          * If the IPv6 header is not aligned, slurp it up into a new
3449          * mbuf with space for link headers, in the event we forward
3450          * it.  Otherwise, if it is aligned, make sure the entire base
3451          * IPv6 header is in the first mbuf of the chain.
3452          */
3453         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3454                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3455                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3456                             (max_linkhdr + 3) & ~3)) == NULL) {
3457                         /* XXXJRT new stat, please */
3458                         IP6STAT_INC(ip6s_toosmall);
3459                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3460                         goto bad;
3461                 }
3462         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3463                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3464                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3465                         IP6STAT_INC(ip6s_toosmall);
3466                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3467                         goto bad;
3468                 }
3469         }
3470
3471         ip6 = mtod(m, struct ip6_hdr *);
3472
3473         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3474                 IP6STAT_INC(ip6s_badvers);
3475                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3476                 goto bad;
3477         }
3478
3479         /* Checks out, proceed */
3480         *mp = m;
3481         return (0);
3482
3483 bad:
3484         *mp = m;
3485         return (-1);
3486 }
3487 #endif /* INET6 */
3488
3489 /*
3490  * bridge_fragment:
3491  *
3492  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3493  */
3494 static int
3495 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3496     int snap, struct llc *llc)
3497 {
3498         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3499         struct ip *ip;
3500         int error = -1;
3501
3502         if (m->m_len < sizeof(struct ip) &&
3503             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3504                 goto dropit;
3505         ip = mtod(m, struct ip *);
3506
3507         m->m_pkthdr.csum_flags |= CSUM_IP;
3508         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3509         if (error)
3510                 goto dropit;
3511
3512         /*
3513          * Walk the chain and re-add the Ethernet header for
3514          * each mbuf packet.
3515          */
3516         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3517                 nextpkt = mcur->m_nextpkt;
3518                 mcur->m_nextpkt = NULL;
3519                 if (snap) {
3520                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3521                         if (mcur == NULL) {
3522                                 error = ENOBUFS;
3523                                 if (mprev != NULL)
3524                                         mprev->m_nextpkt = nextpkt;
3525                                 goto dropit;
3526                         }
3527                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3528                 }
3529
3530                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3531                 if (mcur == NULL) {
3532                         error = ENOBUFS;
3533                         if (mprev != NULL)
3534                                 mprev->m_nextpkt = nextpkt;
3535                         goto dropit;
3536                 }
3537                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3538
3539                 /*
3540                  * The previous two M_PREPEND could have inserted one or two
3541                  * mbufs in front so we have to update the previous packet's
3542                  * m_nextpkt.
3543                  */
3544                 mcur->m_nextpkt = nextpkt;
3545                 if (mprev != NULL)
3546                         mprev->m_nextpkt = mcur;
3547                 else {
3548                         /* The first mbuf in the original chain needs to be
3549                          * updated. */
3550                         *mp = mcur;
3551                 }
3552                 mprev = mcur;
3553         }
3554
3555         KMOD_IPSTAT_INC(ips_fragmented);
3556         return (error);
3557
3558 dropit:
3559         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3560                 m = mcur->m_nextpkt;
3561                 m_freem(mcur);
3562         }
3563         return (error);
3564 }
3565
3566 static void
3567 bridge_linkstate(struct ifnet *ifp)
3568 {
3569         struct bridge_softc *sc = ifp->if_bridge;
3570         struct bridge_iflist *bif;
3571
3572         BRIDGE_LOCK(sc);
3573         bif = bridge_lookup_member_if(sc, ifp);
3574         if (bif == NULL) {
3575                 BRIDGE_UNLOCK(sc);
3576                 return;
3577         }
3578         bridge_linkcheck(sc);
3579         BRIDGE_UNLOCK(sc);
3580
3581         bstp_linkstate(&bif->bif_stp);
3582 }
3583
3584 static void
3585 bridge_linkcheck(struct bridge_softc *sc)
3586 {
3587         struct bridge_iflist *bif;
3588         int new_link, hasls;
3589
3590         BRIDGE_LOCK_ASSERT(sc);
3591         new_link = LINK_STATE_DOWN;
3592         hasls = 0;
3593         /* Our link is considered up if at least one of our ports is active */
3594         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3595                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3596                         hasls++;
3597                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3598                         new_link = LINK_STATE_UP;
3599                         break;
3600                 }
3601         }
3602         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3603                 /* If no interfaces support link-state then we default to up */
3604                 new_link = LINK_STATE_UP;
3605         }
3606         if_link_state_change(sc->sc_ifp, new_link);
3607 }