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         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 913                 enabled = bif->bif_ifp->if_capenable;
 914                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 915                 /* strip off mask bits and enable them again if allowed */
 916                 enabled &= ~BRIDGE_IFCAPS_MASK;
 917                 enabled |= mask;
 918                 bridge_set_ifcap(sc, bif, enabled);
 919         }
 920
 921 }
 922
 923 static void
 924 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 925 {
 926         struct ifnet *ifp = bif->bif_ifp;
 927         struct ifreq ifr;
 928         int error;
 929
 930         bzero(&ifr, sizeof(ifr));
 931         ifr.ifr_reqcap = set;
 932
 933         if (ifp->if_capenable != set) {
 934                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 935                 if (error)
 936                         if_printf(sc->sc_ifp,
 937                             "error setting interface capabilities on %s\n",
 938                             ifp->if_xname);
 939         }
 940 }
 941
 942 /*
 943  * bridge_lookup_member:
 944  *
 945  *      Lookup a bridge member interface.
 946  */
 947 static struct bridge_iflist *
 948 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 949 {
 950         struct bridge_iflist *bif;
 951         struct ifnet *ifp;
 952
 953         BRIDGE_LOCK_ASSERT(sc);
 954
 955         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 956                 ifp = bif->bif_ifp;
 957                 if (strcmp(ifp->if_xname, name) == 0)
 958                         return (bif);
 959         }
 960
 961         return (NULL);
 962 }
 963
 964 /*
 965  * bridge_lookup_member_if:
 966  *
 967  *      Lookup a bridge member interface by ifnet*.
 968  */
 969 static struct bridge_iflist *
 970 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 971 {
 972         struct bridge_iflist *bif;
 973
 974         BRIDGE_LOCK_ASSERT(sc);
 975
 976         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 977                 if (bif->bif_ifp == member_ifp)
 978                         return (bif);
 979         }
 980
 981         return (NULL);
 982 }
 983
 984 /*
 985  * bridge_delete_member:
 986  *
 987  *      Delete the specified member interface.
 988  */
 989 static void
 990 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 991     int gone)
 992 {
 993         struct ifnet *ifs = bif->bif_ifp;
 994         struct ifnet *fif = NULL;
 995
 996         BRIDGE_LOCK_ASSERT(sc);
 997
 998         if (bif->bif_flags & IFBIF_STP)
 999                 bstp_disable(&bif->bif_stp);
1000
1001         ifs->if_bridge = NULL;
1002         BRIDGE_XLOCK(sc);
1003         LIST_REMOVE(bif, bif_next);
1004         BRIDGE_XDROP(sc);
1005
1006         /*
1007          * If removing the interface that gave the bridge its mac address, set
1008          * the mac address of the bridge to the address of the next member, or
1009          * to its default address if no members are left.
1010          */
1011         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1012                 if (LIST_EMPTY(&sc->sc_iflist)) {
1013                         bcopy(sc->sc_defaddr,
1014                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1015                         sc->sc_ifaddr = NULL;
1016                 } else {
1017                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
1018                         bcopy(IF_LLADDR(fif),
1019                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1020                         sc->sc_ifaddr = fif;
1021                 }
1022                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1023         }
1024
1025         bridge_linkcheck(sc);
1026         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1027         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1028         KASSERT(bif->bif_addrcnt == 0,
1029             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1030
1031         BRIDGE_UNLOCK(sc);
1032         if (!gone) {
1033                 switch (ifs->if_type) {
1034                 case IFT_ETHER:
1035                 case IFT_L2VLAN:
1036                         /*
1037                          * Take the interface out of promiscuous mode, but only
1038                          * if it was promiscuous in the first place. It might
1039                          * not be if we're in the bridge_ioctl_add() error path.
1040                          */
1041                         if (ifs->if_flags & IFF_PROMISC)
1042                                 (void) ifpromisc(ifs, 0);
1043                         break;
1044
1045                 case IFT_GIF:
1046                         break;
1047
1048                 default:
1049 #ifdef DIAGNOSTIC
1050                         panic("bridge_delete_member: impossible");
1051 #endif
1052                         break;
1053                 }
1054                 /* reneable any interface capabilities */
1055                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1056         }
1057         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1058         BRIDGE_LOCK(sc);
1059         free(bif, M_DEVBUF);
1060 }
1061
1062 /*
1063  * bridge_delete_span:
1064  *
1065  *      Delete the specified span interface.
1066  */
1067 static void
1068 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1069 {
1070         BRIDGE_LOCK_ASSERT(sc);
1071
1072         KASSERT(bif->bif_ifp->if_bridge == NULL,
1073             ("%s: not a span interface", __func__));
1074
1075         LIST_REMOVE(bif, bif_next);
1076         free(bif, M_DEVBUF);
1077 }
1078
1079 static int
1080 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1081 {
1082         struct ifbreq *req = arg;
1083         struct bridge_iflist *bif = NULL;
1084         struct ifnet *ifs;
1085         int error = 0;
1086
1087         ifs = ifunit(req->ifbr_ifsname);
1088         if (ifs == NULL)
1089                 return (ENOENT);
1090         if (ifs->if_ioctl == NULL)      /* must be supported */
1091                 return (EINVAL);
1092
1093         /* If it's in the span list, it can't be a member. */
1094         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1095                 if (ifs == bif->bif_ifp)
1096                         return (EBUSY);
1097
1098         if (ifs->if_bridge == sc)
1099                 return (EEXIST);
1100
1101         if (ifs->if_bridge != NULL)
1102                 return (EBUSY);
1103
1104         switch (ifs->if_type) {
1105         case IFT_ETHER:
1106         case IFT_L2VLAN:
1107         case IFT_GIF:
1108                 /* permitted interface types */
1109                 break;
1110         default:
1111                 return (EINVAL);
1112         }
1113
1114 #ifdef INET6
1115         /*
1116          * Two valid inet6 addresses with link-local scope must not be
1117          * on the parent interface and the member interfaces at the
1118          * same time.  This restriction is needed to prevent violation
1119          * of link-local scope zone.  Attempts to add a member
1120          * interface which has inet6 addresses when the parent has
1121          * inet6 triggers removal of all inet6 addresses on the member
1122          * interface.
1123          */
1124
1125         /* Check if the parent interface has a link-local scope addr. */
1126         if (V_allow_llz_overlap == 0 &&
1127             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1128                 /*
1129                  * If any, remove all inet6 addresses from the member
1130                  * interfaces.
1131                  */
1132                 BRIDGE_XLOCK(sc);
1133                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1134                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1135                                 BRIDGE_UNLOCK(sc);
1136                                 in6_ifdetach(bif->bif_ifp);
1137                                 BRIDGE_LOCK(sc);
1138                                 if_printf(sc->sc_ifp,
1139                                     "IPv6 addresses on %s have been removed "
1140                                     "before adding it as a member to prevent "
1141                                     "IPv6 address scope violation.\n",
1142                                     bif->bif_ifp->if_xname);
1143                         }
1144                 }
1145                 BRIDGE_XDROP(sc);
1146                 if (in6ifa_llaonifp(ifs)) {
1147                         BRIDGE_UNLOCK(sc);
1148                         in6_ifdetach(ifs);
1149                         BRIDGE_LOCK(sc);
1150                         if_printf(sc->sc_ifp,
1151                             "IPv6 addresses on %s have been removed "
1152                             "before adding it as a member to prevent "
1153                             "IPv6 address scope violation.\n",
1154                             ifs->if_xname);
1155                 }
1156         }
1157 #endif
1158         /* Allow the first Ethernet member to define the MTU */
1159         if (LIST_EMPTY(&sc->sc_iflist))
1160                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1161         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1162                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1163                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1164                 return (EINVAL);
1165         }
1166
1167         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1168         if (bif == NULL)
1169                 return (ENOMEM);
1170
1171         bif->bif_ifp = ifs;
1172         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1173         bif->bif_savedcaps = ifs->if_capenable;
1174
1175         /*
1176          * Assign the interface's MAC address to the bridge if it's the first
1177          * member and the MAC address of the bridge has not been changed from
1178          * the default randomly generated one.
1179          */
1180         if (V_bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1181             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1182                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1183                 sc->sc_ifaddr = ifs;
1184                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1185         }
1186
1187         ifs->if_bridge = sc;
1188         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1189         /*
1190          * XXX: XLOCK HERE!?!
1191          *
1192          * NOTE: insert_***HEAD*** should be safe for the traversals.
1193          */
1194         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1195
1196         /* Set interface capabilities to the intersection set of all members */
1197         bridge_mutecaps(sc);
1198         bridge_linkcheck(sc);
1199
1200         /* Place the interface into promiscuous mode */
1201         switch (ifs->if_type) {
1202                 case IFT_ETHER:
1203                 case IFT_L2VLAN:
1204                         BRIDGE_UNLOCK(sc);
1205                         error = ifpromisc(ifs, 1);
1206                         BRIDGE_LOCK(sc);
1207                         break;
1208         }
1209
1210         if (error)
1211                 bridge_delete_member(sc, bif, 0);
1212         return (error);
1213 }
1214
1215 static int
1216 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1217 {
1218         struct ifbreq *req = arg;
1219         struct bridge_iflist *bif;
1220
1221         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1222         if (bif == NULL)
1223                 return (ENOENT);
1224
1225         bridge_delete_member(sc, bif, 0);
1226
1227         return (0);
1228 }
1229
1230 static int
1231 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1232 {
1233         struct ifbreq *req = arg;
1234         struct bridge_iflist *bif;
1235         struct bstp_port *bp;
1236
1237         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1238         if (bif == NULL)
1239                 return (ENOENT);
1240
1241         bp = &bif->bif_stp;
1242         req->ifbr_ifsflags = bif->bif_flags;
1243         req->ifbr_state = bp->bp_state;
1244         req->ifbr_priority = bp->bp_priority;
1245         req->ifbr_path_cost = bp->bp_path_cost;
1246         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1247         req->ifbr_proto = bp->bp_protover;
1248         req->ifbr_role = bp->bp_role;
1249         req->ifbr_stpflags = bp->bp_flags;
1250         req->ifbr_addrcnt = bif->bif_addrcnt;
1251         req->ifbr_addrmax = bif->bif_addrmax;
1252         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1253
1254         /* Copy STP state options as flags */
1255         if (bp->bp_operedge)
1256                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1257         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1258                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1259         if (bp->bp_ptp_link)
1260                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1261         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1262                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1263         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1264                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1265         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1266                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1267         return (0);
1268 }
1269
1270 static int
1271 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1272 {
1273         struct ifbreq *req = arg;
1274         struct bridge_iflist *bif;
1275         struct bstp_port *bp;
1276         int error;
1277
1278         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1279         if (bif == NULL)
1280                 return (ENOENT);
1281         bp = &bif->bif_stp;
1282
1283         if (req->ifbr_ifsflags & IFBIF_SPAN)
1284                 /* SPAN is readonly */
1285                 return (EINVAL);
1286
1287         if (req->ifbr_ifsflags & IFBIF_STP) {
1288                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1289                         error = bstp_enable(&bif->bif_stp);
1290                         if (error)
1291                                 return (error);
1292                 }
1293         } else {
1294                 if ((bif->bif_flags & IFBIF_STP) != 0)
1295                         bstp_disable(&bif->bif_stp);
1296         }
1297
1298         /* Pass on STP flags */
1299         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1300         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1301         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1302         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1303
1304         /* Save the bits relating to the bridge */
1305         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1306
1307         return (0);
1308 }
1309
1310 static int
1311 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1312 {
1313         struct ifbrparam *param = arg;
1314
1315         sc->sc_brtmax = param->ifbrp_csize;
1316         bridge_rttrim(sc);
1317
1318         return (0);
1319 }
1320
1321 static int
1322 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1323 {
1324         struct ifbrparam *param = arg;
1325
1326         param->ifbrp_csize = sc->sc_brtmax;
1327
1328         return (0);
1329 }
1330
1331 static int
1332 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1333 {
1334         struct ifbifconf *bifc = arg;
1335         struct bridge_iflist *bif;
1336         struct ifbreq breq;
1337         char *buf, *outbuf;
1338         int count, buflen, len, error = 0;
1339
1340         count = 0;
1341         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1342                 count++;
1343         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1344                 count++;
1345
1346         buflen = sizeof(breq) * count;
1347         if (bifc->ifbic_len == 0) {
1348                 bifc->ifbic_len = buflen;
1349                 return (0);
1350         }
1351         BRIDGE_UNLOCK(sc);
1352         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1353         BRIDGE_LOCK(sc);
1354
1355         count = 0;
1356         buf = outbuf;
1357         len = min(bifc->ifbic_len, buflen);
1358         bzero(&breq, sizeof(breq));
1359         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1360                 if (len < sizeof(breq))
1361                         break;
1362
1363                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1364                     sizeof(breq.ifbr_ifsname));
1365                 /* Fill in the ifbreq structure */
1366                 error = bridge_ioctl_gifflags(sc, &breq);
1367                 if (error)
1368                         break;
1369                 memcpy(buf, &breq, sizeof(breq));
1370                 count++;
1371                 buf += sizeof(breq);
1372                 len -= sizeof(breq);
1373         }
1374         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1375                 if (len < sizeof(breq))
1376                         break;
1377
1378                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1379                     sizeof(breq.ifbr_ifsname));
1380                 breq.ifbr_ifsflags = bif->bif_flags;
1381                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1382                 memcpy(buf, &breq, sizeof(breq));
1383                 count++;
1384                 buf += sizeof(breq);
1385                 len -= sizeof(breq);
1386         }
1387
1388         BRIDGE_UNLOCK(sc);
1389         bifc->ifbic_len = sizeof(breq) * count;
1390         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1391         BRIDGE_LOCK(sc);
1392         free(outbuf, M_TEMP);
1393         return (error);
1394 }
1395
1396 static int
1397 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1398 {
1399         struct ifbaconf *bac = arg;
1400         struct bridge_rtnode *brt;
1401         struct ifbareq bareq;
1402         char *buf, *outbuf;
1403         int count, buflen, len, error = 0;
1404
1405         if (bac->ifbac_len == 0)
1406                 return (0);
1407
1408         count = 0;
1409         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1410                 count++;
1411         buflen = sizeof(bareq) * count;
1412
1413         BRIDGE_UNLOCK(sc);
1414         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1415         BRIDGE_LOCK(sc);
1416
1417         count = 0;
1418         buf = outbuf;
1419         len = min(bac->ifbac_len, buflen);
1420         bzero(&bareq, sizeof(bareq));
1421         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1422                 if (len < sizeof(bareq))
1423                         goto out;
1424                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1425                     sizeof(bareq.ifba_ifsname));
1426                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1427                 bareq.ifba_vlan = brt->brt_vlan;
1428                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1429                                 time_uptime < brt->brt_expire)
1430                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1431                 else
1432                         bareq.ifba_expire = 0;
1433                 bareq.ifba_flags = brt->brt_flags;
1434
1435                 memcpy(buf, &bareq, sizeof(bareq));
1436                 count++;
1437                 buf += sizeof(bareq);
1438                 len -= sizeof(bareq);
1439         }
1440 out:
1441         BRIDGE_UNLOCK(sc);
1442         bac->ifbac_len = sizeof(bareq) * count;
1443         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1444         BRIDGE_LOCK(sc);
1445         free(outbuf, M_TEMP);
1446         return (error);
1447 }
1448
1449 static int
1450 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1451 {
1452         struct ifbareq *req = arg;
1453         struct bridge_iflist *bif;
1454         int error;
1455
1456         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1457         if (bif == NULL)
1458                 return (ENOENT);
1459
1460         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1461             req->ifba_flags);
1462
1463         return (error);
1464 }
1465
1466 static int
1467 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1468 {
1469         struct ifbrparam *param = arg;
1470
1471         sc->sc_brttimeout = param->ifbrp_ctime;
1472         return (0);
1473 }
1474
1475 static int
1476 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1477 {
1478         struct ifbrparam *param = arg;
1479
1480         param->ifbrp_ctime = sc->sc_brttimeout;
1481         return (0);
1482 }
1483
1484 static int
1485 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1486 {
1487         struct ifbareq *req = arg;
1488
1489         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1490 }
1491
1492 static int
1493 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1494 {
1495         struct ifbreq *req = arg;
1496
1497         bridge_rtflush(sc, req->ifbr_ifsflags);
1498         return (0);
1499 }
1500
1501 static int
1502 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1503 {
1504         struct ifbrparam *param = arg;
1505         struct bstp_state *bs = &sc->sc_stp;
1506
1507         param->ifbrp_prio = bs->bs_bridge_priority;
1508         return (0);
1509 }
1510
1511 static int
1512 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1513 {
1514         struct ifbrparam *param = arg;
1515
1516         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1517 }
1518
1519 static int
1520 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1521 {
1522         struct ifbrparam *param = arg;
1523         struct bstp_state *bs = &sc->sc_stp;
1524
1525         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1526         return (0);
1527 }
1528
1529 static int
1530 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1531 {
1532         struct ifbrparam *param = arg;
1533
1534         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1535 }
1536
1537 static int
1538 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1539 {
1540         struct ifbrparam *param = arg;
1541         struct bstp_state *bs = &sc->sc_stp;
1542
1543         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1544         return (0);
1545 }
1546
1547 static int
1548 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1549 {
1550         struct ifbrparam *param = arg;
1551
1552         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1553 }
1554
1555 static int
1556 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1557 {
1558         struct ifbrparam *param = arg;
1559         struct bstp_state *bs = &sc->sc_stp;
1560
1561         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1562         return (0);
1563 }
1564
1565 static int
1566 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1567 {
1568         struct ifbrparam *param = arg;
1569
1570         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1571 }
1572
1573 static int
1574 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1575 {
1576         struct ifbreq *req = arg;
1577         struct bridge_iflist *bif;
1578
1579         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1580         if (bif == NULL)
1581                 return (ENOENT);
1582
1583         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1584 }
1585
1586 static int
1587 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1588 {
1589         struct ifbreq *req = arg;
1590         struct bridge_iflist *bif;
1591
1592         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1593         if (bif == NULL)
1594                 return (ENOENT);
1595
1596         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1597 }
1598
1599 static int
1600 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1601 {
1602         struct ifbreq *req = arg;
1603         struct bridge_iflist *bif;
1604
1605         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1606         if (bif == NULL)
1607                 return (ENOENT);
1608
1609         bif->bif_addrmax = req->ifbr_addrmax;
1610         return (0);
1611 }
1612
1613 static int
1614 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1615 {
1616         struct ifbreq *req = arg;
1617         struct bridge_iflist *bif = NULL;
1618         struct ifnet *ifs;
1619
1620         ifs = ifunit(req->ifbr_ifsname);
1621         if (ifs == NULL)
1622                 return (ENOENT);
1623
1624         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1625                 if (ifs == bif->bif_ifp)
1626                         return (EBUSY);
1627
1628         if (ifs->if_bridge != NULL)
1629                 return (EBUSY);
1630
1631         switch (ifs->if_type) {
1632                 case IFT_ETHER:
1633                 case IFT_GIF:
1634                 case IFT_L2VLAN:
1635                         break;
1636                 default:
1637                         return (EINVAL);
1638         }
1639
1640         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1641         if (bif == NULL)
1642                 return (ENOMEM);
1643
1644         bif->bif_ifp = ifs;
1645         bif->bif_flags = IFBIF_SPAN;
1646
1647         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1648
1649         return (0);
1650 }
1651
1652 static int
1653 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1654 {
1655         struct ifbreq *req = arg;
1656         struct bridge_iflist *bif;
1657         struct ifnet *ifs;
1658
1659         ifs = ifunit(req->ifbr_ifsname);
1660         if (ifs == NULL)
1661                 return (ENOENT);
1662
1663         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1664                 if (ifs == bif->bif_ifp)
1665                         break;
1666
1667         if (bif == NULL)
1668                 return (ENOENT);
1669
1670         bridge_delete_span(sc, bif);
1671
1672         return (0);
1673 }
1674
1675 static int
1676 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1677 {
1678         struct ifbropreq *req = arg;
1679         struct bstp_state *bs = &sc->sc_stp;
1680         struct bstp_port *root_port;
1681
1682         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1683         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1684         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1685
1686         root_port = bs->bs_root_port;
1687         if (root_port == NULL)
1688                 req->ifbop_root_port = 0;
1689         else
1690                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1691
1692         req->ifbop_holdcount = bs->bs_txholdcount;
1693         req->ifbop_priority = bs->bs_bridge_priority;
1694         req->ifbop_protocol = bs->bs_protover;
1695         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1696         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1697         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1698         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1699         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1700         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1701
1702         return (0);
1703 }
1704
1705 static int
1706 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1707 {
1708         struct ifbrparam *param = arg;
1709
1710         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1711         return (0);
1712 }
1713
1714 static int
1715 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1716 {
1717         struct ifbpstpconf *bifstp = arg;
1718         struct bridge_iflist *bif;
1719         struct bstp_port *bp;
1720         struct ifbpstpreq bpreq;
1721         char *buf, *outbuf;
1722         int count, buflen, len, error = 0;
1723
1724         count = 0;
1725         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1726                 if ((bif->bif_flags & IFBIF_STP) != 0)
1727                         count++;
1728         }
1729
1730         buflen = sizeof(bpreq) * count;
1731         if (bifstp->ifbpstp_len == 0) {
1732                 bifstp->ifbpstp_len = buflen;
1733                 return (0);
1734         }
1735
1736         BRIDGE_UNLOCK(sc);
1737         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1738         BRIDGE_LOCK(sc);
1739
1740         count = 0;
1741         buf = outbuf;
1742         len = min(bifstp->ifbpstp_len, buflen);
1743         bzero(&bpreq, sizeof(bpreq));
1744         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1745                 if (len < sizeof(bpreq))
1746                         break;
1747
1748                 if ((bif->bif_flags & IFBIF_STP) == 0)
1749                         continue;
1750
1751                 bp = &bif->bif_stp;
1752                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1753                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1754                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1755                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1756                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1757                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1758
1759                 memcpy(buf, &bpreq, sizeof(bpreq));
1760                 count++;
1761                 buf += sizeof(bpreq);
1762                 len -= sizeof(bpreq);
1763         }
1764
1765         BRIDGE_UNLOCK(sc);
1766         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1767         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1768         BRIDGE_LOCK(sc);
1769         free(outbuf, M_TEMP);
1770         return (error);
1771 }
1772
1773 static int
1774 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1775 {
1776         struct ifbrparam *param = arg;
1777
1778         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1779 }
1780
1781 static int
1782 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1783 {
1784         struct ifbrparam *param = arg;
1785
1786         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1787 }
1788
1789 /*
1790  * bridge_ifdetach:
1791  *
1792  *      Detach an interface from a bridge.  Called when a member
1793  *      interface is detaching.
1794  */
1795 static void
1796 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1797 {
1798         struct bridge_softc *sc = ifp->if_bridge;
1799         struct bridge_iflist *bif;
1800
1801         if (ifp->if_flags & IFF_RENAMING)
1802                 return;
1803         if (V_bridge_cloner == NULL) {
1804                 /*
1805                  * This detach handler can be called after
1806                  * vnet_bridge_uninit().  Just return in that case.
1807                  */
1808                 return;
1809         }
1810         /* Check if the interface is a bridge member */
1811         if (sc != NULL) {
1812                 BRIDGE_LOCK(sc);
1813
1814                 bif = bridge_lookup_member_if(sc, ifp);
1815                 if (bif != NULL)
1816                         bridge_delete_member(sc, bif, 1);
1817
1818                 BRIDGE_UNLOCK(sc);
1819                 return;
1820         }
1821
1822         /* Check if the interface is a span port */
1823         BRIDGE_LIST_LOCK();
1824         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1825                 BRIDGE_LOCK(sc);
1826                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1827                         if (ifp == bif->bif_ifp) {
1828                                 bridge_delete_span(sc, bif);
1829                                 break;
1830                         }
1831
1832                 BRIDGE_UNLOCK(sc);
1833         }
1834         BRIDGE_LIST_UNLOCK();
1835 }
1836
1837 /*
1838  * bridge_init:
1839  *
1840  *      Initialize a bridge interface.
1841  */
1842 static void
1843 bridge_init(void *xsc)
1844 {
1845         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1846         struct ifnet *ifp = sc->sc_ifp;
1847
1848         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1849                 return;
1850
1851         BRIDGE_LOCK(sc);
1852         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1853             bridge_timer, sc);
1854
1855         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1856         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1857
1858         BRIDGE_UNLOCK(sc);
1859 }
1860
1861 /*
1862  * bridge_stop:
1863  *
1864  *      Stop the bridge interface.
1865  */
1866 static void
1867 bridge_stop(struct ifnet *ifp, int disable)
1868 {
1869         struct bridge_softc *sc = ifp->if_softc;
1870
1871         BRIDGE_LOCK_ASSERT(sc);
1872
1873         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1874                 return;
1875
1876         callout_stop(&sc->sc_brcallout);
1877         bstp_stop(&sc->sc_stp);
1878
1879         bridge_rtflush(sc, IFBF_FLUSHDYN);
1880
1881         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1882 }
1883
1884 /*
1885  * bridge_enqueue:
1886  *
1887  *      Enqueue a packet on a bridge member interface.
1888  *
1889  */
1890 static int
1891 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1892 {
1893         int len, err = 0;
1894         short mflags;
1895         struct mbuf *m0;
1896
1897         /* We may be sending a fragment so traverse the mbuf */
1898         for (; m; m = m0) {
1899                 m0 = m->m_nextpkt;
1900                 m->m_nextpkt = NULL;
1901                 len = m->m_pkthdr.len;
1902                 mflags = m->m_flags;
1903
1904                 /*
1905                  * If underlying interface can not do VLAN tag insertion itself
1906                  * then attach a packet tag that holds it.
1907                  */
1908                 if ((m->m_flags & M_VLANTAG) &&
1909                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1910                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1911                         if (m == NULL) {
1912                                 if_printf(dst_ifp,
1913                                     "unable to prepend VLAN header\n");
1914                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
1915                                 continue;
1916                         }
1917                         m->m_flags &= ~M_VLANTAG;
1918                 }
1919
1920                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
1921                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1922                         m_freem(m0);
1923                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
1924                         break;
1925                 }
1926
1927                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
1928                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
1929                 if (mflags & M_MCAST)
1930                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
1931         }
1932
1933         return (err);
1934 }
1935
1936 /*
1937  * bridge_dummynet:
1938  *
1939  *      Receive a queued packet from dummynet and pass it on to the output
1940  *      interface.
1941  *
1942  *      The mbuf has the Ethernet header already attached.
1943  */
1944 static void
1945 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1946 {
1947         struct bridge_softc *sc;
1948
1949         sc = ifp->if_bridge;
1950
1951         /*
1952          * The packet didnt originate from a member interface. This should only
1953          * ever happen if a member interface is removed while packets are
1954          * queued for it.
1955          */
1956         if (sc == NULL) {
1957                 m_freem(m);
1958                 return;
1959         }
1960
1961         if (PFIL_HOOKED(&V_inet_pfil_hook)
1962 #ifdef INET6
1963             || PFIL_HOOKED(&V_inet6_pfil_hook)
1964 #endif
1965             ) {
1966                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1967                         return;
1968                 if (m == NULL)
1969                         return;
1970         }
1971
1972         bridge_enqueue(sc, ifp, m);
1973 }
1974
1975 /*
1976  * bridge_output:
1977  *
1978  *      Send output from a bridge member interface.  This
1979  *      performs the bridging function for locally originated
1980  *      packets.
1981  *
1982  *      The mbuf has the Ethernet header already attached.  We must
1983  *      enqueue or free the mbuf before returning.
1984  */
1985 static int
1986 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1987     struct rtentry *rt)
1988 {
1989         struct ether_header *eh;
1990         struct ifnet *dst_if;
1991         struct bridge_softc *sc;
1992         uint16_t vlan;
1993
1994         if (m->m_len < ETHER_HDR_LEN) {
1995                 m = m_pullup(m, ETHER_HDR_LEN);
1996                 if (m == NULL)
1997                         return (0);
1998         }
1999
2000         eh = mtod(m, struct ether_header *);
2001         sc = ifp->if_bridge;
2002         vlan = VLANTAGOF(m);
2003
2004         BRIDGE_LOCK(sc);
2005
2006         /*
2007          * If bridge is down, but the original output interface is up,
2008          * go ahead and send out that interface.  Otherwise, the packet
2009          * is dropped below.
2010          */
2011         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2012                 dst_if = ifp;
2013                 goto sendunicast;
2014         }
2015
2016         /*
2017          * If the packet is a multicast, or we don't know a better way to
2018          * get there, send to all interfaces.
2019          */
2020         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2021                 dst_if = NULL;
2022         else
2023                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2024         if (dst_if == NULL) {
2025                 struct bridge_iflist *bif;
2026                 struct mbuf *mc;
2027                 int error = 0, used = 0;
2028
2029                 bridge_span(sc, m);
2030
2031                 BRIDGE_LOCK2REF(sc, error);
2032                 if (error) {
2033                         m_freem(m);
2034                         return (0);
2035                 }
2036
2037                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2038                         dst_if = bif->bif_ifp;
2039
2040                         if (dst_if->if_type == IFT_GIF)
2041                                 continue;
2042                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2043                                 continue;
2044
2045                         /*
2046                          * If this is not the original output interface,
2047                          * and the interface is participating in spanning
2048                          * tree, make sure the port is in a state that
2049                          * allows forwarding.
2050                          */
2051                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2052                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2053                                 continue;
2054
2055                         if (LIST_NEXT(bif, bif_next) == NULL) {
2056                                 used = 1;
2057                                 mc = m;
2058                         } else {
2059                                 mc = m_copypacket(m, M_NOWAIT);
2060                                 if (mc == NULL) {
2061                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2062                                         continue;
2063                                 }
2064                         }
2065
2066                         bridge_enqueue(sc, dst_if, mc);
2067                 }
2068                 if (used == 0)
2069                         m_freem(m);
2070                 BRIDGE_UNREF(sc);
2071                 return (0);
2072         }
2073
2074 sendunicast:
2075         /*
2076          * XXX Spanning tree consideration here?
2077          */
2078
2079         bridge_span(sc, m);
2080         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2081                 m_freem(m);
2082                 BRIDGE_UNLOCK(sc);
2083                 return (0);
2084         }
2085
2086         BRIDGE_UNLOCK(sc);
2087         bridge_enqueue(sc, dst_if, m);
2088         return (0);
2089 }
2090
2091 /*
2092  * bridge_transmit:
2093  *
2094  *      Do output on a bridge.
2095  *
2096  */
2097 static int
2098 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2099 {
2100         struct bridge_softc *sc;
2101         struct ether_header *eh;
2102         struct ifnet *dst_if;
2103         int error = 0;
2104
2105         sc = ifp->if_softc;
2106
2107         ETHER_BPF_MTAP(ifp, m);
2108
2109         eh = mtod(m, struct ether_header *);
2110
2111         BRIDGE_LOCK(sc);
2112         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2113             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2114                 BRIDGE_UNLOCK(sc);
2115                 error = bridge_enqueue(sc, dst_if, m);
2116         } else
2117                 bridge_broadcast(sc, ifp, m, 0);
2118
2119         return (error);
2120 }
2121
2122 /*
2123  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2124  */
2125 static void
2126 bridge_qflush(struct ifnet *ifp __unused)
2127 {
2128 }
2129
2130 /*
2131  * bridge_forward:
2132  *
2133  *      The forwarding function of the bridge.
2134  *
2135  *      NOTE: Releases the lock on return.
2136  */
2137 static void
2138 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2139     struct mbuf *m)
2140 {
2141         struct bridge_iflist *dbif;
2142         struct ifnet *src_if, *dst_if, *ifp;
2143         struct ether_header *eh;
2144         uint16_t vlan;
2145         uint8_t *dst;
2146         int error;
2147
2148         src_if = m->m_pkthdr.rcvif;
2149         ifp = sc->sc_ifp;
2150
2151         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2152         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2153         vlan = VLANTAGOF(m);
2154
2155         if ((sbif->bif_flags & IFBIF_STP) &&
2156             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2157                 goto drop;
2158
2159         eh = mtod(m, struct ether_header *);
2160         dst = eh->ether_dhost;
2161
2162         /* If the interface is learning, record the address. */
2163         if (sbif->bif_flags & IFBIF_LEARNING) {
2164                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2165                     sbif, 0, IFBAF_DYNAMIC);
2166                 /*
2167                  * If the interface has addresses limits then deny any source
2168                  * that is not in the cache.
2169                  */
2170                 if (error && sbif->bif_addrmax)
2171                         goto drop;
2172         }
2173
2174         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2175             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2176                 goto drop;
2177
2178         /*
2179          * At this point, the port either doesn't participate
2180          * in spanning tree or it is in the forwarding state.
2181          */
2182
2183         /*
2184          * If the packet is unicast, destined for someone on
2185          * "this" side of the bridge, drop it.
2186          */
2187         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2188                 dst_if = bridge_rtlookup(sc, dst, vlan);
2189                 if (src_if == dst_if)
2190                         goto drop;
2191         } else {
2192                 /*
2193                  * Check if its a reserved multicast address, any address
2194                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2195                  * bridge.
2196                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2197                  */
2198                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2199                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2200                     dst[4] == 0x00 && dst[5] <= 0x0f)
2201                         goto drop;
2202
2203                 /* ...forward it to all interfaces. */
2204                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2205                 dst_if = NULL;
2206         }
2207
2208         /*
2209          * If we have a destination interface which is a member of our bridge,
2210          * OR this is a unicast packet, push it through the bpf(4) machinery.
2211          * For broadcast or multicast packets, don't bother because it will
2212          * be reinjected into ether_input. We do this before we pass the packets
2213          * through the pfil(9) framework, as it is possible that pfil(9) will
2214          * drop the packet, or possibly modify it, making it difficult to debug
2215          * firewall issues on the bridge.
2216          */
2217         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2218                 ETHER_BPF_MTAP(ifp, m);
2219
2220         /* run the packet filter */
2221         if (PFIL_HOOKED(&V_inet_pfil_hook)
2222 #ifdef INET6
2223             || PFIL_HOOKED(&V_inet6_pfil_hook)
2224 #endif
2225             ) {
2226                 BRIDGE_UNLOCK(sc);
2227                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2228                         return;
2229                 if (m == NULL)
2230                         return;
2231                 BRIDGE_LOCK(sc);
2232         }
2233
2234         if (dst_if == NULL) {
2235                 bridge_broadcast(sc, src_if, m, 1);
2236                 return;
2237         }
2238
2239         /*
2240          * At this point, we're dealing with a unicast frame
2241          * going to a different interface.
2242          */
2243         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2244                 goto drop;
2245
2246         dbif = bridge_lookup_member_if(sc, dst_if);
2247         if (dbif == NULL)
2248                 /* Not a member of the bridge (anymore?) */
2249                 goto drop;
2250
2251         /* Private segments can not talk to each other */
2252         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2253                 goto drop;
2254
2255         if ((dbif->bif_flags & IFBIF_STP) &&
2256             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2257                 goto drop;
2258
2259         BRIDGE_UNLOCK(sc);
2260
2261         if (PFIL_HOOKED(&V_inet_pfil_hook)
2262 #ifdef INET6
2263             || PFIL_HOOKED(&V_inet6_pfil_hook)
2264 #endif
2265             ) {
2266                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2267                         return;
2268                 if (m == NULL)
2269                         return;
2270         }
2271
2272         bridge_enqueue(sc, dst_if, m);
2273         return;
2274
2275 drop:
2276         BRIDGE_UNLOCK(sc);
2277         m_freem(m);
2278 }
2279
2280 /*
2281  * bridge_input:
2282  *
2283  *      Receive input from a member interface.  Queue the packet for
2284  *      bridging if it is not for us.
2285  */
2286 static struct mbuf *
2287 bridge_input(struct ifnet *ifp, struct mbuf *m)
2288 {
2289         struct bridge_softc *sc = ifp->if_bridge;
2290         struct bridge_iflist *bif, *bif2;
2291         struct ifnet *bifp;
2292         struct ether_header *eh;
2293         struct mbuf *mc, *mc2;
2294         uint16_t vlan;
2295         int error;
2296
2297         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2298                 return (m);
2299
2300         bifp = sc->sc_ifp;
2301         vlan = VLANTAGOF(m);
2302
2303         /*
2304          * Implement support for bridge monitoring. If this flag has been
2305          * set on this interface, discard the packet once we push it through
2306          * the bpf(4) machinery, but before we do, increment the byte and
2307          * packet counters associated with this interface.
2308          */
2309         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2310                 m->m_pkthdr.rcvif  = bifp;
2311                 ETHER_BPF_MTAP(bifp, m);
2312                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2313                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2314                 m_freem(m);
2315                 return (NULL);
2316         }
2317         BRIDGE_LOCK(sc);
2318         bif = bridge_lookup_member_if(sc, ifp);
2319         if (bif == NULL) {
2320                 BRIDGE_UNLOCK(sc);
2321                 return (m);
2322         }
2323
2324         eh = mtod(m, struct ether_header *);
2325
2326         bridge_span(sc, m);
2327
2328         if (m->m_flags & (M_BCAST|M_MCAST)) {
2329                 /* Tap off 802.1D packets; they do not get forwarded. */
2330                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2331                     ETHER_ADDR_LEN) == 0) {
2332                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2333                         BRIDGE_UNLOCK(sc);
2334                         return (NULL);
2335                 }
2336
2337                 if ((bif->bif_flags & IFBIF_STP) &&
2338                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2339                         BRIDGE_UNLOCK(sc);
2340                         return (m);
2341                 }
2342
2343                 /*
2344                  * Make a deep copy of the packet and enqueue the copy
2345                  * for bridge processing; return the original packet for
2346                  * local processing.
2347                  */
2348                 mc = m_dup(m, M_NOWAIT);
2349                 if (mc == NULL) {
2350                         BRIDGE_UNLOCK(sc);
2351                         return (m);
2352                 }
2353
2354                 /* Perform the bridge forwarding function with the copy. */
2355                 bridge_forward(sc, bif, mc);
2356
2357                 /*
2358                  * Reinject the mbuf as arriving on the bridge so we have a
2359                  * chance at claiming multicast packets. We can not loop back
2360                  * here from ether_input as a bridge is never a member of a
2361                  * bridge.
2362                  */
2363                 KASSERT(bifp->if_bridge == NULL,
2364                     ("loop created in bridge_input"));
2365                 mc2 = m_dup(m, M_NOWAIT);
2366                 if (mc2 != NULL) {
2367                         /* Keep the layer3 header aligned */
2368                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2369                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2370                 }
2371                 if (mc2 != NULL) {
2372                         mc2->m_pkthdr.rcvif = bifp;
2373                         (*bifp->if_input)(bifp, mc2);
2374                 }
2375
2376                 /* Return the original packet for local processing. */
2377                 return (m);
2378         }
2379
2380         if ((bif->bif_flags & IFBIF_STP) &&
2381             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2382                 BRIDGE_UNLOCK(sc);
2383                 return (m);
2384         }
2385
2386 #if (defined(INET) || defined(INET6))
2387 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2388         || ((iface)->if_carp \
2389             && (*carp_forus_p)((iface), eh->ether_dhost))
2390 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2391         || ((iface)->if_carp \
2392             && (*carp_forus_p)((iface), eh->ether_shost))
2393 #else
2394 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2395 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2396 #endif
2397
2398 #ifdef INET6
2399 #   define OR_PFIL_HOOKED_INET6 \
2400         || PFIL_HOOKED(&V_inet6_pfil_hook)
2401 #else
2402 #   define OR_PFIL_HOOKED_INET6
2403 #endif
2404
2405 #define GRAB_OUR_PACKETS(iface) \
2406         if ((iface)->if_type == IFT_GIF) \
2407                 continue; \
2408         /* It is destined for us. */ \
2409         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2410             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2411             ) {                                                         \
2412                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2413                         ETHER_BPF_MTAP(iface, m);                       \
2414                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2415                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2416                         /* Filter on the physical interface. */         \
2417                         if (V_pfil_local_phys &&                        \
2418                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2419                              OR_PFIL_HOOKED_INET6)) {                   \
2420                                 if (bridge_pfil(&m, NULL, ifp,          \
2421                                     PFIL_IN) != 0 || m == NULL) {       \
2422                                         BRIDGE_UNLOCK(sc);              \
2423                                         return (NULL);                  \
2424                                 }                                       \
2425                                 eh = mtod(m, struct ether_header *);    \
2426                         }                                               \
2427                 }                                                       \
2428                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2429                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2430                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2431                         if (error && bif->bif_addrmax) {                \
2432                                 BRIDGE_UNLOCK(sc);                      \
2433                                 m_freem(m);                             \
2434                                 return (NULL);                          \
2435                         }                                               \
2436                 }                                                       \
2437                 m->m_pkthdr.rcvif = iface;                              \
2438                 BRIDGE_UNLOCK(sc);                                      \
2439                 return (m);                                             \
2440         }                                                               \
2441                                                                         \
2442         /* We just received a packet that we sent out. */               \
2443         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2444             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2445             ) {                                                         \
2446                 BRIDGE_UNLOCK(sc);                                      \
2447                 m_freem(m);                                             \
2448                 return (NULL);                                          \
2449         }
2450
2451         /*
2452          * Unicast.  Make sure it's not for the bridge.
2453          */
2454         do { GRAB_OUR_PACKETS(bifp) } while (0);
2455
2456         /*
2457          * Give a chance for ifp at first priority. This will help when the
2458          * packet comes through the interface like VLAN's with the same MACs
2459          * on several interfaces from the same bridge. This also will save
2460          * some CPU cycles in case the destination interface and the input
2461          * interface (eq ifp) are the same.
2462          */
2463         do { GRAB_OUR_PACKETS(ifp) } while (0);
2464
2465         /* Now check the all bridge members. */
2466         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2467                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2468         }
2469
2470 #undef OR_CARP_CHECK_WE_ARE_DST
2471 #undef OR_CARP_CHECK_WE_ARE_SRC
2472 #undef OR_PFIL_HOOKED_INET6
2473 #undef GRAB_OUR_PACKETS
2474
2475         /* Perform the bridge forwarding function. */
2476         bridge_forward(sc, bif, m);
2477
2478         return (NULL);
2479 }
2480
2481 /*
2482  * bridge_broadcast:
2483  *
2484  *      Send a frame to all interfaces that are members of
2485  *      the bridge, except for the one on which the packet
2486  *      arrived.
2487  *
2488  *      NOTE: Releases the lock on return.
2489  */
2490 static void
2491 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2492     struct mbuf *m, int runfilt)
2493 {
2494         struct bridge_iflist *dbif, *sbif;
2495         struct mbuf *mc;
2496         struct ifnet *dst_if;
2497         int error = 0, used = 0, i;
2498
2499         sbif = bridge_lookup_member_if(sc, src_if);
2500
2501         BRIDGE_LOCK2REF(sc, error);
2502         if (error) {
2503                 m_freem(m);
2504                 return;
2505         }
2506
2507         /* Filter on the bridge interface before broadcasting */
2508         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2509 #ifdef INET6
2510             || PFIL_HOOKED(&V_inet6_pfil_hook)
2511 #endif
2512             )) {
2513                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2514                         goto out;
2515                 if (m == NULL)
2516                         goto out;
2517         }
2518
2519         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2520                 dst_if = dbif->bif_ifp;
2521                 if (dst_if == src_if)
2522                         continue;
2523
2524                 /* Private segments can not talk to each other */
2525                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2526                         continue;
2527
2528                 if ((dbif->bif_flags & IFBIF_STP) &&
2529                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2530                         continue;
2531
2532                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2533                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2534                         continue;
2535
2536                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2537                         continue;
2538
2539                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2540                         mc = m;
2541                         used = 1;
2542                 } else {
2543                         mc = m_dup(m, M_NOWAIT);
2544                         if (mc == NULL) {
2545                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2546                                 continue;
2547                         }
2548                 }
2549
2550                 /*
2551                  * Filter on the output interface. Pass a NULL bridge interface
2552                  * pointer so we do not redundantly filter on the bridge for
2553                  * each interface we broadcast on.
2554                  */
2555                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2556 #ifdef INET6
2557                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2558 #endif
2559                     )) {
2560                         if (used == 0) {
2561                                 /* Keep the layer3 header aligned */
2562                                 i = min(mc->m_pkthdr.len, max_protohdr);
2563                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2564                                 if (mc == NULL) {
2565                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2566                                         continue;
2567                                 }
2568                         }
2569                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2570                                 continue;
2571                         if (mc == NULL)
2572                                 continue;
2573                 }
2574
2575                 bridge_enqueue(sc, dst_if, mc);
2576         }
2577         if (used == 0)
2578                 m_freem(m);
2579
2580 out:
2581         BRIDGE_UNREF(sc);
2582 }
2583
2584 /*
2585  * bridge_span:
2586  *
2587  *      Duplicate a packet out one or more interfaces that are in span mode,
2588  *      the original mbuf is unmodified.
2589  */
2590 static void
2591 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2592 {
2593         struct bridge_iflist *bif;
2594         struct ifnet *dst_if;
2595         struct mbuf *mc;
2596
2597         if (LIST_EMPTY(&sc->sc_spanlist))
2598                 return;
2599
2600         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2601                 dst_if = bif->bif_ifp;
2602
2603                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2604                         continue;
2605
2606                 mc = m_copypacket(m, M_NOWAIT);
2607                 if (mc == NULL) {
2608                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2609                         continue;
2610                 }
2611
2612                 bridge_enqueue(sc, dst_if, mc);
2613         }
2614 }
2615
2616 /*
2617  * bridge_rtupdate:
2618  *
2619  *      Add a bridge routing entry.
2620  */
2621 static int
2622 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2623     struct bridge_iflist *bif, int setflags, uint8_t flags)
2624 {
2625         struct bridge_rtnode *brt;
2626         int error;
2627
2628         BRIDGE_LOCK_ASSERT(sc);
2629
2630         /* Check the source address is valid and not multicast. */
2631         if (ETHER_IS_MULTICAST(dst) ||
2632             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2633              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2634                 return (EINVAL);
2635
2636         /* 802.1p frames map to vlan 1 */
2637         if (vlan == 0)
2638                 vlan = 1;
2639
2640         /*
2641          * A route for this destination might already exist.  If so,
2642          * update it, otherwise create a new one.
2643          */
2644         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2645                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2646                         sc->sc_brtexceeded++;
2647                         return (ENOSPC);
2648                 }
2649                 /* Check per interface address limits (if enabled) */
2650                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2651                         bif->bif_addrexceeded++;
2652                         return (ENOSPC);
2653                 }
2654
2655                 /*
2656                  * Allocate a new bridge forwarding node, and
2657                  * initialize the expiration time and Ethernet
2658                  * address.
2659                  */
2660                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2661                 if (brt == NULL)
2662                         return (ENOMEM);
2663
2664                 if (bif->bif_flags & IFBIF_STICKY)
2665                         brt->brt_flags = IFBAF_STICKY;
2666                 else
2667                         brt->brt_flags = IFBAF_DYNAMIC;
2668
2669                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2670                 brt->brt_vlan = vlan;
2671
2672                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2673                         uma_zfree(bridge_rtnode_zone, brt);
2674                         return (error);
2675                 }
2676                 brt->brt_dst = bif;
2677                 bif->bif_addrcnt++;
2678         }
2679
2680         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2681             brt->brt_dst != bif) {
2682                 brt->brt_dst->bif_addrcnt--;
2683                 brt->brt_dst = bif;
2684                 brt->brt_dst->bif_addrcnt++;
2685         }
2686
2687         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2688                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2689         if (setflags)
2690                 brt->brt_flags = flags;
2691
2692         return (0);
2693 }
2694
2695 /*
2696  * bridge_rtlookup:
2697  *
2698  *      Lookup the destination interface for an address.
2699  */
2700 static struct ifnet *
2701 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2702 {
2703         struct bridge_rtnode *brt;
2704
2705         BRIDGE_LOCK_ASSERT(sc);
2706
2707         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2708                 return (NULL);
2709
2710         return (brt->brt_ifp);
2711 }
2712
2713 /*
2714  * bridge_rttrim:
2715  *
2716  *      Trim the routine table so that we have a number
2717  *      of routing entries less than or equal to the
2718  *      maximum number.
2719  */
2720 static void
2721 bridge_rttrim(struct bridge_softc *sc)
2722 {
2723         struct bridge_rtnode *brt, *nbrt;
2724
2725         BRIDGE_LOCK_ASSERT(sc);
2726
2727         /* Make sure we actually need to do this. */
2728         if (sc->sc_brtcnt <= sc->sc_brtmax)
2729                 return;
2730
2731         /* Force an aging cycle; this might trim enough addresses. */
2732         bridge_rtage(sc);
2733         if (sc->sc_brtcnt <= sc->sc_brtmax)
2734                 return;
2735
2736         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2737                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2738                         bridge_rtnode_destroy(sc, brt);
2739                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2740                                 return;
2741                 }
2742         }
2743 }
2744
2745 /*
2746  * bridge_timer:
2747  *
2748  *      Aging timer for the bridge.
2749  */
2750 static void
2751 bridge_timer(void *arg)
2752 {
2753         struct bridge_softc *sc = arg;
2754
2755         BRIDGE_LOCK_ASSERT(sc);
2756
2757         bridge_rtage(sc);
2758
2759         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2760                 callout_reset(&sc->sc_brcallout,
2761                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2762 }
2763
2764 /*
2765  * bridge_rtage:
2766  *
2767  *      Perform an aging cycle.
2768  */
2769 static void
2770 bridge_rtage(struct bridge_softc *sc)
2771 {
2772         struct bridge_rtnode *brt, *nbrt;
2773
2774         BRIDGE_LOCK_ASSERT(sc);
2775
2776         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2777                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2778                         if (time_uptime >= brt->brt_expire)
2779                                 bridge_rtnode_destroy(sc, brt);
2780                 }
2781         }
2782 }
2783
2784 /*
2785  * bridge_rtflush:
2786  *
2787  *      Remove all dynamic addresses from the bridge.
2788  */
2789 static void
2790 bridge_rtflush(struct bridge_softc *sc, int full)
2791 {
2792         struct bridge_rtnode *brt, *nbrt;
2793
2794         BRIDGE_LOCK_ASSERT(sc);
2795
2796         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2797                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2798                         bridge_rtnode_destroy(sc, brt);
2799         }
2800 }
2801
2802 /*
2803  * bridge_rtdaddr:
2804  *
2805  *      Remove an address from the table.
2806  */
2807 static int
2808 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2809 {
2810         struct bridge_rtnode *brt;
2811         int found = 0;
2812
2813         BRIDGE_LOCK_ASSERT(sc);
2814
2815         /*
2816          * If vlan is zero then we want to delete for all vlans so the lookup
2817          * may return more than one.
2818          */
2819         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2820                 bridge_rtnode_destroy(sc, brt);
2821                 found = 1;
2822         }
2823
2824         return (found ? 0 : ENOENT);
2825 }
2826
2827 /*
2828  * bridge_rtdelete:
2829  *
2830  *      Delete routes to a speicifc member interface.
2831  */
2832 static void
2833 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2834 {
2835         struct bridge_rtnode *brt, *nbrt;
2836
2837         BRIDGE_LOCK_ASSERT(sc);
2838
2839         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2840                 if (brt->brt_ifp == ifp && (full ||
2841                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2842                         bridge_rtnode_destroy(sc, brt);
2843         }
2844 }
2845
2846 /*
2847  * bridge_rtable_init:
2848  *
2849  *      Initialize the route table for this bridge.
2850  */
2851 static void
2852 bridge_rtable_init(struct bridge_softc *sc)
2853 {
2854         int i;
2855
2856         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2857             M_DEVBUF, M_WAITOK);
2858
2859         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2860                 LIST_INIT(&sc->sc_rthash[i]);
2861
2862         sc->sc_rthash_key = arc4random();
2863         LIST_INIT(&sc->sc_rtlist);
2864 }
2865
2866 /*
2867  * bridge_rtable_fini:
2868  *
2869  *      Deconstruct the route table for this bridge.
2870  */
2871 static void
2872 bridge_rtable_fini(struct bridge_softc *sc)
2873 {
2874
2875         KASSERT(sc->sc_brtcnt == 0,
2876             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2877         free(sc->sc_rthash, M_DEVBUF);
2878 }
2879
2880 /*
2881  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2882  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2883  */
2884 #define mix(a, b, c)                                                    \
2885 do {                                                                    \
2886         a -= b; a -= c; a ^= (c >> 13);                                 \
2887         b -= c; b -= a; b ^= (a << 8);                                  \
2888         c -= a; c -= b; c ^= (b >> 13);                                 \
2889         a -= b; a -= c; a ^= (c >> 12);                                 \
2890         b -= c; b -= a; b ^= (a << 16);                                 \
2891         c -= a; c -= b; c ^= (b >> 5);                                  \
2892         a -= b; a -= c; a ^= (c >> 3);                                  \
2893         b -= c; b -= a; b ^= (a << 10);                                 \
2894         c -= a; c -= b; c ^= (b >> 15);                                 \
2895 } while (/*CONSTCOND*/0)
2896
2897 static __inline uint32_t
2898 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2899 {
2900         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2901
2902         b += addr[5] << 8;
2903         b += addr[4];
2904         a += addr[3] << 24;
2905         a += addr[2] << 16;
2906         a += addr[1] << 8;
2907         a += addr[0];
2908
2909         mix(a, b, c);
2910
2911         return (c & BRIDGE_RTHASH_MASK);
2912 }
2913
2914 #undef mix
2915
2916 static int
2917 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2918 {
2919         int i, d;
2920
2921         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2922                 d = ((int)a[i]) - ((int)b[i]);
2923         }
2924
2925         return (d);
2926 }
2927
2928 /*
2929  * bridge_rtnode_lookup:
2930  *
2931  *      Look up a bridge route node for the specified destination. Compare the
2932  *      vlan id or if zero then just return the first match.
2933  */
2934 static struct bridge_rtnode *
2935 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2936 {
2937         struct bridge_rtnode *brt;
2938         uint32_t hash;
2939         int dir;
2940
2941         BRIDGE_LOCK_ASSERT(sc);
2942
2943         hash = bridge_rthash(sc, addr);
2944         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2945                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2946                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2947                         return (brt);
2948                 if (dir > 0)
2949                         return (NULL);
2950         }
2951
2952         return (NULL);
2953 }
2954
2955 /*
2956  * bridge_rtnode_insert:
2957  *
2958  *      Insert the specified bridge node into the route table.  We
2959  *      assume the entry is not already in the table.
2960  */
2961 static int
2962 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2963 {
2964         struct bridge_rtnode *lbrt;
2965         uint32_t hash;
2966         int dir;
2967
2968         BRIDGE_LOCK_ASSERT(sc);
2969
2970         hash = bridge_rthash(sc, brt->brt_addr);
2971
2972         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2973         if (lbrt == NULL) {
2974                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2975                 goto out;
2976         }
2977
2978         do {
2979                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2980                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2981                         return (EEXIST);
2982                 if (dir > 0) {
2983                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2984                         goto out;
2985                 }
2986                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2987                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2988                         goto out;
2989                 }
2990                 lbrt = LIST_NEXT(lbrt, brt_hash);
2991         } while (lbrt != NULL);
2992
2993 #ifdef DIAGNOSTIC
2994         panic("bridge_rtnode_insert: impossible");
2995 #endif
2996
2997 out:
2998         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2999         sc->sc_brtcnt++;
3000
3001         return (0);
3002 }
3003
3004 /*
3005  * bridge_rtnode_destroy:
3006  *
3007  *      Destroy a bridge rtnode.
3008  */
3009 static void
3010 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3011 {
3012         BRIDGE_LOCK_ASSERT(sc);
3013
3014         LIST_REMOVE(brt, brt_hash);
3015
3016         LIST_REMOVE(brt, brt_list);
3017         sc->sc_brtcnt--;
3018         brt->brt_dst->bif_addrcnt--;
3019         uma_zfree(bridge_rtnode_zone, brt);
3020 }
3021
3022 /*
3023  * bridge_rtable_expire:
3024  *
3025  *      Set the expiry time for all routes on an interface.
3026  */
3027 static void
3028 bridge_rtable_expire(struct ifnet *ifp, int age)
3029 {
3030         struct bridge_softc *sc = ifp->if_bridge;
3031         struct bridge_rtnode *brt;
3032
3033         BRIDGE_LOCK(sc);
3034
3035         /*
3036          * If the age is zero then flush, otherwise set all the expiry times to
3037          * age for the interface
3038          */
3039         if (age == 0)
3040                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3041         else {
3042                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3043                         /* Cap the expiry time to 'age' */
3044                         if (brt->brt_ifp == ifp &&
3045                             brt->brt_expire > time_uptime + age &&
3046                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3047                                 brt->brt_expire = time_uptime + age;
3048                 }
3049         }
3050         BRIDGE_UNLOCK(sc);
3051 }
3052
3053 /*
3054  * bridge_state_change:
3055  *
3056  *      Callback from the bridgestp code when a port changes states.
3057  */
3058 static void
3059 bridge_state_change(struct ifnet *ifp, int state)
3060 {
3061         struct bridge_softc *sc = ifp->if_bridge;
3062         static const char *stpstates[] = {
3063                 "disabled",
3064                 "listening",
3065                 "learning",
3066                 "forwarding",
3067                 "blocking",
3068                 "discarding"
3069         };
3070
3071         CURVNET_SET(ifp->if_vnet);
3072         if (V_log_stp)
3073                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3074                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3075         CURVNET_RESTORE();
3076 }
3077
3078 /*
3079  * Send bridge packets through pfil if they are one of the types pfil can deal
3080  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3081  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3082  * that interface.
3083  */
3084 static int
3085 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3086 {
3087         int snap, error, i, hlen;
3088         struct ether_header *eh1, eh2;
3089         struct ip *ip;
3090         struct llc llc1;
3091         u_int16_t ether_type;
3092
3093         snap = 0;
3094         error = -1;     /* Default error if not error == 0 */
3095
3096 #if 0
3097         /* we may return with the IP fields swapped, ensure its not shared */
3098         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3099 #endif
3100
3101         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3102                 return (0); /* filtering is disabled */
3103
3104         i = min((*mp)->m_pkthdr.len, max_protohdr);
3105         if ((*mp)->m_len < i) {
3106             *mp = m_pullup(*mp, i);
3107             if (*mp == NULL) {
3108                 printf("%s: m_pullup failed\n", __func__);
3109                 return (-1);
3110             }
3111         }
3112
3113         eh1 = mtod(*mp, struct ether_header *);
3114         ether_type = ntohs(eh1->ether_type);
3115
3116         /*
3117          * Check for SNAP/LLC.
3118          */
3119         if (ether_type < ETHERMTU) {
3120                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3121
3122                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3123                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3124                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3125                     llc2->llc_control == LLC_UI) {
3126                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3127                         snap = 1;
3128                 }
3129         }
3130
3131         /*
3132          * If we're trying to filter bridge traffic, don't look at anything
3133          * other than IP and ARP traffic.  If the filter doesn't understand
3134          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3135          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3136          * but of course we don't have an AppleTalk filter to begin with.
3137          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3138          * ARP traffic.)
3139          */
3140         switch (ether_type) {
3141                 case ETHERTYPE_ARP:
3142                 case ETHERTYPE_REVARP:
3143                         if (V_pfil_ipfw_arp == 0)
3144                                 return (0); /* Automatically pass */
3145                         break;
3146
3147                 case ETHERTYPE_IP:
3148 #ifdef INET6
3149                 case ETHERTYPE_IPV6:
3150 #endif /* INET6 */
3151                         break;
3152                 default:
3153                         /*
3154                          * Check to see if the user wants to pass non-ip
3155                          * packets, these will not be checked by pfil(9) and
3156                          * passed unconditionally so the default is to drop.
3157                          */
3158                         if (V_pfil_onlyip)
3159                                 goto bad;
3160         }
3161
3162         /* Run the packet through pfil before stripping link headers */
3163         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3164                         dir == PFIL_OUT && ifp != NULL) {
3165
3166                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, NULL);
3167
3168                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3169                         return (error);
3170         }
3171
3172         /* Strip off the Ethernet header and keep a copy. */
3173         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3174         m_adj(*mp, ETHER_HDR_LEN);
3175
3176         /* Strip off snap header, if present */
3177         if (snap) {
3178                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3179                 m_adj(*mp, sizeof(struct llc));
3180         }
3181
3182         /*
3183          * Check the IP header for alignment and errors
3184          */
3185         if (dir == PFIL_IN) {
3186                 switch (ether_type) {
3187                         case ETHERTYPE_IP:
3188                                 error = bridge_ip_checkbasic(mp);
3189                                 break;
3190 #ifdef INET6
3191                         case ETHERTYPE_IPV6:
3192                                 error = bridge_ip6_checkbasic(mp);
3193                                 break;
3194 #endif /* INET6 */
3195                         default:
3196                                 error = 0;
3197                 }
3198                 if (error)
3199                         goto bad;
3200         }
3201
3202         error = 0;
3203
3204         /*
3205          * Run the packet through pfil
3206          */
3207         switch (ether_type) {
3208         case ETHERTYPE_IP:
3209                 /*
3210                  * Run pfil on the member interface and the bridge, both can
3211                  * be skipped by clearing pfil_member or pfil_bridge.
3212                  *
3213                  * Keep the order:
3214                  *   in_if -> bridge_if -> out_if
3215                  */
3216                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3217                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3218                                         dir, NULL);
3219
3220                 if (*mp == NULL || error != 0) /* filter may consume */
3221                         break;
3222
3223                 if (V_pfil_member && ifp != NULL)
3224                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3225                                         dir, NULL);
3226
3227                 if (*mp == NULL || error != 0) /* filter may consume */
3228                         break;
3229
3230                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3231                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3232                                         dir, NULL);
3233
3234                 if (*mp == NULL || error != 0) /* filter may consume */
3235                         break;
3236
3237                 /* check if we need to fragment the packet */
3238                 /* bridge_fragment generates a mbuf chain of packets */
3239                 /* that already include eth headers */
3240                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3241                         i = (*mp)->m_pkthdr.len;
3242                         if (i > ifp->if_mtu) {
3243                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3244                                             &llc1);
3245                                 return (error);
3246                         }
3247                 }
3248
3249                 /* Recalculate the ip checksum. */
3250                 ip = mtod(*mp, struct ip *);
3251                 hlen = ip->ip_hl << 2;
3252                 if (hlen < sizeof(struct ip))
3253                         goto bad;
3254                 if (hlen > (*mp)->m_len) {
3255                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3256                                 goto bad;
3257                         ip = mtod(*mp, struct ip *);
3258                         if (ip == NULL)
3259                                 goto bad;
3260                 }
3261                 ip->ip_sum = 0;
3262                 if (hlen == sizeof(struct ip))
3263                         ip->ip_sum = in_cksum_hdr(ip);
3264                 else
3265                         ip->ip_sum = in_cksum(*mp, hlen);
3266
3267                 break;
3268 #ifdef INET6
3269         case ETHERTYPE_IPV6:
3270                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3271                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3272                                         dir, NULL);
3273
3274                 if (*mp == NULL || error != 0) /* filter may consume */
3275                         break;
3276
3277                 if (V_pfil_member && ifp != NULL)
3278                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3279                                         dir, NULL);
3280
3281                 if (*mp == NULL || error != 0) /* filter may consume */
3282                         break;
3283
3284                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3285                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3286                                         dir, NULL);
3287                 break;
3288 #endif
3289         default:
3290                 error = 0;
3291                 break;
3292         }
3293
3294         if (*mp == NULL)
3295                 return (error);
3296         if (error != 0)
3297                 goto bad;
3298
3299         error = -1;
3300
3301         /*
3302          * Finally, put everything back the way it was and return
3303          */
3304         if (snap) {
3305                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3306                 if (*mp == NULL)
3307                         return (error);
3308                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3309         }
3310
3311         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3312         if (*mp == NULL)
3313                 return (error);
3314         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3315
3316         return (0);
3317
3318 bad:
3319         m_freem(*mp);
3320         *mp = NULL;
3321         return (error);
3322 }
3323
3324 /*
3325  * Perform basic checks on header size since
3326  * pfil assumes ip_input has already processed
3327  * it for it.  Cut-and-pasted from ip_input.c.
3328  * Given how simple the IPv6 version is,
3329  * does the IPv4 version really need to be
3330  * this complicated?
3331  *
3332  * XXX Should we update ipstat here, or not?
3333  * XXX Right now we update ipstat but not
3334  * XXX csum_counter.
3335  */
3336 static int
3337 bridge_ip_checkbasic(struct mbuf **mp)
3338 {
3339         struct mbuf *m = *mp;
3340         struct ip *ip;
3341         int len, hlen;
3342         u_short sum;
3343
3344         if (*mp == NULL)
3345                 return (-1);
3346
3347         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3348                 if ((m = m_copyup(m, sizeof(struct ip),
3349                         (max_linkhdr + 3) & ~3)) == NULL) {
3350                         /* XXXJRT new stat, please */
3351                         KMOD_IPSTAT_INC(ips_toosmall);
3352                         goto bad;
3353                 }
3354         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3355                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3356                         KMOD_IPSTAT_INC(ips_toosmall);
3357                         goto bad;
3358                 }
3359         }
3360         ip = mtod(m, struct ip *);
3361         if (ip == NULL) goto bad;
3362
3363         if (ip->ip_v != IPVERSION) {
3364                 KMOD_IPSTAT_INC(ips_badvers);
3365                 goto bad;
3366         }
3367         hlen = ip->ip_hl << 2;
3368         if (hlen < sizeof(struct ip)) { /* minimum header length */
3369                 KMOD_IPSTAT_INC(ips_badhlen);
3370                 goto bad;
3371         }
3372         if (hlen > m->m_len) {
3373                 if ((m = m_pullup(m, hlen)) == NULL) {
3374                         KMOD_IPSTAT_INC(ips_badhlen);
3375                         goto bad;
3376                 }
3377                 ip = mtod(m, struct ip *);
3378                 if (ip == NULL) goto bad;
3379         }
3380
3381         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3382                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3383         } else {
3384                 if (hlen == sizeof(struct ip)) {
3385                         sum = in_cksum_hdr(ip);
3386                 } else {
3387                         sum = in_cksum(m, hlen);
3388                 }
3389         }
3390         if (sum) {
3391                 KMOD_IPSTAT_INC(ips_badsum);
3392                 goto bad;
3393         }
3394
3395         /* Retrieve the packet length. */
3396         len = ntohs(ip->ip_len);
3397
3398         /*
3399          * Check for additional length bogosity
3400          */
3401         if (len < hlen) {
3402                 KMOD_IPSTAT_INC(ips_badlen);
3403                 goto bad;
3404         }
3405
3406         /*
3407          * Check that the amount of data in the buffers
3408          * is as at least much as the IP header would have us expect.
3409          * Drop packet if shorter than we expect.
3410          */
3411         if (m->m_pkthdr.len < len) {
3412                 KMOD_IPSTAT_INC(ips_tooshort);
3413                 goto bad;
3414         }
3415
3416         /* Checks out, proceed */
3417         *mp = m;
3418         return (0);
3419
3420 bad:
3421         *mp = m;
3422         return (-1);
3423 }
3424
3425 #ifdef INET6
3426 /*
3427  * Same as above, but for IPv6.
3428  * Cut-and-pasted from ip6_input.c.
3429  * XXX Should we update ip6stat, or not?
3430  */
3431 static int
3432 bridge_ip6_checkbasic(struct mbuf **mp)
3433 {
3434         struct mbuf *m = *mp;
3435         struct ip6_hdr *ip6;
3436
3437         /*
3438          * If the IPv6 header is not aligned, slurp it up into a new
3439          * mbuf with space for link headers, in the event we forward
3440          * it.  Otherwise, if it is aligned, make sure the entire base
3441          * IPv6 header is in the first mbuf of the chain.
3442          */
3443         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3444                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3445                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3446                             (max_linkhdr + 3) & ~3)) == NULL) {
3447                         /* XXXJRT new stat, please */
3448                         IP6STAT_INC(ip6s_toosmall);
3449                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3450                         goto bad;
3451                 }
3452         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3453                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3454                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3455                         IP6STAT_INC(ip6s_toosmall);
3456                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3457                         goto bad;
3458                 }
3459         }
3460
3461         ip6 = mtod(m, struct ip6_hdr *);
3462
3463         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3464                 IP6STAT_INC(ip6s_badvers);
3465                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3466                 goto bad;
3467         }
3468
3469         /* Checks out, proceed */
3470         *mp = m;
3471         return (0);
3472
3473 bad:
3474         *mp = m;
3475         return (-1);
3476 }
3477 #endif /* INET6 */
3478
3479 /*
3480  * bridge_fragment:
3481  *
3482  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3483  */
3484 static int
3485 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3486     int snap, struct llc *llc)
3487 {
3488         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3489         struct ip *ip;
3490         int error = -1;
3491
3492         if (m->m_len < sizeof(struct ip) &&
3493             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3494                 goto dropit;
3495         ip = mtod(m, struct ip *);
3496
3497         m->m_pkthdr.csum_flags |= CSUM_IP;
3498         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3499         if (error)
3500                 goto dropit;
3501
3502         /*
3503          * Walk the chain and re-add the Ethernet header for
3504          * each mbuf packet.
3505          */
3506         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3507                 nextpkt = mcur->m_nextpkt;
3508                 mcur->m_nextpkt = NULL;
3509                 if (snap) {
3510                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3511                         if (mcur == NULL) {
3512                                 error = ENOBUFS;
3513                                 if (mprev != NULL)
3514                                         mprev->m_nextpkt = nextpkt;
3515                                 goto dropit;
3516                         }
3517                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3518                 }
3519
3520                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3521                 if (mcur == NULL) {
3522                         error = ENOBUFS;
3523                         if (mprev != NULL)
3524                                 mprev->m_nextpkt = nextpkt;
3525                         goto dropit;
3526                 }
3527                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3528
3529                 /*
3530                  * The previous two M_PREPEND could have inserted one or two
3531                  * mbufs in front so we have to update the previous packet's
3532                  * m_nextpkt.
3533                  */
3534                 mcur->m_nextpkt = nextpkt;
3535                 if (mprev != NULL)
3536                         mprev->m_nextpkt = mcur;
3537                 else {
3538                         /* The first mbuf in the original chain needs to be
3539                          * updated. */
3540                         *mp = mcur;
3541                 }
3542                 mprev = mcur;
3543         }
3544
3545         KMOD_IPSTAT_INC(ips_fragmented);
3546         return (error);
3547
3548 dropit:
3549         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3550                 m = mcur->m_nextpkt;
3551                 m_freem(mcur);
3552         }
3553         return (error);
3554 }
3555
3556 static void
3557 bridge_linkstate(struct ifnet *ifp)
3558 {
3559         struct bridge_softc *sc = ifp->if_bridge;
3560         struct bridge_iflist *bif;
3561
3562         BRIDGE_LOCK(sc);
3563         bif = bridge_lookup_member_if(sc, ifp);
3564         if (bif == NULL) {
3565                 BRIDGE_UNLOCK(sc);
3566                 return;
3567         }
3568         bridge_linkcheck(sc);
3569         BRIDGE_UNLOCK(sc);
3570
3571         bstp_linkstate(&bif->bif_stp);
3572 }
3573
3574 static void
3575 bridge_linkcheck(struct bridge_softc *sc)
3576 {
3577         struct bridge_iflist *bif;
3578         int new_link, hasls;
3579
3580         BRIDGE_LOCK_ASSERT(sc);
3581         new_link = LINK_STATE_DOWN;
3582         hasls = 0;
3583         /* Our link is considered up if at least one of our ports is active */
3584         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3585                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3586                         hasls++;
3587                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3588                         new_link = LINK_STATE_UP;
3589                         break;
3590                 }
3591         }
3592         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3593                 /* If no interfaces support link-state then we default to up */
3594                 new_link = LINK_STATE_UP;
3595         }
3596         if_link_state_change(sc->sc_ifp, new_link);
3597 }