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_VERSION(if_bridge, 1);
 588 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 589
 590 /*
 591  * handler for net.link.bridge.ipfw
 592  */
 593 static int
 594 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 595 {
 596         int enable = V_pfil_ipfw;
 597         int error;
 598
 599         error = sysctl_handle_int(oidp, &enable, 0, req);
 600         enable &= 1;
 601
 602         if (enable != V_pfil_ipfw) {
 603                 V_pfil_ipfw = enable;
 604
 605                 /*
 606                  * Disable pfil so that ipfw doesnt run twice, if the user
 607                  * really wants both then they can re-enable pfil_bridge and/or
 608                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 609                  * layer2 type.
 610                  */
 611                 if (V_pfil_ipfw) {
 612                         V_pfil_onlyip = 0;
 613                         V_pfil_bridge = 0;
 614                         V_pfil_member = 0;
 615                 }
 616         }
 617
 618         return (error);
 619 }
 620 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 621     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 622     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 623     "Layer2 filter with IPFW");
 624
 625 /*
 626  * bridge_clone_create:
 627  *
 628  *      Create a new bridge instance.
 629  */
 630 static int
 631 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 632 {
 633         struct bridge_softc *sc, *sc2;
 634         struct ifnet *bifp, *ifp;
 635         int fb, retry;
 636         unsigned long hostid;
 637
 638         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 639         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 640         if (ifp == NULL) {
 641                 free(sc, M_DEVBUF);
 642                 return (ENOSPC);
 643         }
 644
 645         BRIDGE_LOCK_INIT(sc);
 646         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 647         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 648
 649         /* Initialize our routing table. */
 650         bridge_rtable_init(sc);
 651
 652         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 653
 654         LIST_INIT(&sc->sc_iflist);
 655         LIST_INIT(&sc->sc_spanlist);
 656
 657         ifp->if_softc = sc;
 658         if_initname(ifp, bridge_name, unit);
 659         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 660         ifp->if_ioctl = bridge_ioctl;
 661         ifp->if_transmit = bridge_transmit;
 662         ifp->if_qflush = bridge_qflush;
 663         ifp->if_init = bridge_init;
 664         ifp->if_type = IFT_BRIDGE;
 665
 666         /*
 667          * Generate an ethernet address with a locally administered address.
 668          *
 669          * Since we are using random ethernet addresses for the bridge, it is
 670          * possible that we might have address collisions, so make sure that
 671          * this hardware address isn't already in use on another bridge.
 672          * The first try uses the hostid and falls back to arc4rand().
 673          */
 674         fb = 0;
 675         getcredhostid(curthread->td_ucred, &hostid);
 676         do {
 677                 if (fb || hostid == 0) {
 678                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 679                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 680                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 681                 } else {
 682                         sc->sc_defaddr[0] = 0x2;
 683                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 684                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 685                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 686                         sc->sc_defaddr[4] =  hostid        & 0xff;
 687                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 688                 }
 689
 690                 fb = 1;
 691                 retry = 0;
 692                 BRIDGE_LIST_LOCK();
 693                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 694                         bifp = sc2->sc_ifp;
 695                         if (memcmp(sc->sc_defaddr,
 696                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 697                                 retry = 1;
 698                                 break;
 699                         }
 700                 }
 701                 BRIDGE_LIST_UNLOCK();
 702         } while (retry == 1);
 703
 704         bstp_attach(&sc->sc_stp, &bridge_ops);
 705         ether_ifattach(ifp, sc->sc_defaddr);
 706         /* Now undo some of the damage... */
 707         ifp->if_baudrate = 0;
 708         ifp->if_type = IFT_BRIDGE;
 709
 710         BRIDGE_LIST_LOCK();
 711         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 712         BRIDGE_LIST_UNLOCK();
 713
 714         return (0);
 715 }
 716
 717 /*
 718  * bridge_clone_destroy:
 719  *
 720  *      Destroy a bridge instance.
 721  */
 722 static void
 723 bridge_clone_destroy(struct ifnet *ifp)
 724 {
 725         struct bridge_softc *sc = ifp->if_softc;
 726         struct bridge_iflist *bif;
 727
 728         BRIDGE_LOCK(sc);
 729
 730         bridge_stop(ifp, 1);
 731         ifp->if_flags &= ~IFF_UP;
 732
 733         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 734                 bridge_delete_member(sc, bif, 0);
 735
 736         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 737                 bridge_delete_span(sc, bif);
 738         }
 739
 740         BRIDGE_UNLOCK(sc);
 741
 742         callout_drain(&sc->sc_brcallout);
 743
 744         BRIDGE_LIST_LOCK();
 745         LIST_REMOVE(sc, sc_list);
 746         BRIDGE_LIST_UNLOCK();
 747
 748         bstp_detach(&sc->sc_stp);
 749         ether_ifdetach(ifp);
 750         if_free(ifp);
 751
 752         /* Tear down the routing table. */
 753         bridge_rtable_fini(sc);
 754
 755         BRIDGE_LOCK_DESTROY(sc);
 756         free(sc, M_DEVBUF);
 757 }
 758
 759 /*
 760  * bridge_ioctl:
 761  *
 762  *      Handle a control request from the operator.
 763  */
 764 static int
 765 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 766 {
 767         struct bridge_softc *sc = ifp->if_softc;
 768         struct ifreq *ifr = (struct ifreq *)data;
 769         struct bridge_iflist *bif;
 770         struct thread *td = curthread;
 771         union {
 772                 struct ifbreq ifbreq;
 773                 struct ifbifconf ifbifconf;
 774                 struct ifbareq ifbareq;
 775                 struct ifbaconf ifbaconf;
 776                 struct ifbrparam ifbrparam;
 777                 struct ifbropreq ifbropreq;
 778         } args;
 779         struct ifdrv *ifd = (struct ifdrv *) data;
 780         const struct bridge_control *bc;
 781         int error = 0;
 782
 783         switch (cmd) {
 784
 785         case SIOCADDMULTI:
 786         case SIOCDELMULTI:
 787                 break;
 788
 789         case SIOCGDRVSPEC:
 790         case SIOCSDRVSPEC:
 791                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 792                         error = EINVAL;
 793                         break;
 794                 }
 795                 bc = &bridge_control_table[ifd->ifd_cmd];
 796
 797                 if (cmd == SIOCGDRVSPEC &&
 798                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 799                         error = EINVAL;
 800                         break;
 801                 }
 802                 else if (cmd == SIOCSDRVSPEC &&
 803                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 804                         error = EINVAL;
 805                         break;
 806                 }
 807
 808                 if (bc->bc_flags & BC_F_SUSER) {
 809                         error = priv_check(td, PRIV_NET_BRIDGE);
 810                         if (error)
 811                                 break;
 812                 }
 813
 814                 if (ifd->ifd_len != bc->bc_argsize ||
 815                     ifd->ifd_len > sizeof(args)) {
 816                         error = EINVAL;
 817                         break;
 818                 }
 819
 820                 bzero(&args, sizeof(args));
 821                 if (bc->bc_flags & BC_F_COPYIN) {
 822                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 823                         if (error)
 824                                 break;
 825                 }
 826
 827                 BRIDGE_LOCK(sc);
 828                 error = (*bc->bc_func)(sc, &args);
 829                 BRIDGE_UNLOCK(sc);
 830                 if (error)
 831                         break;
 832
 833                 if (bc->bc_flags & BC_F_COPYOUT)
 834                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 835
 836                 break;
 837
 838         case SIOCSIFFLAGS:
 839                 if (!(ifp->if_flags & IFF_UP) &&
 840                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 841                         /*
 842                          * If interface is marked down and it is running,
 843                          * then stop and disable it.
 844                          */
 845                         BRIDGE_LOCK(sc);
 846                         bridge_stop(ifp, 1);
 847                         BRIDGE_UNLOCK(sc);
 848                 } else if ((ifp->if_flags & IFF_UP) &&
 849                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 850                         /*
 851                          * If interface is marked up and it is stopped, then
 852                          * start it.
 853                          */
 854                         (*ifp->if_init)(sc);
 855                 }
 856                 break;
 857
 858         case SIOCSIFMTU:
 859                 if (ifr->ifr_mtu < 576) {
 860                         error = EINVAL;
 861                         break;
 862                 }
 863                 if (LIST_EMPTY(&sc->sc_iflist)) {
 864                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 865                         break;
 866                 }
 867                 BRIDGE_LOCK(sc);
 868                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 869                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 870                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 871                                     " != %d\n", sc->sc_ifp->if_xname,
 872                                     bif->bif_ifp->if_mtu,
 873                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 874                                 error = EINVAL;
 875                                 break;
 876                         }
 877                 }
 878                 if (!error)
 879                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 880                 BRIDGE_UNLOCK(sc);
 881                 break;
 882         default:
 883                 /*
 884                  * drop the lock as ether_ioctl() will call bridge_start() and
 885                  * cause the lock to be recursed.
 886                  */
 887                 error = ether_ioctl(ifp, cmd, data);
 888                 break;
 889         }
 890
 891         return (error);
 892 }
 893
 894 /*
 895  * bridge_mutecaps:
 896  *
 897  *      Clear or restore unwanted capabilities on the member interface
 898  */
 899 static void
 900 bridge_mutecaps(struct bridge_softc *sc)
 901 {
 902         struct bridge_iflist *bif;
 903         int enabled, mask;
 904
 905         /* Initial bitmask of capabilities to test */
 906         mask = BRIDGE_IFCAPS_MASK;
 907
 908         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 909                 /* Every member must support it or its disabled */
 910                 mask &= bif->bif_savedcaps;
 911         }
 912
 913         BRIDGE_XLOCK(sc);
 914         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 915                 enabled = bif->bif_ifp->if_capenable;
 916                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 917                 /* strip off mask bits and enable them again if allowed */
 918                 enabled &= ~BRIDGE_IFCAPS_MASK;
 919                 enabled |= mask;
 920                 BRIDGE_UNLOCK(sc);
 921                 bridge_set_ifcap(sc, bif, enabled);
 922                 BRIDGE_LOCK(sc);
 923         }
 924         BRIDGE_XDROP(sc);
 925
 926 }
 927
 928 static void
 929 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 930 {
 931         struct ifnet *ifp = bif->bif_ifp;
 932         struct ifreq ifr;
 933         int error;
 934
 935         BRIDGE_UNLOCK_ASSERT(sc);
 936
 937         bzero(&ifr, sizeof(ifr));
 938         ifr.ifr_reqcap = set;
 939
 940         if (ifp->if_capenable != set) {
 941                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 942                 if (error)
 943                         if_printf(sc->sc_ifp,
 944                             "error setting capabilities on %s: %d\n",
 945                             ifp->if_xname, error);
 946                 if ((ifp->if_capenable & ~set) != 0)
 947                         if_printf(sc->sc_ifp,
 948                             "can't disable some capabilities on %s: 0x%x\n",
 949                             ifp->if_xname, ifp->if_capenable & ~set);
 950         }
 951 }
 952
 953 /*
 954  * bridge_lookup_member:
 955  *
 956  *      Lookup a bridge member interface.
 957  */
 958 static struct bridge_iflist *
 959 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 960 {
 961         struct bridge_iflist *bif;
 962         struct ifnet *ifp;
 963
 964         BRIDGE_LOCK_ASSERT(sc);
 965
 966         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 967                 ifp = bif->bif_ifp;
 968                 if (strcmp(ifp->if_xname, name) == 0)
 969                         return (bif);
 970         }
 971
 972         return (NULL);
 973 }
 974
 975 /*
 976  * bridge_lookup_member_if:
 977  *
 978  *      Lookup a bridge member interface by ifnet*.
 979  */
 980 static struct bridge_iflist *
 981 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 982 {
 983         struct bridge_iflist *bif;
 984
 985         BRIDGE_LOCK_ASSERT(sc);
 986
 987         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 988                 if (bif->bif_ifp == member_ifp)
 989                         return (bif);
 990         }
 991
 992         return (NULL);
 993 }
 994
 995 /*
 996  * bridge_delete_member:
 997  *
 998  *      Delete the specified member interface.
 999  */
1000 static void
1001 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1002     int gone)
1003 {
1004         struct ifnet *ifs = bif->bif_ifp;
1005         struct ifnet *fif = NULL;
1006
1007         BRIDGE_LOCK_ASSERT(sc);
1008
1009         if (bif->bif_flags & IFBIF_STP)
1010                 bstp_disable(&bif->bif_stp);
1011
1012         ifs->if_bridge = NULL;
1013         BRIDGE_XLOCK(sc);
1014         LIST_REMOVE(bif, bif_next);
1015         BRIDGE_XDROP(sc);
1016
1017         /*
1018          * If removing the interface that gave the bridge its mac address, set
1019          * the mac address of the bridge to the address of the next member, or
1020          * to its default address if no members are left.
1021          */
1022         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1023                 if (LIST_EMPTY(&sc->sc_iflist)) {
1024                         bcopy(sc->sc_defaddr,
1025                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1026                         sc->sc_ifaddr = NULL;
1027                 } else {
1028                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
1029                         bcopy(IF_LLADDR(fif),
1030                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1031                         sc->sc_ifaddr = fif;
1032                 }
1033                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1034         }
1035
1036         bridge_linkcheck(sc);
1037         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1038         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1039         KASSERT(bif->bif_addrcnt == 0,
1040             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1041
1042         BRIDGE_UNLOCK(sc);
1043         if (!gone) {
1044                 switch (ifs->if_type) {
1045                 case IFT_ETHER:
1046                 case IFT_L2VLAN:
1047                         /*
1048                          * Take the interface out of promiscuous mode, but only
1049                          * if it was promiscuous in the first place. It might
1050                          * not be if we're in the bridge_ioctl_add() error path.
1051                          */
1052                         if (ifs->if_flags & IFF_PROMISC)
1053                                 (void) ifpromisc(ifs, 0);
1054                         break;
1055
1056                 case IFT_GIF:
1057                         break;
1058
1059                 default:
1060 #ifdef DIAGNOSTIC
1061                         panic("bridge_delete_member: impossible");
1062 #endif
1063                         break;
1064                 }
1065                 /* reneable any interface capabilities */
1066                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1067         }
1068         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1069         BRIDGE_LOCK(sc);
1070         free(bif, M_DEVBUF);
1071 }
1072
1073 /*
1074  * bridge_delete_span:
1075  *
1076  *      Delete the specified span interface.
1077  */
1078 static void
1079 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1080 {
1081         BRIDGE_LOCK_ASSERT(sc);
1082
1083         KASSERT(bif->bif_ifp->if_bridge == NULL,
1084             ("%s: not a span interface", __func__));
1085
1086         LIST_REMOVE(bif, bif_next);
1087         free(bif, M_DEVBUF);
1088 }
1089
1090 static int
1091 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1092 {
1093         struct ifbreq *req = arg;
1094         struct bridge_iflist *bif = NULL;
1095         struct ifnet *ifs;
1096         int error = 0;
1097
1098         ifs = ifunit(req->ifbr_ifsname);
1099         if (ifs == NULL)
1100                 return (ENOENT);
1101         if (ifs->if_ioctl == NULL)      /* must be supported */
1102                 return (EINVAL);
1103
1104         /* If it's in the span list, it can't be a member. */
1105         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1106                 if (ifs == bif->bif_ifp)
1107                         return (EBUSY);
1108
1109         if (ifs->if_bridge == sc)
1110                 return (EEXIST);
1111
1112         if (ifs->if_bridge != NULL)
1113                 return (EBUSY);
1114
1115         switch (ifs->if_type) {
1116         case IFT_ETHER:
1117         case IFT_L2VLAN:
1118         case IFT_GIF:
1119                 /* permitted interface types */
1120                 break;
1121         default:
1122                 return (EINVAL);
1123         }
1124
1125 #ifdef INET6
1126         /*
1127          * Two valid inet6 addresses with link-local scope must not be
1128          * on the parent interface and the member interfaces at the
1129          * same time.  This restriction is needed to prevent violation
1130          * of link-local scope zone.  Attempts to add a member
1131          * interface which has inet6 addresses when the parent has
1132          * inet6 triggers removal of all inet6 addresses on the member
1133          * interface.
1134          */
1135
1136         /* Check if the parent interface has a link-local scope addr. */
1137         if (V_allow_llz_overlap == 0 &&
1138             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1139                 /*
1140                  * If any, remove all inet6 addresses from the member
1141                  * interfaces.
1142                  */
1143                 BRIDGE_XLOCK(sc);
1144                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1145                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1146                                 BRIDGE_UNLOCK(sc);
1147                                 in6_ifdetach(bif->bif_ifp);
1148                                 BRIDGE_LOCK(sc);
1149                                 if_printf(sc->sc_ifp,
1150                                     "IPv6 addresses on %s have been removed "
1151                                     "before adding it as a member to prevent "
1152                                     "IPv6 address scope violation.\n",
1153                                     bif->bif_ifp->if_xname);
1154                         }
1155                 }
1156                 BRIDGE_XDROP(sc);
1157                 if (in6ifa_llaonifp(ifs)) {
1158                         BRIDGE_UNLOCK(sc);
1159                         in6_ifdetach(ifs);
1160                         BRIDGE_LOCK(sc);
1161                         if_printf(sc->sc_ifp,
1162                             "IPv6 addresses on %s have been removed "
1163                             "before adding it as a member to prevent "
1164                             "IPv6 address scope violation.\n",
1165                             ifs->if_xname);
1166                 }
1167         }
1168 #endif
1169         /* Allow the first Ethernet member to define the MTU */
1170         if (LIST_EMPTY(&sc->sc_iflist))
1171                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1172         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1173                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1174                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1175                 return (EINVAL);
1176         }
1177
1178         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1179         if (bif == NULL)
1180                 return (ENOMEM);
1181
1182         bif->bif_ifp = ifs;
1183         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1184         bif->bif_savedcaps = ifs->if_capenable;
1185
1186         /*
1187          * Assign the interface's MAC address to the bridge if it's the first
1188          * member and the MAC address of the bridge has not been changed from
1189          * the default randomly generated one.
1190          */
1191         if (V_bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1192             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1193                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1194                 sc->sc_ifaddr = ifs;
1195                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1196         }
1197
1198         ifs->if_bridge = sc;
1199         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1200         /*
1201          * XXX: XLOCK HERE!?!
1202          *
1203          * NOTE: insert_***HEAD*** should be safe for the traversals.
1204          */
1205         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1206
1207         /* Set interface capabilities to the intersection set of all members */
1208         bridge_mutecaps(sc);
1209         bridge_linkcheck(sc);
1210
1211         /* Place the interface into promiscuous mode */
1212         switch (ifs->if_type) {
1213                 case IFT_ETHER:
1214                 case IFT_L2VLAN:
1215                         BRIDGE_UNLOCK(sc);
1216                         error = ifpromisc(ifs, 1);
1217                         BRIDGE_LOCK(sc);
1218                         break;
1219         }
1220
1221         if (error)
1222                 bridge_delete_member(sc, bif, 0);
1223         return (error);
1224 }
1225
1226 static int
1227 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1228 {
1229         struct ifbreq *req = arg;
1230         struct bridge_iflist *bif;
1231
1232         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1233         if (bif == NULL)
1234                 return (ENOENT);
1235
1236         bridge_delete_member(sc, bif, 0);
1237
1238         return (0);
1239 }
1240
1241 static int
1242 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1243 {
1244         struct ifbreq *req = arg;
1245         struct bridge_iflist *bif;
1246         struct bstp_port *bp;
1247
1248         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1249         if (bif == NULL)
1250                 return (ENOENT);
1251
1252         bp = &bif->bif_stp;
1253         req->ifbr_ifsflags = bif->bif_flags;
1254         req->ifbr_state = bp->bp_state;
1255         req->ifbr_priority = bp->bp_priority;
1256         req->ifbr_path_cost = bp->bp_path_cost;
1257         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1258         req->ifbr_proto = bp->bp_protover;
1259         req->ifbr_role = bp->bp_role;
1260         req->ifbr_stpflags = bp->bp_flags;
1261         req->ifbr_addrcnt = bif->bif_addrcnt;
1262         req->ifbr_addrmax = bif->bif_addrmax;
1263         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1264
1265         /* Copy STP state options as flags */
1266         if (bp->bp_operedge)
1267                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1268         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1269                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1270         if (bp->bp_ptp_link)
1271                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1272         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1273                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1274         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1275                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1276         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1277                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1278         return (0);
1279 }
1280
1281 static int
1282 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1283 {
1284         struct ifbreq *req = arg;
1285         struct bridge_iflist *bif;
1286         struct bstp_port *bp;
1287         int error;
1288
1289         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1290         if (bif == NULL)
1291                 return (ENOENT);
1292         bp = &bif->bif_stp;
1293
1294         if (req->ifbr_ifsflags & IFBIF_SPAN)
1295                 /* SPAN is readonly */
1296                 return (EINVAL);
1297
1298         if (req->ifbr_ifsflags & IFBIF_STP) {
1299                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1300                         error = bstp_enable(&bif->bif_stp);
1301                         if (error)
1302                                 return (error);
1303                 }
1304         } else {
1305                 if ((bif->bif_flags & IFBIF_STP) != 0)
1306                         bstp_disable(&bif->bif_stp);
1307         }
1308
1309         /* Pass on STP flags */
1310         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1311         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1312         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1313         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1314
1315         /* Save the bits relating to the bridge */
1316         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1317
1318         return (0);
1319 }
1320
1321 static int
1322 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1323 {
1324         struct ifbrparam *param = arg;
1325
1326         sc->sc_brtmax = param->ifbrp_csize;
1327         bridge_rttrim(sc);
1328
1329         return (0);
1330 }
1331
1332 static int
1333 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1334 {
1335         struct ifbrparam *param = arg;
1336
1337         param->ifbrp_csize = sc->sc_brtmax;
1338
1339         return (0);
1340 }
1341
1342 static int
1343 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1344 {
1345         struct ifbifconf *bifc = arg;
1346         struct bridge_iflist *bif;
1347         struct ifbreq breq;
1348         char *buf, *outbuf;
1349         int count, buflen, len, error = 0;
1350
1351         count = 0;
1352         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1353                 count++;
1354         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1355                 count++;
1356
1357         buflen = sizeof(breq) * count;
1358         if (bifc->ifbic_len == 0) {
1359                 bifc->ifbic_len = buflen;
1360                 return (0);
1361         }
1362         BRIDGE_UNLOCK(sc);
1363         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1364         BRIDGE_LOCK(sc);
1365
1366         count = 0;
1367         buf = outbuf;
1368         len = min(bifc->ifbic_len, buflen);
1369         bzero(&breq, sizeof(breq));
1370         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1371                 if (len < sizeof(breq))
1372                         break;
1373
1374                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1375                     sizeof(breq.ifbr_ifsname));
1376                 /* Fill in the ifbreq structure */
1377                 error = bridge_ioctl_gifflags(sc, &breq);
1378                 if (error)
1379                         break;
1380                 memcpy(buf, &breq, sizeof(breq));
1381                 count++;
1382                 buf += sizeof(breq);
1383                 len -= sizeof(breq);
1384         }
1385         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1386                 if (len < sizeof(breq))
1387                         break;
1388
1389                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1390                     sizeof(breq.ifbr_ifsname));
1391                 breq.ifbr_ifsflags = bif->bif_flags;
1392                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1393                 memcpy(buf, &breq, sizeof(breq));
1394                 count++;
1395                 buf += sizeof(breq);
1396                 len -= sizeof(breq);
1397         }
1398
1399         BRIDGE_UNLOCK(sc);
1400         bifc->ifbic_len = sizeof(breq) * count;
1401         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1402         BRIDGE_LOCK(sc);
1403         free(outbuf, M_TEMP);
1404         return (error);
1405 }
1406
1407 static int
1408 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1409 {
1410         struct ifbaconf *bac = arg;
1411         struct bridge_rtnode *brt;
1412         struct ifbareq bareq;
1413         char *buf, *outbuf;
1414         int count, buflen, len, error = 0;
1415
1416         if (bac->ifbac_len == 0)
1417                 return (0);
1418
1419         count = 0;
1420         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1421                 count++;
1422         buflen = sizeof(bareq) * count;
1423
1424         BRIDGE_UNLOCK(sc);
1425         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1426         BRIDGE_LOCK(sc);
1427
1428         count = 0;
1429         buf = outbuf;
1430         len = min(bac->ifbac_len, buflen);
1431         bzero(&bareq, sizeof(bareq));
1432         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1433                 if (len < sizeof(bareq))
1434                         goto out;
1435                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1436                     sizeof(bareq.ifba_ifsname));
1437                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1438                 bareq.ifba_vlan = brt->brt_vlan;
1439                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1440                                 time_uptime < brt->brt_expire)
1441                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1442                 else
1443                         bareq.ifba_expire = 0;
1444                 bareq.ifba_flags = brt->brt_flags;
1445
1446                 memcpy(buf, &bareq, sizeof(bareq));
1447                 count++;
1448                 buf += sizeof(bareq);
1449                 len -= sizeof(bareq);
1450         }
1451 out:
1452         BRIDGE_UNLOCK(sc);
1453         bac->ifbac_len = sizeof(bareq) * count;
1454         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1455         BRIDGE_LOCK(sc);
1456         free(outbuf, M_TEMP);
1457         return (error);
1458 }
1459
1460 static int
1461 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1462 {
1463         struct ifbareq *req = arg;
1464         struct bridge_iflist *bif;
1465         int error;
1466
1467         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1468         if (bif == NULL)
1469                 return (ENOENT);
1470
1471         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1472             req->ifba_flags);
1473
1474         return (error);
1475 }
1476
1477 static int
1478 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1479 {
1480         struct ifbrparam *param = arg;
1481
1482         sc->sc_brttimeout = param->ifbrp_ctime;
1483         return (0);
1484 }
1485
1486 static int
1487 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1488 {
1489         struct ifbrparam *param = arg;
1490
1491         param->ifbrp_ctime = sc->sc_brttimeout;
1492         return (0);
1493 }
1494
1495 static int
1496 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1497 {
1498         struct ifbareq *req = arg;
1499
1500         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1501 }
1502
1503 static int
1504 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1505 {
1506         struct ifbreq *req = arg;
1507
1508         bridge_rtflush(sc, req->ifbr_ifsflags);
1509         return (0);
1510 }
1511
1512 static int
1513 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1514 {
1515         struct ifbrparam *param = arg;
1516         struct bstp_state *bs = &sc->sc_stp;
1517
1518         param->ifbrp_prio = bs->bs_bridge_priority;
1519         return (0);
1520 }
1521
1522 static int
1523 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1524 {
1525         struct ifbrparam *param = arg;
1526
1527         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1528 }
1529
1530 static int
1531 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1532 {
1533         struct ifbrparam *param = arg;
1534         struct bstp_state *bs = &sc->sc_stp;
1535
1536         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1537         return (0);
1538 }
1539
1540 static int
1541 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1542 {
1543         struct ifbrparam *param = arg;
1544
1545         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1546 }
1547
1548 static int
1549 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1550 {
1551         struct ifbrparam *param = arg;
1552         struct bstp_state *bs = &sc->sc_stp;
1553
1554         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1555         return (0);
1556 }
1557
1558 static int
1559 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1560 {
1561         struct ifbrparam *param = arg;
1562
1563         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1564 }
1565
1566 static int
1567 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1568 {
1569         struct ifbrparam *param = arg;
1570         struct bstp_state *bs = &sc->sc_stp;
1571
1572         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1573         return (0);
1574 }
1575
1576 static int
1577 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1578 {
1579         struct ifbrparam *param = arg;
1580
1581         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1582 }
1583
1584 static int
1585 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1586 {
1587         struct ifbreq *req = arg;
1588         struct bridge_iflist *bif;
1589
1590         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1591         if (bif == NULL)
1592                 return (ENOENT);
1593
1594         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1595 }
1596
1597 static int
1598 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1599 {
1600         struct ifbreq *req = arg;
1601         struct bridge_iflist *bif;
1602
1603         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1604         if (bif == NULL)
1605                 return (ENOENT);
1606
1607         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1608 }
1609
1610 static int
1611 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1612 {
1613         struct ifbreq *req = arg;
1614         struct bridge_iflist *bif;
1615
1616         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1617         if (bif == NULL)
1618                 return (ENOENT);
1619
1620         bif->bif_addrmax = req->ifbr_addrmax;
1621         return (0);
1622 }
1623
1624 static int
1625 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1626 {
1627         struct ifbreq *req = arg;
1628         struct bridge_iflist *bif = NULL;
1629         struct ifnet *ifs;
1630
1631         ifs = ifunit(req->ifbr_ifsname);
1632         if (ifs == NULL)
1633                 return (ENOENT);
1634
1635         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1636                 if (ifs == bif->bif_ifp)
1637                         return (EBUSY);
1638
1639         if (ifs->if_bridge != NULL)
1640                 return (EBUSY);
1641
1642         switch (ifs->if_type) {
1643                 case IFT_ETHER:
1644                 case IFT_GIF:
1645                 case IFT_L2VLAN:
1646                         break;
1647                 default:
1648                         return (EINVAL);
1649         }
1650
1651         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1652         if (bif == NULL)
1653                 return (ENOMEM);
1654
1655         bif->bif_ifp = ifs;
1656         bif->bif_flags = IFBIF_SPAN;
1657
1658         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1659
1660         return (0);
1661 }
1662
1663 static int
1664 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1665 {
1666         struct ifbreq *req = arg;
1667         struct bridge_iflist *bif;
1668         struct ifnet *ifs;
1669
1670         ifs = ifunit(req->ifbr_ifsname);
1671         if (ifs == NULL)
1672                 return (ENOENT);
1673
1674         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1675                 if (ifs == bif->bif_ifp)
1676                         break;
1677
1678         if (bif == NULL)
1679                 return (ENOENT);
1680
1681         bridge_delete_span(sc, bif);
1682
1683         return (0);
1684 }
1685
1686 static int
1687 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1688 {
1689         struct ifbropreq *req = arg;
1690         struct bstp_state *bs = &sc->sc_stp;
1691         struct bstp_port *root_port;
1692
1693         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1694         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1695         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1696
1697         root_port = bs->bs_root_port;
1698         if (root_port == NULL)
1699                 req->ifbop_root_port = 0;
1700         else
1701                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1702
1703         req->ifbop_holdcount = bs->bs_txholdcount;
1704         req->ifbop_priority = bs->bs_bridge_priority;
1705         req->ifbop_protocol = bs->bs_protover;
1706         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1707         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1708         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1709         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1710         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1711         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1712
1713         return (0);
1714 }
1715
1716 static int
1717 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1718 {
1719         struct ifbrparam *param = arg;
1720
1721         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1722         return (0);
1723 }
1724
1725 static int
1726 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1727 {
1728         struct ifbpstpconf *bifstp = arg;
1729         struct bridge_iflist *bif;
1730         struct bstp_port *bp;
1731         struct ifbpstpreq bpreq;
1732         char *buf, *outbuf;
1733         int count, buflen, len, error = 0;
1734
1735         count = 0;
1736         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1737                 if ((bif->bif_flags & IFBIF_STP) != 0)
1738                         count++;
1739         }
1740
1741         buflen = sizeof(bpreq) * count;
1742         if (bifstp->ifbpstp_len == 0) {
1743                 bifstp->ifbpstp_len = buflen;
1744                 return (0);
1745         }
1746
1747         BRIDGE_UNLOCK(sc);
1748         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1749         BRIDGE_LOCK(sc);
1750
1751         count = 0;
1752         buf = outbuf;
1753         len = min(bifstp->ifbpstp_len, buflen);
1754         bzero(&bpreq, sizeof(bpreq));
1755         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1756                 if (len < sizeof(bpreq))
1757                         break;
1758
1759                 if ((bif->bif_flags & IFBIF_STP) == 0)
1760                         continue;
1761
1762                 bp = &bif->bif_stp;
1763                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1764                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1765                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1766                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1767                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1768                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1769
1770                 memcpy(buf, &bpreq, sizeof(bpreq));
1771                 count++;
1772                 buf += sizeof(bpreq);
1773                 len -= sizeof(bpreq);
1774         }
1775
1776         BRIDGE_UNLOCK(sc);
1777         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1778         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1779         BRIDGE_LOCK(sc);
1780         free(outbuf, M_TEMP);
1781         return (error);
1782 }
1783
1784 static int
1785 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1786 {
1787         struct ifbrparam *param = arg;
1788
1789         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1790 }
1791
1792 static int
1793 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1794 {
1795         struct ifbrparam *param = arg;
1796
1797         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1798 }
1799
1800 /*
1801  * bridge_ifdetach:
1802  *
1803  *      Detach an interface from a bridge.  Called when a member
1804  *      interface is detaching.
1805  */
1806 static void
1807 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1808 {
1809         struct bridge_softc *sc = ifp->if_bridge;
1810         struct bridge_iflist *bif;
1811
1812         if (ifp->if_flags & IFF_RENAMING)
1813                 return;
1814         if (V_bridge_cloner == NULL) {
1815                 /*
1816                  * This detach handler can be called after
1817                  * vnet_bridge_uninit().  Just return in that case.
1818                  */
1819                 return;
1820         }
1821         /* Check if the interface is a bridge member */
1822         if (sc != NULL) {
1823                 BRIDGE_LOCK(sc);
1824
1825                 bif = bridge_lookup_member_if(sc, ifp);
1826                 if (bif != NULL)
1827                         bridge_delete_member(sc, bif, 1);
1828
1829                 BRIDGE_UNLOCK(sc);
1830                 return;
1831         }
1832
1833         /* Check if the interface is a span port */
1834         BRIDGE_LIST_LOCK();
1835         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1836                 BRIDGE_LOCK(sc);
1837                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1838                         if (ifp == bif->bif_ifp) {
1839                                 bridge_delete_span(sc, bif);
1840                                 break;
1841                         }
1842
1843                 BRIDGE_UNLOCK(sc);
1844         }
1845         BRIDGE_LIST_UNLOCK();
1846 }
1847
1848 /*
1849  * bridge_init:
1850  *
1851  *      Initialize a bridge interface.
1852  */
1853 static void
1854 bridge_init(void *xsc)
1855 {
1856         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1857         struct ifnet *ifp = sc->sc_ifp;
1858
1859         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1860                 return;
1861
1862         BRIDGE_LOCK(sc);
1863         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1864             bridge_timer, sc);
1865
1866         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1867         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1868
1869         BRIDGE_UNLOCK(sc);
1870 }
1871
1872 /*
1873  * bridge_stop:
1874  *
1875  *      Stop the bridge interface.
1876  */
1877 static void
1878 bridge_stop(struct ifnet *ifp, int disable)
1879 {
1880         struct bridge_softc *sc = ifp->if_softc;
1881
1882         BRIDGE_LOCK_ASSERT(sc);
1883
1884         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1885                 return;
1886
1887         callout_stop(&sc->sc_brcallout);
1888         bstp_stop(&sc->sc_stp);
1889
1890         bridge_rtflush(sc, IFBF_FLUSHDYN);
1891
1892         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1893 }
1894
1895 /*
1896  * bridge_enqueue:
1897  *
1898  *      Enqueue a packet on a bridge member interface.
1899  *
1900  */
1901 static int
1902 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1903 {
1904         int len, err = 0;
1905         short mflags;
1906         struct mbuf *m0;
1907
1908         /* We may be sending a fragment so traverse the mbuf */
1909         for (; m; m = m0) {
1910                 m0 = m->m_nextpkt;
1911                 m->m_nextpkt = NULL;
1912                 len = m->m_pkthdr.len;
1913                 mflags = m->m_flags;
1914
1915                 /*
1916                  * If underlying interface can not do VLAN tag insertion itself
1917                  * then attach a packet tag that holds it.
1918                  */
1919                 if ((m->m_flags & M_VLANTAG) &&
1920                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1921                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1922                         if (m == NULL) {
1923                                 if_printf(dst_ifp,
1924                                     "unable to prepend VLAN header\n");
1925                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
1926                                 continue;
1927                         }
1928                         m->m_flags &= ~M_VLANTAG;
1929                 }
1930
1931                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
1932                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1933                         m_freem(m0);
1934                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
1935                         break;
1936                 }
1937
1938                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
1939                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
1940                 if (mflags & M_MCAST)
1941                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
1942         }
1943
1944         return (err);
1945 }
1946
1947 /*
1948  * bridge_dummynet:
1949  *
1950  *      Receive a queued packet from dummynet and pass it on to the output
1951  *      interface.
1952  *
1953  *      The mbuf has the Ethernet header already attached.
1954  */
1955 static void
1956 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1957 {
1958         struct bridge_softc *sc;
1959
1960         sc = ifp->if_bridge;
1961
1962         /*
1963          * The packet didnt originate from a member interface. This should only
1964          * ever happen if a member interface is removed while packets are
1965          * queued for it.
1966          */
1967         if (sc == NULL) {
1968                 m_freem(m);
1969                 return;
1970         }
1971
1972         if (PFIL_HOOKED(&V_inet_pfil_hook)
1973 #ifdef INET6
1974             || PFIL_HOOKED(&V_inet6_pfil_hook)
1975 #endif
1976             ) {
1977                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1978                         return;
1979                 if (m == NULL)
1980                         return;
1981         }
1982
1983         bridge_enqueue(sc, ifp, m);
1984 }
1985
1986 /*
1987  * bridge_output:
1988  *
1989  *      Send output from a bridge member interface.  This
1990  *      performs the bridging function for locally originated
1991  *      packets.
1992  *
1993  *      The mbuf has the Ethernet header already attached.  We must
1994  *      enqueue or free the mbuf before returning.
1995  */
1996 static int
1997 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1998     struct rtentry *rt)
1999 {
2000         struct ether_header *eh;
2001         struct ifnet *dst_if;
2002         struct bridge_softc *sc;
2003         uint16_t vlan;
2004
2005         if (m->m_len < ETHER_HDR_LEN) {
2006                 m = m_pullup(m, ETHER_HDR_LEN);
2007                 if (m == NULL)
2008                         return (0);
2009         }
2010
2011         eh = mtod(m, struct ether_header *);
2012         sc = ifp->if_bridge;
2013         vlan = VLANTAGOF(m);
2014
2015         BRIDGE_LOCK(sc);
2016
2017         /*
2018          * If bridge is down, but the original output interface is up,
2019          * go ahead and send out that interface.  Otherwise, the packet
2020          * is dropped below.
2021          */
2022         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2023                 dst_if = ifp;
2024                 goto sendunicast;
2025         }
2026
2027         /*
2028          * If the packet is a multicast, or we don't know a better way to
2029          * get there, send to all interfaces.
2030          */
2031         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2032                 dst_if = NULL;
2033         else
2034                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2035         if (dst_if == NULL) {
2036                 struct bridge_iflist *bif;
2037                 struct mbuf *mc;
2038                 int error = 0, used = 0;
2039
2040                 bridge_span(sc, m);
2041
2042                 BRIDGE_LOCK2REF(sc, error);
2043                 if (error) {
2044                         m_freem(m);
2045                         return (0);
2046                 }
2047
2048                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2049                         dst_if = bif->bif_ifp;
2050
2051                         if (dst_if->if_type == IFT_GIF)
2052                                 continue;
2053                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2054                                 continue;
2055
2056                         /*
2057                          * If this is not the original output interface,
2058                          * and the interface is participating in spanning
2059                          * tree, make sure the port is in a state that
2060                          * allows forwarding.
2061                          */
2062                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2063                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2064                                 continue;
2065
2066                         if (LIST_NEXT(bif, bif_next) == NULL) {
2067                                 used = 1;
2068                                 mc = m;
2069                         } else {
2070                                 mc = m_copypacket(m, M_NOWAIT);
2071                                 if (mc == NULL) {
2072                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2073                                         continue;
2074                                 }
2075                         }
2076
2077                         bridge_enqueue(sc, dst_if, mc);
2078                 }
2079                 if (used == 0)
2080                         m_freem(m);
2081                 BRIDGE_UNREF(sc);
2082                 return (0);
2083         }
2084
2085 sendunicast:
2086         /*
2087          * XXX Spanning tree consideration here?
2088          */
2089
2090         bridge_span(sc, m);
2091         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2092                 m_freem(m);
2093                 BRIDGE_UNLOCK(sc);
2094                 return (0);
2095         }
2096
2097         BRIDGE_UNLOCK(sc);
2098         bridge_enqueue(sc, dst_if, m);
2099         return (0);
2100 }
2101
2102 /*
2103  * bridge_transmit:
2104  *
2105  *      Do output on a bridge.
2106  *
2107  */
2108 static int
2109 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2110 {
2111         struct bridge_softc *sc;
2112         struct ether_header *eh;
2113         struct ifnet *dst_if;
2114         int error = 0;
2115
2116         sc = ifp->if_softc;
2117
2118         ETHER_BPF_MTAP(ifp, m);
2119
2120         eh = mtod(m, struct ether_header *);
2121
2122         BRIDGE_LOCK(sc);
2123         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2124             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2125                 BRIDGE_UNLOCK(sc);
2126                 error = bridge_enqueue(sc, dst_if, m);
2127         } else
2128                 bridge_broadcast(sc, ifp, m, 0);
2129
2130         return (error);
2131 }
2132
2133 /*
2134  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2135  */
2136 static void
2137 bridge_qflush(struct ifnet *ifp __unused)
2138 {
2139 }
2140
2141 /*
2142  * bridge_forward:
2143  *
2144  *      The forwarding function of the bridge.
2145  *
2146  *      NOTE: Releases the lock on return.
2147  */
2148 static void
2149 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2150     struct mbuf *m)
2151 {
2152         struct bridge_iflist *dbif;
2153         struct ifnet *src_if, *dst_if, *ifp;
2154         struct ether_header *eh;
2155         uint16_t vlan;
2156         uint8_t *dst;
2157         int error;
2158
2159         src_if = m->m_pkthdr.rcvif;
2160         ifp = sc->sc_ifp;
2161
2162         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2163         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2164         vlan = VLANTAGOF(m);
2165
2166         if ((sbif->bif_flags & IFBIF_STP) &&
2167             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2168                 goto drop;
2169
2170         eh = mtod(m, struct ether_header *);
2171         dst = eh->ether_dhost;
2172
2173         /* If the interface is learning, record the address. */
2174         if (sbif->bif_flags & IFBIF_LEARNING) {
2175                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2176                     sbif, 0, IFBAF_DYNAMIC);
2177                 /*
2178                  * If the interface has addresses limits then deny any source
2179                  * that is not in the cache.
2180                  */
2181                 if (error && sbif->bif_addrmax)
2182                         goto drop;
2183         }
2184
2185         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2186             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2187                 goto drop;
2188
2189         /*
2190          * At this point, the port either doesn't participate
2191          * in spanning tree or it is in the forwarding state.
2192          */
2193
2194         /*
2195          * If the packet is unicast, destined for someone on
2196          * "this" side of the bridge, drop it.
2197          */
2198         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2199                 dst_if = bridge_rtlookup(sc, dst, vlan);
2200                 if (src_if == dst_if)
2201                         goto drop;
2202         } else {
2203                 /*
2204                  * Check if its a reserved multicast address, any address
2205                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2206                  * bridge.
2207                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2208                  */
2209                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2210                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2211                     dst[4] == 0x00 && dst[5] <= 0x0f)
2212                         goto drop;
2213
2214                 /* ...forward it to all interfaces. */
2215                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2216                 dst_if = NULL;
2217         }
2218
2219         /*
2220          * If we have a destination interface which is a member of our bridge,
2221          * OR this is a unicast packet, push it through the bpf(4) machinery.
2222          * For broadcast or multicast packets, don't bother because it will
2223          * be reinjected into ether_input. We do this before we pass the packets
2224          * through the pfil(9) framework, as it is possible that pfil(9) will
2225          * drop the packet, or possibly modify it, making it difficult to debug
2226          * firewall issues on the bridge.
2227          */
2228         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2229                 ETHER_BPF_MTAP(ifp, m);
2230
2231         /* run the packet filter */
2232         if (PFIL_HOOKED(&V_inet_pfil_hook)
2233 #ifdef INET6
2234             || PFIL_HOOKED(&V_inet6_pfil_hook)
2235 #endif
2236             ) {
2237                 BRIDGE_UNLOCK(sc);
2238                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2239                         return;
2240                 if (m == NULL)
2241                         return;
2242                 BRIDGE_LOCK(sc);
2243         }
2244
2245         if (dst_if == NULL) {
2246                 bridge_broadcast(sc, src_if, m, 1);
2247                 return;
2248         }
2249
2250         /*
2251          * At this point, we're dealing with a unicast frame
2252          * going to a different interface.
2253          */
2254         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2255                 goto drop;
2256
2257         dbif = bridge_lookup_member_if(sc, dst_if);
2258         if (dbif == NULL)
2259                 /* Not a member of the bridge (anymore?) */
2260                 goto drop;
2261
2262         /* Private segments can not talk to each other */
2263         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2264                 goto drop;
2265
2266         if ((dbif->bif_flags & IFBIF_STP) &&
2267             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2268                 goto drop;
2269
2270         BRIDGE_UNLOCK(sc);
2271
2272         if (PFIL_HOOKED(&V_inet_pfil_hook)
2273 #ifdef INET6
2274             || PFIL_HOOKED(&V_inet6_pfil_hook)
2275 #endif
2276             ) {
2277                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2278                         return;
2279                 if (m == NULL)
2280                         return;
2281         }
2282
2283         bridge_enqueue(sc, dst_if, m);
2284         return;
2285
2286 drop:
2287         BRIDGE_UNLOCK(sc);
2288         m_freem(m);
2289 }
2290
2291 /*
2292  * bridge_input:
2293  *
2294  *      Receive input from a member interface.  Queue the packet for
2295  *      bridging if it is not for us.
2296  */
2297 static struct mbuf *
2298 bridge_input(struct ifnet *ifp, struct mbuf *m)
2299 {
2300         struct bridge_softc *sc = ifp->if_bridge;
2301         struct bridge_iflist *bif, *bif2;
2302         struct ifnet *bifp;
2303         struct ether_header *eh;
2304         struct mbuf *mc, *mc2;
2305         uint16_t vlan;
2306         int error;
2307
2308         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2309                 return (m);
2310
2311         bifp = sc->sc_ifp;
2312         vlan = VLANTAGOF(m);
2313
2314         /*
2315          * Implement support for bridge monitoring. If this flag has been
2316          * set on this interface, discard the packet once we push it through
2317          * the bpf(4) machinery, but before we do, increment the byte and
2318          * packet counters associated with this interface.
2319          */
2320         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2321                 m->m_pkthdr.rcvif  = bifp;
2322                 ETHER_BPF_MTAP(bifp, m);
2323                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2324                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2325                 m_freem(m);
2326                 return (NULL);
2327         }
2328         BRIDGE_LOCK(sc);
2329         bif = bridge_lookup_member_if(sc, ifp);
2330         if (bif == NULL) {
2331                 BRIDGE_UNLOCK(sc);
2332                 return (m);
2333         }
2334
2335         eh = mtod(m, struct ether_header *);
2336
2337         bridge_span(sc, m);
2338
2339         if (m->m_flags & (M_BCAST|M_MCAST)) {
2340                 /* Tap off 802.1D packets; they do not get forwarded. */
2341                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2342                     ETHER_ADDR_LEN) == 0) {
2343                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2344                         BRIDGE_UNLOCK(sc);
2345                         return (NULL);
2346                 }
2347
2348                 if ((bif->bif_flags & IFBIF_STP) &&
2349                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2350                         BRIDGE_UNLOCK(sc);
2351                         return (m);
2352                 }
2353
2354                 /*
2355                  * Make a deep copy of the packet and enqueue the copy
2356                  * for bridge processing; return the original packet for
2357                  * local processing.
2358                  */
2359                 mc = m_dup(m, M_NOWAIT);
2360                 if (mc == NULL) {
2361                         BRIDGE_UNLOCK(sc);
2362                         return (m);
2363                 }
2364
2365                 /* Perform the bridge forwarding function with the copy. */
2366                 bridge_forward(sc, bif, mc);
2367
2368                 /*
2369                  * Reinject the mbuf as arriving on the bridge so we have a
2370                  * chance at claiming multicast packets. We can not loop back
2371                  * here from ether_input as a bridge is never a member of a
2372                  * bridge.
2373                  */
2374                 KASSERT(bifp->if_bridge == NULL,
2375                     ("loop created in bridge_input"));
2376                 mc2 = m_dup(m, M_NOWAIT);
2377                 if (mc2 != NULL) {
2378                         /* Keep the layer3 header aligned */
2379                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2380                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2381                 }
2382                 if (mc2 != NULL) {
2383                         mc2->m_pkthdr.rcvif = bifp;
2384                         (*bifp->if_input)(bifp, mc2);
2385                 }
2386
2387                 /* Return the original packet for local processing. */
2388                 return (m);
2389         }
2390
2391         if ((bif->bif_flags & IFBIF_STP) &&
2392             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2393                 BRIDGE_UNLOCK(sc);
2394                 return (m);
2395         }
2396
2397 #if (defined(INET) || defined(INET6))
2398 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2399         || ((iface)->if_carp \
2400             && (*carp_forus_p)((iface), eh->ether_dhost))
2401 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2402         || ((iface)->if_carp \
2403             && (*carp_forus_p)((iface), eh->ether_shost))
2404 #else
2405 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2406 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2407 #endif
2408
2409 #ifdef INET6
2410 #   define OR_PFIL_HOOKED_INET6 \
2411         || PFIL_HOOKED(&V_inet6_pfil_hook)
2412 #else
2413 #   define OR_PFIL_HOOKED_INET6
2414 #endif
2415
2416 #define GRAB_OUR_PACKETS(iface) \
2417         if ((iface)->if_type == IFT_GIF) \
2418                 continue; \
2419         /* It is destined for us. */ \
2420         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2421             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2422             ) {                                                         \
2423                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2424                         ETHER_BPF_MTAP(iface, m);                       \
2425                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2426                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2427                         /* Filter on the physical interface. */         \
2428                         if (V_pfil_local_phys &&                        \
2429                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2430                              OR_PFIL_HOOKED_INET6)) {                   \
2431                                 if (bridge_pfil(&m, NULL, ifp,          \
2432                                     PFIL_IN) != 0 || m == NULL) {       \
2433                                         BRIDGE_UNLOCK(sc);              \
2434                                         return (NULL);                  \
2435                                 }                                       \
2436                                 eh = mtod(m, struct ether_header *);    \
2437                         }                                               \
2438                 }                                                       \
2439                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2440                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2441                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2442                         if (error && bif->bif_addrmax) {                \
2443                                 BRIDGE_UNLOCK(sc);                      \
2444                                 m_freem(m);                             \
2445                                 return (NULL);                          \
2446                         }                                               \
2447                 }                                                       \
2448                 m->m_pkthdr.rcvif = iface;                              \
2449                 BRIDGE_UNLOCK(sc);                                      \
2450                 return (m);                                             \
2451         }                                                               \
2452                                                                         \
2453         /* We just received a packet that we sent out. */               \
2454         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2455             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2456             ) {                                                         \
2457                 BRIDGE_UNLOCK(sc);                                      \
2458                 m_freem(m);                                             \
2459                 return (NULL);                                          \
2460         }
2461
2462         /*
2463          * Unicast.  Make sure it's not for the bridge.
2464          */
2465         do { GRAB_OUR_PACKETS(bifp) } while (0);
2466
2467         /*
2468          * Give a chance for ifp at first priority. This will help when the
2469          * packet comes through the interface like VLAN's with the same MACs
2470          * on several interfaces from the same bridge. This also will save
2471          * some CPU cycles in case the destination interface and the input
2472          * interface (eq ifp) are the same.
2473          */
2474         do { GRAB_OUR_PACKETS(ifp) } while (0);
2475
2476         /* Now check the all bridge members. */
2477         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2478                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2479         }
2480
2481 #undef OR_CARP_CHECK_WE_ARE_DST
2482 #undef OR_CARP_CHECK_WE_ARE_SRC
2483 #undef OR_PFIL_HOOKED_INET6
2484 #undef GRAB_OUR_PACKETS
2485
2486         /* Perform the bridge forwarding function. */
2487         bridge_forward(sc, bif, m);
2488
2489         return (NULL);
2490 }
2491
2492 /*
2493  * bridge_broadcast:
2494  *
2495  *      Send a frame to all interfaces that are members of
2496  *      the bridge, except for the one on which the packet
2497  *      arrived.
2498  *
2499  *      NOTE: Releases the lock on return.
2500  */
2501 static void
2502 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2503     struct mbuf *m, int runfilt)
2504 {
2505         struct bridge_iflist *dbif, *sbif;
2506         struct mbuf *mc;
2507         struct ifnet *dst_if;
2508         int error = 0, used = 0, i;
2509
2510         sbif = bridge_lookup_member_if(sc, src_if);
2511
2512         BRIDGE_LOCK2REF(sc, error);
2513         if (error) {
2514                 m_freem(m);
2515                 return;
2516         }
2517
2518         /* Filter on the bridge interface before broadcasting */
2519         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2520 #ifdef INET6
2521             || PFIL_HOOKED(&V_inet6_pfil_hook)
2522 #endif
2523             )) {
2524                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2525                         goto out;
2526                 if (m == NULL)
2527                         goto out;
2528         }
2529
2530         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2531                 dst_if = dbif->bif_ifp;
2532                 if (dst_if == src_if)
2533                         continue;
2534
2535                 /* Private segments can not talk to each other */
2536                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2537                         continue;
2538
2539                 if ((dbif->bif_flags & IFBIF_STP) &&
2540                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2541                         continue;
2542
2543                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2544                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2545                         continue;
2546
2547                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2548                         continue;
2549
2550                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2551                         mc = m;
2552                         used = 1;
2553                 } else {
2554                         mc = m_dup(m, M_NOWAIT);
2555                         if (mc == NULL) {
2556                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2557                                 continue;
2558                         }
2559                 }
2560
2561                 /*
2562                  * Filter on the output interface. Pass a NULL bridge interface
2563                  * pointer so we do not redundantly filter on the bridge for
2564                  * each interface we broadcast on.
2565                  */
2566                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2567 #ifdef INET6
2568                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2569 #endif
2570                     )) {
2571                         if (used == 0) {
2572                                 /* Keep the layer3 header aligned */
2573                                 i = min(mc->m_pkthdr.len, max_protohdr);
2574                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2575                                 if (mc == NULL) {
2576                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2577                                         continue;
2578                                 }
2579                         }
2580                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2581                                 continue;
2582                         if (mc == NULL)
2583                                 continue;
2584                 }
2585
2586                 bridge_enqueue(sc, dst_if, mc);
2587         }
2588         if (used == 0)
2589                 m_freem(m);
2590
2591 out:
2592         BRIDGE_UNREF(sc);
2593 }
2594
2595 /*
2596  * bridge_span:
2597  *
2598  *      Duplicate a packet out one or more interfaces that are in span mode,
2599  *      the original mbuf is unmodified.
2600  */
2601 static void
2602 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2603 {
2604         struct bridge_iflist *bif;
2605         struct ifnet *dst_if;
2606         struct mbuf *mc;
2607
2608         if (LIST_EMPTY(&sc->sc_spanlist))
2609                 return;
2610
2611         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2612                 dst_if = bif->bif_ifp;
2613
2614                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2615                         continue;
2616
2617                 mc = m_copypacket(m, M_NOWAIT);
2618                 if (mc == NULL) {
2619                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2620                         continue;
2621                 }
2622
2623                 bridge_enqueue(sc, dst_if, mc);
2624         }
2625 }
2626
2627 /*
2628  * bridge_rtupdate:
2629  *
2630  *      Add a bridge routing entry.
2631  */
2632 static int
2633 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2634     struct bridge_iflist *bif, int setflags, uint8_t flags)
2635 {
2636         struct bridge_rtnode *brt;
2637         int error;
2638
2639         BRIDGE_LOCK_ASSERT(sc);
2640
2641         /* Check the source address is valid and not multicast. */
2642         if (ETHER_IS_MULTICAST(dst) ||
2643             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2644              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2645                 return (EINVAL);
2646
2647         /* 802.1p frames map to vlan 1 */
2648         if (vlan == 0)
2649                 vlan = 1;
2650
2651         /*
2652          * A route for this destination might already exist.  If so,
2653          * update it, otherwise create a new one.
2654          */
2655         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2656                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2657                         sc->sc_brtexceeded++;
2658                         return (ENOSPC);
2659                 }
2660                 /* Check per interface address limits (if enabled) */
2661                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2662                         bif->bif_addrexceeded++;
2663                         return (ENOSPC);
2664                 }
2665
2666                 /*
2667                  * Allocate a new bridge forwarding node, and
2668                  * initialize the expiration time and Ethernet
2669                  * address.
2670                  */
2671                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2672                 if (brt == NULL)
2673                         return (ENOMEM);
2674
2675                 if (bif->bif_flags & IFBIF_STICKY)
2676                         brt->brt_flags = IFBAF_STICKY;
2677                 else
2678                         brt->brt_flags = IFBAF_DYNAMIC;
2679
2680                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2681                 brt->brt_vlan = vlan;
2682
2683                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2684                         uma_zfree(bridge_rtnode_zone, brt);
2685                         return (error);
2686                 }
2687                 brt->brt_dst = bif;
2688                 bif->bif_addrcnt++;
2689         }
2690
2691         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2692             brt->brt_dst != bif) {
2693                 brt->brt_dst->bif_addrcnt--;
2694                 brt->brt_dst = bif;
2695                 brt->brt_dst->bif_addrcnt++;
2696         }
2697
2698         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2699                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2700         if (setflags)
2701                 brt->brt_flags = flags;
2702
2703         return (0);
2704 }
2705
2706 /*
2707  * bridge_rtlookup:
2708  *
2709  *      Lookup the destination interface for an address.
2710  */
2711 static struct ifnet *
2712 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2713 {
2714         struct bridge_rtnode *brt;
2715
2716         BRIDGE_LOCK_ASSERT(sc);
2717
2718         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2719                 return (NULL);
2720
2721         return (brt->brt_ifp);
2722 }
2723
2724 /*
2725  * bridge_rttrim:
2726  *
2727  *      Trim the routine table so that we have a number
2728  *      of routing entries less than or equal to the
2729  *      maximum number.
2730  */
2731 static void
2732 bridge_rttrim(struct bridge_softc *sc)
2733 {
2734         struct bridge_rtnode *brt, *nbrt;
2735
2736         BRIDGE_LOCK_ASSERT(sc);
2737
2738         /* Make sure we actually need to do this. */
2739         if (sc->sc_brtcnt <= sc->sc_brtmax)
2740                 return;
2741
2742         /* Force an aging cycle; this might trim enough addresses. */
2743         bridge_rtage(sc);
2744         if (sc->sc_brtcnt <= sc->sc_brtmax)
2745                 return;
2746
2747         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2748                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2749                         bridge_rtnode_destroy(sc, brt);
2750                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2751                                 return;
2752                 }
2753         }
2754 }
2755
2756 /*
2757  * bridge_timer:
2758  *
2759  *      Aging timer for the bridge.
2760  */
2761 static void
2762 bridge_timer(void *arg)
2763 {
2764         struct bridge_softc *sc = arg;
2765
2766         BRIDGE_LOCK_ASSERT(sc);
2767
2768         bridge_rtage(sc);
2769
2770         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2771                 callout_reset(&sc->sc_brcallout,
2772                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2773 }
2774
2775 /*
2776  * bridge_rtage:
2777  *
2778  *      Perform an aging cycle.
2779  */
2780 static void
2781 bridge_rtage(struct bridge_softc *sc)
2782 {
2783         struct bridge_rtnode *brt, *nbrt;
2784
2785         BRIDGE_LOCK_ASSERT(sc);
2786
2787         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2788                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2789                         if (time_uptime >= brt->brt_expire)
2790                                 bridge_rtnode_destroy(sc, brt);
2791                 }
2792         }
2793 }
2794
2795 /*
2796  * bridge_rtflush:
2797  *
2798  *      Remove all dynamic addresses from the bridge.
2799  */
2800 static void
2801 bridge_rtflush(struct bridge_softc *sc, int full)
2802 {
2803         struct bridge_rtnode *brt, *nbrt;
2804
2805         BRIDGE_LOCK_ASSERT(sc);
2806
2807         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2808                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2809                         bridge_rtnode_destroy(sc, brt);
2810         }
2811 }
2812
2813 /*
2814  * bridge_rtdaddr:
2815  *
2816  *      Remove an address from the table.
2817  */
2818 static int
2819 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2820 {
2821         struct bridge_rtnode *brt;
2822         int found = 0;
2823
2824         BRIDGE_LOCK_ASSERT(sc);
2825
2826         /*
2827          * If vlan is zero then we want to delete for all vlans so the lookup
2828          * may return more than one.
2829          */
2830         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2831                 bridge_rtnode_destroy(sc, brt);
2832                 found = 1;
2833         }
2834
2835         return (found ? 0 : ENOENT);
2836 }
2837
2838 /*
2839  * bridge_rtdelete:
2840  *
2841  *      Delete routes to a speicifc member interface.
2842  */
2843 static void
2844 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2845 {
2846         struct bridge_rtnode *brt, *nbrt;
2847
2848         BRIDGE_LOCK_ASSERT(sc);
2849
2850         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2851                 if (brt->brt_ifp == ifp && (full ||
2852                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2853                         bridge_rtnode_destroy(sc, brt);
2854         }
2855 }
2856
2857 /*
2858  * bridge_rtable_init:
2859  *
2860  *      Initialize the route table for this bridge.
2861  */
2862 static void
2863 bridge_rtable_init(struct bridge_softc *sc)
2864 {
2865         int i;
2866
2867         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2868             M_DEVBUF, M_WAITOK);
2869
2870         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2871                 LIST_INIT(&sc->sc_rthash[i]);
2872
2873         sc->sc_rthash_key = arc4random();
2874         LIST_INIT(&sc->sc_rtlist);
2875 }
2876
2877 /*
2878  * bridge_rtable_fini:
2879  *
2880  *      Deconstruct the route table for this bridge.
2881  */
2882 static void
2883 bridge_rtable_fini(struct bridge_softc *sc)
2884 {
2885
2886         KASSERT(sc->sc_brtcnt == 0,
2887             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2888         free(sc->sc_rthash, M_DEVBUF);
2889 }
2890
2891 /*
2892  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2893  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2894  */
2895 #define mix(a, b, c)                                                    \
2896 do {                                                                    \
2897         a -= b; a -= c; a ^= (c >> 13);                                 \
2898         b -= c; b -= a; b ^= (a << 8);                                  \
2899         c -= a; c -= b; c ^= (b >> 13);                                 \
2900         a -= b; a -= c; a ^= (c >> 12);                                 \
2901         b -= c; b -= a; b ^= (a << 16);                                 \
2902         c -= a; c -= b; c ^= (b >> 5);                                  \
2903         a -= b; a -= c; a ^= (c >> 3);                                  \
2904         b -= c; b -= a; b ^= (a << 10);                                 \
2905         c -= a; c -= b; c ^= (b >> 15);                                 \
2906 } while (/*CONSTCOND*/0)
2907
2908 static __inline uint32_t
2909 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2910 {
2911         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2912
2913         b += addr[5] << 8;
2914         b += addr[4];
2915         a += addr[3] << 24;
2916         a += addr[2] << 16;
2917         a += addr[1] << 8;
2918         a += addr[0];
2919
2920         mix(a, b, c);
2921
2922         return (c & BRIDGE_RTHASH_MASK);
2923 }
2924
2925 #undef mix
2926
2927 static int
2928 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2929 {
2930         int i, d;
2931
2932         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2933                 d = ((int)a[i]) - ((int)b[i]);
2934         }
2935
2936         return (d);
2937 }
2938
2939 /*
2940  * bridge_rtnode_lookup:
2941  *
2942  *      Look up a bridge route node for the specified destination. Compare the
2943  *      vlan id or if zero then just return the first match.
2944  */
2945 static struct bridge_rtnode *
2946 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2947 {
2948         struct bridge_rtnode *brt;
2949         uint32_t hash;
2950         int dir;
2951
2952         BRIDGE_LOCK_ASSERT(sc);
2953
2954         hash = bridge_rthash(sc, addr);
2955         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2956                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2957                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2958                         return (brt);
2959                 if (dir > 0)
2960                         return (NULL);
2961         }
2962
2963         return (NULL);
2964 }
2965
2966 /*
2967  * bridge_rtnode_insert:
2968  *
2969  *      Insert the specified bridge node into the route table.  We
2970  *      assume the entry is not already in the table.
2971  */
2972 static int
2973 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2974 {
2975         struct bridge_rtnode *lbrt;
2976         uint32_t hash;
2977         int dir;
2978
2979         BRIDGE_LOCK_ASSERT(sc);
2980
2981         hash = bridge_rthash(sc, brt->brt_addr);
2982
2983         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2984         if (lbrt == NULL) {
2985                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2986                 goto out;
2987         }
2988
2989         do {
2990                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2991                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2992                         return (EEXIST);
2993                 if (dir > 0) {
2994                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2995                         goto out;
2996                 }
2997                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2998                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2999                         goto out;
3000                 }
3001                 lbrt = LIST_NEXT(lbrt, brt_hash);
3002         } while (lbrt != NULL);
3003
3004 #ifdef DIAGNOSTIC
3005         panic("bridge_rtnode_insert: impossible");
3006 #endif
3007
3008 out:
3009         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3010         sc->sc_brtcnt++;
3011
3012         return (0);
3013 }
3014
3015 /*
3016  * bridge_rtnode_destroy:
3017  *
3018  *      Destroy a bridge rtnode.
3019  */
3020 static void
3021 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3022 {
3023         BRIDGE_LOCK_ASSERT(sc);
3024
3025         LIST_REMOVE(brt, brt_hash);
3026
3027         LIST_REMOVE(brt, brt_list);
3028         sc->sc_brtcnt--;
3029         brt->brt_dst->bif_addrcnt--;
3030         uma_zfree(bridge_rtnode_zone, brt);
3031 }
3032
3033 /*
3034  * bridge_rtable_expire:
3035  *
3036  *      Set the expiry time for all routes on an interface.
3037  */
3038 static void
3039 bridge_rtable_expire(struct ifnet *ifp, int age)
3040 {
3041         struct bridge_softc *sc = ifp->if_bridge;
3042         struct bridge_rtnode *brt;
3043
3044         BRIDGE_LOCK(sc);
3045
3046         /*
3047          * If the age is zero then flush, otherwise set all the expiry times to
3048          * age for the interface
3049          */
3050         if (age == 0)
3051                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3052         else {
3053                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3054                         /* Cap the expiry time to 'age' */
3055                         if (brt->brt_ifp == ifp &&
3056                             brt->brt_expire > time_uptime + age &&
3057                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3058                                 brt->brt_expire = time_uptime + age;
3059                 }
3060         }
3061         BRIDGE_UNLOCK(sc);
3062 }
3063
3064 /*
3065  * bridge_state_change:
3066  *
3067  *      Callback from the bridgestp code when a port changes states.
3068  */
3069 static void
3070 bridge_state_change(struct ifnet *ifp, int state)
3071 {
3072         struct bridge_softc *sc = ifp->if_bridge;
3073         static const char *stpstates[] = {
3074                 "disabled",
3075                 "listening",
3076                 "learning",
3077                 "forwarding",
3078                 "blocking",
3079                 "discarding"
3080         };
3081
3082         CURVNET_SET(ifp->if_vnet);
3083         if (V_log_stp)
3084                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3085                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3086         CURVNET_RESTORE();
3087 }
3088
3089 /*
3090  * Send bridge packets through pfil if they are one of the types pfil can deal
3091  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3092  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3093  * that interface.
3094  */
3095 static int
3096 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3097 {
3098         int snap, error, i, hlen;
3099         struct ether_header *eh1, eh2;
3100         struct ip *ip;
3101         struct llc llc1;
3102         u_int16_t ether_type;
3103
3104         snap = 0;
3105         error = -1;     /* Default error if not error == 0 */
3106
3107 #if 0
3108         /* we may return with the IP fields swapped, ensure its not shared */
3109         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3110 #endif
3111
3112         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3113                 return (0); /* filtering is disabled */
3114
3115         i = min((*mp)->m_pkthdr.len, max_protohdr);
3116         if ((*mp)->m_len < i) {
3117             *mp = m_pullup(*mp, i);
3118             if (*mp == NULL) {
3119                 printf("%s: m_pullup failed\n", __func__);
3120                 return (-1);
3121             }
3122         }
3123
3124         eh1 = mtod(*mp, struct ether_header *);
3125         ether_type = ntohs(eh1->ether_type);
3126
3127         /*
3128          * Check for SNAP/LLC.
3129          */
3130         if (ether_type < ETHERMTU) {
3131                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3132
3133                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3134                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3135                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3136                     llc2->llc_control == LLC_UI) {
3137                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3138                         snap = 1;
3139                 }
3140         }
3141
3142         /*
3143          * If we're trying to filter bridge traffic, don't look at anything
3144          * other than IP and ARP traffic.  If the filter doesn't understand
3145          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3146          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3147          * but of course we don't have an AppleTalk filter to begin with.
3148          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3149          * ARP traffic.)
3150          */
3151         switch (ether_type) {
3152                 case ETHERTYPE_ARP:
3153                 case ETHERTYPE_REVARP:
3154                         if (V_pfil_ipfw_arp == 0)
3155                                 return (0); /* Automatically pass */
3156                         break;
3157
3158                 case ETHERTYPE_IP:
3159 #ifdef INET6
3160                 case ETHERTYPE_IPV6:
3161 #endif /* INET6 */
3162                         break;
3163                 default:
3164                         /*
3165                          * Check to see if the user wants to pass non-ip
3166                          * packets, these will not be checked by pfil(9) and
3167                          * passed unconditionally so the default is to drop.
3168                          */
3169                         if (V_pfil_onlyip)
3170                                 goto bad;
3171         }
3172
3173         /* Run the packet through pfil before stripping link headers */
3174         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3175                         dir == PFIL_OUT && ifp != NULL) {
3176
3177                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, NULL);
3178
3179                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3180                         return (error);
3181         }
3182
3183         /* Strip off the Ethernet header and keep a copy. */
3184         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3185         m_adj(*mp, ETHER_HDR_LEN);
3186
3187         /* Strip off snap header, if present */
3188         if (snap) {
3189                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3190                 m_adj(*mp, sizeof(struct llc));
3191         }
3192
3193         /*
3194          * Check the IP header for alignment and errors
3195          */
3196         if (dir == PFIL_IN) {
3197                 switch (ether_type) {
3198                         case ETHERTYPE_IP:
3199                                 error = bridge_ip_checkbasic(mp);
3200                                 break;
3201 #ifdef INET6
3202                         case ETHERTYPE_IPV6:
3203                                 error = bridge_ip6_checkbasic(mp);
3204                                 break;
3205 #endif /* INET6 */
3206                         default:
3207                                 error = 0;
3208                 }
3209                 if (error)
3210                         goto bad;
3211         }
3212
3213         error = 0;
3214
3215         /*
3216          * Run the packet through pfil
3217          */
3218         switch (ether_type) {
3219         case ETHERTYPE_IP:
3220                 /*
3221                  * Run pfil on the member interface and the bridge, both can
3222                  * be skipped by clearing pfil_member or pfil_bridge.
3223                  *
3224                  * Keep the order:
3225                  *   in_if -> bridge_if -> out_if
3226                  */
3227                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3228                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3229                                         dir, NULL);
3230
3231                 if (*mp == NULL || error != 0) /* filter may consume */
3232                         break;
3233
3234                 if (V_pfil_member && ifp != NULL)
3235                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3236                                         dir, NULL);
3237
3238                 if (*mp == NULL || error != 0) /* filter may consume */
3239                         break;
3240
3241                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3242                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3243                                         dir, NULL);
3244
3245                 if (*mp == NULL || error != 0) /* filter may consume */
3246                         break;
3247
3248                 /* check if we need to fragment the packet */
3249                 /* bridge_fragment generates a mbuf chain of packets */
3250                 /* that already include eth headers */
3251                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3252                         i = (*mp)->m_pkthdr.len;
3253                         if (i > ifp->if_mtu) {
3254                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3255                                             &llc1);
3256                                 return (error);
3257                         }
3258                 }
3259
3260                 /* Recalculate the ip checksum. */
3261                 ip = mtod(*mp, struct ip *);
3262                 hlen = ip->ip_hl << 2;
3263                 if (hlen < sizeof(struct ip))
3264                         goto bad;
3265                 if (hlen > (*mp)->m_len) {
3266                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3267                                 goto bad;
3268                         ip = mtod(*mp, struct ip *);
3269                         if (ip == NULL)
3270                                 goto bad;
3271                 }
3272                 ip->ip_sum = 0;
3273                 if (hlen == sizeof(struct ip))
3274                         ip->ip_sum = in_cksum_hdr(ip);
3275                 else
3276                         ip->ip_sum = in_cksum(*mp, hlen);
3277
3278                 break;
3279 #ifdef INET6
3280         case ETHERTYPE_IPV6:
3281                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3282                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3283                                         dir, NULL);
3284
3285                 if (*mp == NULL || error != 0) /* filter may consume */
3286                         break;
3287
3288                 if (V_pfil_member && ifp != NULL)
3289                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3290                                         dir, NULL);
3291
3292                 if (*mp == NULL || error != 0) /* filter may consume */
3293                         break;
3294
3295                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3296                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3297                                         dir, NULL);
3298                 break;
3299 #endif
3300         default:
3301                 error = 0;
3302                 break;
3303         }
3304
3305         if (*mp == NULL)
3306                 return (error);
3307         if (error != 0)
3308                 goto bad;
3309
3310         error = -1;
3311
3312         /*
3313          * Finally, put everything back the way it was and return
3314          */
3315         if (snap) {
3316                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3317                 if (*mp == NULL)
3318                         return (error);
3319                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3320         }
3321
3322         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3323         if (*mp == NULL)
3324                 return (error);
3325         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3326
3327         return (0);
3328
3329 bad:
3330         m_freem(*mp);
3331         *mp = NULL;
3332         return (error);
3333 }
3334
3335 /*
3336  * Perform basic checks on header size since
3337  * pfil assumes ip_input has already processed
3338  * it for it.  Cut-and-pasted from ip_input.c.
3339  * Given how simple the IPv6 version is,
3340  * does the IPv4 version really need to be
3341  * this complicated?
3342  *
3343  * XXX Should we update ipstat here, or not?
3344  * XXX Right now we update ipstat but not
3345  * XXX csum_counter.
3346  */
3347 static int
3348 bridge_ip_checkbasic(struct mbuf **mp)
3349 {
3350         struct mbuf *m = *mp;
3351         struct ip *ip;
3352         int len, hlen;
3353         u_short sum;
3354
3355         if (*mp == NULL)
3356                 return (-1);
3357
3358         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3359                 if ((m = m_copyup(m, sizeof(struct ip),
3360                         (max_linkhdr + 3) & ~3)) == NULL) {
3361                         /* XXXJRT new stat, please */
3362                         KMOD_IPSTAT_INC(ips_toosmall);
3363                         goto bad;
3364                 }
3365         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3366                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3367                         KMOD_IPSTAT_INC(ips_toosmall);
3368                         goto bad;
3369                 }
3370         }
3371         ip = mtod(m, struct ip *);
3372         if (ip == NULL) goto bad;
3373
3374         if (ip->ip_v != IPVERSION) {
3375                 KMOD_IPSTAT_INC(ips_badvers);
3376                 goto bad;
3377         }
3378         hlen = ip->ip_hl << 2;
3379         if (hlen < sizeof(struct ip)) { /* minimum header length */
3380                 KMOD_IPSTAT_INC(ips_badhlen);
3381                 goto bad;
3382         }
3383         if (hlen > m->m_len) {
3384                 if ((m = m_pullup(m, hlen)) == NULL) {
3385                         KMOD_IPSTAT_INC(ips_badhlen);
3386                         goto bad;
3387                 }
3388                 ip = mtod(m, struct ip *);
3389                 if (ip == NULL) goto bad;
3390         }
3391
3392         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3393                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3394         } else {
3395                 if (hlen == sizeof(struct ip)) {
3396                         sum = in_cksum_hdr(ip);
3397                 } else {
3398                         sum = in_cksum(m, hlen);
3399                 }
3400         }
3401         if (sum) {
3402                 KMOD_IPSTAT_INC(ips_badsum);
3403                 goto bad;
3404         }
3405
3406         /* Retrieve the packet length. */
3407         len = ntohs(ip->ip_len);
3408
3409         /*
3410          * Check for additional length bogosity
3411          */
3412         if (len < hlen) {
3413                 KMOD_IPSTAT_INC(ips_badlen);
3414                 goto bad;
3415         }
3416
3417         /*
3418          * Check that the amount of data in the buffers
3419          * is as at least much as the IP header would have us expect.
3420          * Drop packet if shorter than we expect.
3421          */
3422         if (m->m_pkthdr.len < len) {
3423                 KMOD_IPSTAT_INC(ips_tooshort);
3424                 goto bad;
3425         }
3426
3427         /* Checks out, proceed */
3428         *mp = m;
3429         return (0);
3430
3431 bad:
3432         *mp = m;
3433         return (-1);
3434 }
3435
3436 #ifdef INET6
3437 /*
3438  * Same as above, but for IPv6.
3439  * Cut-and-pasted from ip6_input.c.
3440  * XXX Should we update ip6stat, or not?
3441  */
3442 static int
3443 bridge_ip6_checkbasic(struct mbuf **mp)
3444 {
3445         struct mbuf *m = *mp;
3446         struct ip6_hdr *ip6;
3447
3448         /*
3449          * If the IPv6 header is not aligned, slurp it up into a new
3450          * mbuf with space for link headers, in the event we forward
3451          * it.  Otherwise, if it is aligned, make sure the entire base
3452          * IPv6 header is in the first mbuf of the chain.
3453          */
3454         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3455                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3456                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3457                             (max_linkhdr + 3) & ~3)) == NULL) {
3458                         /* XXXJRT new stat, please */
3459                         IP6STAT_INC(ip6s_toosmall);
3460                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3461                         goto bad;
3462                 }
3463         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3464                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3465                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3466                         IP6STAT_INC(ip6s_toosmall);
3467                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3468                         goto bad;
3469                 }
3470         }
3471
3472         ip6 = mtod(m, struct ip6_hdr *);
3473
3474         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3475                 IP6STAT_INC(ip6s_badvers);
3476                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3477                 goto bad;
3478         }
3479
3480         /* Checks out, proceed */
3481         *mp = m;
3482         return (0);
3483
3484 bad:
3485         *mp = m;
3486         return (-1);
3487 }
3488 #endif /* INET6 */
3489
3490 /*
3491  * bridge_fragment:
3492  *
3493  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3494  */
3495 static int
3496 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3497     int snap, struct llc *llc)
3498 {
3499         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3500         struct ip *ip;
3501         int error = -1;
3502
3503         if (m->m_len < sizeof(struct ip) &&
3504             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3505                 goto dropit;
3506         ip = mtod(m, struct ip *);
3507
3508         m->m_pkthdr.csum_flags |= CSUM_IP;
3509         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3510         if (error)
3511                 goto dropit;
3512
3513         /*
3514          * Walk the chain and re-add the Ethernet header for
3515          * each mbuf packet.
3516          */
3517         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3518                 nextpkt = mcur->m_nextpkt;
3519                 mcur->m_nextpkt = NULL;
3520                 if (snap) {
3521                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3522                         if (mcur == NULL) {
3523                                 error = ENOBUFS;
3524                                 if (mprev != NULL)
3525                                         mprev->m_nextpkt = nextpkt;
3526                                 goto dropit;
3527                         }
3528                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3529                 }
3530
3531                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3532                 if (mcur == NULL) {
3533                         error = ENOBUFS;
3534                         if (mprev != NULL)
3535                                 mprev->m_nextpkt = nextpkt;
3536                         goto dropit;
3537                 }
3538                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3539
3540                 /*
3541                  * The previous two M_PREPEND could have inserted one or two
3542                  * mbufs in front so we have to update the previous packet's
3543                  * m_nextpkt.
3544                  */
3545                 mcur->m_nextpkt = nextpkt;
3546                 if (mprev != NULL)
3547                         mprev->m_nextpkt = mcur;
3548                 else {
3549                         /* The first mbuf in the original chain needs to be
3550                          * updated. */
3551                         *mp = mcur;
3552                 }
3553                 mprev = mcur;
3554         }
3555
3556         KMOD_IPSTAT_INC(ips_fragmented);
3557         return (error);
3558
3559 dropit:
3560         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3561                 m = mcur->m_nextpkt;
3562                 m_freem(mcur);
3563         }
3564         return (error);
3565 }
3566
3567 static void
3568 bridge_linkstate(struct ifnet *ifp)
3569 {
3570         struct bridge_softc *sc = ifp->if_bridge;
3571         struct bridge_iflist *bif;
3572
3573         BRIDGE_LOCK(sc);
3574         bif = bridge_lookup_member_if(sc, ifp);
3575         if (bif == NULL) {
3576                 BRIDGE_UNLOCK(sc);
3577                 return;
3578         }
3579         bridge_linkcheck(sc);
3580         BRIDGE_UNLOCK(sc);
3581
3582         bstp_linkstate(&bif->bif_stp);
3583 }
3584
3585 static void
3586 bridge_linkcheck(struct bridge_softc *sc)
3587 {
3588         struct bridge_iflist *bif;
3589         int new_link, hasls;
3590
3591         BRIDGE_LOCK_ASSERT(sc);
3592         new_link = LINK_STATE_DOWN;
3593         hasls = 0;
3594         /* Our link is considered up if at least one of our ports is active */
3595         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3596                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3597                         hasls++;
3598                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3599                         new_link = LINK_STATE_UP;
3600                         break;
3601                 }
3602         }
3603         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3604                 /* If no interfaces support link-state then we default to up */
3605                 new_link = LINK_STATE_UP;
3606         }
3607         if_link_state_change(sc->sc_ifp, new_link);
3608 }