sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*-
   4  * SPDX-License-Identifier: BSD-4-Clause
   5  *
   6  * Copyright 2001 Wasabi Systems, Inc.
   7  * All rights reserved.
   8  *
   9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed for the NetBSD Project by
  22  *      Wasabi Systems, Inc.
  23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  24  *    or promote products derived from this software without specific prior
  25  *    written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 /*
  41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  42  * All rights reserved.
  43  *
  44  * Redistribution and use in source and binary forms, with or without
  45  * modification, are permitted provided that the following conditions
  46  * are met:
  47  * 1. Redistributions of source code must retain the above copyright
  48  *    notice, this list of conditions and the following disclaimer.
  49  * 2. Redistributions in binary form must reproduce the above copyright
  50  *    notice, this list of conditions and the following disclaimer in the
  51  *    documentation and/or other materials provided with the distribution.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  63  * POSSIBILITY OF SUCH DAMAGE.
  64  *
  65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  66  */
  67
  68 /*
  69  * Network interface bridge support.
  70  *
  71  * TODO:
  72  *
  73  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  74  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  75  *        to bridge other types of interfaces (maybe consider
  76  *        heterogeneous bridges).
  77  */
  78
  79 #include <sys/cdefs.h>
  80 __FBSDID("$FreeBSD$");
  81
  82 #include "opt_inet.h"
  83 #include "opt_inet6.h"
  84
  85 #include <sys/param.h>
  86 #include <sys/eventhandler.h>
  87 #include <sys/mbuf.h>
  88 #include <sys/malloc.h>
  89 #include <sys/protosw.h>
  90 #include <sys/systm.h>
  91 #include <sys/jail.h>
  92 #include <sys/time.h>
  93 #include <sys/socket.h> /* for net/if.h */
  94 #include <sys/sockio.h>
  95 #include <sys/ctype.h>  /* string functions */
  96 #include <sys/kernel.h>
  97 #include <sys/random.h>
  98 #include <sys/syslog.h>
  99 #include <sys/sysctl.h>
 100 #include <vm/uma.h>
 101 #include <sys/module.h>
 102 #include <sys/priv.h>
 103 #include <sys/proc.h>
 104 #include <sys/lock.h>
 105 #include <sys/mutex.h>
 106
 107 #include <net/bpf.h>
 108 #include <net/if.h>
 109 #include <net/if_clone.h>
 110 #include <net/if_dl.h>
 111 #include <net/if_types.h>
 112 #include <net/if_var.h>
 113 #include <net/pfil.h>
 114 #include <net/vnet.h>
 115
 116 #include <netinet/in.h>
 117 #include <netinet/in_systm.h>
 118 #include <netinet/in_var.h>
 119 #include <netinet/ip.h>
 120 #include <netinet/ip_var.h>
 121 #ifdef INET6
 122 #include <netinet/ip6.h>
 123 #include <netinet6/ip6_var.h>
 124 #include <netinet6/in6_ifattach.h>
 125 #endif
 126 #if defined(INET) || defined(INET6)
 127 #include <netinet/ip_carp.h>
 128 #endif
 129 #include <machine/in_cksum.h>
 130 #include <netinet/if_ether.h>
 131 #include <net/bridgestp.h>
 132 #include <net/if_bridgevar.h>
 133 #include <net/if_llc.h>
 134 #include <net/if_vlan_var.h>
 135
 136 #include <net/route.h>
 137
 138 #ifdef INET6
 139 /*
 140  * XXX: declare here to avoid to include many inet6 related files..
 141  * should be more generalized?
 142  */
 143 extern void     nd6_setmtu(struct ifnet *);
 144 #endif
 145
 146 /*
 147  * Size of the route hash table.  Must be a power of two.
 148  */
 149 #ifndef BRIDGE_RTHASH_SIZE
 150 #define BRIDGE_RTHASH_SIZE              1024
 151 #endif
 152
 153 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 154
 155 /*
 156  * Default maximum number of addresses to cache.
 157  */
 158 #ifndef BRIDGE_RTABLE_MAX
 159 #define BRIDGE_RTABLE_MAX               2000
 160 #endif
 161
 162 /*
 163  * Timeout (in seconds) for entries learned dynamically.
 164  */
 165 #ifndef BRIDGE_RTABLE_TIMEOUT
 166 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 167 #endif
 168
 169 /*
 170  * Number of seconds between walks of the route list.
 171  */
 172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 173 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 174 #endif
 175
 176 /*
 177  * List of capabilities to possibly mask on the member interface.
 178  */
 179 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 180                                          IFCAP_TXCSUM_IPV6)
 181
 182 /*
 183  * List of capabilities to strip
 184  */
 185 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 186
 187 /*
 188  * Bridge locking
 189  *
 190  * The bridge relies heavily on the epoch(9) system to protect its data
 191  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 192  * must ensure there is only one writer at a time.
 193  *
 194  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 195  * accesses we must hold:
 196  *
 197  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 198  *  - BRIDGE_LOCK, for any other change
 199  *
 200  * The BRIDGE_LOCK is a sleepable lock, because it is held accross ioctl()
 201  * calls to bridge member interfaces and these ioctl()s can sleep.
 202  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 203  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 204  */
 205 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 206         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 207         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 208 } while (0)
 209 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 210         sx_destroy(&(_sc)->sc_sx);              \
 211         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 212 } while (0)
 213 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 214 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 215 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 216 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 217             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 218 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 219 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 220 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 221 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 222 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 223             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 224
 225 /*
 226  * Bridge interface list entry.
 227  */
 228 struct bridge_iflist {
 229         CK_LIST_ENTRY(bridge_iflist) bif_next;
 230         struct ifnet            *bif_ifp;       /* member if */
 231         struct bstp_port        bif_stp;        /* STP state */
 232         uint32_t                bif_flags;      /* member if flags */
 233         int                     bif_savedcaps;  /* saved capabilities */
 234         uint32_t                bif_addrmax;    /* max # of addresses */
 235         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 236         uint32_t                bif_addrexceeded;/* # of address violations */
 237         struct epoch_context    bif_epoch_ctx;
 238 };
 239
 240 /*
 241  * Bridge route node.
 242  */
 243 struct bridge_rtnode {
 244         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 245         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 246         struct bridge_iflist    *brt_dst;       /* destination if */
 247         unsigned long           brt_expire;     /* expiration time */
 248         uint8_t                 brt_flags;      /* address flags */
 249         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 250         uint16_t                brt_vlan;       /* vlan id */
 251         struct  vnet            *brt_vnet;
 252         struct  epoch_context   brt_epoch_ctx;
 253 };
 254 #define brt_ifp                 brt_dst->bif_ifp
 255
 256 /*
 257  * Software state for each bridge.
 258  */
 259 struct bridge_softc {
 260         struct ifnet            *sc_ifp;        /* make this an interface */
 261         LIST_ENTRY(bridge_softc) sc_list;
 262         struct sx               sc_sx;
 263         struct mtx              sc_rt_mtx;
 264         uint32_t                sc_brtmax;      /* max # of addresses */
 265         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 266         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 267         struct callout          sc_brcallout;   /* bridge callout */
 268         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 269         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 270         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 271         uint32_t                sc_rthash_key;  /* key for hash */
 272         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 273         struct bstp_state       sc_stp;         /* STP state */
 274         uint32_t                sc_brtexceeded; /* # of cache drops */
 275         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 276         struct ether_addr       sc_defaddr;     /* Default MAC address */
 277         struct epoch_context    sc_epoch_ctx;
 278 };
 279
 280 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 281 #define V_bridge_list_sx        VNET(bridge_list_sx)
 282 static eventhandler_tag bridge_detach_cookie;
 283
 284 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 285
 286 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 287 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 288
 289 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 290 static void     bridge_clone_destroy(struct ifnet *);
 291
 292 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 293 static void     bridge_mutecaps(struct bridge_softc *);
 294 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 295                     int);
 296 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 297 static void     bridge_init(void *);
 298 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 299 static void     bridge_stop(struct ifnet *, int);
 300 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 301 #ifdef ALTQ
 302 static void     bridge_altq_start(if_t);
 303 static int      bridge_altq_transmit(if_t, struct mbuf *);
 304 #endif
 305 static void     bridge_qflush(struct ifnet *);
 306 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 307 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 308                     struct rtentry *);
 309 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 310                     struct mbuf *);
 311 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 312
 313 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 314                     struct mbuf *m);
 315
 316 static void     bridge_timer(void *);
 317
 318 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 319                     struct mbuf *, int);
 320 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 321
 322 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 323                     uint16_t, struct bridge_iflist *, int, uint8_t);
 324 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 325                     uint16_t);
 326 static void     bridge_rttrim(struct bridge_softc *);
 327 static void     bridge_rtage(struct bridge_softc *);
 328 static void     bridge_rtflush(struct bridge_softc *, int);
 329 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 330                     uint16_t);
 331
 332 static void     bridge_rtable_init(struct bridge_softc *);
 333 static void     bridge_rtable_fini(struct bridge_softc *);
 334
 335 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 336 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 337                     const uint8_t *, uint16_t);
 338 static int      bridge_rtnode_insert(struct bridge_softc *,
 339                     struct bridge_rtnode *);
 340 static void     bridge_rtnode_destroy(struct bridge_softc *,
 341                     struct bridge_rtnode *);
 342 static void     bridge_rtable_expire(struct ifnet *, int);
 343 static void     bridge_state_change(struct ifnet *, int);
 344
 345 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 346                     const char *name);
 347 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 348                     struct ifnet *ifp);
 349 static void     bridge_delete_member(struct bridge_softc *,
 350                     struct bridge_iflist *, int);
 351 static void     bridge_delete_span(struct bridge_softc *,
 352                     struct bridge_iflist *);
 353
 354 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 381 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 382 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 383 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 384 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 385 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 386                     int);
 387 static int      bridge_ip_checkbasic(struct mbuf **mp);
 388 #ifdef INET6
 389 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 390 #endif /* INET6 */
 391 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 392                     struct ether_header *, int, struct llc *);
 393 static void     bridge_linkstate(struct ifnet *ifp);
 394 static void     bridge_linkcheck(struct bridge_softc *sc);
 395
 396 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 397 #define VLANTAGOF(_m)   \
 398     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 399
 400 static struct bstp_cb_ops bridge_ops = {
 401         .bcb_state = bridge_state_change,
 402         .bcb_rtage = bridge_rtable_expire
 403 };
 404
 405 SYSCTL_DECL(_net_link);
 406 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 407     "Bridge");
 408
 409 /* only pass IP[46] packets when pfil is enabled */
 410 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 411 #define V_pfil_onlyip   VNET(pfil_onlyip)
 412 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 413     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 414     "Only pass IP packets when pfil is enabled");
 415
 416 /* run pfil hooks on the bridge interface */
 417 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 418 #define V_pfil_bridge   VNET(pfil_bridge)
 419 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 420     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 421     "Packet filter on the bridge interface");
 422
 423 /* layer2 filter with ipfw */
 424 VNET_DEFINE_STATIC(int, pfil_ipfw);
 425 #define V_pfil_ipfw     VNET(pfil_ipfw)
 426
 427 /* layer2 ARP filter with ipfw */
 428 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 429 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 430 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 431     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 432     "Filter ARP packets through IPFW layer2");
 433
 434 /* run pfil hooks on the member interface */
 435 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 436 #define V_pfil_member   VNET(pfil_member)
 437 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 438     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 439     "Packet filter on the member interface");
 440
 441 /* run pfil hooks on the physical interface for locally destined packets */
 442 VNET_DEFINE_STATIC(int, pfil_local_phys);
 443 #define V_pfil_local_phys       VNET(pfil_local_phys)
 444 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 445     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 446     "Packet filter on the physical interface for locally destined packets");
 447
 448 /* log STP state changes */
 449 VNET_DEFINE_STATIC(int, log_stp);
 450 #define V_log_stp       VNET(log_stp)
 451 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 452     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 453     "Log STP state changes");
 454
 455 /* share MAC with first bridge member */
 456 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 457 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 458 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 459     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 460     "Inherit MAC address from the first bridge member");
 461
 462 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 463 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 464 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 465     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 466     "Allow overlap of link-local scope "
 467     "zones of a bridge interface and the member interfaces");
 468
 469 struct bridge_control {
 470         int     (*bc_func)(struct bridge_softc *, void *);
 471         int     bc_argsize;
 472         int     bc_flags;
 473 };
 474
 475 #define BC_F_COPYIN             0x01    /* copy arguments in */
 476 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 477 #define BC_F_SUSER              0x04    /* do super-user check */
 478
 479 const struct bridge_control bridge_control_table[] = {
 480         { bridge_ioctl_add,             sizeof(struct ifbreq),
 481           BC_F_COPYIN|BC_F_SUSER },
 482         { bridge_ioctl_del,             sizeof(struct ifbreq),
 483           BC_F_COPYIN|BC_F_SUSER },
 484
 485         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 486           BC_F_COPYIN|BC_F_COPYOUT },
 487         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 488           BC_F_COPYIN|BC_F_SUSER },
 489
 490         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 491           BC_F_COPYIN|BC_F_SUSER },
 492         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 493           BC_F_COPYOUT },
 494
 495         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 496           BC_F_COPYIN|BC_F_COPYOUT },
 497         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 498           BC_F_COPYIN|BC_F_COPYOUT },
 499
 500         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 501           BC_F_COPYIN|BC_F_SUSER },
 502
 503         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 504           BC_F_COPYIN|BC_F_SUSER },
 505         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 506           BC_F_COPYOUT },
 507
 508         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 509           BC_F_COPYIN|BC_F_SUSER },
 510
 511         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 512           BC_F_COPYIN|BC_F_SUSER },
 513
 514         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 515           BC_F_COPYOUT },
 516         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 517           BC_F_COPYIN|BC_F_SUSER },
 518
 519         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 520           BC_F_COPYOUT },
 521         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 522           BC_F_COPYIN|BC_F_SUSER },
 523
 524         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 525           BC_F_COPYOUT },
 526         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 527           BC_F_COPYIN|BC_F_SUSER },
 528
 529         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 530           BC_F_COPYOUT },
 531         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 532           BC_F_COPYIN|BC_F_SUSER },
 533
 534         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 535           BC_F_COPYIN|BC_F_SUSER },
 536
 537         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 538           BC_F_COPYIN|BC_F_SUSER },
 539
 540         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 541           BC_F_COPYIN|BC_F_SUSER },
 542         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 543           BC_F_COPYIN|BC_F_SUSER },
 544
 545         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 546           BC_F_COPYOUT },
 547
 548         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 549           BC_F_COPYOUT },
 550
 551         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 552           BC_F_COPYIN|BC_F_COPYOUT },
 553
 554         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 555           BC_F_COPYIN|BC_F_SUSER },
 556
 557         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 558           BC_F_COPYIN|BC_F_SUSER },
 559
 560         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 561           BC_F_COPYIN|BC_F_SUSER },
 562
 563 };
 564 const int bridge_control_table_size = nitems(bridge_control_table);
 565
 566 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 567 #define V_bridge_list   VNET(bridge_list)
 568 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 569                                             "if_bridge list")
 570 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 571 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 572 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 573
 574 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 575 #define V_bridge_cloner VNET(bridge_cloner)
 576
 577 static const char bridge_name[] = "bridge";
 578
 579 static void
 580 vnet_bridge_init(const void *unused __unused)
 581 {
 582
 583         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 584             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 585             UMA_ALIGN_PTR, 0);
 586         BRIDGE_LIST_LOCK_INIT();
 587         LIST_INIT(&V_bridge_list);
 588         V_bridge_cloner = if_clone_simple(bridge_name,
 589             bridge_clone_create, bridge_clone_destroy, 0);
 590 }
 591 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 592     vnet_bridge_init, NULL);
 593
 594 static void
 595 vnet_bridge_uninit(const void *unused __unused)
 596 {
 597
 598         if_clone_detach(V_bridge_cloner);
 599         V_bridge_cloner = NULL;
 600         BRIDGE_LIST_LOCK_DESTROY();
 601
 602         /* Callbacks may use the UMA zone. */
 603         epoch_drain_callbacks(net_epoch_preempt);
 604
 605         uma_zdestroy(V_bridge_rtnode_zone);
 606 }
 607 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 608     vnet_bridge_uninit, NULL);
 609
 610 static int
 611 bridge_modevent(module_t mod, int type, void *data)
 612 {
 613
 614         switch (type) {
 615         case MOD_LOAD:
 616                 bridge_dn_p = bridge_dummynet;
 617                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 618                     ifnet_departure_event, bridge_ifdetach, NULL,
 619                     EVENTHANDLER_PRI_ANY);
 620                 break;
 621         case MOD_UNLOAD:
 622                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 623                     bridge_detach_cookie);
 624                 bridge_dn_p = NULL;
 625                 break;
 626         default:
 627                 return (EOPNOTSUPP);
 628         }
 629         return (0);
 630 }
 631
 632 static moduledata_t bridge_mod = {
 633         "if_bridge",
 634         bridge_modevent,
 635         0
 636 };
 637
 638 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 639 MODULE_VERSION(if_bridge, 1);
 640 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 641
 642 /*
 643  * handler for net.link.bridge.ipfw
 644  */
 645 static int
 646 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 647 {
 648         int enable = V_pfil_ipfw;
 649         int error;
 650
 651         error = sysctl_handle_int(oidp, &enable, 0, req);
 652         enable &= 1;
 653
 654         if (enable != V_pfil_ipfw) {
 655                 V_pfil_ipfw = enable;
 656
 657                 /*
 658                  * Disable pfil so that ipfw doesnt run twice, if the user
 659                  * really wants both then they can re-enable pfil_bridge and/or
 660                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 661                  * layer2 type.
 662                  */
 663                 if (V_pfil_ipfw) {
 664                         V_pfil_onlyip = 0;
 665                         V_pfil_bridge = 0;
 666                         V_pfil_member = 0;
 667                 }
 668         }
 669
 670         return (error);
 671 }
 672 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 673     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
 674     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 675     "Layer2 filter with IPFW");
 676
 677 #ifdef VIMAGE
 678 static void
 679 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 680 {
 681         struct bridge_softc *sc = ifp->if_softc;
 682         struct bridge_iflist *bif;
 683
 684         BRIDGE_LOCK(sc);
 685
 686         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 687                 bridge_delete_member(sc, bif, 0);
 688
 689         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 690                 bridge_delete_span(sc, bif);
 691         }
 692
 693         BRIDGE_UNLOCK(sc);
 694
 695         ether_reassign(ifp, newvnet, arg);
 696 }
 697 #endif
 698
 699 /*
 700  * bridge_clone_create:
 701  *
 702  *      Create a new bridge instance.
 703  */
 704 static int
 705 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 706 {
 707         struct bridge_softc *sc;
 708         struct ifnet *ifp;
 709
 710         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 711         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 712         if (ifp == NULL) {
 713                 free(sc, M_DEVBUF);
 714                 return (ENOSPC);
 715         }
 716
 717         BRIDGE_LOCK_INIT(sc);
 718         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 719         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 720
 721         /* Initialize our routing table. */
 722         bridge_rtable_init(sc);
 723
 724         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 725
 726         CK_LIST_INIT(&sc->sc_iflist);
 727         CK_LIST_INIT(&sc->sc_spanlist);
 728
 729         ifp->if_softc = sc;
 730         if_initname(ifp, bridge_name, unit);
 731         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 732         ifp->if_ioctl = bridge_ioctl;
 733 #ifdef ALTQ
 734         ifp->if_start = bridge_altq_start;
 735         ifp->if_transmit = bridge_altq_transmit;
 736         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 737         ifp->if_snd.ifq_drv_maxlen = 0;
 738         IFQ_SET_READY(&ifp->if_snd);
 739 #else
 740         ifp->if_transmit = bridge_transmit;
 741 #endif
 742         ifp->if_qflush = bridge_qflush;
 743         ifp->if_init = bridge_init;
 744         ifp->if_type = IFT_BRIDGE;
 745
 746         ether_gen_addr(ifp, &sc->sc_defaddr);
 747
 748         bstp_attach(&sc->sc_stp, &bridge_ops);
 749         ether_ifattach(ifp, sc->sc_defaddr.octet);
 750         /* Now undo some of the damage... */
 751         ifp->if_baudrate = 0;
 752         ifp->if_type = IFT_BRIDGE;
 753 #ifdef VIMAGE
 754         ifp->if_reassign = bridge_reassign;
 755 #endif
 756
 757         BRIDGE_LIST_LOCK();
 758         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 759         BRIDGE_LIST_UNLOCK();
 760
 761         return (0);
 762 }
 763
 764 static void
 765 bridge_clone_destroy_cb(struct epoch_context *ctx)
 766 {
 767         struct bridge_softc *sc;
 768
 769         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 770
 771         BRIDGE_LOCK_DESTROY(sc);
 772         free(sc, M_DEVBUF);
 773 }
 774
 775 /*
 776  * bridge_clone_destroy:
 777  *
 778  *      Destroy a bridge instance.
 779  */
 780 static void
 781 bridge_clone_destroy(struct ifnet *ifp)
 782 {
 783         struct bridge_softc *sc = ifp->if_softc;
 784         struct bridge_iflist *bif;
 785         struct epoch_tracker et;
 786
 787         BRIDGE_LOCK(sc);
 788
 789         bridge_stop(ifp, 1);
 790         ifp->if_flags &= ~IFF_UP;
 791
 792         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 793                 bridge_delete_member(sc, bif, 0);
 794
 795         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 796                 bridge_delete_span(sc, bif);
 797         }
 798
 799         /* Tear down the routing table. */
 800         bridge_rtable_fini(sc);
 801
 802         BRIDGE_UNLOCK(sc);
 803
 804         NET_EPOCH_ENTER(et);
 805
 806         callout_drain(&sc->sc_brcallout);
 807
 808         BRIDGE_LIST_LOCK();
 809         LIST_REMOVE(sc, sc_list);
 810         BRIDGE_LIST_UNLOCK();
 811
 812         bstp_detach(&sc->sc_stp);
 813 #ifdef ALTQ
 814         IFQ_PURGE(&ifp->if_snd);
 815 #endif
 816         NET_EPOCH_EXIT(et);
 817
 818         ether_ifdetach(ifp);
 819         if_free(ifp);
 820
 821         NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
 822 }
 823
 824 /*
 825  * bridge_ioctl:
 826  *
 827  *      Handle a control request from the operator.
 828  */
 829 static int
 830 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 831 {
 832         struct bridge_softc *sc = ifp->if_softc;
 833         struct ifreq *ifr = (struct ifreq *)data;
 834         struct bridge_iflist *bif;
 835         struct thread *td = curthread;
 836         union {
 837                 struct ifbreq ifbreq;
 838                 struct ifbifconf ifbifconf;
 839                 struct ifbareq ifbareq;
 840                 struct ifbaconf ifbaconf;
 841                 struct ifbrparam ifbrparam;
 842                 struct ifbropreq ifbropreq;
 843         } args;
 844         struct ifdrv *ifd = (struct ifdrv *) data;
 845         const struct bridge_control *bc;
 846         int error = 0, oldmtu;
 847
 848         BRIDGE_LOCK(sc);
 849
 850         switch (cmd) {
 851         case SIOCADDMULTI:
 852         case SIOCDELMULTI:
 853                 break;
 854
 855         case SIOCGDRVSPEC:
 856         case SIOCSDRVSPEC:
 857                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 858                         error = EINVAL;
 859                         break;
 860                 }
 861                 bc = &bridge_control_table[ifd->ifd_cmd];
 862
 863                 if (cmd == SIOCGDRVSPEC &&
 864                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 865                         error = EINVAL;
 866                         break;
 867                 }
 868                 else if (cmd == SIOCSDRVSPEC &&
 869                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 870                         error = EINVAL;
 871                         break;
 872                 }
 873
 874                 if (bc->bc_flags & BC_F_SUSER) {
 875                         error = priv_check(td, PRIV_NET_BRIDGE);
 876                         if (error)
 877                                 break;
 878                 }
 879
 880                 if (ifd->ifd_len != bc->bc_argsize ||
 881                     ifd->ifd_len > sizeof(args)) {
 882                         error = EINVAL;
 883                         break;
 884                 }
 885
 886                 bzero(&args, sizeof(args));
 887                 if (bc->bc_flags & BC_F_COPYIN) {
 888                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 889                         if (error)
 890                                 break;
 891                 }
 892
 893                 oldmtu = ifp->if_mtu;
 894                 error = (*bc->bc_func)(sc, &args);
 895                 if (error)
 896                         break;
 897
 898                 /*
 899                  * Bridge MTU may change during addition of the first port.
 900                  * If it did, do network layer specific procedure.
 901                  */
 902                 if (ifp->if_mtu != oldmtu) {
 903 #ifdef INET6
 904                         nd6_setmtu(ifp);
 905 #endif
 906                         rt_updatemtu(ifp);
 907                 }
 908
 909                 if (bc->bc_flags & BC_F_COPYOUT)
 910                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 911
 912                 break;
 913
 914         case SIOCSIFFLAGS:
 915                 if (!(ifp->if_flags & IFF_UP) &&
 916                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 917                         /*
 918                          * If interface is marked down and it is running,
 919                          * then stop and disable it.
 920                          */
 921                         bridge_stop(ifp, 1);
 922                 } else if ((ifp->if_flags & IFF_UP) &&
 923                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 924                         /*
 925                          * If interface is marked up and it is stopped, then
 926                          * start it.
 927                          */
 928                         BRIDGE_UNLOCK(sc);
 929                         (*ifp->if_init)(sc);
 930                         BRIDGE_LOCK(sc);
 931                 }
 932                 break;
 933
 934         case SIOCSIFMTU:
 935                 oldmtu = sc->sc_ifp->if_mtu;
 936
 937                 if (ifr->ifr_mtu < 576) {
 938                         error = EINVAL;
 939                         break;
 940                 }
 941                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 942                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 943                         break;
 944                 }
 945                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 946                         error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 947                             SIOCSIFMTU, (caddr_t)ifr);
 948                         if (error != 0) {
 949                                 log(LOG_NOTICE, "%s: invalid MTU: %u for"
 950                                     " member %s\n", sc->sc_ifp->if_xname,
 951                                     ifr->ifr_mtu,
 952                                     bif->bif_ifp->if_xname);
 953                                 error = EINVAL;
 954                                 break;
 955                         }
 956                 }
 957                 if (error) {
 958                         /* Restore the previous MTU on all member interfaces. */
 959                         ifr->ifr_mtu = oldmtu;
 960                         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 961                                 (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 962                                     SIOCSIFMTU, (caddr_t)ifr);
 963                         }
 964                 } else {
 965                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 966                 }
 967                 break;
 968         default:
 969                 /*
 970                  * drop the lock as ether_ioctl() will call bridge_start() and
 971                  * cause the lock to be recursed.
 972                  */
 973                 BRIDGE_UNLOCK(sc);
 974                 error = ether_ioctl(ifp, cmd, data);
 975                 BRIDGE_LOCK(sc);
 976                 break;
 977         }
 978
 979         BRIDGE_UNLOCK(sc);
 980
 981         return (error);
 982 }
 983
 984 /*
 985  * bridge_mutecaps:
 986  *
 987  *      Clear or restore unwanted capabilities on the member interface
 988  */
 989 static void
 990 bridge_mutecaps(struct bridge_softc *sc)
 991 {
 992         struct bridge_iflist *bif;
 993         int enabled, mask;
 994
 995         BRIDGE_LOCK_ASSERT(sc);
 996
 997         /* Initial bitmask of capabilities to test */
 998         mask = BRIDGE_IFCAPS_MASK;
 999
1000         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1001                 /* Every member must support it or its disabled */
1002                 mask &= bif->bif_savedcaps;
1003         }
1004
1005         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1006                 enabled = bif->bif_ifp->if_capenable;
1007                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1008                 /* strip off mask bits and enable them again if allowed */
1009                 enabled &= ~BRIDGE_IFCAPS_MASK;
1010                 enabled |= mask;
1011                 bridge_set_ifcap(sc, bif, enabled);
1012         }
1013 }
1014
1015 static void
1016 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1017 {
1018         struct ifnet *ifp = bif->bif_ifp;
1019         struct ifreq ifr;
1020         int error, mask, stuck;
1021
1022         bzero(&ifr, sizeof(ifr));
1023         ifr.ifr_reqcap = set;
1024
1025         if (ifp->if_capenable != set) {
1026                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1027                 if (error)
1028                         if_printf(sc->sc_ifp,
1029                             "error setting capabilities on %s: %d\n",
1030                             ifp->if_xname, error);
1031                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1032                 stuck = ifp->if_capenable & mask & ~set;
1033                 if (stuck != 0)
1034                         if_printf(sc->sc_ifp,
1035                             "can't disable some capabilities on %s: 0x%x\n",
1036                             ifp->if_xname, stuck);
1037         }
1038 }
1039
1040 /*
1041  * bridge_lookup_member:
1042  *
1043  *      Lookup a bridge member interface.
1044  */
1045 static struct bridge_iflist *
1046 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1047 {
1048         struct bridge_iflist *bif;
1049         struct ifnet *ifp;
1050
1051         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1052
1053         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1054                 ifp = bif->bif_ifp;
1055                 if (strcmp(ifp->if_xname, name) == 0)
1056                         return (bif);
1057         }
1058
1059         return (NULL);
1060 }
1061
1062 /*
1063  * bridge_lookup_member_if:
1064  *
1065  *      Lookup a bridge member interface by ifnet*.
1066  */
1067 static struct bridge_iflist *
1068 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1069 {
1070         struct bridge_iflist *bif;
1071
1072         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1073
1074         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1075                 if (bif->bif_ifp == member_ifp)
1076                         return (bif);
1077         }
1078
1079         return (NULL);
1080 }
1081
1082 static void
1083 bridge_delete_member_cb(struct epoch_context *ctx)
1084 {
1085         struct bridge_iflist *bif;
1086
1087         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1088
1089         free(bif, M_DEVBUF);
1090 }
1091
1092 /*
1093  * bridge_delete_member:
1094  *
1095  *      Delete the specified member interface.
1096  */
1097 static void
1098 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1099     int gone)
1100 {
1101         struct ifnet *ifs = bif->bif_ifp;
1102         struct ifnet *fif = NULL;
1103         struct bridge_iflist *bifl;
1104
1105         BRIDGE_LOCK_ASSERT(sc);
1106
1107         if (bif->bif_flags & IFBIF_STP)
1108                 bstp_disable(&bif->bif_stp);
1109
1110         ifs->if_bridge = NULL;
1111         CK_LIST_REMOVE(bif, bif_next);
1112
1113         /*
1114          * If removing the interface that gave the bridge its mac address, set
1115          * the mac address of the bridge to the address of the next member, or
1116          * to its default address if no members are left.
1117          */
1118         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1119                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1120                         bcopy(&sc->sc_defaddr,
1121                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1122                         sc->sc_ifaddr = NULL;
1123                 } else {
1124                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1125                         fif = bifl->bif_ifp;
1126                         bcopy(IF_LLADDR(fif),
1127                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1128                         sc->sc_ifaddr = fif;
1129                 }
1130                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1131         }
1132
1133         bridge_linkcheck(sc);
1134         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1135         BRIDGE_RT_LOCK(sc);
1136         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1137         BRIDGE_RT_UNLOCK(sc);
1138         KASSERT(bif->bif_addrcnt == 0,
1139             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1140
1141         ifs->if_bridge_output = NULL;
1142         ifs->if_bridge_input = NULL;
1143         ifs->if_bridge_linkstate = NULL;
1144         if (!gone) {
1145                 switch (ifs->if_type) {
1146                 case IFT_ETHER:
1147                 case IFT_L2VLAN:
1148                         /*
1149                          * Take the interface out of promiscuous mode, but only
1150                          * if it was promiscuous in the first place. It might
1151                          * not be if we're in the bridge_ioctl_add() error path.
1152                          */
1153                         if (ifs->if_flags & IFF_PROMISC)
1154                                 (void) ifpromisc(ifs, 0);
1155                         break;
1156
1157                 case IFT_GIF:
1158                         break;
1159
1160                 default:
1161 #ifdef DIAGNOSTIC
1162                         panic("bridge_delete_member: impossible");
1163 #endif
1164                         break;
1165                 }
1166                 /* reneable any interface capabilities */
1167                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1168         }
1169         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1170
1171         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1172 }
1173
1174 /*
1175  * bridge_delete_span:
1176  *
1177  *      Delete the specified span interface.
1178  */
1179 static void
1180 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1181 {
1182         BRIDGE_LOCK_ASSERT(sc);
1183
1184         KASSERT(bif->bif_ifp->if_bridge == NULL,
1185             ("%s: not a span interface", __func__));
1186
1187         CK_LIST_REMOVE(bif, bif_next);
1188
1189         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1190 }
1191
1192 static int
1193 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1194 {
1195         struct ifbreq *req = arg;
1196         struct bridge_iflist *bif = NULL;
1197         struct ifnet *ifs;
1198         int error = 0;
1199
1200         ifs = ifunit(req->ifbr_ifsname);
1201         if (ifs == NULL)
1202                 return (ENOENT);
1203         if (ifs->if_ioctl == NULL)      /* must be supported */
1204                 return (EINVAL);
1205
1206         /* If it's in the span list, it can't be a member. */
1207         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1208                 if (ifs == bif->bif_ifp)
1209                         return (EBUSY);
1210
1211         if (ifs->if_bridge == sc)
1212                 return (EEXIST);
1213
1214         if (ifs->if_bridge != NULL)
1215                 return (EBUSY);
1216
1217         switch (ifs->if_type) {
1218         case IFT_ETHER:
1219         case IFT_L2VLAN:
1220         case IFT_GIF:
1221                 /* permitted interface types */
1222                 break;
1223         default:
1224                 return (EINVAL);
1225         }
1226
1227 #ifdef INET6
1228         /*
1229          * Two valid inet6 addresses with link-local scope must not be
1230          * on the parent interface and the member interfaces at the
1231          * same time.  This restriction is needed to prevent violation
1232          * of link-local scope zone.  Attempts to add a member
1233          * interface which has inet6 addresses when the parent has
1234          * inet6 triggers removal of all inet6 addresses on the member
1235          * interface.
1236          */
1237
1238         /* Check if the parent interface has a link-local scope addr. */
1239         if (V_allow_llz_overlap == 0 &&
1240             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1241                 /*
1242                  * If any, remove all inet6 addresses from the member
1243                  * interfaces.
1244                  */
1245                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1246                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1247                                 in6_ifdetach(bif->bif_ifp);
1248                                 if_printf(sc->sc_ifp,
1249                                     "IPv6 addresses on %s have been removed "
1250                                     "before adding it as a member to prevent "
1251                                     "IPv6 address scope violation.\n",
1252                                     bif->bif_ifp->if_xname);
1253                         }
1254                 }
1255                 if (in6ifa_llaonifp(ifs)) {
1256                         in6_ifdetach(ifs);
1257                         if_printf(sc->sc_ifp,
1258                             "IPv6 addresses on %s have been removed "
1259                             "before adding it as a member to prevent "
1260                             "IPv6 address scope violation.\n",
1261                             ifs->if_xname);
1262                 }
1263         }
1264 #endif
1265         /* Allow the first Ethernet member to define the MTU */
1266         if (CK_LIST_EMPTY(&sc->sc_iflist))
1267                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1268         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1269                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1270                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1271                 return (EINVAL);
1272         }
1273
1274         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1275         if (bif == NULL)
1276                 return (ENOMEM);
1277
1278         bif->bif_ifp = ifs;
1279         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1280         bif->bif_savedcaps = ifs->if_capenable;
1281
1282         /*
1283          * Assign the interface's MAC address to the bridge if it's the first
1284          * member and the MAC address of the bridge has not been changed from
1285          * the default randomly generated one.
1286          */
1287         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1288             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1289                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1290                 sc->sc_ifaddr = ifs;
1291                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1292         }
1293
1294         ifs->if_bridge = sc;
1295         ifs->if_bridge_output = bridge_output;
1296         ifs->if_bridge_input = bridge_input;
1297         ifs->if_bridge_linkstate = bridge_linkstate;
1298         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1299         /*
1300          * XXX: XLOCK HERE!?!
1301          *
1302          * NOTE: insert_***HEAD*** should be safe for the traversals.
1303          */
1304         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1305
1306         /* Set interface capabilities to the intersection set of all members */
1307         bridge_mutecaps(sc);
1308         bridge_linkcheck(sc);
1309
1310         /* Place the interface into promiscuous mode */
1311         switch (ifs->if_type) {
1312                 case IFT_ETHER:
1313                 case IFT_L2VLAN:
1314                         error = ifpromisc(ifs, 1);
1315                         break;
1316         }
1317
1318         if (error)
1319                 bridge_delete_member(sc, bif, 0);
1320         return (error);
1321 }
1322
1323 static int
1324 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1325 {
1326         struct ifbreq *req = arg;
1327         struct bridge_iflist *bif;
1328
1329         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1330         if (bif == NULL)
1331                 return (ENOENT);
1332
1333         bridge_delete_member(sc, bif, 0);
1334
1335         return (0);
1336 }
1337
1338 static int
1339 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1340 {
1341         struct ifbreq *req = arg;
1342         struct bridge_iflist *bif;
1343         struct bstp_port *bp;
1344
1345         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1346         if (bif == NULL)
1347                 return (ENOENT);
1348
1349         bp = &bif->bif_stp;
1350         req->ifbr_ifsflags = bif->bif_flags;
1351         req->ifbr_state = bp->bp_state;
1352         req->ifbr_priority = bp->bp_priority;
1353         req->ifbr_path_cost = bp->bp_path_cost;
1354         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1355         req->ifbr_proto = bp->bp_protover;
1356         req->ifbr_role = bp->bp_role;
1357         req->ifbr_stpflags = bp->bp_flags;
1358         req->ifbr_addrcnt = bif->bif_addrcnt;
1359         req->ifbr_addrmax = bif->bif_addrmax;
1360         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1361
1362         /* Copy STP state options as flags */
1363         if (bp->bp_operedge)
1364                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1365         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1366                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1367         if (bp->bp_ptp_link)
1368                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1369         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1370                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1371         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1372                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1373         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1374                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1375         return (0);
1376 }
1377
1378 static int
1379 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1380 {
1381         struct epoch_tracker et;
1382         struct ifbreq *req = arg;
1383         struct bridge_iflist *bif;
1384         struct bstp_port *bp;
1385         int error;
1386
1387         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1388         if (bif == NULL)
1389                 return (ENOENT);
1390         bp = &bif->bif_stp;
1391
1392         if (req->ifbr_ifsflags & IFBIF_SPAN)
1393                 /* SPAN is readonly */
1394                 return (EINVAL);
1395
1396         NET_EPOCH_ENTER(et);
1397
1398         if (req->ifbr_ifsflags & IFBIF_STP) {
1399                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1400                         error = bstp_enable(&bif->bif_stp);
1401                         if (error) {
1402                                 NET_EPOCH_EXIT(et);
1403                                 return (error);
1404                         }
1405                 }
1406         } else {
1407                 if ((bif->bif_flags & IFBIF_STP) != 0)
1408                         bstp_disable(&bif->bif_stp);
1409         }
1410
1411         /* Pass on STP flags */
1412         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1413         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1414         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1415         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1416
1417         /* Save the bits relating to the bridge */
1418         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1419
1420         NET_EPOCH_EXIT(et);
1421
1422         return (0);
1423 }
1424
1425 static int
1426 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1427 {
1428         struct ifbrparam *param = arg;
1429
1430         sc->sc_brtmax = param->ifbrp_csize;
1431         bridge_rttrim(sc);
1432
1433         return (0);
1434 }
1435
1436 static int
1437 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1438 {
1439         struct ifbrparam *param = arg;
1440
1441         param->ifbrp_csize = sc->sc_brtmax;
1442
1443         return (0);
1444 }
1445
1446 static int
1447 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1448 {
1449         struct ifbifconf *bifc = arg;
1450         struct bridge_iflist *bif;
1451         struct ifbreq breq;
1452         char *buf, *outbuf;
1453         int count, buflen, len, error = 0;
1454
1455         count = 0;
1456         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1457                 count++;
1458         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1459                 count++;
1460
1461         buflen = sizeof(breq) * count;
1462         if (bifc->ifbic_len == 0) {
1463                 bifc->ifbic_len = buflen;
1464                 return (0);
1465         }
1466         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1467         if (outbuf == NULL)
1468                 return (ENOMEM);
1469
1470         count = 0;
1471         buf = outbuf;
1472         len = min(bifc->ifbic_len, buflen);
1473         bzero(&breq, sizeof(breq));
1474         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1475                 if (len < sizeof(breq))
1476                         break;
1477
1478                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1479                     sizeof(breq.ifbr_ifsname));
1480                 /* Fill in the ifbreq structure */
1481                 error = bridge_ioctl_gifflags(sc, &breq);
1482                 if (error)
1483                         break;
1484                 memcpy(buf, &breq, sizeof(breq));
1485                 count++;
1486                 buf += sizeof(breq);
1487                 len -= sizeof(breq);
1488         }
1489         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1490                 if (len < sizeof(breq))
1491                         break;
1492
1493                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1494                     sizeof(breq.ifbr_ifsname));
1495                 breq.ifbr_ifsflags = bif->bif_flags;
1496                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1497                 memcpy(buf, &breq, sizeof(breq));
1498                 count++;
1499                 buf += sizeof(breq);
1500                 len -= sizeof(breq);
1501         }
1502
1503         bifc->ifbic_len = sizeof(breq) * count;
1504         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1505         free(outbuf, M_TEMP);
1506         return (error);
1507 }
1508
1509 static int
1510 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1511 {
1512         struct ifbaconf *bac = arg;
1513         struct bridge_rtnode *brt;
1514         struct ifbareq bareq;
1515         char *buf, *outbuf;
1516         int count, buflen, len, error = 0;
1517
1518         if (bac->ifbac_len == 0)
1519                 return (0);
1520
1521         count = 0;
1522         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1523                 count++;
1524         buflen = sizeof(bareq) * count;
1525
1526         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1527         if (outbuf == NULL)
1528                 return (ENOMEM);
1529
1530         count = 0;
1531         buf = outbuf;
1532         len = min(bac->ifbac_len, buflen);
1533         bzero(&bareq, sizeof(bareq));
1534         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1535                 if (len < sizeof(bareq))
1536                         goto out;
1537                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1538                     sizeof(bareq.ifba_ifsname));
1539                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1540                 bareq.ifba_vlan = brt->brt_vlan;
1541                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1542                                 time_uptime < brt->brt_expire)
1543                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1544                 else
1545                         bareq.ifba_expire = 0;
1546                 bareq.ifba_flags = brt->brt_flags;
1547
1548                 memcpy(buf, &bareq, sizeof(bareq));
1549                 count++;
1550                 buf += sizeof(bareq);
1551                 len -= sizeof(bareq);
1552         }
1553 out:
1554         bac->ifbac_len = sizeof(bareq) * count;
1555         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1556         free(outbuf, M_TEMP);
1557         return (error);
1558 }
1559
1560 static int
1561 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1562 {
1563         struct ifbareq *req = arg;
1564         struct bridge_iflist *bif;
1565         struct epoch_tracker et;
1566         int error;
1567
1568         NET_EPOCH_ENTER(et);
1569         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1570         if (bif == NULL) {
1571                 NET_EPOCH_EXIT(et);
1572                 return (ENOENT);
1573         }
1574
1575         /* bridge_rtupdate() may acquire the lock. */
1576         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1577             req->ifba_flags);
1578         NET_EPOCH_EXIT(et);
1579
1580         return (error);
1581 }
1582
1583 static int
1584 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1585 {
1586         struct ifbrparam *param = arg;
1587
1588         sc->sc_brttimeout = param->ifbrp_ctime;
1589         return (0);
1590 }
1591
1592 static int
1593 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1594 {
1595         struct ifbrparam *param = arg;
1596
1597         param->ifbrp_ctime = sc->sc_brttimeout;
1598         return (0);
1599 }
1600
1601 static int
1602 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1603 {
1604         struct ifbareq *req = arg;
1605
1606         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1607 }
1608
1609 static int
1610 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1611 {
1612         struct ifbreq *req = arg;
1613
1614         BRIDGE_RT_LOCK(sc);
1615         bridge_rtflush(sc, req->ifbr_ifsflags);
1616         BRIDGE_RT_UNLOCK(sc);
1617
1618         return (0);
1619 }
1620
1621 static int
1622 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1623 {
1624         struct ifbrparam *param = arg;
1625         struct bstp_state *bs = &sc->sc_stp;
1626
1627         param->ifbrp_prio = bs->bs_bridge_priority;
1628         return (0);
1629 }
1630
1631 static int
1632 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1633 {
1634         struct ifbrparam *param = arg;
1635
1636         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1637 }
1638
1639 static int
1640 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1641 {
1642         struct ifbrparam *param = arg;
1643         struct bstp_state *bs = &sc->sc_stp;
1644
1645         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1646         return (0);
1647 }
1648
1649 static int
1650 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1651 {
1652         struct ifbrparam *param = arg;
1653
1654         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1655 }
1656
1657 static int
1658 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1659 {
1660         struct ifbrparam *param = arg;
1661         struct bstp_state *bs = &sc->sc_stp;
1662
1663         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1664         return (0);
1665 }
1666
1667 static int
1668 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1669 {
1670         struct ifbrparam *param = arg;
1671
1672         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1673 }
1674
1675 static int
1676 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1677 {
1678         struct ifbrparam *param = arg;
1679         struct bstp_state *bs = &sc->sc_stp;
1680
1681         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1682         return (0);
1683 }
1684
1685 static int
1686 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1687 {
1688         struct ifbrparam *param = arg;
1689
1690         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1691 }
1692
1693 static int
1694 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1695 {
1696         struct ifbreq *req = arg;
1697         struct bridge_iflist *bif;
1698
1699         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1700         if (bif == NULL)
1701                 return (ENOENT);
1702
1703         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1704 }
1705
1706 static int
1707 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1708 {
1709         struct ifbreq *req = arg;
1710         struct bridge_iflist *bif;
1711
1712         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1713         if (bif == NULL)
1714                 return (ENOENT);
1715
1716         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1717 }
1718
1719 static int
1720 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1721 {
1722         struct ifbreq *req = arg;
1723         struct bridge_iflist *bif;
1724
1725         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1726         if (bif == NULL)
1727                 return (ENOENT);
1728
1729         bif->bif_addrmax = req->ifbr_addrmax;
1730         return (0);
1731 }
1732
1733 static int
1734 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1735 {
1736         struct ifbreq *req = arg;
1737         struct bridge_iflist *bif = NULL;
1738         struct ifnet *ifs;
1739
1740         ifs = ifunit(req->ifbr_ifsname);
1741         if (ifs == NULL)
1742                 return (ENOENT);
1743
1744         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1745                 if (ifs == bif->bif_ifp)
1746                         return (EBUSY);
1747
1748         if (ifs->if_bridge != NULL)
1749                 return (EBUSY);
1750
1751         switch (ifs->if_type) {
1752                 case IFT_ETHER:
1753                 case IFT_GIF:
1754                 case IFT_L2VLAN:
1755                         break;
1756                 default:
1757                         return (EINVAL);
1758         }
1759
1760         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1761         if (bif == NULL)
1762                 return (ENOMEM);
1763
1764         bif->bif_ifp = ifs;
1765         bif->bif_flags = IFBIF_SPAN;
1766
1767         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1768
1769         return (0);
1770 }
1771
1772 static int
1773 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1774 {
1775         struct ifbreq *req = arg;
1776         struct bridge_iflist *bif;
1777         struct ifnet *ifs;
1778
1779         ifs = ifunit(req->ifbr_ifsname);
1780         if (ifs == NULL)
1781                 return (ENOENT);
1782
1783         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1784                 if (ifs == bif->bif_ifp)
1785                         break;
1786
1787         if (bif == NULL)
1788                 return (ENOENT);
1789
1790         bridge_delete_span(sc, bif);
1791
1792         return (0);
1793 }
1794
1795 static int
1796 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1797 {
1798         struct ifbropreq *req = arg;
1799         struct bstp_state *bs = &sc->sc_stp;
1800         struct bstp_port *root_port;
1801
1802         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1803         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1804         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1805
1806         root_port = bs->bs_root_port;
1807         if (root_port == NULL)
1808                 req->ifbop_root_port = 0;
1809         else
1810                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1811
1812         req->ifbop_holdcount = bs->bs_txholdcount;
1813         req->ifbop_priority = bs->bs_bridge_priority;
1814         req->ifbop_protocol = bs->bs_protover;
1815         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1816         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1817         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1818         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1819         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1820         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1821
1822         return (0);
1823 }
1824
1825 static int
1826 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1827 {
1828         struct ifbrparam *param = arg;
1829
1830         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1831         return (0);
1832 }
1833
1834 static int
1835 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1836 {
1837         struct ifbpstpconf *bifstp = arg;
1838         struct bridge_iflist *bif;
1839         struct bstp_port *bp;
1840         struct ifbpstpreq bpreq;
1841         char *buf, *outbuf;
1842         int count, buflen, len, error = 0;
1843
1844         count = 0;
1845         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1846                 if ((bif->bif_flags & IFBIF_STP) != 0)
1847                         count++;
1848         }
1849
1850         buflen = sizeof(bpreq) * count;
1851         if (bifstp->ifbpstp_len == 0) {
1852                 bifstp->ifbpstp_len = buflen;
1853                 return (0);
1854         }
1855
1856         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1857         if (outbuf == NULL)
1858                 return (ENOMEM);
1859
1860         count = 0;
1861         buf = outbuf;
1862         len = min(bifstp->ifbpstp_len, buflen);
1863         bzero(&bpreq, sizeof(bpreq));
1864         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1865                 if (len < sizeof(bpreq))
1866                         break;
1867
1868                 if ((bif->bif_flags & IFBIF_STP) == 0)
1869                         continue;
1870
1871                 bp = &bif->bif_stp;
1872                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1873                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1874                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1875                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1876                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1877                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1878
1879                 memcpy(buf, &bpreq, sizeof(bpreq));
1880                 count++;
1881                 buf += sizeof(bpreq);
1882                 len -= sizeof(bpreq);
1883         }
1884
1885         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1886         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1887         free(outbuf, M_TEMP);
1888         return (error);
1889 }
1890
1891 static int
1892 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1893 {
1894         struct ifbrparam *param = arg;
1895
1896         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1897 }
1898
1899 static int
1900 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1901 {
1902         struct ifbrparam *param = arg;
1903
1904         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1905 }
1906
1907 /*
1908  * bridge_ifdetach:
1909  *
1910  *      Detach an interface from a bridge.  Called when a member
1911  *      interface is detaching.
1912  */
1913 static void
1914 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1915 {
1916         struct bridge_softc *sc = ifp->if_bridge;
1917         struct bridge_iflist *bif;
1918
1919         if (ifp->if_flags & IFF_RENAMING)
1920                 return;
1921         if (V_bridge_cloner == NULL) {
1922                 /*
1923                  * This detach handler can be called after
1924                  * vnet_bridge_uninit().  Just return in that case.
1925                  */
1926                 return;
1927         }
1928         /* Check if the interface is a bridge member */
1929         if (sc != NULL) {
1930                 BRIDGE_LOCK(sc);
1931
1932                 bif = bridge_lookup_member_if(sc, ifp);
1933                 if (bif != NULL)
1934                         bridge_delete_member(sc, bif, 1);
1935
1936                 BRIDGE_UNLOCK(sc);
1937                 return;
1938         }
1939
1940         /* Check if the interface is a span port */
1941         BRIDGE_LIST_LOCK();
1942         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1943                 BRIDGE_LOCK(sc);
1944                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1945                         if (ifp == bif->bif_ifp) {
1946                                 bridge_delete_span(sc, bif);
1947                                 break;
1948                         }
1949
1950                 BRIDGE_UNLOCK(sc);
1951         }
1952         BRIDGE_LIST_UNLOCK();
1953 }
1954
1955 /*
1956  * bridge_init:
1957  *
1958  *      Initialize a bridge interface.
1959  */
1960 static void
1961 bridge_init(void *xsc)
1962 {
1963         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1964         struct ifnet *ifp = sc->sc_ifp;
1965
1966         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1967                 return;
1968
1969         BRIDGE_LOCK(sc);
1970         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1971             bridge_timer, sc);
1972
1973         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1974         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1975
1976         BRIDGE_UNLOCK(sc);
1977 }
1978
1979 /*
1980  * bridge_stop:
1981  *
1982  *      Stop the bridge interface.
1983  */
1984 static void
1985 bridge_stop(struct ifnet *ifp, int disable)
1986 {
1987         struct bridge_softc *sc = ifp->if_softc;
1988
1989         BRIDGE_LOCK_ASSERT(sc);
1990
1991         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1992                 return;
1993
1994         BRIDGE_RT_LOCK(sc);
1995         callout_stop(&sc->sc_brcallout);
1996
1997         bstp_stop(&sc->sc_stp);
1998
1999         bridge_rtflush(sc, IFBF_FLUSHDYN);
2000         BRIDGE_RT_UNLOCK(sc);
2001
2002         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2003 }
2004
2005 /*
2006  * bridge_enqueue:
2007  *
2008  *      Enqueue a packet on a bridge member interface.
2009  *
2010  */
2011 static int
2012 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2013 {
2014         int len, err = 0;
2015         short mflags;
2016         struct mbuf *m0;
2017
2018         /* We may be sending a fragment so traverse the mbuf */
2019         for (; m; m = m0) {
2020                 m0 = m->m_nextpkt;
2021                 m->m_nextpkt = NULL;
2022                 len = m->m_pkthdr.len;
2023                 mflags = m->m_flags;
2024
2025                 /*
2026                  * If underlying interface can not do VLAN tag insertion itself
2027                  * then attach a packet tag that holds it.
2028                  */
2029                 if ((m->m_flags & M_VLANTAG) &&
2030                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2031                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2032                         if (m == NULL) {
2033                                 if_printf(dst_ifp,
2034                                     "unable to prepend VLAN header\n");
2035                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2036                                 continue;
2037                         }
2038                         m->m_flags &= ~M_VLANTAG;
2039                 }
2040
2041                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2042                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2043                         m_freem(m0);
2044                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2045                         break;
2046                 }
2047
2048                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2049                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2050                 if (mflags & M_MCAST)
2051                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2052         }
2053
2054         return (err);
2055 }
2056
2057 /*
2058  * bridge_dummynet:
2059  *
2060  *      Receive a queued packet from dummynet and pass it on to the output
2061  *      interface.
2062  *
2063  *      The mbuf has the Ethernet header already attached.
2064  */
2065 static void
2066 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2067 {
2068         struct bridge_softc *sc;
2069
2070         sc = ifp->if_bridge;
2071
2072         /*
2073          * The packet didnt originate from a member interface. This should only
2074          * ever happen if a member interface is removed while packets are
2075          * queued for it.
2076          */
2077         if (sc == NULL) {
2078                 m_freem(m);
2079                 return;
2080         }
2081
2082         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2083 #ifdef INET6
2084             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2085 #endif
2086             ) {
2087                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2088                         return;
2089                 if (m == NULL)
2090                         return;
2091         }
2092
2093         bridge_enqueue(sc, ifp, m);
2094 }
2095
2096 /*
2097  * bridge_output:
2098  *
2099  *      Send output from a bridge member interface.  This
2100  *      performs the bridging function for locally originated
2101  *      packets.
2102  *
2103  *      The mbuf has the Ethernet header already attached.  We must
2104  *      enqueue or free the mbuf before returning.
2105  */
2106 static int
2107 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2108     struct rtentry *rt)
2109 {
2110         struct ether_header *eh;
2111         struct ifnet *bifp, *dst_if;
2112         struct bridge_softc *sc;
2113         uint16_t vlan;
2114
2115         NET_EPOCH_ASSERT();
2116
2117         if (m->m_len < ETHER_HDR_LEN) {
2118                 m = m_pullup(m, ETHER_HDR_LEN);
2119                 if (m == NULL)
2120                         return (0);
2121         }
2122
2123         eh = mtod(m, struct ether_header *);
2124         sc = ifp->if_bridge;
2125         vlan = VLANTAGOF(m);
2126
2127         bifp = sc->sc_ifp;
2128
2129         /*
2130          * If bridge is down, but the original output interface is up,
2131          * go ahead and send out that interface.  Otherwise, the packet
2132          * is dropped below.
2133          */
2134         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2135                 dst_if = ifp;
2136                 goto sendunicast;
2137         }
2138
2139         /*
2140          * If the packet is a multicast, or we don't know a better way to
2141          * get there, send to all interfaces.
2142          */
2143         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2144                 dst_if = NULL;
2145         else
2146                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2147         /* Tap any traffic not passing back out the originating interface */
2148         if (dst_if != ifp)
2149                 ETHER_BPF_MTAP(bifp, m);
2150         if (dst_if == NULL) {
2151                 struct bridge_iflist *bif;
2152                 struct mbuf *mc;
2153                 int used = 0;
2154
2155                 bridge_span(sc, m);
2156
2157                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2158                         dst_if = bif->bif_ifp;
2159
2160                         if (dst_if->if_type == IFT_GIF)
2161                                 continue;
2162                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2163                                 continue;
2164
2165                         /*
2166                          * If this is not the original output interface,
2167                          * and the interface is participating in spanning
2168                          * tree, make sure the port is in a state that
2169                          * allows forwarding.
2170                          */
2171                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2172                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2173                                 continue;
2174
2175                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2176                                 used = 1;
2177                                 mc = m;
2178                         } else {
2179                                 mc = m_copypacket(m, M_NOWAIT);
2180                                 if (mc == NULL) {
2181                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2182                                         continue;
2183                                 }
2184                         }
2185
2186                         bridge_enqueue(sc, dst_if, mc);
2187                 }
2188                 if (used == 0)
2189                         m_freem(m);
2190                 return (0);
2191         }
2192
2193 sendunicast:
2194         /*
2195          * XXX Spanning tree consideration here?
2196          */
2197
2198         bridge_span(sc, m);
2199         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2200                 m_freem(m);
2201                 return (0);
2202         }
2203
2204         bridge_enqueue(sc, dst_if, m);
2205         return (0);
2206 }
2207
2208 /*
2209  * bridge_transmit:
2210  *
2211  *      Do output on a bridge.
2212  *
2213  */
2214 static int
2215 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2216 {
2217         struct bridge_softc *sc;
2218         struct ether_header *eh;
2219         struct ifnet *dst_if;
2220         int error = 0;
2221
2222         sc = ifp->if_softc;
2223
2224         ETHER_BPF_MTAP(ifp, m);
2225
2226         eh = mtod(m, struct ether_header *);
2227
2228         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2229             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2230                 error = bridge_enqueue(sc, dst_if, m);
2231         } else
2232                 bridge_broadcast(sc, ifp, m, 0);
2233
2234         return (error);
2235 }
2236
2237 #ifdef ALTQ
2238 static void
2239 bridge_altq_start(if_t ifp)
2240 {
2241         struct ifaltq *ifq = &ifp->if_snd;
2242         struct mbuf *m;
2243
2244         IFQ_LOCK(ifq);
2245         IFQ_DEQUEUE_NOLOCK(ifq, m);
2246         while (m != NULL) {
2247                 bridge_transmit(ifp, m);
2248                 IFQ_DEQUEUE_NOLOCK(ifq, m);
2249         }
2250         IFQ_UNLOCK(ifq);
2251 }
2252
2253 static int
2254 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2255 {
2256         int err;
2257
2258         if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2259                 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2260                 if (err == 0)
2261                         bridge_altq_start(ifp);
2262         } else
2263                 err = bridge_transmit(ifp, m);
2264
2265         return (err);
2266 }
2267 #endif  /* ALTQ */
2268
2269 /*
2270  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2271  */
2272 static void
2273 bridge_qflush(struct ifnet *ifp __unused)
2274 {
2275 }
2276
2277 /*
2278  * bridge_forward:
2279  *
2280  *      The forwarding function of the bridge.
2281  *
2282  *      NOTE: Releases the lock on return.
2283  */
2284 static void
2285 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2286     struct mbuf *m)
2287 {
2288         struct bridge_iflist *dbif;
2289         struct ifnet *src_if, *dst_if, *ifp;
2290         struct ether_header *eh;
2291         uint16_t vlan;
2292         uint8_t *dst;
2293         int error;
2294
2295         NET_EPOCH_ASSERT();
2296
2297         src_if = m->m_pkthdr.rcvif;
2298         ifp = sc->sc_ifp;
2299
2300         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2301         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2302         vlan = VLANTAGOF(m);
2303
2304         if ((sbif->bif_flags & IFBIF_STP) &&
2305             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2306                 goto drop;
2307
2308         eh = mtod(m, struct ether_header *);
2309         dst = eh->ether_dhost;
2310
2311         /* If the interface is learning, record the address. */
2312         if (sbif->bif_flags & IFBIF_LEARNING) {
2313                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2314                     sbif, 0, IFBAF_DYNAMIC);
2315                 /*
2316                  * If the interface has addresses limits then deny any source
2317                  * that is not in the cache.
2318                  */
2319                 if (error && sbif->bif_addrmax)
2320                         goto drop;
2321         }
2322
2323         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2324             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2325                 goto drop;
2326
2327         /*
2328          * At this point, the port either doesn't participate
2329          * in spanning tree or it is in the forwarding state.
2330          */
2331
2332         /*
2333          * If the packet is unicast, destined for someone on
2334          * "this" side of the bridge, drop it.
2335          */
2336         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2337                 dst_if = bridge_rtlookup(sc, dst, vlan);
2338                 if (src_if == dst_if)
2339                         goto drop;
2340         } else {
2341                 /*
2342                  * Check if its a reserved multicast address, any address
2343                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2344                  * bridge.
2345                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2346                  */
2347                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2348                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2349                     dst[4] == 0x00 && dst[5] <= 0x0f)
2350                         goto drop;
2351
2352                 /* ...forward it to all interfaces. */
2353                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2354                 dst_if = NULL;
2355         }
2356
2357         /*
2358          * If we have a destination interface which is a member of our bridge,
2359          * OR this is a unicast packet, push it through the bpf(4) machinery.
2360          * For broadcast or multicast packets, don't bother because it will
2361          * be reinjected into ether_input. We do this before we pass the packets
2362          * through the pfil(9) framework, as it is possible that pfil(9) will
2363          * drop the packet, or possibly modify it, making it difficult to debug
2364          * firewall issues on the bridge.
2365          */
2366         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2367                 ETHER_BPF_MTAP(ifp, m);
2368
2369         /* run the packet filter */
2370         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2371 #ifdef INET6
2372             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2373 #endif
2374             ) {
2375                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2376                         return;
2377                 if (m == NULL)
2378                         return;
2379         }
2380
2381         if (dst_if == NULL) {
2382                 bridge_broadcast(sc, src_if, m, 1);
2383                 return;
2384         }
2385
2386         /*
2387          * At this point, we're dealing with a unicast frame
2388          * going to a different interface.
2389          */
2390         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2391                 goto drop;
2392
2393         dbif = bridge_lookup_member_if(sc, dst_if);
2394         if (dbif == NULL)
2395                 /* Not a member of the bridge (anymore?) */
2396                 goto drop;
2397
2398         /* Private segments can not talk to each other */
2399         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2400                 goto drop;
2401
2402         if ((dbif->bif_flags & IFBIF_STP) &&
2403             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2404                 goto drop;
2405
2406         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2407 #ifdef INET6
2408             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2409 #endif
2410             ) {
2411                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2412                         return;
2413                 if (m == NULL)
2414                         return;
2415         }
2416
2417         bridge_enqueue(sc, dst_if, m);
2418         return;
2419
2420 drop:
2421         m_freem(m);
2422 }
2423
2424 /*
2425  * bridge_input:
2426  *
2427  *      Receive input from a member interface.  Queue the packet for
2428  *      bridging if it is not for us.
2429  */
2430 static struct mbuf *
2431 bridge_input(struct ifnet *ifp, struct mbuf *m)
2432 {
2433         struct bridge_softc *sc = ifp->if_bridge;
2434         struct bridge_iflist *bif, *bif2;
2435         struct ifnet *bifp;
2436         struct ether_header *eh;
2437         struct mbuf *mc, *mc2;
2438         uint16_t vlan;
2439         int error;
2440
2441         NET_EPOCH_ASSERT();
2442
2443         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2444                 return (m);
2445
2446         bifp = sc->sc_ifp;
2447         vlan = VLANTAGOF(m);
2448
2449         /*
2450          * Implement support for bridge monitoring. If this flag has been
2451          * set on this interface, discard the packet once we push it through
2452          * the bpf(4) machinery, but before we do, increment the byte and
2453          * packet counters associated with this interface.
2454          */
2455         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2456                 m->m_pkthdr.rcvif  = bifp;
2457                 ETHER_BPF_MTAP(bifp, m);
2458                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2459                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2460                 m_freem(m);
2461                 return (NULL);
2462         }
2463         bif = bridge_lookup_member_if(sc, ifp);
2464         if (bif == NULL) {
2465                 return (m);
2466         }
2467
2468         eh = mtod(m, struct ether_header *);
2469
2470         bridge_span(sc, m);
2471
2472         if (m->m_flags & (M_BCAST|M_MCAST)) {
2473                 /* Tap off 802.1D packets; they do not get forwarded. */
2474                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2475                     ETHER_ADDR_LEN) == 0) {
2476                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2477                         return (NULL);
2478                 }
2479
2480                 if ((bif->bif_flags & IFBIF_STP) &&
2481                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2482                         return (m);
2483                 }
2484
2485                 /*
2486                  * Make a deep copy of the packet and enqueue the copy
2487                  * for bridge processing; return the original packet for
2488                  * local processing.
2489                  */
2490                 mc = m_dup(m, M_NOWAIT);
2491                 if (mc == NULL) {
2492                         return (m);
2493                 }
2494
2495                 /* Perform the bridge forwarding function with the copy. */
2496                 bridge_forward(sc, bif, mc);
2497
2498                 /*
2499                  * Reinject the mbuf as arriving on the bridge so we have a
2500                  * chance at claiming multicast packets. We can not loop back
2501                  * here from ether_input as a bridge is never a member of a
2502                  * bridge.
2503                  */
2504                 KASSERT(bifp->if_bridge == NULL,
2505                     ("loop created in bridge_input"));
2506                 mc2 = m_dup(m, M_NOWAIT);
2507                 if (mc2 != NULL) {
2508                         /* Keep the layer3 header aligned */
2509                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2510                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2511                 }
2512                 if (mc2 != NULL) {
2513                         mc2->m_pkthdr.rcvif = bifp;
2514                         (*bifp->if_input)(bifp, mc2);
2515                 }
2516
2517                 /* Return the original packet for local processing. */
2518                 return (m);
2519         }
2520
2521         if ((bif->bif_flags & IFBIF_STP) &&
2522             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2523                 return (m);
2524         }
2525
2526 #if (defined(INET) || defined(INET6))
2527 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2528         || ((iface)->if_carp \
2529             && (*carp_forus_p)((iface), eh->ether_dhost))
2530 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2531         || ((iface)->if_carp \
2532             && (*carp_forus_p)((iface), eh->ether_shost))
2533 #else
2534 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2535 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2536 #endif
2537
2538 #ifdef INET6
2539 #   define OR_PFIL_HOOKED_INET6 \
2540         || PFIL_HOOKED_IN(V_inet6_pfil_head)
2541 #else
2542 #   define OR_PFIL_HOOKED_INET6
2543 #endif
2544
2545 #define GRAB_OUR_PACKETS(iface) \
2546         if ((iface)->if_type == IFT_GIF) \
2547                 continue; \
2548         /* It is destined for us. */ \
2549         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2550             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2551             ) {                                                         \
2552                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2553                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2554                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2555                         if (error && bif->bif_addrmax) {                \
2556                                 m_freem(m);                             \
2557                                 return (NULL);                          \
2558                         }                                               \
2559                 }                                                       \
2560                 m->m_pkthdr.rcvif = iface;                              \
2561                 if ((iface) == ifp) {                                   \
2562                         /* Skip bridge processing... src == dest */     \
2563                         return (m);                                     \
2564                 }                                                       \
2565                 /* It's passing over or to the bridge, locally. */      \
2566                 ETHER_BPF_MTAP(bifp, m);                                \
2567                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2568                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len); \
2569                 /* Filter on the physical interface. */                 \
2570                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) \
2571                      OR_PFIL_HOOKED_INET6)) {                           \
2572                         if (bridge_pfil(&m, NULL, ifp,                  \
2573                             PFIL_IN) != 0 || m == NULL) {               \
2574                                 return (NULL);                          \
2575                         }                                               \
2576                 }                                                       \
2577                 if ((iface) != bifp)                                    \
2578                         ETHER_BPF_MTAP(iface, m);                       \
2579                 return (m);                                             \
2580         }                                                               \
2581                                                                         \
2582         /* We just received a packet that we sent out. */               \
2583         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2584             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2585             ) {                                                         \
2586                 m_freem(m);                                             \
2587                 return (NULL);                                          \
2588         }
2589
2590         /*
2591          * Unicast.  Make sure it's not for the bridge.
2592          */
2593         do { GRAB_OUR_PACKETS(bifp) } while (0);
2594
2595         /*
2596          * Give a chance for ifp at first priority. This will help when the
2597          * packet comes through the interface like VLAN's with the same MACs
2598          * on several interfaces from the same bridge. This also will save
2599          * some CPU cycles in case the destination interface and the input
2600          * interface (eq ifp) are the same.
2601          */
2602         do { GRAB_OUR_PACKETS(ifp) } while (0);
2603
2604         /* Now check the all bridge members. */
2605         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2606                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2607         }
2608
2609 #undef OR_CARP_CHECK_WE_ARE_DST
2610 #undef OR_CARP_CHECK_WE_ARE_SRC
2611 #undef OR_PFIL_HOOKED_INET6
2612 #undef GRAB_OUR_PACKETS
2613
2614         /* Perform the bridge forwarding function. */
2615         bridge_forward(sc, bif, m);
2616
2617         return (NULL);
2618 }
2619
2620 /*
2621  * bridge_broadcast:
2622  *
2623  *      Send a frame to all interfaces that are members of
2624  *      the bridge, except for the one on which the packet
2625  *      arrived.
2626  *
2627  *      NOTE: Releases the lock on return.
2628  */
2629 static void
2630 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2631     struct mbuf *m, int runfilt)
2632 {
2633         struct bridge_iflist *dbif, *sbif;
2634         struct mbuf *mc;
2635         struct ifnet *dst_if;
2636         int used = 0, i;
2637
2638         NET_EPOCH_ASSERT();
2639
2640         sbif = bridge_lookup_member_if(sc, src_if);
2641
2642         /* Filter on the bridge interface before broadcasting */
2643         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2644 #ifdef INET6
2645             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2646 #endif
2647             )) {
2648                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2649                         return;
2650                 if (m == NULL)
2651                         return;
2652         }
2653
2654         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2655                 dst_if = dbif->bif_ifp;
2656                 if (dst_if == src_if)
2657                         continue;
2658
2659                 /* Private segments can not talk to each other */
2660                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2661                         continue;
2662
2663                 if ((dbif->bif_flags & IFBIF_STP) &&
2664                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2665                         continue;
2666
2667                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2668                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2669                         continue;
2670
2671                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2672                         continue;
2673
2674                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2675                         mc = m;
2676                         used = 1;
2677                 } else {
2678                         mc = m_dup(m, M_NOWAIT);
2679                         if (mc == NULL) {
2680                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2681                                 continue;
2682                         }
2683                 }
2684
2685                 /*
2686                  * Filter on the output interface. Pass a NULL bridge interface
2687                  * pointer so we do not redundantly filter on the bridge for
2688                  * each interface we broadcast on.
2689                  */
2690                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2691 #ifdef INET6
2692                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2693 #endif
2694                     )) {
2695                         if (used == 0) {
2696                                 /* Keep the layer3 header aligned */
2697                                 i = min(mc->m_pkthdr.len, max_protohdr);
2698                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2699                                 if (mc == NULL) {
2700                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2701                                         continue;
2702                                 }
2703                         }
2704                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2705                                 continue;
2706                         if (mc == NULL)
2707                                 continue;
2708                 }
2709
2710                 bridge_enqueue(sc, dst_if, mc);
2711         }
2712         if (used == 0)
2713                 m_freem(m);
2714 }
2715
2716 /*
2717  * bridge_span:
2718  *
2719  *      Duplicate a packet out one or more interfaces that are in span mode,
2720  *      the original mbuf is unmodified.
2721  */
2722 static void
2723 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2724 {
2725         struct bridge_iflist *bif;
2726         struct ifnet *dst_if;
2727         struct mbuf *mc;
2728
2729         NET_EPOCH_ASSERT();
2730
2731         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2732                 return;
2733
2734         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2735                 dst_if = bif->bif_ifp;
2736
2737                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2738                         continue;
2739
2740                 mc = m_copypacket(m, M_NOWAIT);
2741                 if (mc == NULL) {
2742                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2743                         continue;
2744                 }
2745
2746                 bridge_enqueue(sc, dst_if, mc);
2747         }
2748 }
2749
2750 /*
2751  * bridge_rtupdate:
2752  *
2753  *      Add a bridge routing entry.
2754  */
2755 static int
2756 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2757     struct bridge_iflist *bif, int setflags, uint8_t flags)
2758 {
2759         struct bridge_rtnode *brt;
2760         int error;
2761
2762         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2763
2764         /* Check the source address is valid and not multicast. */
2765         if (ETHER_IS_MULTICAST(dst) ||
2766             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2767              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2768                 return (EINVAL);
2769
2770         /* 802.1p frames map to vlan 1 */
2771         if (vlan == 0)
2772                 vlan = 1;
2773
2774         /*
2775          * A route for this destination might already exist.  If so,
2776          * update it, otherwise create a new one.
2777          */
2778         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2779                 BRIDGE_RT_LOCK(sc);
2780
2781                 /* Check again, now that we have the lock. There could have
2782                  * been a race and we only want to insert this once. */
2783                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2784                         BRIDGE_RT_UNLOCK(sc);
2785                         return (0);
2786                 }
2787
2788                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2789                         sc->sc_brtexceeded++;
2790                         BRIDGE_RT_UNLOCK(sc);
2791                         return (ENOSPC);
2792                 }
2793                 /* Check per interface address limits (if enabled) */
2794                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2795                         bif->bif_addrexceeded++;
2796                         BRIDGE_RT_UNLOCK(sc);
2797                         return (ENOSPC);
2798                 }
2799
2800                 /*
2801                  * Allocate a new bridge forwarding node, and
2802                  * initialize the expiration time and Ethernet
2803                  * address.
2804                  */
2805                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2806                 if (brt == NULL) {
2807                         BRIDGE_RT_UNLOCK(sc);
2808                         return (ENOMEM);
2809                 }
2810                 brt->brt_vnet = curvnet;
2811
2812                 if (bif->bif_flags & IFBIF_STICKY)
2813                         brt->brt_flags = IFBAF_STICKY;
2814                 else
2815                         brt->brt_flags = IFBAF_DYNAMIC;
2816
2817                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2818                 brt->brt_vlan = vlan;
2819
2820                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2821                         uma_zfree(V_bridge_rtnode_zone, brt);
2822                         BRIDGE_RT_UNLOCK(sc);
2823                         return (error);
2824                 }
2825                 brt->brt_dst = bif;
2826                 bif->bif_addrcnt++;
2827
2828                 BRIDGE_RT_UNLOCK(sc);
2829         }
2830
2831         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2832             brt->brt_dst != bif) {
2833                 BRIDGE_RT_LOCK(sc);
2834                 brt->brt_dst->bif_addrcnt--;
2835                 brt->brt_dst = bif;
2836                 brt->brt_dst->bif_addrcnt++;
2837                 BRIDGE_RT_UNLOCK(sc);
2838         }
2839
2840         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2841                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2842         if (setflags)
2843                 brt->brt_flags = flags;
2844
2845         return (0);
2846 }
2847
2848 /*
2849  * bridge_rtlookup:
2850  *
2851  *      Lookup the destination interface for an address.
2852  */
2853 static struct ifnet *
2854 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2855 {
2856         struct bridge_rtnode *brt;
2857
2858         NET_EPOCH_ASSERT();
2859
2860         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2861                 return (NULL);
2862
2863         return (brt->brt_ifp);
2864 }
2865
2866 /*
2867  * bridge_rttrim:
2868  *
2869  *      Trim the routine table so that we have a number
2870  *      of routing entries less than or equal to the
2871  *      maximum number.
2872  */
2873 static void
2874 bridge_rttrim(struct bridge_softc *sc)
2875 {
2876         struct bridge_rtnode *brt, *nbrt;
2877
2878         NET_EPOCH_ASSERT();
2879         BRIDGE_RT_LOCK_ASSERT(sc);
2880
2881         /* Make sure we actually need to do this. */
2882         if (sc->sc_brtcnt <= sc->sc_brtmax)
2883                 return;
2884
2885         /* Force an aging cycle; this might trim enough addresses. */
2886         bridge_rtage(sc);
2887         if (sc->sc_brtcnt <= sc->sc_brtmax)
2888                 return;
2889
2890         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2891                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2892                         bridge_rtnode_destroy(sc, brt);
2893                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2894                                 return;
2895                 }
2896         }
2897 }
2898
2899 /*
2900  * bridge_timer:
2901  *
2902  *      Aging timer for the bridge.
2903  */
2904 static void
2905 bridge_timer(void *arg)
2906 {
2907         struct bridge_softc *sc = arg;
2908
2909         BRIDGE_RT_LOCK_ASSERT(sc);
2910
2911         /* Destruction of rtnodes requires a proper vnet context */
2912         CURVNET_SET(sc->sc_ifp->if_vnet);
2913         bridge_rtage(sc);
2914
2915         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2916                 callout_reset(&sc->sc_brcallout,
2917                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2918         CURVNET_RESTORE();
2919 }
2920
2921 /*
2922  * bridge_rtage:
2923  *
2924  *      Perform an aging cycle.
2925  */
2926 static void
2927 bridge_rtage(struct bridge_softc *sc)
2928 {
2929         struct bridge_rtnode *brt, *nbrt;
2930
2931         BRIDGE_RT_LOCK_ASSERT(sc);
2932
2933         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2934                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2935                         if (time_uptime >= brt->brt_expire)
2936                                 bridge_rtnode_destroy(sc, brt);
2937                 }
2938         }
2939 }
2940
2941 /*
2942  * bridge_rtflush:
2943  *
2944  *      Remove all dynamic addresses from the bridge.
2945  */
2946 static void
2947 bridge_rtflush(struct bridge_softc *sc, int full)
2948 {
2949         struct bridge_rtnode *brt, *nbrt;
2950
2951         BRIDGE_RT_LOCK_ASSERT(sc);
2952
2953         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2954                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2955                         bridge_rtnode_destroy(sc, brt);
2956         }
2957 }
2958
2959 /*
2960  * bridge_rtdaddr:
2961  *
2962  *      Remove an address from the table.
2963  */
2964 static int
2965 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2966 {
2967         struct bridge_rtnode *brt;
2968         int found = 0;
2969
2970         BRIDGE_RT_LOCK(sc);
2971
2972         /*
2973          * If vlan is zero then we want to delete for all vlans so the lookup
2974          * may return more than one.
2975          */
2976         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2977                 bridge_rtnode_destroy(sc, brt);
2978                 found = 1;
2979         }
2980
2981         BRIDGE_RT_UNLOCK(sc);
2982
2983         return (found ? 0 : ENOENT);
2984 }
2985
2986 /*
2987  * bridge_rtdelete:
2988  *
2989  *      Delete routes to a speicifc member interface.
2990  */
2991 static void
2992 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2993 {
2994         struct bridge_rtnode *brt, *nbrt;
2995
2996         BRIDGE_RT_LOCK_ASSERT(sc);
2997
2998         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2999                 if (brt->brt_ifp == ifp && (full ||
3000                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3001                         bridge_rtnode_destroy(sc, brt);
3002         }
3003 }
3004
3005 /*
3006  * bridge_rtable_init:
3007  *
3008  *      Initialize the route table for this bridge.
3009  */
3010 static void
3011 bridge_rtable_init(struct bridge_softc *sc)
3012 {
3013         int i;
3014
3015         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3016             M_DEVBUF, M_WAITOK);
3017
3018         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3019                 CK_LIST_INIT(&sc->sc_rthash[i]);
3020
3021         sc->sc_rthash_key = arc4random();
3022         CK_LIST_INIT(&sc->sc_rtlist);
3023 }
3024
3025 /*
3026  * bridge_rtable_fini:
3027  *
3028  *      Deconstruct the route table for this bridge.
3029  */
3030 static void
3031 bridge_rtable_fini(struct bridge_softc *sc)
3032 {
3033
3034         KASSERT(sc->sc_brtcnt == 0,
3035             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3036         free(sc->sc_rthash, M_DEVBUF);
3037 }
3038
3039 /*
3040  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3041  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3042  */
3043 #define mix(a, b, c)                                                    \
3044 do {                                                                    \
3045         a -= b; a -= c; a ^= (c >> 13);                                 \
3046         b -= c; b -= a; b ^= (a << 8);                                  \
3047         c -= a; c -= b; c ^= (b >> 13);                                 \
3048         a -= b; a -= c; a ^= (c >> 12);                                 \
3049         b -= c; b -= a; b ^= (a << 16);                                 \
3050         c -= a; c -= b; c ^= (b >> 5);                                  \
3051         a -= b; a -= c; a ^= (c >> 3);                                  \
3052         b -= c; b -= a; b ^= (a << 10);                                 \
3053         c -= a; c -= b; c ^= (b >> 15);                                 \
3054 } while (/*CONSTCOND*/0)
3055
3056 static __inline uint32_t
3057 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3058 {
3059         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3060
3061         b += addr[5] << 8;
3062         b += addr[4];
3063         a += addr[3] << 24;
3064         a += addr[2] << 16;
3065         a += addr[1] << 8;
3066         a += addr[0];
3067
3068         mix(a, b, c);
3069
3070         return (c & BRIDGE_RTHASH_MASK);
3071 }
3072
3073 #undef mix
3074
3075 static int
3076 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3077 {
3078         int i, d;
3079
3080         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3081                 d = ((int)a[i]) - ((int)b[i]);
3082         }
3083
3084         return (d);
3085 }
3086
3087 /*
3088  * bridge_rtnode_lookup:
3089  *
3090  *      Look up a bridge route node for the specified destination. Compare the
3091  *      vlan id or if zero then just return the first match.
3092  */
3093 static struct bridge_rtnode *
3094 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3095 {
3096         struct bridge_rtnode *brt;
3097         uint32_t hash;
3098         int dir;
3099
3100         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3101
3102         hash = bridge_rthash(sc, addr);
3103         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3104                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3105                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3106                         return (brt);
3107                 if (dir > 0)
3108                         return (NULL);
3109         }
3110
3111         return (NULL);
3112 }
3113
3114 /*
3115  * bridge_rtnode_insert:
3116  *
3117  *      Insert the specified bridge node into the route table.  We
3118  *      assume the entry is not already in the table.
3119  */
3120 static int
3121 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3122 {
3123         struct bridge_rtnode *lbrt;
3124         uint32_t hash;
3125         int dir;
3126
3127         BRIDGE_RT_LOCK_ASSERT(sc);
3128
3129         hash = bridge_rthash(sc, brt->brt_addr);
3130
3131         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3132         if (lbrt == NULL) {
3133                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3134                 goto out;
3135         }
3136
3137         do {
3138                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3139                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3140                         return (EEXIST);
3141                 if (dir > 0) {
3142                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3143                         goto out;
3144                 }
3145                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3146                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3147                         goto out;
3148                 }
3149                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3150         } while (lbrt != NULL);
3151
3152 #ifdef DIAGNOSTIC
3153         panic("bridge_rtnode_insert: impossible");
3154 #endif
3155
3156 out:
3157         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3158         sc->sc_brtcnt++;
3159
3160         return (0);
3161 }
3162
3163 static void
3164 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3165 {
3166         struct bridge_rtnode *brt;
3167
3168         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3169
3170         CURVNET_SET(brt->brt_vnet);
3171         uma_zfree(V_bridge_rtnode_zone, brt);
3172         CURVNET_RESTORE();
3173 }
3174
3175 /*
3176  * bridge_rtnode_destroy:
3177  *
3178  *      Destroy a bridge rtnode.
3179  */
3180 static void
3181 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3182 {
3183         BRIDGE_RT_LOCK_ASSERT(sc);
3184
3185         CK_LIST_REMOVE(brt, brt_hash);
3186
3187         CK_LIST_REMOVE(brt, brt_list);
3188         sc->sc_brtcnt--;
3189         brt->brt_dst->bif_addrcnt--;
3190
3191         NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3192 }
3193
3194 /*
3195  * bridge_rtable_expire:
3196  *
3197  *      Set the expiry time for all routes on an interface.
3198  */
3199 static void
3200 bridge_rtable_expire(struct ifnet *ifp, int age)
3201 {
3202         struct bridge_softc *sc = ifp->if_bridge;
3203         struct bridge_rtnode *brt;
3204
3205         CURVNET_SET(ifp->if_vnet);
3206         BRIDGE_RT_LOCK(sc);
3207
3208         /*
3209          * If the age is zero then flush, otherwise set all the expiry times to
3210          * age for the interface
3211          */
3212         if (age == 0)
3213                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3214         else {
3215                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3216                         /* Cap the expiry time to 'age' */
3217                         if (brt->brt_ifp == ifp &&
3218                             brt->brt_expire > time_uptime + age &&
3219                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3220                                 brt->brt_expire = time_uptime + age;
3221                 }
3222         }
3223         BRIDGE_RT_UNLOCK(sc);
3224         CURVNET_RESTORE();
3225 }
3226
3227 /*
3228  * bridge_state_change:
3229  *
3230  *      Callback from the bridgestp code when a port changes states.
3231  */
3232 static void
3233 bridge_state_change(struct ifnet *ifp, int state)
3234 {
3235         struct bridge_softc *sc = ifp->if_bridge;
3236         static const char *stpstates[] = {
3237                 "disabled",
3238                 "listening",
3239                 "learning",
3240                 "forwarding",
3241                 "blocking",
3242                 "discarding"
3243         };
3244
3245         CURVNET_SET(ifp->if_vnet);
3246         if (V_log_stp)
3247                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3248                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3249         CURVNET_RESTORE();
3250 }
3251
3252 /*
3253  * Send bridge packets through pfil if they are one of the types pfil can deal
3254  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3255  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3256  * that interface.
3257  */
3258 static int
3259 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3260 {
3261         int snap, error, i, hlen;
3262         struct ether_header *eh1, eh2;
3263         struct ip *ip;
3264         struct llc llc1;
3265         u_int16_t ether_type;
3266         pfil_return_t rv;
3267
3268         snap = 0;
3269         error = -1;     /* Default error if not error == 0 */
3270
3271 #if 0
3272         /* we may return with the IP fields swapped, ensure its not shared */
3273         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3274 #endif
3275
3276         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3277                 return (0); /* filtering is disabled */
3278
3279         i = min((*mp)->m_pkthdr.len, max_protohdr);
3280         if ((*mp)->m_len < i) {
3281             *mp = m_pullup(*mp, i);
3282             if (*mp == NULL) {
3283                 printf("%s: m_pullup failed\n", __func__);
3284                 return (-1);
3285             }
3286         }
3287
3288         eh1 = mtod(*mp, struct ether_header *);
3289         ether_type = ntohs(eh1->ether_type);
3290
3291         /*
3292          * Check for SNAP/LLC.
3293          */
3294         if (ether_type < ETHERMTU) {
3295                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3296
3297                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3298                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3299                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3300                     llc2->llc_control == LLC_UI) {
3301                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3302                         snap = 1;
3303                 }
3304         }
3305
3306         /*
3307          * If we're trying to filter bridge traffic, don't look at anything
3308          * other than IP and ARP traffic.  If the filter doesn't understand
3309          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3310          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3311          * but of course we don't have an AppleTalk filter to begin with.
3312          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3313          * ARP traffic.)
3314          */
3315         switch (ether_type) {
3316                 case ETHERTYPE_ARP:
3317                 case ETHERTYPE_REVARP:
3318                         if (V_pfil_ipfw_arp == 0)
3319                                 return (0); /* Automatically pass */
3320                         break;
3321
3322                 case ETHERTYPE_IP:
3323 #ifdef INET6
3324                 case ETHERTYPE_IPV6:
3325 #endif /* INET6 */
3326                         break;
3327                 default:
3328                         /*
3329                          * Check to see if the user wants to pass non-ip
3330                          * packets, these will not be checked by pfil(9) and
3331                          * passed unconditionally so the default is to drop.
3332                          */
3333                         if (V_pfil_onlyip)
3334                                 goto bad;
3335         }
3336
3337         /* Run the packet through pfil before stripping link headers */
3338         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3339             dir == PFIL_OUT && ifp != NULL) {
3340                 switch (pfil_run_hooks(V_link_pfil_head, mp, ifp, dir, NULL)) {
3341                 case PFIL_DROPPED:
3342                         return (EACCES);
3343                 case PFIL_CONSUMED:
3344                         return (0);
3345                 }
3346         }
3347
3348         /* Strip off the Ethernet header and keep a copy. */
3349         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3350         m_adj(*mp, ETHER_HDR_LEN);
3351
3352         /* Strip off snap header, if present */
3353         if (snap) {
3354                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3355                 m_adj(*mp, sizeof(struct llc));
3356         }
3357
3358         /*
3359          * Check the IP header for alignment and errors
3360          */
3361         if (dir == PFIL_IN) {
3362                 switch (ether_type) {
3363                         case ETHERTYPE_IP:
3364                                 error = bridge_ip_checkbasic(mp);
3365                                 break;
3366 #ifdef INET6
3367                         case ETHERTYPE_IPV6:
3368                                 error = bridge_ip6_checkbasic(mp);
3369                                 break;
3370 #endif /* INET6 */
3371                         default:
3372                                 error = 0;
3373                 }
3374                 if (error)
3375                         goto bad;
3376         }
3377
3378         error = 0;
3379
3380         /*
3381          * Run the packet through pfil
3382          */
3383         rv = PFIL_PASS;
3384         switch (ether_type) {
3385         case ETHERTYPE_IP:
3386                 /*
3387                  * Run pfil on the member interface and the bridge, both can
3388                  * be skipped by clearing pfil_member or pfil_bridge.
3389                  *
3390                  * Keep the order:
3391                  *   in_if -> bridge_if -> out_if
3392                  */
3393                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3394                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3395                     PFIL_PASS)
3396                         break;
3397
3398                 if (V_pfil_member && ifp != NULL && (rv =
3399                     pfil_run_hooks(V_inet_pfil_head, mp, ifp, dir, NULL)) !=
3400                     PFIL_PASS)
3401                         break;
3402
3403                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3404                     pfil_run_hooks(V_inet_pfil_head, mp, bifp, dir, NULL)) !=
3405                     PFIL_PASS)
3406                         break;
3407
3408                 /* check if we need to fragment the packet */
3409                 /* bridge_fragment generates a mbuf chain of packets */
3410                 /* that already include eth headers */
3411                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3412                         i = (*mp)->m_pkthdr.len;
3413                         if (i > ifp->if_mtu) {
3414                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3415                                             &llc1);
3416                                 return (error);
3417                         }
3418                 }
3419
3420                 /* Recalculate the ip checksum. */
3421                 ip = mtod(*mp, struct ip *);
3422                 hlen = ip->ip_hl << 2;
3423                 if (hlen < sizeof(struct ip))
3424                         goto bad;
3425                 if (hlen > (*mp)->m_len) {
3426                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3427                                 goto bad;
3428                         ip = mtod(*mp, struct ip *);
3429                         if (ip == NULL)
3430                                 goto bad;
3431                 }
3432                 ip->ip_sum = 0;
3433                 if (hlen == sizeof(struct ip))
3434                         ip->ip_sum = in_cksum_hdr(ip);
3435                 else
3436                         ip->ip_sum = in_cksum(*mp, hlen);
3437
3438                 break;
3439 #ifdef INET6
3440         case ETHERTYPE_IPV6:
3441                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3442                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3443                     PFIL_PASS)
3444                         break;
3445
3446                 if (V_pfil_member && ifp != NULL && (rv =
3447                     pfil_run_hooks(V_inet6_pfil_head, mp, ifp, dir, NULL)) !=
3448                     PFIL_PASS)
3449                         break;
3450
3451                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3452                     pfil_run_hooks(V_inet6_pfil_head, mp, bifp, dir, NULL)) !=
3453                     PFIL_PASS)
3454                         break;
3455                 break;
3456 #endif
3457         }
3458
3459         switch (rv) {
3460         case PFIL_CONSUMED:
3461                 return (0);
3462         case PFIL_DROPPED:
3463                 return (EACCES);
3464         default:
3465                 break;
3466         }
3467
3468         error = -1;
3469
3470         /*
3471          * Finally, put everything back the way it was and return
3472          */
3473         if (snap) {
3474                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3475                 if (*mp == NULL)
3476                         return (error);
3477                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3478         }
3479
3480         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3481         if (*mp == NULL)
3482                 return (error);
3483         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3484
3485         return (0);
3486
3487 bad:
3488         m_freem(*mp);
3489         *mp = NULL;
3490         return (error);
3491 }
3492
3493 /*
3494  * Perform basic checks on header size since
3495  * pfil assumes ip_input has already processed
3496  * it for it.  Cut-and-pasted from ip_input.c.
3497  * Given how simple the IPv6 version is,
3498  * does the IPv4 version really need to be
3499  * this complicated?
3500  *
3501  * XXX Should we update ipstat here, or not?
3502  * XXX Right now we update ipstat but not
3503  * XXX csum_counter.
3504  */
3505 static int
3506 bridge_ip_checkbasic(struct mbuf **mp)
3507 {
3508         struct mbuf *m = *mp;
3509         struct ip *ip;
3510         int len, hlen;
3511         u_short sum;
3512
3513         if (*mp == NULL)
3514                 return (-1);
3515
3516         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3517                 if ((m = m_copyup(m, sizeof(struct ip),
3518                         (max_linkhdr + 3) & ~3)) == NULL) {
3519                         /* XXXJRT new stat, please */
3520                         KMOD_IPSTAT_INC(ips_toosmall);
3521                         goto bad;
3522                 }
3523         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3524                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3525                         KMOD_IPSTAT_INC(ips_toosmall);
3526                         goto bad;
3527                 }
3528         }
3529         ip = mtod(m, struct ip *);
3530         if (ip == NULL) goto bad;
3531
3532         if (ip->ip_v != IPVERSION) {
3533                 KMOD_IPSTAT_INC(ips_badvers);
3534                 goto bad;
3535         }
3536         hlen = ip->ip_hl << 2;
3537         if (hlen < sizeof(struct ip)) { /* minimum header length */
3538                 KMOD_IPSTAT_INC(ips_badhlen);
3539                 goto bad;
3540         }
3541         if (hlen > m->m_len) {
3542                 if ((m = m_pullup(m, hlen)) == NULL) {
3543                         KMOD_IPSTAT_INC(ips_badhlen);
3544                         goto bad;
3545                 }
3546                 ip = mtod(m, struct ip *);
3547                 if (ip == NULL) goto bad;
3548         }
3549
3550         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3551                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3552         } else {
3553                 if (hlen == sizeof(struct ip)) {
3554                         sum = in_cksum_hdr(ip);
3555                 } else {
3556                         sum = in_cksum(m, hlen);
3557                 }
3558         }
3559         if (sum) {
3560                 KMOD_IPSTAT_INC(ips_badsum);
3561                 goto bad;
3562         }
3563
3564         /* Retrieve the packet length. */
3565         len = ntohs(ip->ip_len);
3566
3567         /*
3568          * Check for additional length bogosity
3569          */
3570         if (len < hlen) {
3571                 KMOD_IPSTAT_INC(ips_badlen);
3572                 goto bad;
3573         }
3574
3575         /*
3576          * Check that the amount of data in the buffers
3577          * is as at least much as the IP header would have us expect.
3578          * Drop packet if shorter than we expect.
3579          */
3580         if (m->m_pkthdr.len < len) {
3581                 KMOD_IPSTAT_INC(ips_tooshort);
3582                 goto bad;
3583         }
3584
3585         /* Checks out, proceed */
3586         *mp = m;
3587         return (0);
3588
3589 bad:
3590         *mp = m;
3591         return (-1);
3592 }
3593
3594 #ifdef INET6
3595 /*
3596  * Same as above, but for IPv6.
3597  * Cut-and-pasted from ip6_input.c.
3598  * XXX Should we update ip6stat, or not?
3599  */
3600 static int
3601 bridge_ip6_checkbasic(struct mbuf **mp)
3602 {
3603         struct mbuf *m = *mp;
3604         struct ip6_hdr *ip6;
3605
3606         /*
3607          * If the IPv6 header is not aligned, slurp it up into a new
3608          * mbuf with space for link headers, in the event we forward
3609          * it.  Otherwise, if it is aligned, make sure the entire base
3610          * IPv6 header is in the first mbuf of the chain.
3611          */
3612         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3613                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3614                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3615                             (max_linkhdr + 3) & ~3)) == NULL) {
3616                         /* XXXJRT new stat, please */
3617                         IP6STAT_INC(ip6s_toosmall);
3618                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3619                         goto bad;
3620                 }
3621         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3622                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3623                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3624                         IP6STAT_INC(ip6s_toosmall);
3625                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3626                         goto bad;
3627                 }
3628         }
3629
3630         ip6 = mtod(m, struct ip6_hdr *);
3631
3632         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3633                 IP6STAT_INC(ip6s_badvers);
3634                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3635                 goto bad;
3636         }
3637
3638         /* Checks out, proceed */
3639         *mp = m;
3640         return (0);
3641
3642 bad:
3643         *mp = m;
3644         return (-1);
3645 }
3646 #endif /* INET6 */
3647
3648 /*
3649  * bridge_fragment:
3650  *
3651  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3652  */
3653 static int
3654 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3655     int snap, struct llc *llc)
3656 {
3657         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3658         struct ip *ip;
3659         int error = -1;
3660
3661         if (m->m_len < sizeof(struct ip) &&
3662             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3663                 goto dropit;
3664         ip = mtod(m, struct ip *);
3665
3666         m->m_pkthdr.csum_flags |= CSUM_IP;
3667         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3668         if (error)
3669                 goto dropit;
3670
3671         /*
3672          * Walk the chain and re-add the Ethernet header for
3673          * each mbuf packet.
3674          */
3675         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3676                 nextpkt = mcur->m_nextpkt;
3677                 mcur->m_nextpkt = NULL;
3678                 if (snap) {
3679                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3680                         if (mcur == NULL) {
3681                                 error = ENOBUFS;
3682                                 if (mprev != NULL)
3683                                         mprev->m_nextpkt = nextpkt;
3684                                 goto dropit;
3685                         }
3686                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3687                 }
3688
3689                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3690                 if (mcur == NULL) {
3691                         error = ENOBUFS;
3692                         if (mprev != NULL)
3693                                 mprev->m_nextpkt = nextpkt;
3694                         goto dropit;
3695                 }
3696                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3697
3698                 /*
3699                  * The previous two M_PREPEND could have inserted one or two
3700                  * mbufs in front so we have to update the previous packet's
3701                  * m_nextpkt.
3702                  */
3703                 mcur->m_nextpkt = nextpkt;
3704                 if (mprev != NULL)
3705                         mprev->m_nextpkt = mcur;
3706                 else {
3707                         /* The first mbuf in the original chain needs to be
3708                          * updated. */
3709                         *mp = mcur;
3710                 }
3711                 mprev = mcur;
3712         }
3713
3714         KMOD_IPSTAT_INC(ips_fragmented);
3715         return (error);
3716
3717 dropit:
3718         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3719                 m = mcur->m_nextpkt;
3720                 m_freem(mcur);
3721         }
3722         return (error);
3723 }
3724
3725 static void
3726 bridge_linkstate(struct ifnet *ifp)
3727 {
3728         struct bridge_softc *sc = ifp->if_bridge;
3729         struct bridge_iflist *bif;
3730         struct epoch_tracker et;
3731
3732         NET_EPOCH_ENTER(et);
3733
3734         bif = bridge_lookup_member_if(sc, ifp);
3735         if (bif == NULL) {
3736                 NET_EPOCH_EXIT(et);
3737                 return;
3738         }
3739         bridge_linkcheck(sc);
3740
3741         bstp_linkstate(&bif->bif_stp);
3742
3743         NET_EPOCH_EXIT(et);
3744 }
3745
3746 static void
3747 bridge_linkcheck(struct bridge_softc *sc)
3748 {
3749         struct bridge_iflist *bif;
3750         int new_link, hasls;
3751
3752         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3753
3754         new_link = LINK_STATE_DOWN;
3755         hasls = 0;
3756         /* Our link is considered up if at least one of our ports is active */
3757         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3758                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3759                         hasls++;
3760                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3761                         new_link = LINK_STATE_UP;
3762                         break;
3763                 }
3764         }
3765         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3766                 /* If no interfaces support link-state then we default to up */
3767                 new_link = LINK_STATE_UP;
3768         }
3769         if_link_state_change(sc->sc_ifp, new_link);
3770 }