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/if_private.h>
 114 #include <net/pfil.h>
 115 #include <net/vnet.h>
 116
 117 #include <netinet/in.h>
 118 #include <netinet/in_systm.h>
 119 #include <netinet/in_var.h>
 120 #include <netinet/ip.h>
 121 #include <netinet/ip_var.h>
 122 #ifdef INET6
 123 #include <netinet/ip6.h>
 124 #include <netinet6/ip6_var.h>
 125 #include <netinet6/in6_ifattach.h>
 126 #endif
 127 #if defined(INET) || defined(INET6)
 128 #include <netinet/ip_carp.h>
 129 #endif
 130 #include <machine/in_cksum.h>
 131 #include <netinet/if_ether.h>
 132 #include <net/bridgestp.h>
 133 #include <net/if_bridgevar.h>
 134 #include <net/if_llc.h>
 135 #include <net/if_vlan_var.h>
 136
 137 #include <net/route.h>
 138
 139 /*
 140  * Size of the route hash table.  Must be a power of two.
 141  */
 142 #ifndef BRIDGE_RTHASH_SIZE
 143 #define BRIDGE_RTHASH_SIZE              1024
 144 #endif
 145
 146 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 147
 148 /*
 149  * Default maximum number of addresses to cache.
 150  */
 151 #ifndef BRIDGE_RTABLE_MAX
 152 #define BRIDGE_RTABLE_MAX               2000
 153 #endif
 154
 155 /*
 156  * Timeout (in seconds) for entries learned dynamically.
 157  */
 158 #ifndef BRIDGE_RTABLE_TIMEOUT
 159 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 160 #endif
 161
 162 /*
 163  * Number of seconds between walks of the route list.
 164  */
 165 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 166 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 167 #endif
 168
 169 /*
 170  * List of capabilities to possibly mask on the member interface.
 171  */
 172 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 173                                          IFCAP_TXCSUM_IPV6)
 174
 175 /*
 176  * List of capabilities to strip
 177  */
 178 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 179
 180 /*
 181  * Bridge locking
 182  *
 183  * The bridge relies heavily on the epoch(9) system to protect its data
 184  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 185  * must ensure there is only one writer at a time.
 186  *
 187  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 188  * accesses we must hold:
 189  *
 190  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 191  *  - BRIDGE_LOCK, for any other change
 192  *
 193  * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
 194  * calls to bridge member interfaces and these ioctl()s can sleep.
 195  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 196  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 197  */
 198 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 199         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 200         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 201 } while (0)
 202 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 203         sx_destroy(&(_sc)->sc_sx);              \
 204         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 205 } while (0)
 206 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 207 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 208 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 209 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 210             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 211 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 212 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 213 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 214 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 215 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 216             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 217
 218 /*
 219  * Bridge interface list entry.
 220  */
 221 struct bridge_iflist {
 222         CK_LIST_ENTRY(bridge_iflist) bif_next;
 223         struct ifnet            *bif_ifp;       /* member if */
 224         struct bstp_port        bif_stp;        /* STP state */
 225         uint32_t                bif_flags;      /* member if flags */
 226         int                     bif_savedcaps;  /* saved capabilities */
 227         uint32_t                bif_addrmax;    /* max # of addresses */
 228         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 229         uint32_t                bif_addrexceeded;/* # of address violations */
 230         struct epoch_context    bif_epoch_ctx;
 231 };
 232
 233 /*
 234  * Bridge route node.
 235  */
 236 struct bridge_rtnode {
 237         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 238         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 239         struct bridge_iflist    *brt_dst;       /* destination if */
 240         unsigned long           brt_expire;     /* expiration time */
 241         uint8_t                 brt_flags;      /* address flags */
 242         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 243         uint16_t                brt_vlan;       /* vlan id */
 244         struct  vnet            *brt_vnet;
 245         struct  epoch_context   brt_epoch_ctx;
 246 };
 247 #define brt_ifp                 brt_dst->bif_ifp
 248
 249 /*
 250  * Software state for each bridge.
 251  */
 252 struct bridge_softc {
 253         struct ifnet            *sc_ifp;        /* make this an interface */
 254         LIST_ENTRY(bridge_softc) sc_list;
 255         struct sx               sc_sx;
 256         struct mtx              sc_rt_mtx;
 257         uint32_t                sc_brtmax;      /* max # of addresses */
 258         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 259         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 260         struct callout          sc_brcallout;   /* bridge callout */
 261         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 262         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 263         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 264         uint32_t                sc_rthash_key;  /* key for hash */
 265         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 266         struct bstp_state       sc_stp;         /* STP state */
 267         uint32_t                sc_brtexceeded; /* # of cache drops */
 268         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 269         struct ether_addr       sc_defaddr;     /* Default MAC address */
 270         if_input_fn_t           sc_if_input;    /* Saved copy of if_input */
 271         struct epoch_context    sc_epoch_ctx;
 272 };
 273
 274 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 275 #define V_bridge_list_sx        VNET(bridge_list_sx)
 276 static eventhandler_tag bridge_detach_cookie;
 277
 278 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 279
 280 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 281 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 282
 283 static int      bridge_clone_create(struct if_clone *, char *, size_t,
 284                     struct ifc_data *, struct ifnet **);
 285 static int      bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
 286
 287 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 288 static void     bridge_mutecaps(struct bridge_softc *);
 289 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 290                     int);
 291 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 292 static void     bridge_init(void *);
 293 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 294 static void     bridge_stop(struct ifnet *, int);
 295 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 296 #ifdef ALTQ
 297 static void     bridge_altq_start(if_t);
 298 static int      bridge_altq_transmit(if_t, struct mbuf *);
 299 #endif
 300 static void     bridge_qflush(struct ifnet *);
 301 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 302 static void     bridge_inject(struct ifnet *, struct mbuf *);
 303 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 304                     struct rtentry *);
 305 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 306                     struct mbuf *);
 307 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 308
 309 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 310                     struct mbuf *m);
 311
 312 static void     bridge_timer(void *);
 313
 314 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 315                     struct mbuf *, int);
 316 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 317
 318 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 319                     uint16_t, struct bridge_iflist *, int, uint8_t);
 320 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 321                     uint16_t);
 322 static void     bridge_rttrim(struct bridge_softc *);
 323 static void     bridge_rtage(struct bridge_softc *);
 324 static void     bridge_rtflush(struct bridge_softc *, int);
 325 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 326                     uint16_t);
 327
 328 static void     bridge_rtable_init(struct bridge_softc *);
 329 static void     bridge_rtable_fini(struct bridge_softc *);
 330
 331 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 332 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 333                     const uint8_t *, uint16_t);
 334 static int      bridge_rtnode_insert(struct bridge_softc *,
 335                     struct bridge_rtnode *);
 336 static void     bridge_rtnode_destroy(struct bridge_softc *,
 337                     struct bridge_rtnode *);
 338 static void     bridge_rtable_expire(struct ifnet *, int);
 339 static void     bridge_state_change(struct ifnet *, int);
 340
 341 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 342                     const char *name);
 343 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 344                     struct ifnet *ifp);
 345 static void     bridge_delete_member(struct bridge_softc *,
 346                     struct bridge_iflist *, int);
 347 static void     bridge_delete_span(struct bridge_softc *,
 348                     struct bridge_iflist *);
 349
 350 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 351 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 352 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 353 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 354 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 381 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 382                     int);
 383 static int      bridge_ip_checkbasic(struct mbuf **mp);
 384 #ifdef INET6
 385 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 386 #endif /* INET6 */
 387 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 388                     struct ether_header *, int, struct llc *);
 389 static void     bridge_linkstate(struct ifnet *ifp);
 390 static void     bridge_linkcheck(struct bridge_softc *sc);
 391
 392 /*
 393  * Use the "null" value from IEEE 802.1Q-2014 Table 9-2
 394  * to indicate untagged frames.
 395  */
 396 #define VLANTAGOF(_m)   \
 397     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL
 398
 399 static struct bstp_cb_ops bridge_ops = {
 400         .bcb_state = bridge_state_change,
 401         .bcb_rtage = bridge_rtable_expire
 402 };
 403
 404 SYSCTL_DECL(_net_link);
 405 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 406     "Bridge");
 407
 408 /* only pass IP[46] packets when pfil is enabled */
 409 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 410 #define V_pfil_onlyip   VNET(pfil_onlyip)
 411 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 412     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 413     "Only pass IP packets when pfil is enabled");
 414
 415 /* run pfil hooks on the bridge interface */
 416 VNET_DEFINE_STATIC(int, pfil_bridge) = 0;
 417 #define V_pfil_bridge   VNET(pfil_bridge)
 418 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 419     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 420     "Packet filter on the bridge interface");
 421
 422 /* layer2 filter with ipfw */
 423 VNET_DEFINE_STATIC(int, pfil_ipfw);
 424 #define V_pfil_ipfw     VNET(pfil_ipfw)
 425
 426 /* layer2 ARP filter with ipfw */
 427 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 428 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 429 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 430     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 431     "Filter ARP packets through IPFW layer2");
 432
 433 /* run pfil hooks on the member interface */
 434 VNET_DEFINE_STATIC(int, pfil_member) = 0;
 435 #define V_pfil_member   VNET(pfil_member)
 436 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 437     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 438     "Packet filter on the member interface");
 439
 440 /* run pfil hooks on the physical interface for locally destined packets */
 441 VNET_DEFINE_STATIC(int, pfil_local_phys);
 442 #define V_pfil_local_phys       VNET(pfil_local_phys)
 443 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 444     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 445     "Packet filter on the physical interface for locally destined packets");
 446
 447 /* log STP state changes */
 448 VNET_DEFINE_STATIC(int, log_stp);
 449 #define V_log_stp       VNET(log_stp)
 450 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 451     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 452     "Log STP state changes");
 453
 454 /* share MAC with first bridge member */
 455 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 456 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 457 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 458     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 459     "Inherit MAC address from the first bridge member");
 460
 461 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 462 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 463 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 464     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 465     "Allow overlap of link-local scope "
 466     "zones of a bridge interface and the member interfaces");
 467
 468 /* log MAC address port flapping */
 469 VNET_DEFINE_STATIC(bool, log_mac_flap) = true;
 470 #define V_log_mac_flap  VNET(log_mac_flap)
 471 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
 472     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true,
 473     "Log MAC address port flapping");
 474
 475 VNET_DEFINE_STATIC(int, log_interval) = 5;
 476 VNET_DEFINE_STATIC(int, log_count) = 0;
 477 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 };
 478
 479 #define V_log_interval  VNET(log_interval)
 480 #define V_log_count     VNET(log_count)
 481 #define V_log_last      VNET(log_last)
 482
 483 struct bridge_control {
 484         int     (*bc_func)(struct bridge_softc *, void *);
 485         int     bc_argsize;
 486         int     bc_flags;
 487 };
 488
 489 #define BC_F_COPYIN             0x01    /* copy arguments in */
 490 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 491 #define BC_F_SUSER              0x04    /* do super-user check */
 492
 493 static const struct bridge_control bridge_control_table[] = {
 494         { bridge_ioctl_add,             sizeof(struct ifbreq),
 495           BC_F_COPYIN|BC_F_SUSER },
 496         { bridge_ioctl_del,             sizeof(struct ifbreq),
 497           BC_F_COPYIN|BC_F_SUSER },
 498
 499         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 500           BC_F_COPYIN|BC_F_COPYOUT },
 501         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 502           BC_F_COPYIN|BC_F_SUSER },
 503
 504         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 505           BC_F_COPYIN|BC_F_SUSER },
 506         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 507           BC_F_COPYOUT },
 508
 509         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 510           BC_F_COPYIN|BC_F_COPYOUT },
 511         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 512           BC_F_COPYIN|BC_F_COPYOUT },
 513
 514         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 515           BC_F_COPYIN|BC_F_SUSER },
 516
 517         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 518           BC_F_COPYIN|BC_F_SUSER },
 519         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 520           BC_F_COPYOUT },
 521
 522         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 523           BC_F_COPYIN|BC_F_SUSER },
 524
 525         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 526           BC_F_COPYIN|BC_F_SUSER },
 527
 528         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 529           BC_F_COPYOUT },
 530         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 531           BC_F_COPYIN|BC_F_SUSER },
 532
 533         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 534           BC_F_COPYOUT },
 535         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 536           BC_F_COPYIN|BC_F_SUSER },
 537
 538         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 539           BC_F_COPYOUT },
 540         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 541           BC_F_COPYIN|BC_F_SUSER },
 542
 543         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 544           BC_F_COPYOUT },
 545         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 546           BC_F_COPYIN|BC_F_SUSER },
 547
 548         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 549           BC_F_COPYIN|BC_F_SUSER },
 550
 551         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 552           BC_F_COPYIN|BC_F_SUSER },
 553
 554         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 555           BC_F_COPYIN|BC_F_SUSER },
 556         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 557           BC_F_COPYIN|BC_F_SUSER },
 558
 559         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 560           BC_F_COPYOUT },
 561
 562         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 563           BC_F_COPYOUT },
 564
 565         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 566           BC_F_COPYIN|BC_F_COPYOUT },
 567
 568         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 569           BC_F_COPYIN|BC_F_SUSER },
 570
 571         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 572           BC_F_COPYIN|BC_F_SUSER },
 573
 574         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 575           BC_F_COPYIN|BC_F_SUSER },
 576
 577 };
 578 static const int bridge_control_table_size = nitems(bridge_control_table);
 579
 580 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 581 #define V_bridge_list   VNET(bridge_list)
 582 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 583                                             "if_bridge list")
 584 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 585 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 586 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 587
 588 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 589 #define V_bridge_cloner VNET(bridge_cloner)
 590
 591 static const char bridge_name[] = "bridge";
 592
 593 static void
 594 vnet_bridge_init(const void *unused __unused)
 595 {
 596
 597         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 598             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 599             UMA_ALIGN_PTR, 0);
 600         BRIDGE_LIST_LOCK_INIT();
 601         LIST_INIT(&V_bridge_list);
 602
 603         struct if_clone_addreq req = {
 604                 .create_f = bridge_clone_create,
 605                 .destroy_f = bridge_clone_destroy,
 606                 .flags = IFC_F_AUTOUNIT,
 607         };
 608         V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
 609 }
 610 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 611     vnet_bridge_init, NULL);
 612
 613 static void
 614 vnet_bridge_uninit(const void *unused __unused)
 615 {
 616
 617         ifc_detach_cloner(V_bridge_cloner);
 618         V_bridge_cloner = NULL;
 619         BRIDGE_LIST_LOCK_DESTROY();
 620
 621         /* Callbacks may use the UMA zone. */
 622         NET_EPOCH_DRAIN_CALLBACKS();
 623
 624         uma_zdestroy(V_bridge_rtnode_zone);
 625 }
 626 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 627     vnet_bridge_uninit, NULL);
 628
 629 static int
 630 bridge_modevent(module_t mod, int type, void *data)
 631 {
 632
 633         switch (type) {
 634         case MOD_LOAD:
 635                 bridge_dn_p = bridge_dummynet;
 636                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 637                     ifnet_departure_event, bridge_ifdetach, NULL,
 638                     EVENTHANDLER_PRI_ANY);
 639                 break;
 640         case MOD_UNLOAD:
 641                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 642                     bridge_detach_cookie);
 643                 bridge_dn_p = NULL;
 644                 break;
 645         default:
 646                 return (EOPNOTSUPP);
 647         }
 648         return (0);
 649 }
 650
 651 static moduledata_t bridge_mod = {
 652         "if_bridge",
 653         bridge_modevent,
 654         0
 655 };
 656
 657 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 658 MODULE_VERSION(if_bridge, 1);
 659 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 660
 661 /*
 662  * handler for net.link.bridge.ipfw
 663  */
 664 static int
 665 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 666 {
 667         int enable = V_pfil_ipfw;
 668         int error;
 669
 670         error = sysctl_handle_int(oidp, &enable, 0, req);
 671         enable &= 1;
 672
 673         if (enable != V_pfil_ipfw) {
 674                 V_pfil_ipfw = enable;
 675
 676                 /*
 677                  * Disable pfil so that ipfw doesnt run twice, if the user
 678                  * really wants both then they can re-enable pfil_bridge and/or
 679                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 680                  * layer2 type.
 681                  */
 682                 if (V_pfil_ipfw) {
 683                         V_pfil_onlyip = 0;
 684                         V_pfil_bridge = 0;
 685                         V_pfil_member = 0;
 686                 }
 687         }
 688
 689         return (error);
 690 }
 691 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 692     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
 693     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 694     "Layer2 filter with IPFW");
 695
 696 #ifdef VIMAGE
 697 static void
 698 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 699 {
 700         struct bridge_softc *sc = ifp->if_softc;
 701         struct bridge_iflist *bif;
 702
 703         BRIDGE_LOCK(sc);
 704
 705         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 706                 bridge_delete_member(sc, bif, 0);
 707
 708         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 709                 bridge_delete_span(sc, bif);
 710         }
 711
 712         BRIDGE_UNLOCK(sc);
 713
 714         ether_reassign(ifp, newvnet, arg);
 715 }
 716 #endif
 717
 718 /*
 719  * bridge_clone_create:
 720  *
 721  *      Create a new bridge instance.
 722  */
 723 static int
 724 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
 725     struct ifc_data *ifd, struct ifnet **ifpp)
 726 {
 727         struct bridge_softc *sc;
 728         struct ifnet *ifp;
 729
 730         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 731         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 732         if (ifp == NULL) {
 733                 free(sc, M_DEVBUF);
 734                 return (ENOSPC);
 735         }
 736
 737         BRIDGE_LOCK_INIT(sc);
 738         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 739         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 740
 741         /* Initialize our routing table. */
 742         bridge_rtable_init(sc);
 743
 744         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 745
 746         CK_LIST_INIT(&sc->sc_iflist);
 747         CK_LIST_INIT(&sc->sc_spanlist);
 748
 749         ifp->if_softc = sc;
 750         if_initname(ifp, bridge_name, ifd->unit);
 751         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 752         ifp->if_ioctl = bridge_ioctl;
 753 #ifdef ALTQ
 754         ifp->if_start = bridge_altq_start;
 755         ifp->if_transmit = bridge_altq_transmit;
 756         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 757         ifp->if_snd.ifq_drv_maxlen = 0;
 758         IFQ_SET_READY(&ifp->if_snd);
 759 #else
 760         ifp->if_transmit = bridge_transmit;
 761 #endif
 762         ifp->if_qflush = bridge_qflush;
 763         ifp->if_init = bridge_init;
 764         ifp->if_type = IFT_BRIDGE;
 765
 766         ether_gen_addr(ifp, &sc->sc_defaddr);
 767
 768         bstp_attach(&sc->sc_stp, &bridge_ops);
 769         ether_ifattach(ifp, sc->sc_defaddr.octet);
 770         /* Now undo some of the damage... */
 771         ifp->if_baudrate = 0;
 772         ifp->if_type = IFT_BRIDGE;
 773 #ifdef VIMAGE
 774         ifp->if_reassign = bridge_reassign;
 775 #endif
 776         sc->sc_if_input = ifp->if_input;        /* ether_input */
 777         ifp->if_input = bridge_inject;
 778
 779         /*
 780          * Allow BRIDGE_INPUT() to pass in packets originating from the bridge
 781          * itself via bridge_inject().  This is required for netmap but
 782          * otherwise has no effect.
 783          */
 784         ifp->if_bridge_input = bridge_input;
 785
 786         BRIDGE_LIST_LOCK();
 787         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 788         BRIDGE_LIST_UNLOCK();
 789         *ifpp = ifp;
 790
 791         return (0);
 792 }
 793
 794 static void
 795 bridge_clone_destroy_cb(struct epoch_context *ctx)
 796 {
 797         struct bridge_softc *sc;
 798
 799         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 800
 801         BRIDGE_LOCK_DESTROY(sc);
 802         free(sc, M_DEVBUF);
 803 }
 804
 805 /*
 806  * bridge_clone_destroy:
 807  *
 808  *      Destroy a bridge instance.
 809  */
 810 static int
 811 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
 812 {
 813         struct bridge_softc *sc = ifp->if_softc;
 814         struct bridge_iflist *bif;
 815         struct epoch_tracker et;
 816
 817         BRIDGE_LOCK(sc);
 818
 819         bridge_stop(ifp, 1);
 820         ifp->if_flags &= ~IFF_UP;
 821
 822         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 823                 bridge_delete_member(sc, bif, 0);
 824
 825         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 826                 bridge_delete_span(sc, bif);
 827         }
 828
 829         /* Tear down the routing table. */
 830         bridge_rtable_fini(sc);
 831
 832         BRIDGE_UNLOCK(sc);
 833
 834         NET_EPOCH_ENTER(et);
 835
 836         callout_drain(&sc->sc_brcallout);
 837
 838         BRIDGE_LIST_LOCK();
 839         LIST_REMOVE(sc, sc_list);
 840         BRIDGE_LIST_UNLOCK();
 841
 842         bstp_detach(&sc->sc_stp);
 843 #ifdef ALTQ
 844         IFQ_PURGE(&ifp->if_snd);
 845 #endif
 846         NET_EPOCH_EXIT(et);
 847
 848         ether_ifdetach(ifp);
 849         if_free(ifp);
 850
 851         NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
 852
 853         return (0);
 854 }
 855
 856 /*
 857  * bridge_ioctl:
 858  *
 859  *      Handle a control request from the operator.
 860  */
 861 static int
 862 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 863 {
 864         struct bridge_softc *sc = ifp->if_softc;
 865         struct ifreq *ifr = (struct ifreq *)data;
 866         struct bridge_iflist *bif;
 867         struct thread *td = curthread;
 868         union {
 869                 struct ifbreq ifbreq;
 870                 struct ifbifconf ifbifconf;
 871                 struct ifbareq ifbareq;
 872                 struct ifbaconf ifbaconf;
 873                 struct ifbrparam ifbrparam;
 874                 struct ifbropreq ifbropreq;
 875         } args;
 876         struct ifdrv *ifd = (struct ifdrv *) data;
 877         const struct bridge_control *bc;
 878         int error = 0, oldmtu;
 879
 880         BRIDGE_LOCK(sc);
 881
 882         switch (cmd) {
 883         case SIOCADDMULTI:
 884         case SIOCDELMULTI:
 885                 break;
 886
 887         case SIOCGDRVSPEC:
 888         case SIOCSDRVSPEC:
 889                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 890                         error = EINVAL;
 891                         break;
 892                 }
 893                 bc = &bridge_control_table[ifd->ifd_cmd];
 894
 895                 if (cmd == SIOCGDRVSPEC &&
 896                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 897                         error = EINVAL;
 898                         break;
 899                 }
 900                 else if (cmd == SIOCSDRVSPEC &&
 901                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 902                         error = EINVAL;
 903                         break;
 904                 }
 905
 906                 if (bc->bc_flags & BC_F_SUSER) {
 907                         error = priv_check(td, PRIV_NET_BRIDGE);
 908                         if (error)
 909                                 break;
 910                 }
 911
 912                 if (ifd->ifd_len != bc->bc_argsize ||
 913                     ifd->ifd_len > sizeof(args)) {
 914                         error = EINVAL;
 915                         break;
 916                 }
 917
 918                 bzero(&args, sizeof(args));
 919                 if (bc->bc_flags & BC_F_COPYIN) {
 920                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 921                         if (error)
 922                                 break;
 923                 }
 924
 925                 oldmtu = ifp->if_mtu;
 926                 error = (*bc->bc_func)(sc, &args);
 927                 if (error)
 928                         break;
 929
 930                 /*
 931                  * Bridge MTU may change during addition of the first port.
 932                  * If it did, do network layer specific procedure.
 933                  */
 934                 if (ifp->if_mtu != oldmtu)
 935                         if_notifymtu(ifp);
 936
 937                 if (bc->bc_flags & BC_F_COPYOUT)
 938                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 939
 940                 break;
 941
 942         case SIOCSIFFLAGS:
 943                 if (!(ifp->if_flags & IFF_UP) &&
 944                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 945                         /*
 946                          * If interface is marked down and it is running,
 947                          * then stop and disable it.
 948                          */
 949                         bridge_stop(ifp, 1);
 950                 } else if ((ifp->if_flags & IFF_UP) &&
 951                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 952                         /*
 953                          * If interface is marked up and it is stopped, then
 954                          * start it.
 955                          */
 956                         BRIDGE_UNLOCK(sc);
 957                         (*ifp->if_init)(sc);
 958                         BRIDGE_LOCK(sc);
 959                 }
 960                 break;
 961
 962         case SIOCSIFMTU:
 963                 oldmtu = sc->sc_ifp->if_mtu;
 964
 965                 if (ifr->ifr_mtu < 576) {
 966                         error = EINVAL;
 967                         break;
 968                 }
 969                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 970                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 971                         break;
 972                 }
 973                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 974                         error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 975                             SIOCSIFMTU, (caddr_t)ifr);
 976                         if (error != 0) {
 977                                 log(LOG_NOTICE, "%s: invalid MTU: %u for"
 978                                     " member %s\n", sc->sc_ifp->if_xname,
 979                                     ifr->ifr_mtu,
 980                                     bif->bif_ifp->if_xname);
 981                                 error = EINVAL;
 982                                 break;
 983                         }
 984                 }
 985                 if (error) {
 986                         /* Restore the previous MTU on all member interfaces. */
 987                         ifr->ifr_mtu = oldmtu;
 988                         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 989                                 (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 990                                     SIOCSIFMTU, (caddr_t)ifr);
 991                         }
 992                 } else {
 993                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 994                 }
 995                 break;
 996         default:
 997                 /*
 998                  * drop the lock as ether_ioctl() will call bridge_start() and
 999                  * cause the lock to be recursed.
1000                  */
1001                 BRIDGE_UNLOCK(sc);
1002                 error = ether_ioctl(ifp, cmd, data);
1003                 BRIDGE_LOCK(sc);
1004                 break;
1005         }
1006
1007         BRIDGE_UNLOCK(sc);
1008
1009         return (error);
1010 }
1011
1012 /*
1013  * bridge_mutecaps:
1014  *
1015  *      Clear or restore unwanted capabilities on the member interface
1016  */
1017 static void
1018 bridge_mutecaps(struct bridge_softc *sc)
1019 {
1020         struct bridge_iflist *bif;
1021         int enabled, mask;
1022
1023         BRIDGE_LOCK_ASSERT(sc);
1024
1025         /* Initial bitmask of capabilities to test */
1026         mask = BRIDGE_IFCAPS_MASK;
1027
1028         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1029                 /* Every member must support it or its disabled */
1030                 mask &= bif->bif_savedcaps;
1031         }
1032
1033         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1034                 enabled = bif->bif_ifp->if_capenable;
1035                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1036                 /* strip off mask bits and enable them again if allowed */
1037                 enabled &= ~BRIDGE_IFCAPS_MASK;
1038                 enabled |= mask;
1039                 bridge_set_ifcap(sc, bif, enabled);
1040         }
1041 }
1042
1043 static void
1044 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1045 {
1046         struct ifnet *ifp = bif->bif_ifp;
1047         struct ifreq ifr;
1048         int error, mask, stuck;
1049
1050         bzero(&ifr, sizeof(ifr));
1051         ifr.ifr_reqcap = set;
1052
1053         if (ifp->if_capenable != set) {
1054                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1055                 if (error)
1056                         if_printf(sc->sc_ifp,
1057                             "error setting capabilities on %s: %d\n",
1058                             ifp->if_xname, error);
1059                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1060                 stuck = ifp->if_capenable & mask & ~set;
1061                 if (stuck != 0)
1062                         if_printf(sc->sc_ifp,
1063                             "can't disable some capabilities on %s: 0x%x\n",
1064                             ifp->if_xname, stuck);
1065         }
1066 }
1067
1068 /*
1069  * bridge_lookup_member:
1070  *
1071  *      Lookup a bridge member interface.
1072  */
1073 static struct bridge_iflist *
1074 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1075 {
1076         struct bridge_iflist *bif;
1077         struct ifnet *ifp;
1078
1079         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1080
1081         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1082                 ifp = bif->bif_ifp;
1083                 if (strcmp(ifp->if_xname, name) == 0)
1084                         return (bif);
1085         }
1086
1087         return (NULL);
1088 }
1089
1090 /*
1091  * bridge_lookup_member_if:
1092  *
1093  *      Lookup a bridge member interface by ifnet*.
1094  */
1095 static struct bridge_iflist *
1096 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1097 {
1098         struct bridge_iflist *bif;
1099
1100         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1101
1102         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1103                 if (bif->bif_ifp == member_ifp)
1104                         return (bif);
1105         }
1106
1107         return (NULL);
1108 }
1109
1110 static void
1111 bridge_delete_member_cb(struct epoch_context *ctx)
1112 {
1113         struct bridge_iflist *bif;
1114
1115         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1116
1117         free(bif, M_DEVBUF);
1118 }
1119
1120 /*
1121  * bridge_delete_member:
1122  *
1123  *      Delete the specified member interface.
1124  */
1125 static void
1126 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1127     int gone)
1128 {
1129         struct ifnet *ifs = bif->bif_ifp;
1130         struct ifnet *fif = NULL;
1131         struct bridge_iflist *bifl;
1132
1133         BRIDGE_LOCK_ASSERT(sc);
1134
1135         if (bif->bif_flags & IFBIF_STP)
1136                 bstp_disable(&bif->bif_stp);
1137
1138         ifs->if_bridge = NULL;
1139         CK_LIST_REMOVE(bif, bif_next);
1140
1141         /*
1142          * If removing the interface that gave the bridge its mac address, set
1143          * the mac address of the bridge to the address of the next member, or
1144          * to its default address if no members are left.
1145          */
1146         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1147                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1148                         bcopy(&sc->sc_defaddr,
1149                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1150                         sc->sc_ifaddr = NULL;
1151                 } else {
1152                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1153                         fif = bifl->bif_ifp;
1154                         bcopy(IF_LLADDR(fif),
1155                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1156                         sc->sc_ifaddr = fif;
1157                 }
1158                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1159         }
1160
1161         bridge_linkcheck(sc);
1162         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1163         BRIDGE_RT_LOCK(sc);
1164         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1165         BRIDGE_RT_UNLOCK(sc);
1166         KASSERT(bif->bif_addrcnt == 0,
1167             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1168
1169         ifs->if_bridge_output = NULL;
1170         ifs->if_bridge_input = NULL;
1171         ifs->if_bridge_linkstate = NULL;
1172         if (!gone) {
1173                 switch (ifs->if_type) {
1174                 case IFT_ETHER:
1175                 case IFT_L2VLAN:
1176                         /*
1177                          * Take the interface out of promiscuous mode, but only
1178                          * if it was promiscuous in the first place. It might
1179                          * not be if we're in the bridge_ioctl_add() error path.
1180                          */
1181                         if (ifs->if_flags & IFF_PROMISC)
1182                                 (void) ifpromisc(ifs, 0);
1183                         break;
1184
1185                 case IFT_GIF:
1186                         break;
1187
1188                 default:
1189 #ifdef DIAGNOSTIC
1190                         panic("bridge_delete_member: impossible");
1191 #endif
1192                         break;
1193                 }
1194                 /* reneable any interface capabilities */
1195                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1196         }
1197         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1198
1199         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1200 }
1201
1202 /*
1203  * bridge_delete_span:
1204  *
1205  *      Delete the specified span interface.
1206  */
1207 static void
1208 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1209 {
1210         BRIDGE_LOCK_ASSERT(sc);
1211
1212         KASSERT(bif->bif_ifp->if_bridge == NULL,
1213             ("%s: not a span interface", __func__));
1214
1215         CK_LIST_REMOVE(bif, bif_next);
1216
1217         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1218 }
1219
1220 static int
1221 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1222 {
1223         struct ifbreq *req = arg;
1224         struct bridge_iflist *bif = NULL;
1225         struct ifnet *ifs;
1226         int error = 0;
1227
1228         ifs = ifunit(req->ifbr_ifsname);
1229         if (ifs == NULL)
1230                 return (ENOENT);
1231         if (ifs->if_ioctl == NULL)      /* must be supported */
1232                 return (EINVAL);
1233
1234         /* If it's in the span list, it can't be a member. */
1235         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1236                 if (ifs == bif->bif_ifp)
1237                         return (EBUSY);
1238
1239         if (ifs->if_bridge == sc)
1240                 return (EEXIST);
1241
1242         if (ifs->if_bridge != NULL)
1243                 return (EBUSY);
1244
1245         switch (ifs->if_type) {
1246         case IFT_ETHER:
1247         case IFT_L2VLAN:
1248         case IFT_GIF:
1249                 /* permitted interface types */
1250                 break;
1251         default:
1252                 return (EINVAL);
1253         }
1254
1255 #ifdef INET6
1256         /*
1257          * Two valid inet6 addresses with link-local scope must not be
1258          * on the parent interface and the member interfaces at the
1259          * same time.  This restriction is needed to prevent violation
1260          * of link-local scope zone.  Attempts to add a member
1261          * interface which has inet6 addresses when the parent has
1262          * inet6 triggers removal of all inet6 addresses on the member
1263          * interface.
1264          */
1265
1266         /* Check if the parent interface has a link-local scope addr. */
1267         if (V_allow_llz_overlap == 0 &&
1268             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1269                 /*
1270                  * If any, remove all inet6 addresses from the member
1271                  * interfaces.
1272                  */
1273                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1274                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1275                                 in6_ifdetach(bif->bif_ifp);
1276                                 if_printf(sc->sc_ifp,
1277                                     "IPv6 addresses on %s have been removed "
1278                                     "before adding it as a member to prevent "
1279                                     "IPv6 address scope violation.\n",
1280                                     bif->bif_ifp->if_xname);
1281                         }
1282                 }
1283                 if (in6ifa_llaonifp(ifs)) {
1284                         in6_ifdetach(ifs);
1285                         if_printf(sc->sc_ifp,
1286                             "IPv6 addresses on %s have been removed "
1287                             "before adding it as a member to prevent "
1288                             "IPv6 address scope violation.\n",
1289                             ifs->if_xname);
1290                 }
1291         }
1292 #endif
1293         /* Allow the first Ethernet member to define the MTU */
1294         if (CK_LIST_EMPTY(&sc->sc_iflist))
1295                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1296         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1297                 struct ifreq ifr;
1298
1299                 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1300                     ifs->if_xname);
1301                 ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1302
1303                 error = (*ifs->if_ioctl)(ifs,
1304                     SIOCSIFMTU, (caddr_t)&ifr);
1305                 if (error != 0) {
1306                         log(LOG_NOTICE, "%s: invalid MTU: %u for"
1307                             " new member %s\n", sc->sc_ifp->if_xname,
1308                             ifr.ifr_mtu,
1309                             ifs->if_xname);
1310                         return (EINVAL);
1311                 }
1312         }
1313
1314         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1315         if (bif == NULL)
1316                 return (ENOMEM);
1317
1318         bif->bif_ifp = ifs;
1319         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1320         bif->bif_savedcaps = ifs->if_capenable;
1321
1322         /*
1323          * Assign the interface's MAC address to the bridge if it's the first
1324          * member and the MAC address of the bridge has not been changed from
1325          * the default randomly generated one.
1326          */
1327         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1328             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1329                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1330                 sc->sc_ifaddr = ifs;
1331                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1332         }
1333
1334         ifs->if_bridge = sc;
1335         ifs->if_bridge_output = bridge_output;
1336         ifs->if_bridge_input = bridge_input;
1337         ifs->if_bridge_linkstate = bridge_linkstate;
1338         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1339         /*
1340          * XXX: XLOCK HERE!?!
1341          *
1342          * NOTE: insert_***HEAD*** should be safe for the traversals.
1343          */
1344         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1345
1346         /* Set interface capabilities to the intersection set of all members */
1347         bridge_mutecaps(sc);
1348         bridge_linkcheck(sc);
1349
1350         /* Place the interface into promiscuous mode */
1351         switch (ifs->if_type) {
1352                 case IFT_ETHER:
1353                 case IFT_L2VLAN:
1354                         error = ifpromisc(ifs, 1);
1355                         break;
1356         }
1357
1358         if (error)
1359                 bridge_delete_member(sc, bif, 0);
1360         return (error);
1361 }
1362
1363 static int
1364 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1365 {
1366         struct ifbreq *req = arg;
1367         struct bridge_iflist *bif;
1368
1369         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1370         if (bif == NULL)
1371                 return (ENOENT);
1372
1373         bridge_delete_member(sc, bif, 0);
1374
1375         return (0);
1376 }
1377
1378 static int
1379 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1380 {
1381         struct ifbreq *req = arg;
1382         struct bridge_iflist *bif;
1383         struct bstp_port *bp;
1384
1385         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1386         if (bif == NULL)
1387                 return (ENOENT);
1388
1389         bp = &bif->bif_stp;
1390         req->ifbr_ifsflags = bif->bif_flags;
1391         req->ifbr_state = bp->bp_state;
1392         req->ifbr_priority = bp->bp_priority;
1393         req->ifbr_path_cost = bp->bp_path_cost;
1394         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1395         req->ifbr_proto = bp->bp_protover;
1396         req->ifbr_role = bp->bp_role;
1397         req->ifbr_stpflags = bp->bp_flags;
1398         req->ifbr_addrcnt = bif->bif_addrcnt;
1399         req->ifbr_addrmax = bif->bif_addrmax;
1400         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1401
1402         /* Copy STP state options as flags */
1403         if (bp->bp_operedge)
1404                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1405         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1406                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1407         if (bp->bp_ptp_link)
1408                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1409         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1410                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1411         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1412                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1413         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1414                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1415         return (0);
1416 }
1417
1418 static int
1419 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1420 {
1421         struct epoch_tracker et;
1422         struct ifbreq *req = arg;
1423         struct bridge_iflist *bif;
1424         struct bstp_port *bp;
1425         int error;
1426
1427         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1428         if (bif == NULL)
1429                 return (ENOENT);
1430         bp = &bif->bif_stp;
1431
1432         if (req->ifbr_ifsflags & IFBIF_SPAN)
1433                 /* SPAN is readonly */
1434                 return (EINVAL);
1435
1436         NET_EPOCH_ENTER(et);
1437
1438         if (req->ifbr_ifsflags & IFBIF_STP) {
1439                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1440                         error = bstp_enable(&bif->bif_stp);
1441                         if (error) {
1442                                 NET_EPOCH_EXIT(et);
1443                                 return (error);
1444                         }
1445                 }
1446         } else {
1447                 if ((bif->bif_flags & IFBIF_STP) != 0)
1448                         bstp_disable(&bif->bif_stp);
1449         }
1450
1451         /* Pass on STP flags */
1452         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1453         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1454         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1455         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1456
1457         /* Save the bits relating to the bridge */
1458         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1459
1460         NET_EPOCH_EXIT(et);
1461
1462         return (0);
1463 }
1464
1465 static int
1466 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1467 {
1468         struct ifbrparam *param = arg;
1469
1470         sc->sc_brtmax = param->ifbrp_csize;
1471         bridge_rttrim(sc);
1472
1473         return (0);
1474 }
1475
1476 static int
1477 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1478 {
1479         struct ifbrparam *param = arg;
1480
1481         param->ifbrp_csize = sc->sc_brtmax;
1482
1483         return (0);
1484 }
1485
1486 static int
1487 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1488 {
1489         struct ifbifconf *bifc = arg;
1490         struct bridge_iflist *bif;
1491         struct ifbreq breq;
1492         char *buf, *outbuf;
1493         int count, buflen, len, error = 0;
1494
1495         count = 0;
1496         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1497                 count++;
1498         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1499                 count++;
1500
1501         buflen = sizeof(breq) * count;
1502         if (bifc->ifbic_len == 0) {
1503                 bifc->ifbic_len = buflen;
1504                 return (0);
1505         }
1506         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1507         if (outbuf == NULL)
1508                 return (ENOMEM);
1509
1510         count = 0;
1511         buf = outbuf;
1512         len = min(bifc->ifbic_len, buflen);
1513         bzero(&breq, sizeof(breq));
1514         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1515                 if (len < sizeof(breq))
1516                         break;
1517
1518                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1519                     sizeof(breq.ifbr_ifsname));
1520                 /* Fill in the ifbreq structure */
1521                 error = bridge_ioctl_gifflags(sc, &breq);
1522                 if (error)
1523                         break;
1524                 memcpy(buf, &breq, sizeof(breq));
1525                 count++;
1526                 buf += sizeof(breq);
1527                 len -= sizeof(breq);
1528         }
1529         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1530                 if (len < sizeof(breq))
1531                         break;
1532
1533                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1534                     sizeof(breq.ifbr_ifsname));
1535                 breq.ifbr_ifsflags = bif->bif_flags;
1536                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1537                 memcpy(buf, &breq, sizeof(breq));
1538                 count++;
1539                 buf += sizeof(breq);
1540                 len -= sizeof(breq);
1541         }
1542
1543         bifc->ifbic_len = sizeof(breq) * count;
1544         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1545         free(outbuf, M_TEMP);
1546         return (error);
1547 }
1548
1549 static int
1550 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1551 {
1552         struct ifbaconf *bac = arg;
1553         struct bridge_rtnode *brt;
1554         struct ifbareq bareq;
1555         char *buf, *outbuf;
1556         int count, buflen, len, error = 0;
1557
1558         if (bac->ifbac_len == 0)
1559                 return (0);
1560
1561         count = 0;
1562         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1563                 count++;
1564         buflen = sizeof(bareq) * count;
1565
1566         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1567         if (outbuf == NULL)
1568                 return (ENOMEM);
1569
1570         count = 0;
1571         buf = outbuf;
1572         len = min(bac->ifbac_len, buflen);
1573         bzero(&bareq, sizeof(bareq));
1574         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1575                 if (len < sizeof(bareq))
1576                         goto out;
1577                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1578                     sizeof(bareq.ifba_ifsname));
1579                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1580                 bareq.ifba_vlan = brt->brt_vlan;
1581                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1582                                 time_uptime < brt->brt_expire)
1583                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1584                 else
1585                         bareq.ifba_expire = 0;
1586                 bareq.ifba_flags = brt->brt_flags;
1587
1588                 memcpy(buf, &bareq, sizeof(bareq));
1589                 count++;
1590                 buf += sizeof(bareq);
1591                 len -= sizeof(bareq);
1592         }
1593 out:
1594         bac->ifbac_len = sizeof(bareq) * count;
1595         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1596         free(outbuf, M_TEMP);
1597         return (error);
1598 }
1599
1600 static int
1601 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1602 {
1603         struct ifbareq *req = arg;
1604         struct bridge_iflist *bif;
1605         struct epoch_tracker et;
1606         int error;
1607
1608         NET_EPOCH_ENTER(et);
1609         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1610         if (bif == NULL) {
1611                 NET_EPOCH_EXIT(et);
1612                 return (ENOENT);
1613         }
1614
1615         /* bridge_rtupdate() may acquire the lock. */
1616         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1617             req->ifba_flags);
1618         NET_EPOCH_EXIT(et);
1619
1620         return (error);
1621 }
1622
1623 static int
1624 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1625 {
1626         struct ifbrparam *param = arg;
1627
1628         sc->sc_brttimeout = param->ifbrp_ctime;
1629         return (0);
1630 }
1631
1632 static int
1633 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1634 {
1635         struct ifbrparam *param = arg;
1636
1637         param->ifbrp_ctime = sc->sc_brttimeout;
1638         return (0);
1639 }
1640
1641 static int
1642 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1643 {
1644         struct ifbareq *req = arg;
1645         int vlan = req->ifba_vlan;
1646
1647         /* Userspace uses '0' to mean 'any vlan' */
1648         if (vlan == 0)
1649                 vlan = DOT1Q_VID_RSVD_IMPL;
1650
1651         return (bridge_rtdaddr(sc, req->ifba_dst, vlan));
1652 }
1653
1654 static int
1655 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1656 {
1657         struct ifbreq *req = arg;
1658
1659         BRIDGE_RT_LOCK(sc);
1660         bridge_rtflush(sc, req->ifbr_ifsflags);
1661         BRIDGE_RT_UNLOCK(sc);
1662
1663         return (0);
1664 }
1665
1666 static int
1667 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1668 {
1669         struct ifbrparam *param = arg;
1670         struct bstp_state *bs = &sc->sc_stp;
1671
1672         param->ifbrp_prio = bs->bs_bridge_priority;
1673         return (0);
1674 }
1675
1676 static int
1677 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1678 {
1679         struct ifbrparam *param = arg;
1680
1681         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1682 }
1683
1684 static int
1685 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1686 {
1687         struct ifbrparam *param = arg;
1688         struct bstp_state *bs = &sc->sc_stp;
1689
1690         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1691         return (0);
1692 }
1693
1694 static int
1695 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1696 {
1697         struct ifbrparam *param = arg;
1698
1699         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1700 }
1701
1702 static int
1703 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1704 {
1705         struct ifbrparam *param = arg;
1706         struct bstp_state *bs = &sc->sc_stp;
1707
1708         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1709         return (0);
1710 }
1711
1712 static int
1713 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1714 {
1715         struct ifbrparam *param = arg;
1716
1717         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1718 }
1719
1720 static int
1721 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1722 {
1723         struct ifbrparam *param = arg;
1724         struct bstp_state *bs = &sc->sc_stp;
1725
1726         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1727         return (0);
1728 }
1729
1730 static int
1731 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1732 {
1733         struct ifbrparam *param = arg;
1734
1735         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1736 }
1737
1738 static int
1739 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1740 {
1741         struct ifbreq *req = arg;
1742         struct bridge_iflist *bif;
1743
1744         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1745         if (bif == NULL)
1746                 return (ENOENT);
1747
1748         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1749 }
1750
1751 static int
1752 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1753 {
1754         struct ifbreq *req = arg;
1755         struct bridge_iflist *bif;
1756
1757         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1758         if (bif == NULL)
1759                 return (ENOENT);
1760
1761         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1762 }
1763
1764 static int
1765 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1766 {
1767         struct ifbreq *req = arg;
1768         struct bridge_iflist *bif;
1769
1770         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1771         if (bif == NULL)
1772                 return (ENOENT);
1773
1774         bif->bif_addrmax = req->ifbr_addrmax;
1775         return (0);
1776 }
1777
1778 static int
1779 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1780 {
1781         struct ifbreq *req = arg;
1782         struct bridge_iflist *bif = NULL;
1783         struct ifnet *ifs;
1784
1785         ifs = ifunit(req->ifbr_ifsname);
1786         if (ifs == NULL)
1787                 return (ENOENT);
1788
1789         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1790                 if (ifs == bif->bif_ifp)
1791                         return (EBUSY);
1792
1793         if (ifs->if_bridge != NULL)
1794                 return (EBUSY);
1795
1796         switch (ifs->if_type) {
1797                 case IFT_ETHER:
1798                 case IFT_GIF:
1799                 case IFT_L2VLAN:
1800                         break;
1801                 default:
1802                         return (EINVAL);
1803         }
1804
1805         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1806         if (bif == NULL)
1807                 return (ENOMEM);
1808
1809         bif->bif_ifp = ifs;
1810         bif->bif_flags = IFBIF_SPAN;
1811
1812         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1813
1814         return (0);
1815 }
1816
1817 static int
1818 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1819 {
1820         struct ifbreq *req = arg;
1821         struct bridge_iflist *bif;
1822         struct ifnet *ifs;
1823
1824         ifs = ifunit(req->ifbr_ifsname);
1825         if (ifs == NULL)
1826                 return (ENOENT);
1827
1828         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1829                 if (ifs == bif->bif_ifp)
1830                         break;
1831
1832         if (bif == NULL)
1833                 return (ENOENT);
1834
1835         bridge_delete_span(sc, bif);
1836
1837         return (0);
1838 }
1839
1840 static int
1841 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1842 {
1843         struct ifbropreq *req = arg;
1844         struct bstp_state *bs = &sc->sc_stp;
1845         struct bstp_port *root_port;
1846
1847         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1848         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1849         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1850
1851         root_port = bs->bs_root_port;
1852         if (root_port == NULL)
1853                 req->ifbop_root_port = 0;
1854         else
1855                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1856
1857         req->ifbop_holdcount = bs->bs_txholdcount;
1858         req->ifbop_priority = bs->bs_bridge_priority;
1859         req->ifbop_protocol = bs->bs_protover;
1860         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1861         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1862         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1863         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1864         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1865         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1866
1867         return (0);
1868 }
1869
1870 static int
1871 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1872 {
1873         struct ifbrparam *param = arg;
1874
1875         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1876         return (0);
1877 }
1878
1879 static int
1880 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1881 {
1882         struct ifbpstpconf *bifstp = arg;
1883         struct bridge_iflist *bif;
1884         struct bstp_port *bp;
1885         struct ifbpstpreq bpreq;
1886         char *buf, *outbuf;
1887         int count, buflen, len, error = 0;
1888
1889         count = 0;
1890         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1891                 if ((bif->bif_flags & IFBIF_STP) != 0)
1892                         count++;
1893         }
1894
1895         buflen = sizeof(bpreq) * count;
1896         if (bifstp->ifbpstp_len == 0) {
1897                 bifstp->ifbpstp_len = buflen;
1898                 return (0);
1899         }
1900
1901         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1902         if (outbuf == NULL)
1903                 return (ENOMEM);
1904
1905         count = 0;
1906         buf = outbuf;
1907         len = min(bifstp->ifbpstp_len, buflen);
1908         bzero(&bpreq, sizeof(bpreq));
1909         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1910                 if (len < sizeof(bpreq))
1911                         break;
1912
1913                 if ((bif->bif_flags & IFBIF_STP) == 0)
1914                         continue;
1915
1916                 bp = &bif->bif_stp;
1917                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1918                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1919                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1920                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1921                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1922                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1923
1924                 memcpy(buf, &bpreq, sizeof(bpreq));
1925                 count++;
1926                 buf += sizeof(bpreq);
1927                 len -= sizeof(bpreq);
1928         }
1929
1930         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1931         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1932         free(outbuf, M_TEMP);
1933         return (error);
1934 }
1935
1936 static int
1937 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1938 {
1939         struct ifbrparam *param = arg;
1940
1941         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1942 }
1943
1944 static int
1945 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1946 {
1947         struct ifbrparam *param = arg;
1948
1949         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1950 }
1951
1952 /*
1953  * bridge_ifdetach:
1954  *
1955  *      Detach an interface from a bridge.  Called when a member
1956  *      interface is detaching.
1957  */
1958 static void
1959 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1960 {
1961         struct bridge_softc *sc = ifp->if_bridge;
1962         struct bridge_iflist *bif;
1963
1964         if (ifp->if_flags & IFF_RENAMING)
1965                 return;
1966         if (V_bridge_cloner == NULL) {
1967                 /*
1968                  * This detach handler can be called after
1969                  * vnet_bridge_uninit().  Just return in that case.
1970                  */
1971                 return;
1972         }
1973         /* Check if the interface is a bridge member */
1974         if (sc != NULL) {
1975                 BRIDGE_LOCK(sc);
1976
1977                 bif = bridge_lookup_member_if(sc, ifp);
1978                 if (bif != NULL)
1979                         bridge_delete_member(sc, bif, 1);
1980
1981                 BRIDGE_UNLOCK(sc);
1982                 return;
1983         }
1984
1985         /* Check if the interface is a span port */
1986         BRIDGE_LIST_LOCK();
1987         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1988                 BRIDGE_LOCK(sc);
1989                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1990                         if (ifp == bif->bif_ifp) {
1991                                 bridge_delete_span(sc, bif);
1992                                 break;
1993                         }
1994
1995                 BRIDGE_UNLOCK(sc);
1996         }
1997         BRIDGE_LIST_UNLOCK();
1998 }
1999
2000 /*
2001  * bridge_init:
2002  *
2003  *      Initialize a bridge interface.
2004  */
2005 static void
2006 bridge_init(void *xsc)
2007 {
2008         struct bridge_softc *sc = (struct bridge_softc *)xsc;
2009         struct ifnet *ifp = sc->sc_ifp;
2010
2011         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2012                 return;
2013
2014         BRIDGE_LOCK(sc);
2015         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2016             bridge_timer, sc);
2017
2018         ifp->if_drv_flags |= IFF_DRV_RUNNING;
2019         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
2020
2021         BRIDGE_UNLOCK(sc);
2022 }
2023
2024 /*
2025  * bridge_stop:
2026  *
2027  *      Stop the bridge interface.
2028  */
2029 static void
2030 bridge_stop(struct ifnet *ifp, int disable)
2031 {
2032         struct bridge_softc *sc = ifp->if_softc;
2033
2034         BRIDGE_LOCK_ASSERT(sc);
2035
2036         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2037                 return;
2038
2039         BRIDGE_RT_LOCK(sc);
2040         callout_stop(&sc->sc_brcallout);
2041
2042         bstp_stop(&sc->sc_stp);
2043
2044         bridge_rtflush(sc, IFBF_FLUSHDYN);
2045         BRIDGE_RT_UNLOCK(sc);
2046
2047         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2048 }
2049
2050 /*
2051  * bridge_enqueue:
2052  *
2053  *      Enqueue a packet on a bridge member interface.
2054  *
2055  */
2056 static int
2057 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2058 {
2059         int len, err = 0;
2060         short mflags;
2061         struct mbuf *m0;
2062
2063         /* We may be sending a fragment so traverse the mbuf */
2064         for (; m; m = m0) {
2065                 m0 = m->m_nextpkt;
2066                 m->m_nextpkt = NULL;
2067                 len = m->m_pkthdr.len;
2068                 mflags = m->m_flags;
2069
2070                 /*
2071                  * If underlying interface can not do VLAN tag insertion itself
2072                  * then attach a packet tag that holds it.
2073                  */
2074                 if ((m->m_flags & M_VLANTAG) &&
2075                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2076                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2077                         if (m == NULL) {
2078                                 if_printf(dst_ifp,
2079                                     "unable to prepend VLAN header\n");
2080                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2081                                 continue;
2082                         }
2083                         m->m_flags &= ~M_VLANTAG;
2084                 }
2085
2086                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2087                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2088                         int n;
2089
2090                         for (m = m0, n = 1; m != NULL; m = m0, n++) {
2091                                 m0 = m->m_nextpkt;
2092                                 m_freem(m);
2093                         }
2094                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2095                         break;
2096                 }
2097
2098                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2099                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2100                 if (mflags & M_MCAST)
2101                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2102         }
2103
2104         return (err);
2105 }
2106
2107 /*
2108  * bridge_dummynet:
2109  *
2110  *      Receive a queued packet from dummynet and pass it on to the output
2111  *      interface.
2112  *
2113  *      The mbuf has the Ethernet header already attached.
2114  */
2115 static void
2116 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2117 {
2118         struct bridge_softc *sc;
2119
2120         sc = ifp->if_bridge;
2121
2122         /*
2123          * The packet didnt originate from a member interface. This should only
2124          * ever happen if a member interface is removed while packets are
2125          * queued for it.
2126          */
2127         if (sc == NULL) {
2128                 m_freem(m);
2129                 return;
2130         }
2131
2132         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2133 #ifdef INET6
2134             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2135 #endif
2136             ) {
2137                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2138                         return;
2139                 if (m == NULL)
2140                         return;
2141         }
2142
2143         bridge_enqueue(sc, ifp, m);
2144 }
2145
2146 /*
2147  * bridge_output:
2148  *
2149  *      Send output from a bridge member interface.  This
2150  *      performs the bridging function for locally originated
2151  *      packets.
2152  *
2153  *      The mbuf has the Ethernet header already attached.  We must
2154  *      enqueue or free the mbuf before returning.
2155  */
2156 static int
2157 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2158     struct rtentry *rt)
2159 {
2160         struct ether_header *eh;
2161         struct ifnet *bifp, *dst_if;
2162         struct bridge_softc *sc;
2163         uint16_t vlan;
2164
2165         NET_EPOCH_ASSERT();
2166
2167         if (m->m_len < ETHER_HDR_LEN) {
2168                 m = m_pullup(m, ETHER_HDR_LEN);
2169                 if (m == NULL)
2170                         return (0);
2171         }
2172
2173         eh = mtod(m, struct ether_header *);
2174         sc = ifp->if_bridge;
2175         vlan = VLANTAGOF(m);
2176
2177         bifp = sc->sc_ifp;
2178
2179         /*
2180          * If bridge is down, but the original output interface is up,
2181          * go ahead and send out that interface.  Otherwise, the packet
2182          * is dropped below.
2183          */
2184         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2185                 dst_if = ifp;
2186                 goto sendunicast;
2187         }
2188
2189         /*
2190          * If the packet is a multicast, or we don't know a better way to
2191          * get there, send to all interfaces.
2192          */
2193         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2194                 dst_if = NULL;
2195         else
2196                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2197         /* Tap any traffic not passing back out the originating interface */
2198         if (dst_if != ifp)
2199                 ETHER_BPF_MTAP(bifp, m);
2200         if (dst_if == NULL) {
2201                 struct bridge_iflist *bif;
2202                 struct mbuf *mc;
2203                 int used = 0;
2204
2205                 bridge_span(sc, m);
2206
2207                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2208                         dst_if = bif->bif_ifp;
2209
2210                         if (dst_if->if_type == IFT_GIF)
2211                                 continue;
2212                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2213                                 continue;
2214
2215                         /*
2216                          * If this is not the original output interface,
2217                          * and the interface is participating in spanning
2218                          * tree, make sure the port is in a state that
2219                          * allows forwarding.
2220                          */
2221                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2222                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2223                                 continue;
2224
2225                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2226                                 used = 1;
2227                                 mc = m;
2228                         } else {
2229                                 mc = m_dup(m, M_NOWAIT);
2230                                 if (mc == NULL) {
2231                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2232                                         continue;
2233                                 }
2234                         }
2235
2236                         bridge_enqueue(sc, dst_if, mc);
2237                 }
2238                 if (used == 0)
2239                         m_freem(m);
2240                 return (0);
2241         }
2242
2243 sendunicast:
2244         /*
2245          * XXX Spanning tree consideration here?
2246          */
2247
2248         bridge_span(sc, m);
2249         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2250                 m_freem(m);
2251                 return (0);
2252         }
2253
2254         bridge_enqueue(sc, dst_if, m);
2255         return (0);
2256 }
2257
2258 /*
2259  * bridge_transmit:
2260  *
2261  *      Do output on a bridge.
2262  *
2263  */
2264 static int
2265 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2266 {
2267         struct bridge_softc *sc;
2268         struct ether_header *eh;
2269         struct ifnet *dst_if;
2270         int error = 0;
2271
2272         sc = ifp->if_softc;
2273
2274         ETHER_BPF_MTAP(ifp, m);
2275
2276         eh = mtod(m, struct ether_header *);
2277
2278         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2279             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, DOT1Q_VID_NULL)) !=
2280             NULL) {
2281                 error = bridge_enqueue(sc, dst_if, m);
2282         } else
2283                 bridge_broadcast(sc, ifp, m, 0);
2284
2285         return (error);
2286 }
2287
2288 #ifdef ALTQ
2289 static void
2290 bridge_altq_start(if_t ifp)
2291 {
2292         struct ifaltq *ifq = &ifp->if_snd;
2293         struct mbuf *m;
2294
2295         IFQ_LOCK(ifq);
2296         IFQ_DEQUEUE_NOLOCK(ifq, m);
2297         while (m != NULL) {
2298                 bridge_transmit(ifp, m);
2299                 IFQ_DEQUEUE_NOLOCK(ifq, m);
2300         }
2301         IFQ_UNLOCK(ifq);
2302 }
2303
2304 static int
2305 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2306 {
2307         int err;
2308
2309         if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2310                 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2311                 if (err == 0)
2312                         bridge_altq_start(ifp);
2313         } else
2314                 err = bridge_transmit(ifp, m);
2315
2316         return (err);
2317 }
2318 #endif  /* ALTQ */
2319
2320 /*
2321  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2322  */
2323 static void
2324 bridge_qflush(struct ifnet *ifp __unused)
2325 {
2326 }
2327
2328 /*
2329  * bridge_forward:
2330  *
2331  *      The forwarding function of the bridge.
2332  *
2333  *      NOTE: Releases the lock on return.
2334  */
2335 static void
2336 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2337     struct mbuf *m)
2338 {
2339         struct bridge_iflist *dbif;
2340         struct ifnet *src_if, *dst_if, *ifp;
2341         struct ether_header *eh;
2342         uint16_t vlan;
2343         uint8_t *dst;
2344         int error;
2345
2346         NET_EPOCH_ASSERT();
2347
2348         src_if = m->m_pkthdr.rcvif;
2349         ifp = sc->sc_ifp;
2350
2351         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2352         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2353         vlan = VLANTAGOF(m);
2354
2355         if ((sbif->bif_flags & IFBIF_STP) &&
2356             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2357                 goto drop;
2358
2359         eh = mtod(m, struct ether_header *);
2360         dst = eh->ether_dhost;
2361
2362         /* If the interface is learning, record the address. */
2363         if (sbif->bif_flags & IFBIF_LEARNING) {
2364                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2365                     sbif, 0, IFBAF_DYNAMIC);
2366                 /*
2367                  * If the interface has addresses limits then deny any source
2368                  * that is not in the cache.
2369                  */
2370                 if (error && sbif->bif_addrmax)
2371                         goto drop;
2372         }
2373
2374         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2375             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2376                 goto drop;
2377
2378 #ifdef DEV_NETMAP
2379         /*
2380          * Hand the packet to netmap only if it wasn't injected by netmap
2381          * itself.
2382          */
2383         if ((m->m_flags & M_BRIDGE_INJECT) == 0 &&
2384             (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) {
2385                 ifp->if_input(ifp, m);
2386                 return;
2387         }
2388         m->m_flags &= ~M_BRIDGE_INJECT;
2389 #endif
2390
2391         /*
2392          * At this point, the port either doesn't participate
2393          * in spanning tree or it is in the forwarding state.
2394          */
2395
2396         /*
2397          * If the packet is unicast, destined for someone on
2398          * "this" side of the bridge, drop it.
2399          */
2400         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2401                 dst_if = bridge_rtlookup(sc, dst, vlan);
2402                 if (src_if == dst_if)
2403                         goto drop;
2404         } else {
2405                 /*
2406                  * Check if its a reserved multicast address, any address
2407                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2408                  * bridge.
2409                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2410                  */
2411                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2412                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2413                     dst[4] == 0x00 && dst[5] <= 0x0f)
2414                         goto drop;
2415
2416                 /* ...forward it to all interfaces. */
2417                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2418                 dst_if = NULL;
2419         }
2420
2421         /*
2422          * If we have a destination interface which is a member of our bridge,
2423          * OR this is a unicast packet, push it through the bpf(4) machinery.
2424          * For broadcast or multicast packets, don't bother because it will
2425          * be reinjected into ether_input. We do this before we pass the packets
2426          * through the pfil(9) framework, as it is possible that pfil(9) will
2427          * drop the packet, or possibly modify it, making it difficult to debug
2428          * firewall issues on the bridge.
2429          */
2430         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2431                 ETHER_BPF_MTAP(ifp, m);
2432
2433         /* run the packet filter */
2434         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2435 #ifdef INET6
2436             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2437 #endif
2438             ) {
2439                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2440                         return;
2441                 if (m == NULL)
2442                         return;
2443         }
2444
2445         if (dst_if == NULL) {
2446                 bridge_broadcast(sc, src_if, m, 1);
2447                 return;
2448         }
2449
2450         /*
2451          * At this point, we're dealing with a unicast frame
2452          * going to a different interface.
2453          */
2454         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2455                 goto drop;
2456
2457         dbif = bridge_lookup_member_if(sc, dst_if);
2458         if (dbif == NULL)
2459                 /* Not a member of the bridge (anymore?) */
2460                 goto drop;
2461
2462         /* Private segments can not talk to each other */
2463         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2464                 goto drop;
2465
2466         if ((dbif->bif_flags & IFBIF_STP) &&
2467             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2468                 goto drop;
2469
2470         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2471 #ifdef INET6
2472             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2473 #endif
2474             ) {
2475                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2476                         return;
2477                 if (m == NULL)
2478                         return;
2479         }
2480
2481         bridge_enqueue(sc, dst_if, m);
2482         return;
2483
2484 drop:
2485         m_freem(m);
2486 }
2487
2488 /*
2489  * bridge_input:
2490  *
2491  *      Receive input from a member interface.  Queue the packet for
2492  *      bridging if it is not for us.
2493  */
2494 static struct mbuf *
2495 bridge_input(struct ifnet *ifp, struct mbuf *m)
2496 {
2497         struct bridge_softc *sc;
2498         struct bridge_iflist *bif, *bif2;
2499         struct ifnet *bifp;
2500         struct ether_header *eh;
2501         struct mbuf *mc, *mc2;
2502         uint16_t vlan;
2503         int error;
2504
2505         NET_EPOCH_ASSERT();
2506
2507         eh = mtod(m, struct ether_header *);
2508         vlan = VLANTAGOF(m);
2509
2510         sc = ifp->if_bridge;
2511         if (sc == NULL) {
2512                 /*
2513                  * This packet originated from the bridge itself, so it must
2514                  * have been transmitted by netmap.  Derive the "source"
2515                  * interface from the source address and drop the packet if the
2516                  * source address isn't known.
2517                  */
2518                 KASSERT((m->m_flags & M_BRIDGE_INJECT) != 0,
2519                     ("%s: ifnet %p missing a bridge softc", __func__, ifp));
2520                 sc = if_getsoftc(ifp);
2521                 ifp = bridge_rtlookup(sc, eh->ether_shost, vlan);
2522                 if (ifp == NULL) {
2523                         if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1);
2524                         m_freem(m);
2525                         return (NULL);
2526                 }
2527                 m->m_pkthdr.rcvif = ifp;
2528         }
2529         bifp = sc->sc_ifp;
2530         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2531                 return (m);
2532
2533         /*
2534          * Implement support for bridge monitoring. If this flag has been
2535          * set on this interface, discard the packet once we push it through
2536          * the bpf(4) machinery, but before we do, increment the byte and
2537          * packet counters associated with this interface.
2538          */
2539         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2540                 m->m_pkthdr.rcvif  = bifp;
2541                 ETHER_BPF_MTAP(bifp, m);
2542                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2543                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2544                 m_freem(m);
2545                 return (NULL);
2546         }
2547         bif = bridge_lookup_member_if(sc, ifp);
2548         if (bif == NULL) {
2549                 return (m);
2550         }
2551
2552         bridge_span(sc, m);
2553
2554         if (m->m_flags & (M_BCAST|M_MCAST)) {
2555                 /* Tap off 802.1D packets; they do not get forwarded. */
2556                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2557                     ETHER_ADDR_LEN) == 0) {
2558                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2559                         return (NULL);
2560                 }
2561
2562                 if ((bif->bif_flags & IFBIF_STP) &&
2563                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2564                         return (m);
2565                 }
2566
2567                 /*
2568                  * Make a deep copy of the packet and enqueue the copy
2569                  * for bridge processing; return the original packet for
2570                  * local processing.
2571                  */
2572                 mc = m_dup(m, M_NOWAIT);
2573                 if (mc == NULL) {
2574                         return (m);
2575                 }
2576
2577                 /* Perform the bridge forwarding function with the copy. */
2578                 bridge_forward(sc, bif, mc);
2579
2580 #ifdef DEV_NETMAP
2581                 /*
2582                  * If netmap is enabled and has not already seen this packet,
2583                  * then it will be consumed by bridge_forward().
2584                  */
2585                 if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 &&
2586                     (m->m_flags & M_BRIDGE_INJECT) == 0) {
2587                         m_freem(m);
2588                         return (NULL);
2589                 }
2590 #endif
2591
2592                 /*
2593                  * Reinject the mbuf as arriving on the bridge so we have a
2594                  * chance at claiming multicast packets. We can not loop back
2595                  * here from ether_input as a bridge is never a member of a
2596                  * bridge.
2597                  */
2598                 KASSERT(bifp->if_bridge == NULL,
2599                     ("loop created in bridge_input"));
2600                 mc2 = m_dup(m, M_NOWAIT);
2601                 if (mc2 != NULL) {
2602                         /* Keep the layer3 header aligned */
2603                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2604                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2605                 }
2606                 if (mc2 != NULL) {
2607                         mc2->m_pkthdr.rcvif = bifp;
2608                         mc2->m_flags &= ~M_BRIDGE_INJECT;
2609                         sc->sc_if_input(bifp, mc2);
2610                 }
2611
2612                 /* Return the original packet for local processing. */
2613                 return (m);
2614         }
2615
2616         if ((bif->bif_flags & IFBIF_STP) &&
2617             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2618                 return (m);
2619         }
2620
2621 #if defined(INET) || defined(INET6)
2622 #define CARP_CHECK_WE_ARE_DST(iface) \
2623         ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost))
2624 #define CARP_CHECK_WE_ARE_SRC(iface) \
2625         ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost))
2626 #else
2627 #define CARP_CHECK_WE_ARE_DST(iface)    false
2628 #define CARP_CHECK_WE_ARE_SRC(iface)    false
2629 #endif
2630
2631 #ifdef INET6
2632 #define PFIL_HOOKED_INET6       PFIL_HOOKED_IN(V_inet6_pfil_head)
2633 #else
2634 #define PFIL_HOOKED_INET6       false
2635 #endif
2636
2637 #ifdef DEV_NETMAP
2638 #define GRAB_FOR_NETMAP(ifp, m) do {                                    \
2639         if ((if_getcapenable(ifp) & IFCAP_NETMAP) != 0 &&               \
2640             ((m)->m_flags & M_BRIDGE_INJECT) == 0) {                    \
2641                 (ifp)->if_input(ifp, m);                                \
2642                 return (NULL);                                          \
2643         }                                                               \
2644 } while (0)
2645 #else
2646 #define GRAB_FOR_NETMAP(ifp, m)
2647 #endif
2648
2649 #define GRAB_OUR_PACKETS(iface)                                         \
2650         if ((iface)->if_type == IFT_GIF)                                \
2651                 continue;                                               \
2652         /* It is destined for us. */                                    \
2653         if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \
2654             CARP_CHECK_WE_ARE_DST(iface)) {                             \
2655                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2656                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2657                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2658                         if (error && bif->bif_addrmax) {                \
2659                                 m_freem(m);                             \
2660                                 return (NULL);                          \
2661                         }                                               \
2662                 }                                                       \
2663                 m->m_pkthdr.rcvif = iface;                              \
2664                 if ((iface) == ifp) {                                   \
2665                         /* Skip bridge processing... src == dest */     \
2666                         return (m);                                     \
2667                 }                                                       \
2668                 /* It's passing over or to the bridge, locally. */      \
2669                 ETHER_BPF_MTAP(bifp, m);                                \
2670                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2671                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);\
2672                 /* Hand the packet over to netmap if necessary. */      \
2673                 GRAB_FOR_NETMAP(bifp, m);                               \
2674                 /* Filter on the physical interface. */                 \
2675                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) || \
2676                     PFIL_HOOKED_INET6)) {                               \
2677                         if (bridge_pfil(&m, NULL, ifp,                  \
2678                             PFIL_IN) != 0 || m == NULL) {               \
2679                                 return (NULL);                          \
2680                         }                                               \
2681                 }                                                       \
2682                 if ((iface) != bifp)                                    \
2683                         ETHER_BPF_MTAP(iface, m);                       \
2684                 return (m);                                             \
2685         }                                                               \
2686                                                                         \
2687         /* We just received a packet that we sent out. */               \
2688         if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \
2689             CARP_CHECK_WE_ARE_SRC(iface)) {                             \
2690                 m_freem(m);                                             \
2691                 return (NULL);                                          \
2692         }
2693
2694         /*
2695          * Unicast.  Make sure it's not for the bridge.
2696          */
2697         do { GRAB_OUR_PACKETS(bifp) } while (0);
2698
2699         /*
2700          * Give a chance for ifp at first priority. This will help when the
2701          * packet comes through the interface like VLAN's with the same MACs
2702          * on several interfaces from the same bridge. This also will save
2703          * some CPU cycles in case the destination interface and the input
2704          * interface (eq ifp) are the same.
2705          */
2706         do { GRAB_OUR_PACKETS(ifp) } while (0);
2707
2708         /* Now check the all bridge members. */
2709         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2710                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2711         }
2712
2713 #undef CARP_CHECK_WE_ARE_DST
2714 #undef CARP_CHECK_WE_ARE_SRC
2715 #undef PFIL_HOOKED_INET6
2716 #undef GRAB_FOR_NETMAP
2717 #undef GRAB_OUR_PACKETS
2718
2719         /* Perform the bridge forwarding function. */
2720         bridge_forward(sc, bif, m);
2721
2722         return (NULL);
2723 }
2724
2725 /*
2726  * Inject a packet back into the host ethernet stack.  This will generally only
2727  * be used by netmap when an application writes to the host TX ring.  The
2728  * M_BRIDGE_INJECT flag ensures that the packet is re-routed to the bridge
2729  * interface after ethernet processing.
2730  */
2731 static void
2732 bridge_inject(struct ifnet *ifp, struct mbuf *m)
2733 {
2734         struct bridge_softc *sc;
2735
2736         KASSERT((if_getcapenable(ifp) & IFCAP_NETMAP) != 0,
2737             ("%s: iface %s is not running in netmap mode",
2738             __func__, if_name(ifp)));
2739         KASSERT((m->m_flags & M_BRIDGE_INJECT) == 0,
2740             ("%s: mbuf %p has M_BRIDGE_INJECT set", __func__, m));
2741
2742         m->m_flags |= M_BRIDGE_INJECT;
2743         sc = if_getsoftc(ifp);
2744         sc->sc_if_input(ifp, m);
2745 }
2746
2747 /*
2748  * bridge_broadcast:
2749  *
2750  *      Send a frame to all interfaces that are members of
2751  *      the bridge, except for the one on which the packet
2752  *      arrived.
2753  *
2754  *      NOTE: Releases the lock on return.
2755  */
2756 static void
2757 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2758     struct mbuf *m, int runfilt)
2759 {
2760         struct bridge_iflist *dbif, *sbif;
2761         struct mbuf *mc;
2762         struct ifnet *dst_if;
2763         int used = 0, i;
2764
2765         NET_EPOCH_ASSERT();
2766
2767         sbif = bridge_lookup_member_if(sc, src_if);
2768
2769         /* Filter on the bridge interface before broadcasting */
2770         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2771 #ifdef INET6
2772             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2773 #endif
2774             )) {
2775                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2776                         return;
2777                 if (m == NULL)
2778                         return;
2779         }
2780
2781         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2782                 dst_if = dbif->bif_ifp;
2783                 if (dst_if == src_if)
2784                         continue;
2785
2786                 /* Private segments can not talk to each other */
2787                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2788                         continue;
2789
2790                 if ((dbif->bif_flags & IFBIF_STP) &&
2791                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2792                         continue;
2793
2794                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2795                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2796                         continue;
2797
2798                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2799                         continue;
2800
2801                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2802                         mc = m;
2803                         used = 1;
2804                 } else {
2805                         mc = m_dup(m, M_NOWAIT);
2806                         if (mc == NULL) {
2807                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2808                                 continue;
2809                         }
2810                 }
2811
2812                 /*
2813                  * Filter on the output interface. Pass a NULL bridge interface
2814                  * pointer so we do not redundantly filter on the bridge for
2815                  * each interface we broadcast on.
2816                  */
2817                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2818 #ifdef INET6
2819                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2820 #endif
2821                     )) {
2822                         if (used == 0) {
2823                                 /* Keep the layer3 header aligned */
2824                                 i = min(mc->m_pkthdr.len, max_protohdr);
2825                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2826                                 if (mc == NULL) {
2827                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2828                                         continue;
2829                                 }
2830                         }
2831                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2832                                 continue;
2833                         if (mc == NULL)
2834                                 continue;
2835                 }
2836
2837                 bridge_enqueue(sc, dst_if, mc);
2838         }
2839         if (used == 0)
2840                 m_freem(m);
2841 }
2842
2843 /*
2844  * bridge_span:
2845  *
2846  *      Duplicate a packet out one or more interfaces that are in span mode,
2847  *      the original mbuf is unmodified.
2848  */
2849 static void
2850 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2851 {
2852         struct bridge_iflist *bif;
2853         struct ifnet *dst_if;
2854         struct mbuf *mc;
2855
2856         NET_EPOCH_ASSERT();
2857
2858         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2859                 return;
2860
2861         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2862                 dst_if = bif->bif_ifp;
2863
2864                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2865                         continue;
2866
2867                 mc = m_dup(m, M_NOWAIT);
2868                 if (mc == NULL) {
2869                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2870                         continue;
2871                 }
2872
2873                 bridge_enqueue(sc, dst_if, mc);
2874         }
2875 }
2876
2877 /*
2878  * bridge_rtupdate:
2879  *
2880  *      Add a bridge routing entry.
2881  */
2882 static int
2883 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2884     struct bridge_iflist *bif, int setflags, uint8_t flags)
2885 {
2886         struct bridge_rtnode *brt;
2887         struct bridge_iflist *obif;
2888         int error;
2889
2890         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2891
2892         /* Check the source address is valid and not multicast. */
2893         if (ETHER_IS_MULTICAST(dst) ||
2894             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2895              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2896                 return (EINVAL);
2897
2898         /*
2899          * A route for this destination might already exist.  If so,
2900          * update it, otherwise create a new one.
2901          */
2902         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2903                 BRIDGE_RT_LOCK(sc);
2904
2905                 /* Check again, now that we have the lock. There could have
2906                  * been a race and we only want to insert this once. */
2907                 if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) {
2908                         BRIDGE_RT_UNLOCK(sc);
2909                         return (0);
2910                 }
2911
2912                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2913                         sc->sc_brtexceeded++;
2914                         BRIDGE_RT_UNLOCK(sc);
2915                         return (ENOSPC);
2916                 }
2917                 /* Check per interface address limits (if enabled) */
2918                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2919                         bif->bif_addrexceeded++;
2920                         BRIDGE_RT_UNLOCK(sc);
2921                         return (ENOSPC);
2922                 }
2923
2924                 /*
2925                  * Allocate a new bridge forwarding node, and
2926                  * initialize the expiration time and Ethernet
2927                  * address.
2928                  */
2929                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2930                 if (brt == NULL) {
2931                         BRIDGE_RT_UNLOCK(sc);
2932                         return (ENOMEM);
2933                 }
2934                 brt->brt_vnet = curvnet;
2935
2936                 if (bif->bif_flags & IFBIF_STICKY)
2937                         brt->brt_flags = IFBAF_STICKY;
2938                 else
2939                         brt->brt_flags = IFBAF_DYNAMIC;
2940
2941                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2942                 brt->brt_vlan = vlan;
2943
2944                 brt->brt_dst = bif;
2945                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2946                         uma_zfree(V_bridge_rtnode_zone, brt);
2947                         BRIDGE_RT_UNLOCK(sc);
2948                         return (error);
2949                 }
2950                 bif->bif_addrcnt++;
2951
2952                 BRIDGE_RT_UNLOCK(sc);
2953         }
2954
2955         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2956             (obif = brt->brt_dst) != bif) {
2957                 MPASS(obif != NULL);
2958
2959                 BRIDGE_RT_LOCK(sc);
2960                 brt->brt_dst->bif_addrcnt--;
2961                 brt->brt_dst = bif;
2962                 brt->brt_dst->bif_addrcnt++;
2963                 BRIDGE_RT_UNLOCK(sc);
2964
2965                 if (V_log_mac_flap &&
2966                     ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
2967                         log(LOG_NOTICE,
2968                             "%s: mac address %6D vlan %d moved from %s to %s\n",
2969                             sc->sc_ifp->if_xname,
2970                             &brt->brt_addr[0], ":",
2971                             brt->brt_vlan,
2972                             obif->bif_ifp->if_xname,
2973                             bif->bif_ifp->if_xname);
2974                 }
2975         }
2976
2977         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2978                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2979         if (setflags)
2980                 brt->brt_flags = flags;
2981
2982         return (0);
2983 }
2984
2985 /*
2986  * bridge_rtlookup:
2987  *
2988  *      Lookup the destination interface for an address.
2989  */
2990 static struct ifnet *
2991 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2992 {
2993         struct bridge_rtnode *brt;
2994
2995         NET_EPOCH_ASSERT();
2996
2997         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2998                 return (NULL);
2999
3000         return (brt->brt_ifp);
3001 }
3002
3003 /*
3004  * bridge_rttrim:
3005  *
3006  *      Trim the routine table so that we have a number
3007  *      of routing entries less than or equal to the
3008  *      maximum number.
3009  */
3010 static void
3011 bridge_rttrim(struct bridge_softc *sc)
3012 {
3013         struct bridge_rtnode *brt, *nbrt;
3014
3015         NET_EPOCH_ASSERT();
3016         BRIDGE_RT_LOCK_ASSERT(sc);
3017
3018         /* Make sure we actually need to do this. */
3019         if (sc->sc_brtcnt <= sc->sc_brtmax)
3020                 return;
3021
3022         /* Force an aging cycle; this might trim enough addresses. */
3023         bridge_rtage(sc);
3024         if (sc->sc_brtcnt <= sc->sc_brtmax)
3025                 return;
3026
3027         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3028                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3029                         bridge_rtnode_destroy(sc, brt);
3030                         if (sc->sc_brtcnt <= sc->sc_brtmax)
3031                                 return;
3032                 }
3033         }
3034 }
3035
3036 /*
3037  * bridge_timer:
3038  *
3039  *      Aging timer for the bridge.
3040  */
3041 static void
3042 bridge_timer(void *arg)
3043 {
3044         struct bridge_softc *sc = arg;
3045
3046         BRIDGE_RT_LOCK_ASSERT(sc);
3047
3048         /* Destruction of rtnodes requires a proper vnet context */
3049         CURVNET_SET(sc->sc_ifp->if_vnet);
3050         bridge_rtage(sc);
3051
3052         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
3053                 callout_reset(&sc->sc_brcallout,
3054                     bridge_rtable_prune_period * hz, bridge_timer, sc);
3055         CURVNET_RESTORE();
3056 }
3057
3058 /*
3059  * bridge_rtage:
3060  *
3061  *      Perform an aging cycle.
3062  */
3063 static void
3064 bridge_rtage(struct bridge_softc *sc)
3065 {
3066         struct bridge_rtnode *brt, *nbrt;
3067
3068         BRIDGE_RT_LOCK_ASSERT(sc);
3069
3070         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3071                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3072                         if (time_uptime >= brt->brt_expire)
3073                                 bridge_rtnode_destroy(sc, brt);
3074                 }
3075         }
3076 }
3077
3078 /*
3079  * bridge_rtflush:
3080  *
3081  *      Remove all dynamic addresses from the bridge.
3082  */
3083 static void
3084 bridge_rtflush(struct bridge_softc *sc, int full)
3085 {
3086         struct bridge_rtnode *brt, *nbrt;
3087
3088         BRIDGE_RT_LOCK_ASSERT(sc);
3089
3090         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3091                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3092                         bridge_rtnode_destroy(sc, brt);
3093         }
3094 }
3095
3096 /*
3097  * bridge_rtdaddr:
3098  *
3099  *      Remove an address from the table.
3100  */
3101 static int
3102 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3103 {
3104         struct bridge_rtnode *brt;
3105         int found = 0;
3106
3107         BRIDGE_RT_LOCK(sc);
3108
3109         /*
3110          * If vlan is DOT1Q_VID_RSVD_IMPL then we want to delete for all vlans
3111          * so the lookup may return more than one.
3112          */
3113         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3114                 bridge_rtnode_destroy(sc, brt);
3115                 found = 1;
3116         }
3117
3118         BRIDGE_RT_UNLOCK(sc);
3119
3120         return (found ? 0 : ENOENT);
3121 }
3122
3123 /*
3124  * bridge_rtdelete:
3125  *
3126  *      Delete routes to a speicifc member interface.
3127  */
3128 static void
3129 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3130 {
3131         struct bridge_rtnode *brt, *nbrt;
3132
3133         BRIDGE_RT_LOCK_ASSERT(sc);
3134
3135         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3136                 if (brt->brt_ifp == ifp && (full ||
3137                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3138                         bridge_rtnode_destroy(sc, brt);
3139         }
3140 }
3141
3142 /*
3143  * bridge_rtable_init:
3144  *
3145  *      Initialize the route table for this bridge.
3146  */
3147 static void
3148 bridge_rtable_init(struct bridge_softc *sc)
3149 {
3150         int i;
3151
3152         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3153             M_DEVBUF, M_WAITOK);
3154
3155         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3156                 CK_LIST_INIT(&sc->sc_rthash[i]);
3157
3158         sc->sc_rthash_key = arc4random();
3159         CK_LIST_INIT(&sc->sc_rtlist);
3160 }
3161
3162 /*
3163  * bridge_rtable_fini:
3164  *
3165  *      Deconstruct the route table for this bridge.
3166  */
3167 static void
3168 bridge_rtable_fini(struct bridge_softc *sc)
3169 {
3170
3171         KASSERT(sc->sc_brtcnt == 0,
3172             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3173         free(sc->sc_rthash, M_DEVBUF);
3174 }
3175
3176 /*
3177  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3178  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3179  */
3180 #define mix(a, b, c)                                                    \
3181 do {                                                                    \
3182         a -= b; a -= c; a ^= (c >> 13);                                 \
3183         b -= c; b -= a; b ^= (a << 8);                                  \
3184         c -= a; c -= b; c ^= (b >> 13);                                 \
3185         a -= b; a -= c; a ^= (c >> 12);                                 \
3186         b -= c; b -= a; b ^= (a << 16);                                 \
3187         c -= a; c -= b; c ^= (b >> 5);                                  \
3188         a -= b; a -= c; a ^= (c >> 3);                                  \
3189         b -= c; b -= a; b ^= (a << 10);                                 \
3190         c -= a; c -= b; c ^= (b >> 15);                                 \
3191 } while (/*CONSTCOND*/0)
3192
3193 static __inline uint32_t
3194 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3195 {
3196         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3197
3198         b += addr[5] << 8;
3199         b += addr[4];
3200         a += addr[3] << 24;
3201         a += addr[2] << 16;
3202         a += addr[1] << 8;
3203         a += addr[0];
3204
3205         mix(a, b, c);
3206
3207         return (c & BRIDGE_RTHASH_MASK);
3208 }
3209
3210 #undef mix
3211
3212 static int
3213 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3214 {
3215         int i, d;
3216
3217         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3218                 d = ((int)a[i]) - ((int)b[i]);
3219         }
3220
3221         return (d);
3222 }
3223
3224 /*
3225  * bridge_rtnode_lookup:
3226  *
3227  *      Look up a bridge route node for the specified destination. Compare the
3228  *      vlan id or if zero then just return the first match.
3229  */
3230 static struct bridge_rtnode *
3231 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3232 {
3233         struct bridge_rtnode *brt;
3234         uint32_t hash;
3235         int dir;
3236
3237         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3238
3239         hash = bridge_rthash(sc, addr);
3240         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3241                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3242                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == DOT1Q_VID_RSVD_IMPL))
3243                         return (brt);
3244                 if (dir > 0)
3245                         return (NULL);
3246         }
3247
3248         return (NULL);
3249 }
3250
3251 /*
3252  * bridge_rtnode_insert:
3253  *
3254  *      Insert the specified bridge node into the route table.  We
3255  *      assume the entry is not already in the table.
3256  */
3257 static int
3258 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3259 {
3260         struct bridge_rtnode *lbrt;
3261         uint32_t hash;
3262         int dir;
3263
3264         BRIDGE_RT_LOCK_ASSERT(sc);
3265
3266         hash = bridge_rthash(sc, brt->brt_addr);
3267
3268         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3269         if (lbrt == NULL) {
3270                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3271                 goto out;
3272         }
3273
3274         do {
3275                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3276                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3277                         return (EEXIST);
3278                 if (dir > 0) {
3279                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3280                         goto out;
3281                 }
3282                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3283                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3284                         goto out;
3285                 }
3286                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3287         } while (lbrt != NULL);
3288
3289 #ifdef DIAGNOSTIC
3290         panic("bridge_rtnode_insert: impossible");
3291 #endif
3292
3293 out:
3294         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3295         sc->sc_brtcnt++;
3296
3297         return (0);
3298 }
3299
3300 static void
3301 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3302 {
3303         struct bridge_rtnode *brt;
3304
3305         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3306
3307         CURVNET_SET(brt->brt_vnet);
3308         uma_zfree(V_bridge_rtnode_zone, brt);
3309         CURVNET_RESTORE();
3310 }
3311
3312 /*
3313  * bridge_rtnode_destroy:
3314  *
3315  *      Destroy a bridge rtnode.
3316  */
3317 static void
3318 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3319 {
3320         BRIDGE_RT_LOCK_ASSERT(sc);
3321
3322         CK_LIST_REMOVE(brt, brt_hash);
3323
3324         CK_LIST_REMOVE(brt, brt_list);
3325         sc->sc_brtcnt--;
3326         brt->brt_dst->bif_addrcnt--;
3327
3328         NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3329 }
3330
3331 /*
3332  * bridge_rtable_expire:
3333  *
3334  *      Set the expiry time for all routes on an interface.
3335  */
3336 static void
3337 bridge_rtable_expire(struct ifnet *ifp, int age)
3338 {
3339         struct bridge_softc *sc = ifp->if_bridge;
3340         struct bridge_rtnode *brt;
3341
3342         CURVNET_SET(ifp->if_vnet);
3343         BRIDGE_RT_LOCK(sc);
3344
3345         /*
3346          * If the age is zero then flush, otherwise set all the expiry times to
3347          * age for the interface
3348          */
3349         if (age == 0)
3350                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3351         else {
3352                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3353                         /* Cap the expiry time to 'age' */
3354                         if (brt->brt_ifp == ifp &&
3355                             brt->brt_expire > time_uptime + age &&
3356                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3357                                 brt->brt_expire = time_uptime + age;
3358                 }
3359         }
3360         BRIDGE_RT_UNLOCK(sc);
3361         CURVNET_RESTORE();
3362 }
3363
3364 /*
3365  * bridge_state_change:
3366  *
3367  *      Callback from the bridgestp code when a port changes states.
3368  */
3369 static void
3370 bridge_state_change(struct ifnet *ifp, int state)
3371 {
3372         struct bridge_softc *sc = ifp->if_bridge;
3373         static const char *stpstates[] = {
3374                 "disabled",
3375                 "listening",
3376                 "learning",
3377                 "forwarding",
3378                 "blocking",
3379                 "discarding"
3380         };
3381
3382         CURVNET_SET(ifp->if_vnet);
3383         if (V_log_stp)
3384                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3385                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3386         CURVNET_RESTORE();
3387 }
3388
3389 /*
3390  * Send bridge packets through pfil if they are one of the types pfil can deal
3391  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3392  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3393  * that interface.
3394  */
3395 static int
3396 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3397 {
3398         int snap, error, i, hlen;
3399         struct ether_header *eh1, eh2;
3400         struct ip *ip;
3401         struct llc llc1;
3402         u_int16_t ether_type;
3403         pfil_return_t rv;
3404
3405         snap = 0;
3406         error = -1;     /* Default error if not error == 0 */
3407
3408 #if 0
3409         /* we may return with the IP fields swapped, ensure its not shared */
3410         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3411 #endif
3412
3413         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3414                 return (0); /* filtering is disabled */
3415
3416         i = min((*mp)->m_pkthdr.len, max_protohdr);
3417         if ((*mp)->m_len < i) {
3418             *mp = m_pullup(*mp, i);
3419             if (*mp == NULL) {
3420                 printf("%s: m_pullup failed\n", __func__);
3421                 return (-1);
3422             }
3423         }
3424
3425         eh1 = mtod(*mp, struct ether_header *);
3426         ether_type = ntohs(eh1->ether_type);
3427
3428         /*
3429          * Check for SNAP/LLC.
3430          */
3431         if (ether_type < ETHERMTU) {
3432                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3433
3434                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3435                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3436                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3437                     llc2->llc_control == LLC_UI) {
3438                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3439                         snap = 1;
3440                 }
3441         }
3442
3443         /*
3444          * If we're trying to filter bridge traffic, don't look at anything
3445          * other than IP and ARP traffic.  If the filter doesn't understand
3446          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3447          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3448          * but of course we don't have an AppleTalk filter to begin with.
3449          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3450          * ARP traffic.)
3451          */
3452         switch (ether_type) {
3453                 case ETHERTYPE_ARP:
3454                 case ETHERTYPE_REVARP:
3455                         if (V_pfil_ipfw_arp == 0)
3456                                 return (0); /* Automatically pass */
3457                         break;
3458
3459                 case ETHERTYPE_IP:
3460 #ifdef INET6
3461                 case ETHERTYPE_IPV6:
3462 #endif /* INET6 */
3463                         break;
3464                 default:
3465                         /*
3466                          * Check to see if the user wants to pass non-ip
3467                          * packets, these will not be checked by pfil(9) and
3468                          * passed unconditionally so the default is to drop.
3469                          */
3470                         if (V_pfil_onlyip)
3471                                 goto bad;
3472         }
3473
3474         /* Run the packet through pfil before stripping link headers */
3475         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3476             dir == PFIL_OUT && ifp != NULL) {
3477                 switch (pfil_mbuf_out(V_link_pfil_head, mp, ifp, NULL)) {
3478                 case PFIL_DROPPED:
3479                         return (EACCES);
3480                 case PFIL_CONSUMED:
3481                         return (0);
3482                 }
3483         }
3484
3485         /* Strip off the Ethernet header and keep a copy. */
3486         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3487         m_adj(*mp, ETHER_HDR_LEN);
3488
3489         /* Strip off snap header, if present */
3490         if (snap) {
3491                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3492                 m_adj(*mp, sizeof(struct llc));
3493         }
3494
3495         /*
3496          * Check the IP header for alignment and errors
3497          */
3498         if (dir == PFIL_IN) {
3499                 switch (ether_type) {
3500                         case ETHERTYPE_IP:
3501                                 error = bridge_ip_checkbasic(mp);
3502                                 break;
3503 #ifdef INET6
3504                         case ETHERTYPE_IPV6:
3505                                 error = bridge_ip6_checkbasic(mp);
3506                                 break;
3507 #endif /* INET6 */
3508                         default:
3509                                 error = 0;
3510                 }
3511                 if (error)
3512                         goto bad;
3513         }
3514
3515         error = 0;
3516
3517         /*
3518          * Run the packet through pfil
3519          */
3520         rv = PFIL_PASS;
3521         switch (ether_type) {
3522         case ETHERTYPE_IP:
3523                 /*
3524                  * Run pfil on the member interface and the bridge, both can
3525                  * be skipped by clearing pfil_member or pfil_bridge.
3526                  *
3527                  * Keep the order:
3528                  *   in_if -> bridge_if -> out_if
3529                  */
3530                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3531                     pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) !=
3532                     PFIL_PASS)
3533                         break;
3534
3535                 if (V_pfil_member && ifp != NULL) {
3536                         rv = (dir == PFIL_OUT) ?
3537                             pfil_mbuf_out(V_inet_pfil_head, mp, ifp, NULL) :
3538                             pfil_mbuf_in(V_inet_pfil_head, mp, ifp, NULL);
3539                         if (rv != PFIL_PASS)
3540                                 break;
3541                 }
3542
3543                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3544                     pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) !=
3545                     PFIL_PASS)
3546                         break;
3547
3548                 /* check if we need to fragment the packet */
3549                 /* bridge_fragment generates a mbuf chain of packets */
3550                 /* that already include eth headers */
3551                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3552                         i = (*mp)->m_pkthdr.len;
3553                         if (i > ifp->if_mtu) {
3554                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3555                                             &llc1);
3556                                 return (error);
3557                         }
3558                 }
3559
3560                 /* Recalculate the ip checksum. */
3561                 ip = mtod(*mp, struct ip *);
3562                 hlen = ip->ip_hl << 2;
3563                 if (hlen < sizeof(struct ip))
3564                         goto bad;
3565                 if (hlen > (*mp)->m_len) {
3566                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3567                                 goto bad;
3568                         ip = mtod(*mp, struct ip *);
3569                         if (ip == NULL)
3570                                 goto bad;
3571                 }
3572                 ip->ip_sum = 0;
3573                 if (hlen == sizeof(struct ip))
3574                         ip->ip_sum = in_cksum_hdr(ip);
3575                 else
3576                         ip->ip_sum = in_cksum(*mp, hlen);
3577
3578                 break;
3579 #ifdef INET6
3580         case ETHERTYPE_IPV6:
3581                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3582                     pfil_mbuf_out(V_inet6_pfil_head, mp, bifp, NULL)) !=
3583                     PFIL_PASS)
3584                         break;
3585
3586                 if (V_pfil_member && ifp != NULL) {
3587                         rv = (dir == PFIL_OUT) ?
3588                             pfil_mbuf_out(V_inet6_pfil_head, mp, ifp, NULL) :
3589                             pfil_mbuf_in(V_inet6_pfil_head, mp, ifp, NULL);
3590                         if (rv != PFIL_PASS)
3591                                 break;
3592                 }
3593
3594                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3595                     pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) !=
3596                     PFIL_PASS)
3597                         break;
3598                 break;
3599 #endif
3600         }
3601
3602         switch (rv) {
3603         case PFIL_CONSUMED:
3604                 return (0);
3605         case PFIL_DROPPED:
3606                 return (EACCES);
3607         default:
3608                 break;
3609         }
3610
3611         error = -1;
3612
3613         /*
3614          * Finally, put everything back the way it was and return
3615          */
3616         if (snap) {
3617                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3618                 if (*mp == NULL)
3619                         return (error);
3620                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3621         }
3622
3623         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3624         if (*mp == NULL)
3625                 return (error);
3626         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3627
3628         return (0);
3629
3630 bad:
3631         m_freem(*mp);
3632         *mp = NULL;
3633         return (error);
3634 }
3635
3636 /*
3637  * Perform basic checks on header size since
3638  * pfil assumes ip_input has already processed
3639  * it for it.  Cut-and-pasted from ip_input.c.
3640  * Given how simple the IPv6 version is,
3641  * does the IPv4 version really need to be
3642  * this complicated?
3643  *
3644  * XXX Should we update ipstat here, or not?
3645  * XXX Right now we update ipstat but not
3646  * XXX csum_counter.
3647  */
3648 static int
3649 bridge_ip_checkbasic(struct mbuf **mp)
3650 {
3651         struct mbuf *m = *mp;
3652         struct ip *ip;
3653         int len, hlen;
3654         u_short sum;
3655
3656         if (*mp == NULL)
3657                 return (-1);
3658
3659         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3660                 if ((m = m_copyup(m, sizeof(struct ip),
3661                         (max_linkhdr + 3) & ~3)) == NULL) {
3662                         /* XXXJRT new stat, please */
3663                         KMOD_IPSTAT_INC(ips_toosmall);
3664                         goto bad;
3665                 }
3666         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3667                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3668                         KMOD_IPSTAT_INC(ips_toosmall);
3669                         goto bad;
3670                 }
3671         }
3672         ip = mtod(m, struct ip *);
3673         if (ip == NULL) goto bad;
3674
3675         if (ip->ip_v != IPVERSION) {
3676                 KMOD_IPSTAT_INC(ips_badvers);
3677                 goto bad;
3678         }
3679         hlen = ip->ip_hl << 2;
3680         if (hlen < sizeof(struct ip)) { /* minimum header length */
3681                 KMOD_IPSTAT_INC(ips_badhlen);
3682                 goto bad;
3683         }
3684         if (hlen > m->m_len) {
3685                 if ((m = m_pullup(m, hlen)) == NULL) {
3686                         KMOD_IPSTAT_INC(ips_badhlen);
3687                         goto bad;
3688                 }
3689                 ip = mtod(m, struct ip *);
3690                 if (ip == NULL) goto bad;
3691         }
3692
3693         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3694                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3695         } else {
3696                 if (hlen == sizeof(struct ip)) {
3697                         sum = in_cksum_hdr(ip);
3698                 } else {
3699                         sum = in_cksum(m, hlen);
3700                 }
3701         }
3702         if (sum) {
3703                 KMOD_IPSTAT_INC(ips_badsum);
3704                 goto bad;
3705         }
3706
3707         /* Retrieve the packet length. */
3708         len = ntohs(ip->ip_len);
3709
3710         /*
3711          * Check for additional length bogosity
3712          */
3713         if (len < hlen) {
3714                 KMOD_IPSTAT_INC(ips_badlen);
3715                 goto bad;
3716         }
3717
3718         /*
3719          * Check that the amount of data in the buffers
3720          * is as at least much as the IP header would have us expect.
3721          * Drop packet if shorter than we expect.
3722          */
3723         if (m->m_pkthdr.len < len) {
3724                 KMOD_IPSTAT_INC(ips_tooshort);
3725                 goto bad;
3726         }
3727
3728         /* Checks out, proceed */
3729         *mp = m;
3730         return (0);
3731
3732 bad:
3733         *mp = m;
3734         return (-1);
3735 }
3736
3737 #ifdef INET6
3738 /*
3739  * Same as above, but for IPv6.
3740  * Cut-and-pasted from ip6_input.c.
3741  * XXX Should we update ip6stat, or not?
3742  */
3743 static int
3744 bridge_ip6_checkbasic(struct mbuf **mp)
3745 {
3746         struct mbuf *m = *mp;
3747         struct ip6_hdr *ip6;
3748
3749         /*
3750          * If the IPv6 header is not aligned, slurp it up into a new
3751          * mbuf with space for link headers, in the event we forward
3752          * it.  Otherwise, if it is aligned, make sure the entire base
3753          * IPv6 header is in the first mbuf of the chain.
3754          */
3755         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3756                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3757                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3758                             (max_linkhdr + 3) & ~3)) == NULL) {
3759                         /* XXXJRT new stat, please */
3760                         IP6STAT_INC(ip6s_toosmall);
3761                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3762                         goto bad;
3763                 }
3764         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3765                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3766                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3767                         IP6STAT_INC(ip6s_toosmall);
3768                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3769                         goto bad;
3770                 }
3771         }
3772
3773         ip6 = mtod(m, struct ip6_hdr *);
3774
3775         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3776                 IP6STAT_INC(ip6s_badvers);
3777                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3778                 goto bad;
3779         }
3780
3781         /* Checks out, proceed */
3782         *mp = m;
3783         return (0);
3784
3785 bad:
3786         *mp = m;
3787         return (-1);
3788 }
3789 #endif /* INET6 */
3790
3791 /*
3792  * bridge_fragment:
3793  *
3794  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3795  */
3796 static int
3797 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3798     int snap, struct llc *llc)
3799 {
3800         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3801         struct ip *ip;
3802         int error = -1;
3803
3804         if (m->m_len < sizeof(struct ip) &&
3805             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3806                 goto dropit;
3807         ip = mtod(m, struct ip *);
3808
3809         m->m_pkthdr.csum_flags |= CSUM_IP;
3810         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3811         if (error)
3812                 goto dropit;
3813
3814         /*
3815          * Walk the chain and re-add the Ethernet header for
3816          * each mbuf packet.
3817          */
3818         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3819                 nextpkt = mcur->m_nextpkt;
3820                 mcur->m_nextpkt = NULL;
3821                 if (snap) {
3822                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3823                         if (mcur == NULL) {
3824                                 error = ENOBUFS;
3825                                 if (mprev != NULL)
3826                                         mprev->m_nextpkt = nextpkt;
3827                                 goto dropit;
3828                         }
3829                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3830                 }
3831
3832                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3833                 if (mcur == NULL) {
3834                         error = ENOBUFS;
3835                         if (mprev != NULL)
3836                                 mprev->m_nextpkt = nextpkt;
3837                         goto dropit;
3838                 }
3839                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3840
3841                 /*
3842                  * The previous two M_PREPEND could have inserted one or two
3843                  * mbufs in front so we have to update the previous packet's
3844                  * m_nextpkt.
3845                  */
3846                 mcur->m_nextpkt = nextpkt;
3847                 if (mprev != NULL)
3848                         mprev->m_nextpkt = mcur;
3849                 else {
3850                         /* The first mbuf in the original chain needs to be
3851                          * updated. */
3852                         *mp = mcur;
3853                 }
3854                 mprev = mcur;
3855         }
3856
3857         KMOD_IPSTAT_INC(ips_fragmented);
3858         return (error);
3859
3860 dropit:
3861         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3862                 m = mcur->m_nextpkt;
3863                 m_freem(mcur);
3864         }
3865         return (error);
3866 }
3867
3868 static void
3869 bridge_linkstate(struct ifnet *ifp)
3870 {
3871         struct bridge_softc *sc = ifp->if_bridge;
3872         struct bridge_iflist *bif;
3873         struct epoch_tracker et;
3874
3875         NET_EPOCH_ENTER(et);
3876
3877         bif = bridge_lookup_member_if(sc, ifp);
3878         if (bif == NULL) {
3879                 NET_EPOCH_EXIT(et);
3880                 return;
3881         }
3882         bridge_linkcheck(sc);
3883
3884         bstp_linkstate(&bif->bif_stp);
3885
3886         NET_EPOCH_EXIT(et);
3887 }
3888
3889 static void
3890 bridge_linkcheck(struct bridge_softc *sc)
3891 {
3892         struct bridge_iflist *bif;
3893         int new_link, hasls;
3894
3895         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3896
3897         new_link = LINK_STATE_DOWN;
3898         hasls = 0;
3899         /* Our link is considered up if at least one of our ports is active */
3900         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3901                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3902                         hasls++;
3903                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3904                         new_link = LINK_STATE_UP;
3905                         break;
3906                 }
3907         }
3908         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3909                 /* If no interfaces support link-state then we default to up */
3910                 new_link = LINK_STATE_UP;
3911         }
3912         if_link_state_change(sc->sc_ifp, new_link);
3913 }