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 #include "opt_inet.h"
  81 #include "opt_inet6.h"
  82
  83 #include <sys/param.h>
  84 #include <sys/eventhandler.h>
  85 #include <sys/mbuf.h>
  86 #include <sys/malloc.h>
  87 #include <sys/protosw.h>
  88 #include <sys/systm.h>
  89 #include <sys/jail.h>
  90 #include <sys/time.h>
  91 #include <sys/socket.h> /* for net/if.h */
  92 #include <sys/sockio.h>
  93 #include <sys/ctype.h>  /* string functions */
  94 #include <sys/kernel.h>
  95 #include <sys/random.h>
  96 #include <sys/syslog.h>
  97 #include <sys/sysctl.h>
  98 #include <vm/uma.h>
  99 #include <sys/module.h>
 100 #include <sys/priv.h>
 101 #include <sys/proc.h>
 102 #include <sys/lock.h>
 103 #include <sys/mutex.h>
 104
 105 #include <net/bpf.h>
 106 #include <net/if.h>
 107 #include <net/if_clone.h>
 108 #include <net/if_dl.h>
 109 #include <net/if_types.h>
 110 #include <net/if_var.h>
 111 #include <net/if_private.h>
 112 #include <net/pfil.h>
 113 #include <net/vnet.h>
 114
 115 #include <netinet/in.h>
 116 #include <netinet/in_systm.h>
 117 #include <netinet/in_var.h>
 118 #include <netinet/ip.h>
 119 #include <netinet/ip_var.h>
 120 #ifdef INET6
 121 #include <netinet/ip6.h>
 122 #include <netinet6/ip6_var.h>
 123 #include <netinet6/in6_ifattach.h>
 124 #endif
 125 #if defined(INET) || defined(INET6)
 126 #include <netinet/ip_carp.h>
 127 #endif
 128 #include <machine/in_cksum.h>
 129 #include <netinet/if_ether.h>
 130 #include <net/bridgestp.h>
 131 #include <net/if_bridgevar.h>
 132 #include <net/if_llc.h>
 133 #include <net/if_vlan_var.h>
 134
 135 #include <net/route.h>
 136
 137 /*
 138  * Size of the route hash table.  Must be a power of two.
 139  */
 140 #ifndef BRIDGE_RTHASH_SIZE
 141 #define BRIDGE_RTHASH_SIZE              1024
 142 #endif
 143
 144 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 145
 146 /*
 147  * Default maximum number of addresses to cache.
 148  */
 149 #ifndef BRIDGE_RTABLE_MAX
 150 #define BRIDGE_RTABLE_MAX               2000
 151 #endif
 152
 153 /*
 154  * Timeout (in seconds) for entries learned dynamically.
 155  */
 156 #ifndef BRIDGE_RTABLE_TIMEOUT
 157 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 158 #endif
 159
 160 /*
 161  * Number of seconds between walks of the route list.
 162  */
 163 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 164 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 165 #endif
 166
 167 /*
 168  * List of capabilities to possibly mask on the member interface.
 169  */
 170 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 171                                          IFCAP_TXCSUM_IPV6)
 172
 173 /*
 174  * List of capabilities to strip
 175  */
 176 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 177
 178 /*
 179  * Bridge locking
 180  *
 181  * The bridge relies heavily on the epoch(9) system to protect its data
 182  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 183  * must ensure there is only one writer at a time.
 184  *
 185  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 186  * accesses we must hold:
 187  *
 188  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 189  *  - BRIDGE_LOCK, for any other change
 190  *
 191  * The BRIDGE_LOCK is a sleepable lock, because it is held across ioctl()
 192  * calls to bridge member interfaces and these ioctl()s can sleep.
 193  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 194  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 195  */
 196 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 197         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 198         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 199 } while (0)
 200 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 201         sx_destroy(&(_sc)->sc_sx);              \
 202         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 203 } while (0)
 204 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 205 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 206 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 207 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 208             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 209 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 210 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 211 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 212 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 213 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 214             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 215
 216 /*
 217  * Bridge interface list entry.
 218  */
 219 struct bridge_iflist {
 220         CK_LIST_ENTRY(bridge_iflist) bif_next;
 221         struct ifnet            *bif_ifp;       /* member if */
 222         struct bstp_port        bif_stp;        /* STP state */
 223         uint32_t                bif_flags;      /* member if flags */
 224         int                     bif_savedcaps;  /* saved capabilities */
 225         uint32_t                bif_addrmax;    /* max # of addresses */
 226         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 227         uint32_t                bif_addrexceeded;/* # of address violations */
 228         struct epoch_context    bif_epoch_ctx;
 229 };
 230
 231 /*
 232  * Bridge route node.
 233  */
 234 struct bridge_rtnode {
 235         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 236         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 237         struct bridge_iflist    *brt_dst;       /* destination if */
 238         unsigned long           brt_expire;     /* expiration time */
 239         uint8_t                 brt_flags;      /* address flags */
 240         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 241         uint16_t                brt_vlan;       /* vlan id */
 242         struct  vnet            *brt_vnet;
 243         struct  epoch_context   brt_epoch_ctx;
 244 };
 245 #define brt_ifp                 brt_dst->bif_ifp
 246
 247 /*
 248  * Software state for each bridge.
 249  */
 250 struct bridge_softc {
 251         struct ifnet            *sc_ifp;        /* make this an interface */
 252         LIST_ENTRY(bridge_softc) sc_list;
 253         struct sx               sc_sx;
 254         struct mtx              sc_rt_mtx;
 255         uint32_t                sc_brtmax;      /* max # of addresses */
 256         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 257         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 258         struct callout          sc_brcallout;   /* bridge callout */
 259         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 260         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 261         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 262         uint32_t                sc_rthash_key;  /* key for hash */
 263         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 264         struct bstp_state       sc_stp;         /* STP state */
 265         uint32_t                sc_brtexceeded; /* # of cache drops */
 266         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 267         struct ether_addr       sc_defaddr;     /* Default MAC address */
 268         if_input_fn_t           sc_if_input;    /* Saved copy of if_input */
 269         struct epoch_context    sc_epoch_ctx;
 270 };
 271
 272 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 273 #define V_bridge_list_sx        VNET(bridge_list_sx)
 274 static eventhandler_tag bridge_detach_cookie;
 275
 276 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 277
 278 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 279 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 280
 281 static int      bridge_clone_create(struct if_clone *, char *, size_t,
 282                     struct ifc_data *, struct ifnet **);
 283 static int      bridge_clone_destroy(struct if_clone *, struct ifnet *, uint32_t);
 284
 285 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 286 static void     bridge_mutecaps(struct bridge_softc *);
 287 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 288                     int);
 289 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 290 static void     bridge_init(void *);
 291 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 292 static void     bridge_stop(struct ifnet *, int);
 293 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 294 #ifdef ALTQ
 295 static void     bridge_altq_start(if_t);
 296 static int      bridge_altq_transmit(if_t, struct mbuf *);
 297 #endif
 298 static void     bridge_qflush(struct ifnet *);
 299 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 300 static void     bridge_inject(struct ifnet *, struct mbuf *);
 301 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 302                     struct rtentry *);
 303 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 304                     struct mbuf *);
 305 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 306
 307 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 308                     struct mbuf *m);
 309
 310 static void     bridge_timer(void *);
 311
 312 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 313                     struct mbuf *, int);
 314 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 315
 316 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 317                     uint16_t, struct bridge_iflist *, int, uint8_t);
 318 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 319                     uint16_t);
 320 static void     bridge_rttrim(struct bridge_softc *);
 321 static void     bridge_rtage(struct bridge_softc *);
 322 static void     bridge_rtflush(struct bridge_softc *, int);
 323 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 324                     uint16_t);
 325
 326 static void     bridge_rtable_init(struct bridge_softc *);
 327 static void     bridge_rtable_fini(struct bridge_softc *);
 328
 329 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 330 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 331                     const uint8_t *, uint16_t);
 332 static int      bridge_rtnode_insert(struct bridge_softc *,
 333                     struct bridge_rtnode *);
 334 static void     bridge_rtnode_destroy(struct bridge_softc *,
 335                     struct bridge_rtnode *);
 336 static void     bridge_rtable_expire(struct ifnet *, int);
 337 static void     bridge_state_change(struct ifnet *, int);
 338
 339 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 340                     const char *name);
 341 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 342                     struct ifnet *ifp);
 343 static void     bridge_delete_member(struct bridge_softc *,
 344                     struct bridge_iflist *, int);
 345 static void     bridge_delete_span(struct bridge_softc *,
 346                     struct bridge_iflist *);
 347
 348 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 349 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 350 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 351 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 352 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 353 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 354 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 379 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 380                     int);
 381 static int      bridge_ip_checkbasic(struct mbuf **mp);
 382 #ifdef INET6
 383 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 384 #endif /* INET6 */
 385 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 386                     struct ether_header *, int, struct llc *);
 387 static void     bridge_linkstate(struct ifnet *ifp);
 388 static void     bridge_linkcheck(struct bridge_softc *sc);
 389
 390 /*
 391  * Use the "null" value from IEEE 802.1Q-2014 Table 9-2
 392  * to indicate untagged frames.
 393  */
 394 #define VLANTAGOF(_m)   \
 395     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : DOT1Q_VID_NULL
 396
 397 static struct bstp_cb_ops bridge_ops = {
 398         .bcb_state = bridge_state_change,
 399         .bcb_rtage = bridge_rtable_expire
 400 };
 401
 402 SYSCTL_DECL(_net_link);
 403 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
 404     "Bridge");
 405
 406 /* only pass IP[46] packets when pfil is enabled */
 407 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 408 #define V_pfil_onlyip   VNET(pfil_onlyip)
 409 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 410     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 411     "Only pass IP packets when pfil is enabled");
 412
 413 /* run pfil hooks on the bridge interface */
 414 VNET_DEFINE_STATIC(int, pfil_bridge) = 0;
 415 #define V_pfil_bridge   VNET(pfil_bridge)
 416 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 417     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 418     "Packet filter on the bridge interface");
 419
 420 /* layer2 filter with ipfw */
 421 VNET_DEFINE_STATIC(int, pfil_ipfw);
 422 #define V_pfil_ipfw     VNET(pfil_ipfw)
 423
 424 /* layer2 ARP filter with ipfw */
 425 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 426 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 427 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 428     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 429     "Filter ARP packets through IPFW layer2");
 430
 431 /* run pfil hooks on the member interface */
 432 VNET_DEFINE_STATIC(int, pfil_member) = 0;
 433 #define V_pfil_member   VNET(pfil_member)
 434 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 435     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 436     "Packet filter on the member interface");
 437
 438 /* run pfil hooks on the physical interface for locally destined packets */
 439 VNET_DEFINE_STATIC(int, pfil_local_phys);
 440 #define V_pfil_local_phys       VNET(pfil_local_phys)
 441 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 442     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 443     "Packet filter on the physical interface for locally destined packets");
 444
 445 /* log STP state changes */
 446 VNET_DEFINE_STATIC(int, log_stp);
 447 #define V_log_stp       VNET(log_stp)
 448 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 449     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 450     "Log STP state changes");
 451
 452 /* share MAC with first bridge member */
 453 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 454 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 455 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 456     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 457     "Inherit MAC address from the first bridge member");
 458
 459 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 460 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 461 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 462     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 463     "Allow overlap of link-local scope "
 464     "zones of a bridge interface and the member interfaces");
 465
 466 /* log MAC address port flapping */
 467 VNET_DEFINE_STATIC(bool, log_mac_flap) = true;
 468 #define V_log_mac_flap  VNET(log_mac_flap)
 469 SYSCTL_BOOL(_net_link_bridge, OID_AUTO, log_mac_flap,
 470     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(log_mac_flap), true,
 471     "Log MAC address port flapping");
 472
 473 VNET_DEFINE_STATIC(int, log_interval) = 5;
 474 VNET_DEFINE_STATIC(int, log_count) = 0;
 475 VNET_DEFINE_STATIC(struct timeval, log_last) = { 0 };
 476
 477 #define V_log_interval  VNET(log_interval)
 478 #define V_log_count     VNET(log_count)
 479 #define V_log_last      VNET(log_last)
 480
 481 struct bridge_control {
 482         int     (*bc_func)(struct bridge_softc *, void *);
 483         int     bc_argsize;
 484         int     bc_flags;
 485 };
 486
 487 #define BC_F_COPYIN             0x01    /* copy arguments in */
 488 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 489 #define BC_F_SUSER              0x04    /* do super-user check */
 490
 491 static const struct bridge_control bridge_control_table[] = {
 492         { bridge_ioctl_add,             sizeof(struct ifbreq),
 493           BC_F_COPYIN|BC_F_SUSER },
 494         { bridge_ioctl_del,             sizeof(struct ifbreq),
 495           BC_F_COPYIN|BC_F_SUSER },
 496
 497         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 498           BC_F_COPYIN|BC_F_COPYOUT },
 499         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 500           BC_F_COPYIN|BC_F_SUSER },
 501
 502         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 503           BC_F_COPYIN|BC_F_SUSER },
 504         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 505           BC_F_COPYOUT },
 506
 507         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 508           BC_F_COPYIN|BC_F_COPYOUT },
 509         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 510           BC_F_COPYIN|BC_F_COPYOUT },
 511
 512         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 513           BC_F_COPYIN|BC_F_SUSER },
 514
 515         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 516           BC_F_COPYIN|BC_F_SUSER },
 517         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 518           BC_F_COPYOUT },
 519
 520         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 521           BC_F_COPYIN|BC_F_SUSER },
 522
 523         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 524           BC_F_COPYIN|BC_F_SUSER },
 525
 526         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 527           BC_F_COPYOUT },
 528         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 529           BC_F_COPYIN|BC_F_SUSER },
 530
 531         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 532           BC_F_COPYOUT },
 533         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 534           BC_F_COPYIN|BC_F_SUSER },
 535
 536         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 537           BC_F_COPYOUT },
 538         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 539           BC_F_COPYIN|BC_F_SUSER },
 540
 541         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 542           BC_F_COPYOUT },
 543         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 544           BC_F_COPYIN|BC_F_SUSER },
 545
 546         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 547           BC_F_COPYIN|BC_F_SUSER },
 548
 549         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 550           BC_F_COPYIN|BC_F_SUSER },
 551
 552         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 553           BC_F_COPYIN|BC_F_SUSER },
 554         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 555           BC_F_COPYIN|BC_F_SUSER },
 556
 557         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 558           BC_F_COPYOUT },
 559
 560         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 561           BC_F_COPYOUT },
 562
 563         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 564           BC_F_COPYIN|BC_F_COPYOUT },
 565
 566         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 567           BC_F_COPYIN|BC_F_SUSER },
 568
 569         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 570           BC_F_COPYIN|BC_F_SUSER },
 571
 572         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 573           BC_F_COPYIN|BC_F_SUSER },
 574
 575 };
 576 static const int bridge_control_table_size = nitems(bridge_control_table);
 577
 578 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 579 #define V_bridge_list   VNET(bridge_list)
 580 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 581                                             "if_bridge list")
 582 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 583 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 584 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 585
 586 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 587 #define V_bridge_cloner VNET(bridge_cloner)
 588
 589 static const char bridge_name[] = "bridge";
 590
 591 static void
 592 vnet_bridge_init(const void *unused __unused)
 593 {
 594
 595         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 596             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 597             UMA_ALIGN_PTR, 0);
 598         BRIDGE_LIST_LOCK_INIT();
 599         LIST_INIT(&V_bridge_list);
 600
 601         struct if_clone_addreq req = {
 602                 .create_f = bridge_clone_create,
 603                 .destroy_f = bridge_clone_destroy,
 604                 .flags = IFC_F_AUTOUNIT,
 605         };
 606         V_bridge_cloner = ifc_attach_cloner(bridge_name, &req);
 607 }
 608 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 609     vnet_bridge_init, NULL);
 610
 611 static void
 612 vnet_bridge_uninit(const void *unused __unused)
 613 {
 614
 615         ifc_detach_cloner(V_bridge_cloner);
 616         V_bridge_cloner = NULL;
 617         BRIDGE_LIST_LOCK_DESTROY();
 618
 619         /* Callbacks may use the UMA zone. */
 620         NET_EPOCH_DRAIN_CALLBACKS();
 621
 622         uma_zdestroy(V_bridge_rtnode_zone);
 623 }
 624 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 625     vnet_bridge_uninit, NULL);
 626
 627 static int
 628 bridge_modevent(module_t mod, int type, void *data)
 629 {
 630
 631         switch (type) {
 632         case MOD_LOAD:
 633                 bridge_dn_p = bridge_dummynet;
 634                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 635                     ifnet_departure_event, bridge_ifdetach, NULL,
 636                     EVENTHANDLER_PRI_ANY);
 637                 break;
 638         case MOD_UNLOAD:
 639                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 640                     bridge_detach_cookie);
 641                 bridge_dn_p = NULL;
 642                 break;
 643         default:
 644                 return (EOPNOTSUPP);
 645         }
 646         return (0);
 647 }
 648
 649 static moduledata_t bridge_mod = {
 650         "if_bridge",
 651         bridge_modevent,
 652         0
 653 };
 654
 655 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 656 MODULE_VERSION(if_bridge, 1);
 657 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 658
 659 /*
 660  * handler for net.link.bridge.ipfw
 661  */
 662 static int
 663 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 664 {
 665         int enable = V_pfil_ipfw;
 666         int error;
 667
 668         error = sysctl_handle_int(oidp, &enable, 0, req);
 669         enable &= 1;
 670
 671         if (enable != V_pfil_ipfw) {
 672                 V_pfil_ipfw = enable;
 673
 674                 /*
 675                  * Disable pfil so that ipfw doesnt run twice, if the user
 676                  * really wants both then they can re-enable pfil_bridge and/or
 677                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 678                  * layer2 type.
 679                  */
 680                 if (V_pfil_ipfw) {
 681                         V_pfil_onlyip = 0;
 682                         V_pfil_bridge = 0;
 683                         V_pfil_member = 0;
 684                 }
 685         }
 686
 687         return (error);
 688 }
 689 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 690     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET | CTLFLAG_NEEDGIANT,
 691     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 692     "Layer2 filter with IPFW");
 693
 694 #ifdef VIMAGE
 695 static void
 696 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 697 {
 698         struct bridge_softc *sc = ifp->if_softc;
 699         struct bridge_iflist *bif;
 700
 701         BRIDGE_LOCK(sc);
 702
 703         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 704                 bridge_delete_member(sc, bif, 0);
 705
 706         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 707                 bridge_delete_span(sc, bif);
 708         }
 709
 710         BRIDGE_UNLOCK(sc);
 711
 712         ether_reassign(ifp, newvnet, arg);
 713 }
 714 #endif
 715
 716 /*
 717  * bridge_clone_create:
 718  *
 719  *      Create a new bridge instance.
 720  */
 721 static int
 722 bridge_clone_create(struct if_clone *ifc, char *name, size_t len,
 723     struct ifc_data *ifd, struct ifnet **ifpp)
 724 {
 725         struct bridge_softc *sc;
 726         struct ifnet *ifp;
 727
 728         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 729         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 730         if (ifp == NULL) {
 731                 free(sc, M_DEVBUF);
 732                 return (ENOSPC);
 733         }
 734
 735         BRIDGE_LOCK_INIT(sc);
 736         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 737         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 738
 739         /* Initialize our routing table. */
 740         bridge_rtable_init(sc);
 741
 742         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 743
 744         CK_LIST_INIT(&sc->sc_iflist);
 745         CK_LIST_INIT(&sc->sc_spanlist);
 746
 747         ifp->if_softc = sc;
 748         if_initname(ifp, bridge_name, ifd->unit);
 749         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 750         ifp->if_ioctl = bridge_ioctl;
 751 #ifdef ALTQ
 752         ifp->if_start = bridge_altq_start;
 753         ifp->if_transmit = bridge_altq_transmit;
 754         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 755         ifp->if_snd.ifq_drv_maxlen = 0;
 756         IFQ_SET_READY(&ifp->if_snd);
 757 #else
 758         ifp->if_transmit = bridge_transmit;
 759 #endif
 760         ifp->if_qflush = bridge_qflush;
 761         ifp->if_init = bridge_init;
 762         ifp->if_type = IFT_BRIDGE;
 763
 764         ether_gen_addr(ifp, &sc->sc_defaddr);
 765
 766         bstp_attach(&sc->sc_stp, &bridge_ops);
 767         ether_ifattach(ifp, sc->sc_defaddr.octet);
 768         /* Now undo some of the damage... */
 769         ifp->if_baudrate = 0;
 770         ifp->if_type = IFT_BRIDGE;
 771 #ifdef VIMAGE
 772         ifp->if_reassign = bridge_reassign;
 773 #endif
 774         sc->sc_if_input = ifp->if_input;        /* ether_input */
 775         ifp->if_input = bridge_inject;
 776
 777         /*
 778          * Allow BRIDGE_INPUT() to pass in packets originating from the bridge
 779          * itself via bridge_inject().  This is required for netmap but
 780          * otherwise has no effect.
 781          */
 782         ifp->if_bridge_input = bridge_input;
 783
 784         BRIDGE_LIST_LOCK();
 785         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 786         BRIDGE_LIST_UNLOCK();
 787         *ifpp = ifp;
 788
 789         return (0);
 790 }
 791
 792 static void
 793 bridge_clone_destroy_cb(struct epoch_context *ctx)
 794 {
 795         struct bridge_softc *sc;
 796
 797         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 798
 799         BRIDGE_LOCK_DESTROY(sc);
 800         free(sc, M_DEVBUF);
 801 }
 802
 803 /*
 804  * bridge_clone_destroy:
 805  *
 806  *      Destroy a bridge instance.
 807  */
 808 static int
 809 bridge_clone_destroy(struct if_clone *ifc, struct ifnet *ifp, uint32_t flags)
 810 {
 811         struct bridge_softc *sc = ifp->if_softc;
 812         struct bridge_iflist *bif;
 813         struct epoch_tracker et;
 814
 815         BRIDGE_LOCK(sc);
 816
 817         bridge_stop(ifp, 1);
 818         ifp->if_flags &= ~IFF_UP;
 819
 820         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 821                 bridge_delete_member(sc, bif, 0);
 822
 823         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 824                 bridge_delete_span(sc, bif);
 825         }
 826
 827         /* Tear down the routing table. */
 828         bridge_rtable_fini(sc);
 829
 830         BRIDGE_UNLOCK(sc);
 831
 832         NET_EPOCH_ENTER(et);
 833
 834         callout_drain(&sc->sc_brcallout);
 835
 836         BRIDGE_LIST_LOCK();
 837         LIST_REMOVE(sc, sc_list);
 838         BRIDGE_LIST_UNLOCK();
 839
 840         bstp_detach(&sc->sc_stp);
 841 #ifdef ALTQ
 842         IFQ_PURGE(&ifp->if_snd);
 843 #endif
 844         NET_EPOCH_EXIT(et);
 845
 846         ether_ifdetach(ifp);
 847         if_free(ifp);
 848
 849         NET_EPOCH_CALL(bridge_clone_destroy_cb, &sc->sc_epoch_ctx);
 850
 851         return (0);
 852 }
 853
 854 /*
 855  * bridge_ioctl:
 856  *
 857  *      Handle a control request from the operator.
 858  */
 859 static int
 860 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 861 {
 862         struct bridge_softc *sc = ifp->if_softc;
 863         struct ifreq *ifr = (struct ifreq *)data;
 864         struct bridge_iflist *bif;
 865         struct thread *td = curthread;
 866         union {
 867                 struct ifbreq ifbreq;
 868                 struct ifbifconf ifbifconf;
 869                 struct ifbareq ifbareq;
 870                 struct ifbaconf ifbaconf;
 871                 struct ifbrparam ifbrparam;
 872                 struct ifbropreq ifbropreq;
 873         } args;
 874         struct ifdrv *ifd = (struct ifdrv *) data;
 875         const struct bridge_control *bc;
 876         int error = 0, oldmtu;
 877
 878         BRIDGE_LOCK(sc);
 879
 880         switch (cmd) {
 881         case SIOCADDMULTI:
 882         case SIOCDELMULTI:
 883                 break;
 884
 885         case SIOCGDRVSPEC:
 886         case SIOCSDRVSPEC:
 887                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 888                         error = EINVAL;
 889                         break;
 890                 }
 891                 bc = &bridge_control_table[ifd->ifd_cmd];
 892
 893                 if (cmd == SIOCGDRVSPEC &&
 894                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 895                         error = EINVAL;
 896                         break;
 897                 }
 898                 else if (cmd == SIOCSDRVSPEC &&
 899                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 900                         error = EINVAL;
 901                         break;
 902                 }
 903
 904                 if (bc->bc_flags & BC_F_SUSER) {
 905                         error = priv_check(td, PRIV_NET_BRIDGE);
 906                         if (error)
 907                                 break;
 908                 }
 909
 910                 if (ifd->ifd_len != bc->bc_argsize ||
 911                     ifd->ifd_len > sizeof(args)) {
 912                         error = EINVAL;
 913                         break;
 914                 }
 915
 916                 bzero(&args, sizeof(args));
 917                 if (bc->bc_flags & BC_F_COPYIN) {
 918                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 919                         if (error)
 920                                 break;
 921                 }
 922
 923                 oldmtu = ifp->if_mtu;
 924                 error = (*bc->bc_func)(sc, &args);
 925                 if (error)
 926                         break;
 927
 928                 /*
 929                  * Bridge MTU may change during addition of the first port.
 930                  * If it did, do network layer specific procedure.
 931                  */
 932                 if (ifp->if_mtu != oldmtu)
 933                         if_notifymtu(ifp);
 934
 935                 if (bc->bc_flags & BC_F_COPYOUT)
 936                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 937
 938                 break;
 939
 940         case SIOCSIFFLAGS:
 941                 if (!(ifp->if_flags & IFF_UP) &&
 942                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 943                         /*
 944                          * If interface is marked down and it is running,
 945                          * then stop and disable it.
 946                          */
 947                         bridge_stop(ifp, 1);
 948                 } else if ((ifp->if_flags & IFF_UP) &&
 949                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 950                         /*
 951                          * If interface is marked up and it is stopped, then
 952                          * start it.
 953                          */
 954                         BRIDGE_UNLOCK(sc);
 955                         (*ifp->if_init)(sc);
 956                         BRIDGE_LOCK(sc);
 957                 }
 958                 break;
 959
 960         case SIOCSIFMTU:
 961                 oldmtu = sc->sc_ifp->if_mtu;
 962
 963                 if (ifr->ifr_mtu < 576) {
 964                         error = EINVAL;
 965                         break;
 966                 }
 967                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 968                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 969                         break;
 970                 }
 971                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 972                         error = (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 973                             SIOCSIFMTU, (caddr_t)ifr);
 974                         if (error != 0) {
 975                                 log(LOG_NOTICE, "%s: invalid MTU: %u for"
 976                                     " member %s\n", sc->sc_ifp->if_xname,
 977                                     ifr->ifr_mtu,
 978                                     bif->bif_ifp->if_xname);
 979                                 error = EINVAL;
 980                                 break;
 981                         }
 982                 }
 983                 if (error) {
 984                         /* Restore the previous MTU on all member interfaces. */
 985                         ifr->ifr_mtu = oldmtu;
 986                         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 987                                 (*bif->bif_ifp->if_ioctl)(bif->bif_ifp,
 988                                     SIOCSIFMTU, (caddr_t)ifr);
 989                         }
 990                 } else {
 991                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 992                 }
 993                 break;
 994         default:
 995                 /*
 996                  * drop the lock as ether_ioctl() will call bridge_start() and
 997                  * cause the lock to be recursed.
 998                  */
 999                 BRIDGE_UNLOCK(sc);
1000                 error = ether_ioctl(ifp, cmd, data);
1001                 BRIDGE_LOCK(sc);
1002                 break;
1003         }
1004
1005         BRIDGE_UNLOCK(sc);
1006
1007         return (error);
1008 }
1009
1010 /*
1011  * bridge_mutecaps:
1012  *
1013  *      Clear or restore unwanted capabilities on the member interface
1014  */
1015 static void
1016 bridge_mutecaps(struct bridge_softc *sc)
1017 {
1018         struct bridge_iflist *bif;
1019         int enabled, mask;
1020
1021         BRIDGE_LOCK_ASSERT(sc);
1022
1023         /* Initial bitmask of capabilities to test */
1024         mask = BRIDGE_IFCAPS_MASK;
1025
1026         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1027                 /* Every member must support it or its disabled */
1028                 mask &= bif->bif_savedcaps;
1029         }
1030
1031         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1032                 enabled = bif->bif_ifp->if_capenable;
1033                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1034                 /* strip off mask bits and enable them again if allowed */
1035                 enabled &= ~BRIDGE_IFCAPS_MASK;
1036                 enabled |= mask;
1037                 bridge_set_ifcap(sc, bif, enabled);
1038         }
1039 }
1040
1041 static void
1042 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1043 {
1044         struct ifnet *ifp = bif->bif_ifp;
1045         struct ifreq ifr;
1046         int error, mask, stuck;
1047
1048         bzero(&ifr, sizeof(ifr));
1049         ifr.ifr_reqcap = set;
1050
1051         if (ifp->if_capenable != set) {
1052                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1053                 if (error)
1054                         if_printf(sc->sc_ifp,
1055                             "error setting capabilities on %s: %d\n",
1056                             ifp->if_xname, error);
1057                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1058                 stuck = ifp->if_capenable & mask & ~set;
1059                 if (stuck != 0)
1060                         if_printf(sc->sc_ifp,
1061                             "can't disable some capabilities on %s: 0x%x\n",
1062                             ifp->if_xname, stuck);
1063         }
1064 }
1065
1066 /*
1067  * bridge_lookup_member:
1068  *
1069  *      Lookup a bridge member interface.
1070  */
1071 static struct bridge_iflist *
1072 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1073 {
1074         struct bridge_iflist *bif;
1075         struct ifnet *ifp;
1076
1077         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1078
1079         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1080                 ifp = bif->bif_ifp;
1081                 if (strcmp(ifp->if_xname, name) == 0)
1082                         return (bif);
1083         }
1084
1085         return (NULL);
1086 }
1087
1088 /*
1089  * bridge_lookup_member_if:
1090  *
1091  *      Lookup a bridge member interface by ifnet*.
1092  */
1093 static struct bridge_iflist *
1094 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1095 {
1096         struct bridge_iflist *bif;
1097
1098         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1099
1100         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1101                 if (bif->bif_ifp == member_ifp)
1102                         return (bif);
1103         }
1104
1105         return (NULL);
1106 }
1107
1108 static void
1109 bridge_delete_member_cb(struct epoch_context *ctx)
1110 {
1111         struct bridge_iflist *bif;
1112
1113         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1114
1115         free(bif, M_DEVBUF);
1116 }
1117
1118 /*
1119  * bridge_delete_member:
1120  *
1121  *      Delete the specified member interface.
1122  */
1123 static void
1124 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1125     int gone)
1126 {
1127         struct ifnet *ifs = bif->bif_ifp;
1128         struct ifnet *fif = NULL;
1129         struct bridge_iflist *bifl;
1130
1131         BRIDGE_LOCK_ASSERT(sc);
1132
1133         if (bif->bif_flags & IFBIF_STP)
1134                 bstp_disable(&bif->bif_stp);
1135
1136         ifs->if_bridge = NULL;
1137         CK_LIST_REMOVE(bif, bif_next);
1138
1139         /*
1140          * If removing the interface that gave the bridge its mac address, set
1141          * the mac address of the bridge to the address of the next member, or
1142          * to its default address if no members are left.
1143          */
1144         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1145                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1146                         bcopy(&sc->sc_defaddr,
1147                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1148                         sc->sc_ifaddr = NULL;
1149                 } else {
1150                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1151                         fif = bifl->bif_ifp;
1152                         bcopy(IF_LLADDR(fif),
1153                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1154                         sc->sc_ifaddr = fif;
1155                 }
1156                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1157         }
1158
1159         bridge_linkcheck(sc);
1160         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1161         BRIDGE_RT_LOCK(sc);
1162         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1163         BRIDGE_RT_UNLOCK(sc);
1164         KASSERT(bif->bif_addrcnt == 0,
1165             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1166
1167         ifs->if_bridge_output = NULL;
1168         ifs->if_bridge_input = NULL;
1169         ifs->if_bridge_linkstate = NULL;
1170         if (!gone) {
1171                 switch (ifs->if_type) {
1172                 case IFT_ETHER:
1173                 case IFT_L2VLAN:
1174                         /*
1175                          * Take the interface out of promiscuous mode, but only
1176                          * if it was promiscuous in the first place. It might
1177                          * not be if we're in the bridge_ioctl_add() error path.
1178                          */
1179                         if (ifs->if_flags & IFF_PROMISC)
1180                                 (void) ifpromisc(ifs, 0);
1181                         break;
1182
1183                 case IFT_GIF:
1184                         break;
1185
1186                 default:
1187 #ifdef DIAGNOSTIC
1188                         panic("bridge_delete_member: impossible");
1189 #endif
1190                         break;
1191                 }
1192                 /* reneable any interface capabilities */
1193                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1194         }
1195         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1196
1197         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1198 }
1199
1200 /*
1201  * bridge_delete_span:
1202  *
1203  *      Delete the specified span interface.
1204  */
1205 static void
1206 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1207 {
1208         BRIDGE_LOCK_ASSERT(sc);
1209
1210         KASSERT(bif->bif_ifp->if_bridge == NULL,
1211             ("%s: not a span interface", __func__));
1212
1213         CK_LIST_REMOVE(bif, bif_next);
1214
1215         NET_EPOCH_CALL(bridge_delete_member_cb, &bif->bif_epoch_ctx);
1216 }
1217
1218 static int
1219 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1220 {
1221         struct ifbreq *req = arg;
1222         struct bridge_iflist *bif = NULL;
1223         struct ifnet *ifs;
1224         int error = 0;
1225
1226         ifs = ifunit(req->ifbr_ifsname);
1227         if (ifs == NULL)
1228                 return (ENOENT);
1229         if (ifs->if_ioctl == NULL)      /* must be supported */
1230                 return (EINVAL);
1231
1232         /* If it's in the span list, it can't be a member. */
1233         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1234                 if (ifs == bif->bif_ifp)
1235                         return (EBUSY);
1236
1237         if (ifs->if_bridge == sc)
1238                 return (EEXIST);
1239
1240         if (ifs->if_bridge != NULL)
1241                 return (EBUSY);
1242
1243         switch (ifs->if_type) {
1244         case IFT_ETHER:
1245         case IFT_L2VLAN:
1246         case IFT_GIF:
1247                 /* permitted interface types */
1248                 break;
1249         default:
1250                 return (EINVAL);
1251         }
1252
1253 #ifdef INET6
1254         /*
1255          * Two valid inet6 addresses with link-local scope must not be
1256          * on the parent interface and the member interfaces at the
1257          * same time.  This restriction is needed to prevent violation
1258          * of link-local scope zone.  Attempts to add a member
1259          * interface which has inet6 addresses when the parent has
1260          * inet6 triggers removal of all inet6 addresses on the member
1261          * interface.
1262          */
1263
1264         /* Check if the parent interface has a link-local scope addr. */
1265         if (V_allow_llz_overlap == 0 &&
1266             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1267                 /*
1268                  * If any, remove all inet6 addresses from the member
1269                  * interfaces.
1270                  */
1271                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1272                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1273                                 in6_ifdetach(bif->bif_ifp);
1274                                 if_printf(sc->sc_ifp,
1275                                     "IPv6 addresses on %s have been removed "
1276                                     "before adding it as a member to prevent "
1277                                     "IPv6 address scope violation.\n",
1278                                     bif->bif_ifp->if_xname);
1279                         }
1280                 }
1281                 if (in6ifa_llaonifp(ifs)) {
1282                         in6_ifdetach(ifs);
1283                         if_printf(sc->sc_ifp,
1284                             "IPv6 addresses on %s have been removed "
1285                             "before adding it as a member to prevent "
1286                             "IPv6 address scope violation.\n",
1287                             ifs->if_xname);
1288                 }
1289         }
1290 #endif
1291         /* Allow the first Ethernet member to define the MTU */
1292         if (CK_LIST_EMPTY(&sc->sc_iflist))
1293                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1294         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1295                 struct ifreq ifr;
1296
1297                 snprintf(ifr.ifr_name, sizeof(ifr.ifr_name), "%s",
1298                     ifs->if_xname);
1299                 ifr.ifr_mtu = sc->sc_ifp->if_mtu;
1300
1301                 error = (*ifs->if_ioctl)(ifs,
1302                     SIOCSIFMTU, (caddr_t)&ifr);
1303                 if (error != 0) {
1304                         log(LOG_NOTICE, "%s: invalid MTU: %u for"
1305                             " new member %s\n", sc->sc_ifp->if_xname,
1306                             ifr.ifr_mtu,
1307                             ifs->if_xname);
1308                         return (EINVAL);
1309                 }
1310         }
1311
1312         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1313         if (bif == NULL)
1314                 return (ENOMEM);
1315
1316         bif->bif_ifp = ifs;
1317         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1318         bif->bif_savedcaps = ifs->if_capenable;
1319
1320         /*
1321          * Assign the interface's MAC address to the bridge if it's the first
1322          * member and the MAC address of the bridge has not been changed from
1323          * the default randomly generated one.
1324          */
1325         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1326             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1327                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1328                 sc->sc_ifaddr = ifs;
1329                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1330         }
1331
1332         ifs->if_bridge = sc;
1333         ifs->if_bridge_output = bridge_output;
1334         ifs->if_bridge_input = bridge_input;
1335         ifs->if_bridge_linkstate = bridge_linkstate;
1336         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1337         /*
1338          * XXX: XLOCK HERE!?!
1339          *
1340          * NOTE: insert_***HEAD*** should be safe for the traversals.
1341          */
1342         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1343
1344         /* Set interface capabilities to the intersection set of all members */
1345         bridge_mutecaps(sc);
1346         bridge_linkcheck(sc);
1347
1348         /* Place the interface into promiscuous mode */
1349         switch (ifs->if_type) {
1350                 case IFT_ETHER:
1351                 case IFT_L2VLAN:
1352                         error = ifpromisc(ifs, 1);
1353                         break;
1354         }
1355
1356         if (error)
1357                 bridge_delete_member(sc, bif, 0);
1358         return (error);
1359 }
1360
1361 static int
1362 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1363 {
1364         struct ifbreq *req = arg;
1365         struct bridge_iflist *bif;
1366
1367         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1368         if (bif == NULL)
1369                 return (ENOENT);
1370
1371         bridge_delete_member(sc, bif, 0);
1372
1373         return (0);
1374 }
1375
1376 static int
1377 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1378 {
1379         struct ifbreq *req = arg;
1380         struct bridge_iflist *bif;
1381         struct bstp_port *bp;
1382
1383         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1384         if (bif == NULL)
1385                 return (ENOENT);
1386
1387         bp = &bif->bif_stp;
1388         req->ifbr_ifsflags = bif->bif_flags;
1389         req->ifbr_state = bp->bp_state;
1390         req->ifbr_priority = bp->bp_priority;
1391         req->ifbr_path_cost = bp->bp_path_cost;
1392         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1393         req->ifbr_proto = bp->bp_protover;
1394         req->ifbr_role = bp->bp_role;
1395         req->ifbr_stpflags = bp->bp_flags;
1396         req->ifbr_addrcnt = bif->bif_addrcnt;
1397         req->ifbr_addrmax = bif->bif_addrmax;
1398         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1399
1400         /* Copy STP state options as flags */
1401         if (bp->bp_operedge)
1402                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1403         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1404                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1405         if (bp->bp_ptp_link)
1406                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1407         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1408                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1409         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1410                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1411         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1412                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1413         return (0);
1414 }
1415
1416 static int
1417 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1418 {
1419         struct epoch_tracker et;
1420         struct ifbreq *req = arg;
1421         struct bridge_iflist *bif;
1422         struct bstp_port *bp;
1423         int error;
1424
1425         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1426         if (bif == NULL)
1427                 return (ENOENT);
1428         bp = &bif->bif_stp;
1429
1430         if (req->ifbr_ifsflags & IFBIF_SPAN)
1431                 /* SPAN is readonly */
1432                 return (EINVAL);
1433
1434         NET_EPOCH_ENTER(et);
1435
1436         if (req->ifbr_ifsflags & IFBIF_STP) {
1437                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1438                         error = bstp_enable(&bif->bif_stp);
1439                         if (error) {
1440                                 NET_EPOCH_EXIT(et);
1441                                 return (error);
1442                         }
1443                 }
1444         } else {
1445                 if ((bif->bif_flags & IFBIF_STP) != 0)
1446                         bstp_disable(&bif->bif_stp);
1447         }
1448
1449         /* Pass on STP flags */
1450         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1451         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1452         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1453         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1454
1455         /* Save the bits relating to the bridge */
1456         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1457
1458         NET_EPOCH_EXIT(et);
1459
1460         return (0);
1461 }
1462
1463 static int
1464 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1465 {
1466         struct ifbrparam *param = arg;
1467
1468         sc->sc_brtmax = param->ifbrp_csize;
1469         bridge_rttrim(sc);
1470
1471         return (0);
1472 }
1473
1474 static int
1475 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1476 {
1477         struct ifbrparam *param = arg;
1478
1479         param->ifbrp_csize = sc->sc_brtmax;
1480
1481         return (0);
1482 }
1483
1484 static int
1485 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1486 {
1487         struct ifbifconf *bifc = arg;
1488         struct bridge_iflist *bif;
1489         struct ifbreq breq;
1490         char *buf, *outbuf;
1491         int count, buflen, len, error = 0;
1492
1493         count = 0;
1494         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1495                 count++;
1496         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1497                 count++;
1498
1499         buflen = sizeof(breq) * count;
1500         if (bifc->ifbic_len == 0) {
1501                 bifc->ifbic_len = buflen;
1502                 return (0);
1503         }
1504         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1505         if (outbuf == NULL)
1506                 return (ENOMEM);
1507
1508         count = 0;
1509         buf = outbuf;
1510         len = min(bifc->ifbic_len, buflen);
1511         bzero(&breq, sizeof(breq));
1512         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1513                 if (len < sizeof(breq))
1514                         break;
1515
1516                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1517                     sizeof(breq.ifbr_ifsname));
1518                 /* Fill in the ifbreq structure */
1519                 error = bridge_ioctl_gifflags(sc, &breq);
1520                 if (error)
1521                         break;
1522                 memcpy(buf, &breq, sizeof(breq));
1523                 count++;
1524                 buf += sizeof(breq);
1525                 len -= sizeof(breq);
1526         }
1527         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1528                 if (len < sizeof(breq))
1529                         break;
1530
1531                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1532                     sizeof(breq.ifbr_ifsname));
1533                 breq.ifbr_ifsflags = bif->bif_flags;
1534                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1535                 memcpy(buf, &breq, sizeof(breq));
1536                 count++;
1537                 buf += sizeof(breq);
1538                 len -= sizeof(breq);
1539         }
1540
1541         bifc->ifbic_len = sizeof(breq) * count;
1542         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1543         free(outbuf, M_TEMP);
1544         return (error);
1545 }
1546
1547 static int
1548 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1549 {
1550         struct ifbaconf *bac = arg;
1551         struct bridge_rtnode *brt;
1552         struct ifbareq bareq;
1553         char *buf, *outbuf;
1554         int count, buflen, len, error = 0;
1555
1556         if (bac->ifbac_len == 0)
1557                 return (0);
1558
1559         count = 0;
1560         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1561                 count++;
1562         buflen = sizeof(bareq) * count;
1563
1564         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1565         if (outbuf == NULL)
1566                 return (ENOMEM);
1567
1568         count = 0;
1569         buf = outbuf;
1570         len = min(bac->ifbac_len, buflen);
1571         bzero(&bareq, sizeof(bareq));
1572         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1573                 if (len < sizeof(bareq))
1574                         goto out;
1575                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1576                     sizeof(bareq.ifba_ifsname));
1577                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1578                 bareq.ifba_vlan = brt->brt_vlan;
1579                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1580                                 time_uptime < brt->brt_expire)
1581                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1582                 else
1583                         bareq.ifba_expire = 0;
1584                 bareq.ifba_flags = brt->brt_flags;
1585
1586                 memcpy(buf, &bareq, sizeof(bareq));
1587                 count++;
1588                 buf += sizeof(bareq);
1589                 len -= sizeof(bareq);
1590         }
1591 out:
1592         bac->ifbac_len = sizeof(bareq) * count;
1593         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1594         free(outbuf, M_TEMP);
1595         return (error);
1596 }
1597
1598 static int
1599 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1600 {
1601         struct ifbareq *req = arg;
1602         struct bridge_iflist *bif;
1603         struct epoch_tracker et;
1604         int error;
1605
1606         NET_EPOCH_ENTER(et);
1607         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1608         if (bif == NULL) {
1609                 NET_EPOCH_EXIT(et);
1610                 return (ENOENT);
1611         }
1612
1613         /* bridge_rtupdate() may acquire the lock. */
1614         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1615             req->ifba_flags);
1616         NET_EPOCH_EXIT(et);
1617
1618         return (error);
1619 }
1620
1621 static int
1622 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1623 {
1624         struct ifbrparam *param = arg;
1625
1626         sc->sc_brttimeout = param->ifbrp_ctime;
1627         return (0);
1628 }
1629
1630 static int
1631 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1632 {
1633         struct ifbrparam *param = arg;
1634
1635         param->ifbrp_ctime = sc->sc_brttimeout;
1636         return (0);
1637 }
1638
1639 static int
1640 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1641 {
1642         struct ifbareq *req = arg;
1643         int vlan = req->ifba_vlan;
1644
1645         /* Userspace uses '0' to mean 'any vlan' */
1646         if (vlan == 0)
1647                 vlan = DOT1Q_VID_RSVD_IMPL;
1648
1649         return (bridge_rtdaddr(sc, req->ifba_dst, vlan));
1650 }
1651
1652 static int
1653 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1654 {
1655         struct ifbreq *req = arg;
1656
1657         BRIDGE_RT_LOCK(sc);
1658         bridge_rtflush(sc, req->ifbr_ifsflags);
1659         BRIDGE_RT_UNLOCK(sc);
1660
1661         return (0);
1662 }
1663
1664 static int
1665 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1666 {
1667         struct ifbrparam *param = arg;
1668         struct bstp_state *bs = &sc->sc_stp;
1669
1670         param->ifbrp_prio = bs->bs_bridge_priority;
1671         return (0);
1672 }
1673
1674 static int
1675 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1676 {
1677         struct ifbrparam *param = arg;
1678
1679         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1680 }
1681
1682 static int
1683 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1684 {
1685         struct ifbrparam *param = arg;
1686         struct bstp_state *bs = &sc->sc_stp;
1687
1688         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1689         return (0);
1690 }
1691
1692 static int
1693 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1694 {
1695         struct ifbrparam *param = arg;
1696
1697         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1698 }
1699
1700 static int
1701 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1702 {
1703         struct ifbrparam *param = arg;
1704         struct bstp_state *bs = &sc->sc_stp;
1705
1706         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1707         return (0);
1708 }
1709
1710 static int
1711 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1712 {
1713         struct ifbrparam *param = arg;
1714
1715         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1716 }
1717
1718 static int
1719 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1720 {
1721         struct ifbrparam *param = arg;
1722         struct bstp_state *bs = &sc->sc_stp;
1723
1724         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1725         return (0);
1726 }
1727
1728 static int
1729 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1730 {
1731         struct ifbrparam *param = arg;
1732
1733         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1734 }
1735
1736 static int
1737 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1738 {
1739         struct ifbreq *req = arg;
1740         struct bridge_iflist *bif;
1741
1742         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1743         if (bif == NULL)
1744                 return (ENOENT);
1745
1746         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1747 }
1748
1749 static int
1750 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1751 {
1752         struct ifbreq *req = arg;
1753         struct bridge_iflist *bif;
1754
1755         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1756         if (bif == NULL)
1757                 return (ENOENT);
1758
1759         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1760 }
1761
1762 static int
1763 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1764 {
1765         struct ifbreq *req = arg;
1766         struct bridge_iflist *bif;
1767
1768         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1769         if (bif == NULL)
1770                 return (ENOENT);
1771
1772         bif->bif_addrmax = req->ifbr_addrmax;
1773         return (0);
1774 }
1775
1776 static int
1777 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1778 {
1779         struct ifbreq *req = arg;
1780         struct bridge_iflist *bif = NULL;
1781         struct ifnet *ifs;
1782
1783         ifs = ifunit(req->ifbr_ifsname);
1784         if (ifs == NULL)
1785                 return (ENOENT);
1786
1787         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1788                 if (ifs == bif->bif_ifp)
1789                         return (EBUSY);
1790
1791         if (ifs->if_bridge != NULL)
1792                 return (EBUSY);
1793
1794         switch (ifs->if_type) {
1795                 case IFT_ETHER:
1796                 case IFT_GIF:
1797                 case IFT_L2VLAN:
1798                         break;
1799                 default:
1800                         return (EINVAL);
1801         }
1802
1803         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1804         if (bif == NULL)
1805                 return (ENOMEM);
1806
1807         bif->bif_ifp = ifs;
1808         bif->bif_flags = IFBIF_SPAN;
1809
1810         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1811
1812         return (0);
1813 }
1814
1815 static int
1816 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1817 {
1818         struct ifbreq *req = arg;
1819         struct bridge_iflist *bif;
1820         struct ifnet *ifs;
1821
1822         ifs = ifunit(req->ifbr_ifsname);
1823         if (ifs == NULL)
1824                 return (ENOENT);
1825
1826         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1827                 if (ifs == bif->bif_ifp)
1828                         break;
1829
1830         if (bif == NULL)
1831                 return (ENOENT);
1832
1833         bridge_delete_span(sc, bif);
1834
1835         return (0);
1836 }
1837
1838 static int
1839 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1840 {
1841         struct ifbropreq *req = arg;
1842         struct bstp_state *bs = &sc->sc_stp;
1843         struct bstp_port *root_port;
1844
1845         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1846         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1847         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1848
1849         root_port = bs->bs_root_port;
1850         if (root_port == NULL)
1851                 req->ifbop_root_port = 0;
1852         else
1853                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1854
1855         req->ifbop_holdcount = bs->bs_txholdcount;
1856         req->ifbop_priority = bs->bs_bridge_priority;
1857         req->ifbop_protocol = bs->bs_protover;
1858         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1859         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1860         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1861         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1862         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1863         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1864
1865         return (0);
1866 }
1867
1868 static int
1869 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1870 {
1871         struct ifbrparam *param = arg;
1872
1873         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1874         return (0);
1875 }
1876
1877 static int
1878 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1879 {
1880         struct ifbpstpconf *bifstp = arg;
1881         struct bridge_iflist *bif;
1882         struct bstp_port *bp;
1883         struct ifbpstpreq bpreq;
1884         char *buf, *outbuf;
1885         int count, buflen, len, error = 0;
1886
1887         count = 0;
1888         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1889                 if ((bif->bif_flags & IFBIF_STP) != 0)
1890                         count++;
1891         }
1892
1893         buflen = sizeof(bpreq) * count;
1894         if (bifstp->ifbpstp_len == 0) {
1895                 bifstp->ifbpstp_len = buflen;
1896                 return (0);
1897         }
1898
1899         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1900         if (outbuf == NULL)
1901                 return (ENOMEM);
1902
1903         count = 0;
1904         buf = outbuf;
1905         len = min(bifstp->ifbpstp_len, buflen);
1906         bzero(&bpreq, sizeof(bpreq));
1907         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1908                 if (len < sizeof(bpreq))
1909                         break;
1910
1911                 if ((bif->bif_flags & IFBIF_STP) == 0)
1912                         continue;
1913
1914                 bp = &bif->bif_stp;
1915                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1916                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1917                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1918                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1919                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1920                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1921
1922                 memcpy(buf, &bpreq, sizeof(bpreq));
1923                 count++;
1924                 buf += sizeof(bpreq);
1925                 len -= sizeof(bpreq);
1926         }
1927
1928         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1929         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1930         free(outbuf, M_TEMP);
1931         return (error);
1932 }
1933
1934 static int
1935 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1936 {
1937         struct ifbrparam *param = arg;
1938
1939         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1940 }
1941
1942 static int
1943 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1944 {
1945         struct ifbrparam *param = arg;
1946
1947         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1948 }
1949
1950 /*
1951  * bridge_ifdetach:
1952  *
1953  *      Detach an interface from a bridge.  Called when a member
1954  *      interface is detaching.
1955  */
1956 static void
1957 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1958 {
1959         struct bridge_softc *sc = ifp->if_bridge;
1960         struct bridge_iflist *bif;
1961
1962         if (ifp->if_flags & IFF_RENAMING)
1963                 return;
1964         if (V_bridge_cloner == NULL) {
1965                 /*
1966                  * This detach handler can be called after
1967                  * vnet_bridge_uninit().  Just return in that case.
1968                  */
1969                 return;
1970         }
1971         /* Check if the interface is a bridge member */
1972         if (sc != NULL) {
1973                 BRIDGE_LOCK(sc);
1974
1975                 bif = bridge_lookup_member_if(sc, ifp);
1976                 if (bif != NULL)
1977                         bridge_delete_member(sc, bif, 1);
1978
1979                 BRIDGE_UNLOCK(sc);
1980                 return;
1981         }
1982
1983         /* Check if the interface is a span port */
1984         BRIDGE_LIST_LOCK();
1985         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1986                 BRIDGE_LOCK(sc);
1987                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1988                         if (ifp == bif->bif_ifp) {
1989                                 bridge_delete_span(sc, bif);
1990                                 break;
1991                         }
1992
1993                 BRIDGE_UNLOCK(sc);
1994         }
1995         BRIDGE_LIST_UNLOCK();
1996 }
1997
1998 /*
1999  * bridge_init:
2000  *
2001  *      Initialize a bridge interface.
2002  */
2003 static void
2004 bridge_init(void *xsc)
2005 {
2006         struct bridge_softc *sc = (struct bridge_softc *)xsc;
2007         struct ifnet *ifp = sc->sc_ifp;
2008
2009         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
2010                 return;
2011
2012         BRIDGE_LOCK(sc);
2013         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2014             bridge_timer, sc);
2015
2016         ifp->if_drv_flags |= IFF_DRV_RUNNING;
2017         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
2018
2019         BRIDGE_UNLOCK(sc);
2020 }
2021
2022 /*
2023  * bridge_stop:
2024  *
2025  *      Stop the bridge interface.
2026  */
2027 static void
2028 bridge_stop(struct ifnet *ifp, int disable)
2029 {
2030         struct bridge_softc *sc = ifp->if_softc;
2031
2032         BRIDGE_LOCK_ASSERT(sc);
2033
2034         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2035                 return;
2036
2037         BRIDGE_RT_LOCK(sc);
2038         callout_stop(&sc->sc_brcallout);
2039
2040         bstp_stop(&sc->sc_stp);
2041
2042         bridge_rtflush(sc, IFBF_FLUSHDYN);
2043         BRIDGE_RT_UNLOCK(sc);
2044
2045         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2046 }
2047
2048 /*
2049  * bridge_enqueue:
2050  *
2051  *      Enqueue a packet on a bridge member interface.
2052  *
2053  */
2054 static int
2055 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2056 {
2057         int len, err = 0;
2058         short mflags;
2059         struct mbuf *m0;
2060
2061         /* We may be sending a fragment so traverse the mbuf */
2062         for (; m; m = m0) {
2063                 m0 = m->m_nextpkt;
2064                 m->m_nextpkt = NULL;
2065                 len = m->m_pkthdr.len;
2066                 mflags = m->m_flags;
2067
2068                 /*
2069                  * If underlying interface can not do VLAN tag insertion itself
2070                  * then attach a packet tag that holds it.
2071                  */
2072                 if ((m->m_flags & M_VLANTAG) &&
2073                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2074                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2075                         if (m == NULL) {
2076                                 if_printf(dst_ifp,
2077                                     "unable to prepend VLAN header\n");
2078                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2079                                 continue;
2080                         }
2081                         m->m_flags &= ~M_VLANTAG;
2082                 }
2083
2084                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2085                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2086                         int n;
2087
2088                         for (m = m0, n = 1; m != NULL; m = m0, n++) {
2089                                 m0 = m->m_nextpkt;
2090                                 m_freem(m);
2091                         }
2092                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, n);
2093                         break;
2094                 }
2095
2096                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2097                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2098                 if (mflags & M_MCAST)
2099                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2100         }
2101
2102         return (err);
2103 }
2104
2105 /*
2106  * bridge_dummynet:
2107  *
2108  *      Receive a queued packet from dummynet and pass it on to the output
2109  *      interface.
2110  *
2111  *      The mbuf has the Ethernet header already attached.
2112  */
2113 static void
2114 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2115 {
2116         struct bridge_softc *sc;
2117
2118         sc = ifp->if_bridge;
2119
2120         /*
2121          * The packet didnt originate from a member interface. This should only
2122          * ever happen if a member interface is removed while packets are
2123          * queued for it.
2124          */
2125         if (sc == NULL) {
2126                 m_freem(m);
2127                 return;
2128         }
2129
2130         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2131 #ifdef INET6
2132             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2133 #endif
2134             ) {
2135                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2136                         return;
2137                 if (m == NULL)
2138                         return;
2139         }
2140
2141         bridge_enqueue(sc, ifp, m);
2142 }
2143
2144 /*
2145  * bridge_output:
2146  *
2147  *      Send output from a bridge member interface.  This
2148  *      performs the bridging function for locally originated
2149  *      packets.
2150  *
2151  *      The mbuf has the Ethernet header already attached.  We must
2152  *      enqueue or free the mbuf before returning.
2153  */
2154 static int
2155 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2156     struct rtentry *rt)
2157 {
2158         struct ether_header *eh;
2159         struct ifnet *bifp, *dst_if;
2160         struct bridge_softc *sc;
2161         uint16_t vlan;
2162
2163         NET_EPOCH_ASSERT();
2164
2165         if (m->m_len < ETHER_HDR_LEN) {
2166                 m = m_pullup(m, ETHER_HDR_LEN);
2167                 if (m == NULL)
2168                         return (0);
2169         }
2170
2171         eh = mtod(m, struct ether_header *);
2172         sc = ifp->if_bridge;
2173         vlan = VLANTAGOF(m);
2174
2175         bifp = sc->sc_ifp;
2176
2177         /*
2178          * If bridge is down, but the original output interface is up,
2179          * go ahead and send out that interface.  Otherwise, the packet
2180          * is dropped below.
2181          */
2182         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2183                 dst_if = ifp;
2184                 goto sendunicast;
2185         }
2186
2187         /*
2188          * If the packet is a multicast, or we don't know a better way to
2189          * get there, send to all interfaces.
2190          */
2191         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2192                 dst_if = NULL;
2193         else
2194                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2195         /* Tap any traffic not passing back out the originating interface */
2196         if (dst_if != ifp)
2197                 ETHER_BPF_MTAP(bifp, m);
2198         if (dst_if == NULL) {
2199                 struct bridge_iflist *bif;
2200                 struct mbuf *mc;
2201                 int used = 0;
2202
2203                 bridge_span(sc, m);
2204
2205                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2206                         dst_if = bif->bif_ifp;
2207
2208                         if (dst_if->if_type == IFT_GIF)
2209                                 continue;
2210                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2211                                 continue;
2212
2213                         /*
2214                          * If this is not the original output interface,
2215                          * and the interface is participating in spanning
2216                          * tree, make sure the port is in a state that
2217                          * allows forwarding.
2218                          */
2219                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2220                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2221                                 continue;
2222
2223                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2224                                 used = 1;
2225                                 mc = m;
2226                         } else {
2227                                 mc = m_dup(m, M_NOWAIT);
2228                                 if (mc == NULL) {
2229                                         if_inc_counter(bifp, IFCOUNTER_OERRORS, 1);
2230                                         continue;
2231                                 }
2232                         }
2233
2234                         bridge_enqueue(sc, dst_if, mc);
2235                 }
2236                 if (used == 0)
2237                         m_freem(m);
2238                 return (0);
2239         }
2240
2241 sendunicast:
2242         /*
2243          * XXX Spanning tree consideration here?
2244          */
2245
2246         bridge_span(sc, m);
2247         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2248                 m_freem(m);
2249                 return (0);
2250         }
2251
2252         bridge_enqueue(sc, dst_if, m);
2253         return (0);
2254 }
2255
2256 /*
2257  * bridge_transmit:
2258  *
2259  *      Do output on a bridge.
2260  *
2261  */
2262 static int
2263 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2264 {
2265         struct bridge_softc *sc;
2266         struct ether_header *eh;
2267         struct ifnet *dst_if;
2268         int error = 0;
2269
2270         sc = ifp->if_softc;
2271
2272         ETHER_BPF_MTAP(ifp, m);
2273
2274         eh = mtod(m, struct ether_header *);
2275
2276         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2277             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, DOT1Q_VID_NULL)) !=
2278             NULL) {
2279                 error = bridge_enqueue(sc, dst_if, m);
2280         } else
2281                 bridge_broadcast(sc, ifp, m, 0);
2282
2283         return (error);
2284 }
2285
2286 #ifdef ALTQ
2287 static void
2288 bridge_altq_start(if_t ifp)
2289 {
2290         struct ifaltq *ifq = &ifp->if_snd;
2291         struct mbuf *m;
2292
2293         IFQ_LOCK(ifq);
2294         IFQ_DEQUEUE_NOLOCK(ifq, m);
2295         while (m != NULL) {
2296                 bridge_transmit(ifp, m);
2297                 IFQ_DEQUEUE_NOLOCK(ifq, m);
2298         }
2299         IFQ_UNLOCK(ifq);
2300 }
2301
2302 static int
2303 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2304 {
2305         int err;
2306
2307         if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2308                 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2309                 if (err == 0)
2310                         bridge_altq_start(ifp);
2311         } else
2312                 err = bridge_transmit(ifp, m);
2313
2314         return (err);
2315 }
2316 #endif  /* ALTQ */
2317
2318 /*
2319  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2320  */
2321 static void
2322 bridge_qflush(struct ifnet *ifp __unused)
2323 {
2324 }
2325
2326 /*
2327  * bridge_forward:
2328  *
2329  *      The forwarding function of the bridge.
2330  *
2331  *      NOTE: Releases the lock on return.
2332  */
2333 static void
2334 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2335     struct mbuf *m)
2336 {
2337         struct bridge_iflist *dbif;
2338         struct ifnet *src_if, *dst_if, *ifp;
2339         struct ether_header *eh;
2340         uint16_t vlan;
2341         uint8_t *dst;
2342         int error;
2343
2344         NET_EPOCH_ASSERT();
2345
2346         src_if = m->m_pkthdr.rcvif;
2347         ifp = sc->sc_ifp;
2348
2349         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2350         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2351         vlan = VLANTAGOF(m);
2352
2353         if ((sbif->bif_flags & IFBIF_STP) &&
2354             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2355                 goto drop;
2356
2357         eh = mtod(m, struct ether_header *);
2358         dst = eh->ether_dhost;
2359
2360         /* If the interface is learning, record the address. */
2361         if (sbif->bif_flags & IFBIF_LEARNING) {
2362                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2363                     sbif, 0, IFBAF_DYNAMIC);
2364                 /*
2365                  * If the interface has addresses limits then deny any source
2366                  * that is not in the cache.
2367                  */
2368                 if (error && sbif->bif_addrmax)
2369                         goto drop;
2370         }
2371
2372         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2373             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2374                 goto drop;
2375
2376 #ifdef DEV_NETMAP
2377         /*
2378          * Hand the packet to netmap only if it wasn't injected by netmap
2379          * itself.
2380          */
2381         if ((m->m_flags & M_BRIDGE_INJECT) == 0 &&
2382             (if_getcapenable(ifp) & IFCAP_NETMAP) != 0) {
2383                 ifp->if_input(ifp, m);
2384                 return;
2385         }
2386         m->m_flags &= ~M_BRIDGE_INJECT;
2387 #endif
2388
2389         /*
2390          * At this point, the port either doesn't participate
2391          * in spanning tree or it is in the forwarding state.
2392          */
2393
2394         /*
2395          * If the packet is unicast, destined for someone on
2396          * "this" side of the bridge, drop it.
2397          */
2398         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2399                 dst_if = bridge_rtlookup(sc, dst, vlan);
2400                 if (src_if == dst_if)
2401                         goto drop;
2402         } else {
2403                 /*
2404                  * Check if its a reserved multicast address, any address
2405                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2406                  * bridge.
2407                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2408                  */
2409                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2410                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2411                     dst[4] == 0x00 && dst[5] <= 0x0f)
2412                         goto drop;
2413
2414                 /* ...forward it to all interfaces. */
2415                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2416                 dst_if = NULL;
2417         }
2418
2419         /*
2420          * If we have a destination interface which is a member of our bridge,
2421          * OR this is a unicast packet, push it through the bpf(4) machinery.
2422          * For broadcast or multicast packets, don't bother because it will
2423          * be reinjected into ether_input. We do this before we pass the packets
2424          * through the pfil(9) framework, as it is possible that pfil(9) will
2425          * drop the packet, or possibly modify it, making it difficult to debug
2426          * firewall issues on the bridge.
2427          */
2428         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2429                 ETHER_BPF_MTAP(ifp, m);
2430
2431         /* run the packet filter */
2432         if (PFIL_HOOKED_IN(V_inet_pfil_head)
2433 #ifdef INET6
2434             || PFIL_HOOKED_IN(V_inet6_pfil_head)
2435 #endif
2436             ) {
2437                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2438                         return;
2439                 if (m == NULL)
2440                         return;
2441         }
2442
2443         if (dst_if == NULL) {
2444                 bridge_broadcast(sc, src_if, m, 1);
2445                 return;
2446         }
2447
2448         /*
2449          * At this point, we're dealing with a unicast frame
2450          * going to a different interface.
2451          */
2452         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2453                 goto drop;
2454
2455         dbif = bridge_lookup_member_if(sc, dst_if);
2456         if (dbif == NULL)
2457                 /* Not a member of the bridge (anymore?) */
2458                 goto drop;
2459
2460         /* Private segments can not talk to each other */
2461         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2462                 goto drop;
2463
2464         if ((dbif->bif_flags & IFBIF_STP) &&
2465             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2466                 goto drop;
2467
2468         if (PFIL_HOOKED_OUT(V_inet_pfil_head)
2469 #ifdef INET6
2470             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2471 #endif
2472             ) {
2473                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2474                         return;
2475                 if (m == NULL)
2476                         return;
2477         }
2478
2479         bridge_enqueue(sc, dst_if, m);
2480         return;
2481
2482 drop:
2483         m_freem(m);
2484 }
2485
2486 /*
2487  * bridge_input:
2488  *
2489  *      Receive input from a member interface.  Queue the packet for
2490  *      bridging if it is not for us.
2491  */
2492 static struct mbuf *
2493 bridge_input(struct ifnet *ifp, struct mbuf *m)
2494 {
2495         struct bridge_softc *sc;
2496         struct bridge_iflist *bif, *bif2;
2497         struct ifnet *bifp;
2498         struct ether_header *eh;
2499         struct mbuf *mc, *mc2;
2500         uint16_t vlan;
2501         int error;
2502
2503         NET_EPOCH_ASSERT();
2504
2505         eh = mtod(m, struct ether_header *);
2506         vlan = VLANTAGOF(m);
2507
2508         sc = ifp->if_bridge;
2509         if (sc == NULL) {
2510                 /*
2511                  * This packet originated from the bridge itself, so it must
2512                  * have been transmitted by netmap.  Derive the "source"
2513                  * interface from the source address and drop the packet if the
2514                  * source address isn't known.
2515                  */
2516                 KASSERT((m->m_flags & M_BRIDGE_INJECT) != 0,
2517                     ("%s: ifnet %p missing a bridge softc", __func__, ifp));
2518                 sc = if_getsoftc(ifp);
2519                 ifp = bridge_rtlookup(sc, eh->ether_shost, vlan);
2520                 if (ifp == NULL) {
2521                         if_inc_counter(sc->sc_ifp, IFCOUNTER_IERRORS, 1);
2522                         m_freem(m);
2523                         return (NULL);
2524                 }
2525                 m->m_pkthdr.rcvif = ifp;
2526         }
2527         bifp = sc->sc_ifp;
2528         if ((bifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2529                 return (m);
2530
2531         /*
2532          * Implement support for bridge monitoring. If this flag has been
2533          * set on this interface, discard the packet once we push it through
2534          * the bpf(4) machinery, but before we do, increment the byte and
2535          * packet counters associated with this interface.
2536          */
2537         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2538                 m->m_pkthdr.rcvif  = bifp;
2539                 ETHER_BPF_MTAP(bifp, m);
2540                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2541                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2542                 m_freem(m);
2543                 return (NULL);
2544         }
2545         bif = bridge_lookup_member_if(sc, ifp);
2546         if (bif == NULL) {
2547                 return (m);
2548         }
2549
2550         bridge_span(sc, m);
2551
2552         if (m->m_flags & (M_BCAST|M_MCAST)) {
2553                 /* Tap off 802.1D packets; they do not get forwarded. */
2554                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2555                     ETHER_ADDR_LEN) == 0) {
2556                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2557                         return (NULL);
2558                 }
2559
2560                 if ((bif->bif_flags & IFBIF_STP) &&
2561                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2562                         return (m);
2563                 }
2564
2565                 /*
2566                  * Make a deep copy of the packet and enqueue the copy
2567                  * for bridge processing; return the original packet for
2568                  * local processing.
2569                  */
2570                 mc = m_dup(m, M_NOWAIT);
2571                 if (mc == NULL) {
2572                         return (m);
2573                 }
2574
2575                 /* Perform the bridge forwarding function with the copy. */
2576                 bridge_forward(sc, bif, mc);
2577
2578 #ifdef DEV_NETMAP
2579                 /*
2580                  * If netmap is enabled and has not already seen this packet,
2581                  * then it will be consumed by bridge_forward().
2582                  */
2583                 if ((if_getcapenable(bifp) & IFCAP_NETMAP) != 0 &&
2584                     (m->m_flags & M_BRIDGE_INJECT) == 0) {
2585                         m_freem(m);
2586                         return (NULL);
2587                 }
2588 #endif
2589
2590                 /*
2591                  * Reinject the mbuf as arriving on the bridge so we have a
2592                  * chance at claiming multicast packets. We can not loop back
2593                  * here from ether_input as a bridge is never a member of a
2594                  * bridge.
2595                  */
2596                 KASSERT(bifp->if_bridge == NULL,
2597                     ("loop created in bridge_input"));
2598                 mc2 = m_dup(m, M_NOWAIT);
2599                 if (mc2 != NULL) {
2600                         /* Keep the layer3 header aligned */
2601                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2602                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2603                 }
2604                 if (mc2 != NULL) {
2605                         mc2->m_pkthdr.rcvif = bifp;
2606                         mc2->m_flags &= ~M_BRIDGE_INJECT;
2607                         sc->sc_if_input(bifp, mc2);
2608                 }
2609
2610                 /* Return the original packet for local processing. */
2611                 return (m);
2612         }
2613
2614         if ((bif->bif_flags & IFBIF_STP) &&
2615             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2616                 return (m);
2617         }
2618
2619 #if defined(INET) || defined(INET6)
2620 #define CARP_CHECK_WE_ARE_DST(iface) \
2621         ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_dhost))
2622 #define CARP_CHECK_WE_ARE_SRC(iface) \
2623         ((iface)->if_carp && (*carp_forus_p)((iface), eh->ether_shost))
2624 #else
2625 #define CARP_CHECK_WE_ARE_DST(iface)    false
2626 #define CARP_CHECK_WE_ARE_SRC(iface)    false
2627 #endif
2628
2629 #ifdef INET6
2630 #define PFIL_HOOKED_INET6       PFIL_HOOKED_IN(V_inet6_pfil_head)
2631 #else
2632 #define PFIL_HOOKED_INET6       false
2633 #endif
2634
2635 #ifdef DEV_NETMAP
2636 #define GRAB_FOR_NETMAP(ifp, m) do {                                    \
2637         if ((if_getcapenable(ifp) & IFCAP_NETMAP) != 0 &&               \
2638             ((m)->m_flags & M_BRIDGE_INJECT) == 0) {                    \
2639                 (ifp)->if_input(ifp, m);                                \
2640                 return (NULL);                                          \
2641         }                                                               \
2642 } while (0)
2643 #else
2644 #define GRAB_FOR_NETMAP(ifp, m)
2645 #endif
2646
2647 #define GRAB_OUR_PACKETS(iface)                                         \
2648         if ((iface)->if_type == IFT_GIF)                                \
2649                 continue;                                               \
2650         /* It is destined for us. */                                    \
2651         if (memcmp(IF_LLADDR(iface), eh->ether_dhost, ETHER_ADDR_LEN) == 0 || \
2652             CARP_CHECK_WE_ARE_DST(iface)) {                             \
2653                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2654                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2655                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2656                         if (error && bif->bif_addrmax) {                \
2657                                 m_freem(m);                             \
2658                                 return (NULL);                          \
2659                         }                                               \
2660                 }                                                       \
2661                 m->m_pkthdr.rcvif = iface;                              \
2662                 if ((iface) == ifp) {                                   \
2663                         /* Skip bridge processing... src == dest */     \
2664                         return (m);                                     \
2665                 }                                                       \
2666                 /* It's passing over or to the bridge, locally. */      \
2667                 ETHER_BPF_MTAP(bifp, m);                                \
2668                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);            \
2669                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);\
2670                 /* Hand the packet over to netmap if necessary. */      \
2671                 GRAB_FOR_NETMAP(bifp, m);                               \
2672                 /* Filter on the physical interface. */                 \
2673                 if (V_pfil_local_phys && (PFIL_HOOKED_IN(V_inet_pfil_head) || \
2674                     PFIL_HOOKED_INET6)) {                               \
2675                         if (bridge_pfil(&m, NULL, ifp,                  \
2676                             PFIL_IN) != 0 || m == NULL) {               \
2677                                 return (NULL);                          \
2678                         }                                               \
2679                 }                                                       \
2680                 if ((iface) != bifp)                                    \
2681                         ETHER_BPF_MTAP(iface, m);                       \
2682                 return (m);                                             \
2683         }                                                               \
2684                                                                         \
2685         /* We just received a packet that we sent out. */               \
2686         if (memcmp(IF_LLADDR(iface), eh->ether_shost, ETHER_ADDR_LEN) == 0 || \
2687             CARP_CHECK_WE_ARE_SRC(iface)) {                             \
2688                 m_freem(m);                                             \
2689                 return (NULL);                                          \
2690         }
2691
2692         /*
2693          * Unicast.  Make sure it's not for the bridge.
2694          */
2695         do { GRAB_OUR_PACKETS(bifp) } while (0);
2696
2697         /*
2698          * Give a chance for ifp at first priority. This will help when the
2699          * packet comes through the interface like VLAN's with the same MACs
2700          * on several interfaces from the same bridge. This also will save
2701          * some CPU cycles in case the destination interface and the input
2702          * interface (eq ifp) are the same.
2703          */
2704         do { GRAB_OUR_PACKETS(ifp) } while (0);
2705
2706         /* Now check the all bridge members. */
2707         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2708                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2709         }
2710
2711 #undef CARP_CHECK_WE_ARE_DST
2712 #undef CARP_CHECK_WE_ARE_SRC
2713 #undef PFIL_HOOKED_INET6
2714 #undef GRAB_FOR_NETMAP
2715 #undef GRAB_OUR_PACKETS
2716
2717         /* Perform the bridge forwarding function. */
2718         bridge_forward(sc, bif, m);
2719
2720         return (NULL);
2721 }
2722
2723 /*
2724  * Inject a packet back into the host ethernet stack.  This will generally only
2725  * be used by netmap when an application writes to the host TX ring.  The
2726  * M_BRIDGE_INJECT flag ensures that the packet is re-routed to the bridge
2727  * interface after ethernet processing.
2728  */
2729 static void
2730 bridge_inject(struct ifnet *ifp, struct mbuf *m)
2731 {
2732         struct bridge_softc *sc;
2733
2734         KASSERT((if_getcapenable(ifp) & IFCAP_NETMAP) != 0,
2735             ("%s: iface %s is not running in netmap mode",
2736             __func__, if_name(ifp)));
2737         KASSERT((m->m_flags & M_BRIDGE_INJECT) == 0,
2738             ("%s: mbuf %p has M_BRIDGE_INJECT set", __func__, m));
2739
2740         m->m_flags |= M_BRIDGE_INJECT;
2741         sc = if_getsoftc(ifp);
2742         sc->sc_if_input(ifp, m);
2743 }
2744
2745 /*
2746  * bridge_broadcast:
2747  *
2748  *      Send a frame to all interfaces that are members of
2749  *      the bridge, except for the one on which the packet
2750  *      arrived.
2751  *
2752  *      NOTE: Releases the lock on return.
2753  */
2754 static void
2755 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2756     struct mbuf *m, int runfilt)
2757 {
2758         struct bridge_iflist *dbif, *sbif;
2759         struct mbuf *mc;
2760         struct ifnet *dst_if;
2761         int used = 0, i;
2762
2763         NET_EPOCH_ASSERT();
2764
2765         sbif = bridge_lookup_member_if(sc, src_if);
2766
2767         /* Filter on the bridge interface before broadcasting */
2768         if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2769 #ifdef INET6
2770             || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2771 #endif
2772             )) {
2773                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2774                         return;
2775                 if (m == NULL)
2776                         return;
2777         }
2778
2779         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2780                 dst_if = dbif->bif_ifp;
2781                 if (dst_if == src_if)
2782                         continue;
2783
2784                 /* Private segments can not talk to each other */
2785                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2786                         continue;
2787
2788                 if ((dbif->bif_flags & IFBIF_STP) &&
2789                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2790                         continue;
2791
2792                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2793                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2794                         continue;
2795
2796                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2797                         continue;
2798
2799                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2800                         mc = m;
2801                         used = 1;
2802                 } else {
2803                         mc = m_dup(m, M_NOWAIT);
2804                         if (mc == NULL) {
2805                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2806                                 continue;
2807                         }
2808                 }
2809
2810                 /*
2811                  * Filter on the output interface. Pass a NULL bridge interface
2812                  * pointer so we do not redundantly filter on the bridge for
2813                  * each interface we broadcast on.
2814                  */
2815                 if (runfilt && (PFIL_HOOKED_OUT(V_inet_pfil_head)
2816 #ifdef INET6
2817                     || PFIL_HOOKED_OUT(V_inet6_pfil_head)
2818 #endif
2819                     )) {
2820                         if (used == 0) {
2821                                 /* Keep the layer3 header aligned */
2822                                 i = min(mc->m_pkthdr.len, max_protohdr);
2823                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2824                                 if (mc == NULL) {
2825                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2826                                         continue;
2827                                 }
2828                         }
2829                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2830                                 continue;
2831                         if (mc == NULL)
2832                                 continue;
2833                 }
2834
2835                 bridge_enqueue(sc, dst_if, mc);
2836         }
2837         if (used == 0)
2838                 m_freem(m);
2839 }
2840
2841 /*
2842  * bridge_span:
2843  *
2844  *      Duplicate a packet out one or more interfaces that are in span mode,
2845  *      the original mbuf is unmodified.
2846  */
2847 static void
2848 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2849 {
2850         struct bridge_iflist *bif;
2851         struct ifnet *dst_if;
2852         struct mbuf *mc;
2853
2854         NET_EPOCH_ASSERT();
2855
2856         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2857                 return;
2858
2859         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2860                 dst_if = bif->bif_ifp;
2861
2862                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2863                         continue;
2864
2865                 mc = m_dup(m, M_NOWAIT);
2866                 if (mc == NULL) {
2867                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2868                         continue;
2869                 }
2870
2871                 bridge_enqueue(sc, dst_if, mc);
2872         }
2873 }
2874
2875 /*
2876  * bridge_rtupdate:
2877  *
2878  *      Add a bridge routing entry.
2879  */
2880 static int
2881 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2882     struct bridge_iflist *bif, int setflags, uint8_t flags)
2883 {
2884         struct bridge_rtnode *brt;
2885         struct bridge_iflist *obif;
2886         int error;
2887
2888         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2889
2890         /* Check the source address is valid and not multicast. */
2891         if (ETHER_IS_MULTICAST(dst) ||
2892             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2893              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2894                 return (EINVAL);
2895
2896         /*
2897          * A route for this destination might already exist.  If so,
2898          * update it, otherwise create a new one.
2899          */
2900         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2901                 BRIDGE_RT_LOCK(sc);
2902
2903                 /* Check again, now that we have the lock. There could have
2904                  * been a race and we only want to insert this once. */
2905                 if (bridge_rtnode_lookup(sc, dst, vlan) != NULL) {
2906                         BRIDGE_RT_UNLOCK(sc);
2907                         return (0);
2908                 }
2909
2910                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2911                         sc->sc_brtexceeded++;
2912                         BRIDGE_RT_UNLOCK(sc);
2913                         return (ENOSPC);
2914                 }
2915                 /* Check per interface address limits (if enabled) */
2916                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2917                         bif->bif_addrexceeded++;
2918                         BRIDGE_RT_UNLOCK(sc);
2919                         return (ENOSPC);
2920                 }
2921
2922                 /*
2923                  * Allocate a new bridge forwarding node, and
2924                  * initialize the expiration time and Ethernet
2925                  * address.
2926                  */
2927                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2928                 if (brt == NULL) {
2929                         BRIDGE_RT_UNLOCK(sc);
2930                         return (ENOMEM);
2931                 }
2932                 brt->brt_vnet = curvnet;
2933
2934                 if (bif->bif_flags & IFBIF_STICKY)
2935                         brt->brt_flags = IFBAF_STICKY;
2936                 else
2937                         brt->brt_flags = IFBAF_DYNAMIC;
2938
2939                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2940                 brt->brt_vlan = vlan;
2941
2942                 brt->brt_dst = bif;
2943                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2944                         uma_zfree(V_bridge_rtnode_zone, brt);
2945                         BRIDGE_RT_UNLOCK(sc);
2946                         return (error);
2947                 }
2948                 bif->bif_addrcnt++;
2949
2950                 BRIDGE_RT_UNLOCK(sc);
2951         }
2952
2953         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2954             (obif = brt->brt_dst) != bif) {
2955                 MPASS(obif != NULL);
2956
2957                 BRIDGE_RT_LOCK(sc);
2958                 brt->brt_dst->bif_addrcnt--;
2959                 brt->brt_dst = bif;
2960                 brt->brt_dst->bif_addrcnt++;
2961                 BRIDGE_RT_UNLOCK(sc);
2962
2963                 if (V_log_mac_flap &&
2964                     ppsratecheck(&V_log_last, &V_log_count, V_log_interval)) {
2965                         log(LOG_NOTICE,
2966                             "%s: mac address %6D vlan %d moved from %s to %s\n",
2967                             sc->sc_ifp->if_xname,
2968                             &brt->brt_addr[0], ":",
2969                             brt->brt_vlan,
2970                             obif->bif_ifp->if_xname,
2971                             bif->bif_ifp->if_xname);
2972                 }
2973         }
2974
2975         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2976                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2977         if (setflags)
2978                 brt->brt_flags = flags;
2979
2980         return (0);
2981 }
2982
2983 /*
2984  * bridge_rtlookup:
2985  *
2986  *      Lookup the destination interface for an address.
2987  */
2988 static struct ifnet *
2989 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2990 {
2991         struct bridge_rtnode *brt;
2992
2993         NET_EPOCH_ASSERT();
2994
2995         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2996                 return (NULL);
2997
2998         return (brt->brt_ifp);
2999 }
3000
3001 /*
3002  * bridge_rttrim:
3003  *
3004  *      Trim the routine table so that we have a number
3005  *      of routing entries less than or equal to the
3006  *      maximum number.
3007  */
3008 static void
3009 bridge_rttrim(struct bridge_softc *sc)
3010 {
3011         struct bridge_rtnode *brt, *nbrt;
3012
3013         NET_EPOCH_ASSERT();
3014         BRIDGE_RT_LOCK_ASSERT(sc);
3015
3016         /* Make sure we actually need to do this. */
3017         if (sc->sc_brtcnt <= sc->sc_brtmax)
3018                 return;
3019
3020         /* Force an aging cycle; this might trim enough addresses. */
3021         bridge_rtage(sc);
3022         if (sc->sc_brtcnt <= sc->sc_brtmax)
3023                 return;
3024
3025         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3026                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3027                         bridge_rtnode_destroy(sc, brt);
3028                         if (sc->sc_brtcnt <= sc->sc_brtmax)
3029                                 return;
3030                 }
3031         }
3032 }
3033
3034 /*
3035  * bridge_timer:
3036  *
3037  *      Aging timer for the bridge.
3038  */
3039 static void
3040 bridge_timer(void *arg)
3041 {
3042         struct bridge_softc *sc = arg;
3043
3044         BRIDGE_RT_LOCK_ASSERT(sc);
3045
3046         /* Destruction of rtnodes requires a proper vnet context */
3047         CURVNET_SET(sc->sc_ifp->if_vnet);
3048         bridge_rtage(sc);
3049
3050         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
3051                 callout_reset(&sc->sc_brcallout,
3052                     bridge_rtable_prune_period * hz, bridge_timer, sc);
3053         CURVNET_RESTORE();
3054 }
3055
3056 /*
3057  * bridge_rtage:
3058  *
3059  *      Perform an aging cycle.
3060  */
3061 static void
3062 bridge_rtage(struct bridge_softc *sc)
3063 {
3064         struct bridge_rtnode *brt, *nbrt;
3065
3066         BRIDGE_RT_LOCK_ASSERT(sc);
3067
3068         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3069                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
3070                         if (time_uptime >= brt->brt_expire)
3071                                 bridge_rtnode_destroy(sc, brt);
3072                 }
3073         }
3074 }
3075
3076 /*
3077  * bridge_rtflush:
3078  *
3079  *      Remove all dynamic addresses from the bridge.
3080  */
3081 static void
3082 bridge_rtflush(struct bridge_softc *sc, int full)
3083 {
3084         struct bridge_rtnode *brt, *nbrt;
3085
3086         BRIDGE_RT_LOCK_ASSERT(sc);
3087
3088         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3089                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3090                         bridge_rtnode_destroy(sc, brt);
3091         }
3092 }
3093
3094 /*
3095  * bridge_rtdaddr:
3096  *
3097  *      Remove an address from the table.
3098  */
3099 static int
3100 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3101 {
3102         struct bridge_rtnode *brt;
3103         int found = 0;
3104
3105         BRIDGE_RT_LOCK(sc);
3106
3107         /*
3108          * If vlan is DOT1Q_VID_RSVD_IMPL then we want to delete for all vlans
3109          * so the lookup may return more than one.
3110          */
3111         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3112                 bridge_rtnode_destroy(sc, brt);
3113                 found = 1;
3114         }
3115
3116         BRIDGE_RT_UNLOCK(sc);
3117
3118         return (found ? 0 : ENOENT);
3119 }
3120
3121 /*
3122  * bridge_rtdelete:
3123  *
3124  *      Delete routes to a speicifc member interface.
3125  */
3126 static void
3127 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3128 {
3129         struct bridge_rtnode *brt, *nbrt;
3130
3131         BRIDGE_RT_LOCK_ASSERT(sc);
3132
3133         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3134                 if (brt->brt_ifp == ifp && (full ||
3135                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3136                         bridge_rtnode_destroy(sc, brt);
3137         }
3138 }
3139
3140 /*
3141  * bridge_rtable_init:
3142  *
3143  *      Initialize the route table for this bridge.
3144  */
3145 static void
3146 bridge_rtable_init(struct bridge_softc *sc)
3147 {
3148         int i;
3149
3150         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3151             M_DEVBUF, M_WAITOK);
3152
3153         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3154                 CK_LIST_INIT(&sc->sc_rthash[i]);
3155
3156         sc->sc_rthash_key = arc4random();
3157         CK_LIST_INIT(&sc->sc_rtlist);
3158 }
3159
3160 /*
3161  * bridge_rtable_fini:
3162  *
3163  *      Deconstruct the route table for this bridge.
3164  */
3165 static void
3166 bridge_rtable_fini(struct bridge_softc *sc)
3167 {
3168
3169         KASSERT(sc->sc_brtcnt == 0,
3170             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3171         free(sc->sc_rthash, M_DEVBUF);
3172 }
3173
3174 /*
3175  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3176  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3177  */
3178 #define mix(a, b, c)                                                    \
3179 do {                                                                    \
3180         a -= b; a -= c; a ^= (c >> 13);                                 \
3181         b -= c; b -= a; b ^= (a << 8);                                  \
3182         c -= a; c -= b; c ^= (b >> 13);                                 \
3183         a -= b; a -= c; a ^= (c >> 12);                                 \
3184         b -= c; b -= a; b ^= (a << 16);                                 \
3185         c -= a; c -= b; c ^= (b >> 5);                                  \
3186         a -= b; a -= c; a ^= (c >> 3);                                  \
3187         b -= c; b -= a; b ^= (a << 10);                                 \
3188         c -= a; c -= b; c ^= (b >> 15);                                 \
3189 } while (/*CONSTCOND*/0)
3190
3191 static __inline uint32_t
3192 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3193 {
3194         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3195
3196         b += addr[5] << 8;
3197         b += addr[4];
3198         a += addr[3] << 24;
3199         a += addr[2] << 16;
3200         a += addr[1] << 8;
3201         a += addr[0];
3202
3203         mix(a, b, c);
3204
3205         return (c & BRIDGE_RTHASH_MASK);
3206 }
3207
3208 #undef mix
3209
3210 static int
3211 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3212 {
3213         int i, d;
3214
3215         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3216                 d = ((int)a[i]) - ((int)b[i]);
3217         }
3218
3219         return (d);
3220 }
3221
3222 /*
3223  * bridge_rtnode_lookup:
3224  *
3225  *      Look up a bridge route node for the specified destination. Compare the
3226  *      vlan id or if zero then just return the first match.
3227  */
3228 static struct bridge_rtnode *
3229 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3230 {
3231         struct bridge_rtnode *brt;
3232         uint32_t hash;
3233         int dir;
3234
3235         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3236
3237         hash = bridge_rthash(sc, addr);
3238         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3239                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3240                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == DOT1Q_VID_RSVD_IMPL))
3241                         return (brt);
3242                 if (dir > 0)
3243                         return (NULL);
3244         }
3245
3246         return (NULL);
3247 }
3248
3249 /*
3250  * bridge_rtnode_insert:
3251  *
3252  *      Insert the specified bridge node into the route table.  We
3253  *      assume the entry is not already in the table.
3254  */
3255 static int
3256 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3257 {
3258         struct bridge_rtnode *lbrt;
3259         uint32_t hash;
3260         int dir;
3261
3262         BRIDGE_RT_LOCK_ASSERT(sc);
3263
3264         hash = bridge_rthash(sc, brt->brt_addr);
3265
3266         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3267         if (lbrt == NULL) {
3268                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3269                 goto out;
3270         }
3271
3272         do {
3273                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3274                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3275                         return (EEXIST);
3276                 if (dir > 0) {
3277                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3278                         goto out;
3279                 }
3280                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3281                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3282                         goto out;
3283                 }
3284                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3285         } while (lbrt != NULL);
3286
3287 #ifdef DIAGNOSTIC
3288         panic("bridge_rtnode_insert: impossible");
3289 #endif
3290
3291 out:
3292         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3293         sc->sc_brtcnt++;
3294
3295         return (0);
3296 }
3297
3298 static void
3299 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3300 {
3301         struct bridge_rtnode *brt;
3302
3303         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3304
3305         CURVNET_SET(brt->brt_vnet);
3306         uma_zfree(V_bridge_rtnode_zone, brt);
3307         CURVNET_RESTORE();
3308 }
3309
3310 /*
3311  * bridge_rtnode_destroy:
3312  *
3313  *      Destroy a bridge rtnode.
3314  */
3315 static void
3316 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3317 {
3318         BRIDGE_RT_LOCK_ASSERT(sc);
3319
3320         CK_LIST_REMOVE(brt, brt_hash);
3321
3322         CK_LIST_REMOVE(brt, brt_list);
3323         sc->sc_brtcnt--;
3324         brt->brt_dst->bif_addrcnt--;
3325
3326         NET_EPOCH_CALL(bridge_rtnode_destroy_cb, &brt->brt_epoch_ctx);
3327 }
3328
3329 /*
3330  * bridge_rtable_expire:
3331  *
3332  *      Set the expiry time for all routes on an interface.
3333  */
3334 static void
3335 bridge_rtable_expire(struct ifnet *ifp, int age)
3336 {
3337         struct bridge_softc *sc = ifp->if_bridge;
3338         struct bridge_rtnode *brt;
3339
3340         CURVNET_SET(ifp->if_vnet);
3341         BRIDGE_RT_LOCK(sc);
3342
3343         /*
3344          * If the age is zero then flush, otherwise set all the expiry times to
3345          * age for the interface
3346          */
3347         if (age == 0)
3348                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3349         else {
3350                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3351                         /* Cap the expiry time to 'age' */
3352                         if (brt->brt_ifp == ifp &&
3353                             brt->brt_expire > time_uptime + age &&
3354                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3355                                 brt->brt_expire = time_uptime + age;
3356                 }
3357         }
3358         BRIDGE_RT_UNLOCK(sc);
3359         CURVNET_RESTORE();
3360 }
3361
3362 /*
3363  * bridge_state_change:
3364  *
3365  *      Callback from the bridgestp code when a port changes states.
3366  */
3367 static void
3368 bridge_state_change(struct ifnet *ifp, int state)
3369 {
3370         struct bridge_softc *sc = ifp->if_bridge;
3371         static const char *stpstates[] = {
3372                 "disabled",
3373                 "listening",
3374                 "learning",
3375                 "forwarding",
3376                 "blocking",
3377                 "discarding"
3378         };
3379
3380         CURVNET_SET(ifp->if_vnet);
3381         if (V_log_stp)
3382                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3383                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3384         CURVNET_RESTORE();
3385 }
3386
3387 /*
3388  * Send bridge packets through pfil if they are one of the types pfil can deal
3389  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3390  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3391  * that interface.
3392  */
3393 static int
3394 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3395 {
3396         int snap, error, i, hlen;
3397         struct ether_header *eh1, eh2;
3398         struct ip *ip;
3399         struct llc llc1;
3400         u_int16_t ether_type;
3401         pfil_return_t rv;
3402
3403         snap = 0;
3404         error = -1;     /* Default error if not error == 0 */
3405
3406 #if 0
3407         /* we may return with the IP fields swapped, ensure its not shared */
3408         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3409 #endif
3410
3411         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3412                 return (0); /* filtering is disabled */
3413
3414         i = min((*mp)->m_pkthdr.len, max_protohdr);
3415         if ((*mp)->m_len < i) {
3416             *mp = m_pullup(*mp, i);
3417             if (*mp == NULL) {
3418                 printf("%s: m_pullup failed\n", __func__);
3419                 return (-1);
3420             }
3421         }
3422
3423         eh1 = mtod(*mp, struct ether_header *);
3424         ether_type = ntohs(eh1->ether_type);
3425
3426         /*
3427          * Check for SNAP/LLC.
3428          */
3429         if (ether_type < ETHERMTU) {
3430                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3431
3432                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3433                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3434                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3435                     llc2->llc_control == LLC_UI) {
3436                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3437                         snap = 1;
3438                 }
3439         }
3440
3441         /*
3442          * If we're trying to filter bridge traffic, don't look at anything
3443          * other than IP and ARP traffic.  If the filter doesn't understand
3444          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3445          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3446          * but of course we don't have an AppleTalk filter to begin with.
3447          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3448          * ARP traffic.)
3449          */
3450         switch (ether_type) {
3451                 case ETHERTYPE_ARP:
3452                 case ETHERTYPE_REVARP:
3453                         if (V_pfil_ipfw_arp == 0)
3454                                 return (0); /* Automatically pass */
3455                         break;
3456
3457                 case ETHERTYPE_IP:
3458 #ifdef INET6
3459                 case ETHERTYPE_IPV6:
3460 #endif /* INET6 */
3461                         break;
3462                 default:
3463                         /*
3464                          * Check to see if the user wants to pass non-ip
3465                          * packets, these will not be checked by pfil(9) and
3466                          * passed unconditionally so the default is to drop.
3467                          */
3468                         if (V_pfil_onlyip)
3469                                 goto bad;
3470         }
3471
3472         /* Run the packet through pfil before stripping link headers */
3473         if (PFIL_HOOKED_OUT(V_link_pfil_head) && V_pfil_ipfw != 0 &&
3474             dir == PFIL_OUT && ifp != NULL) {
3475                 switch (pfil_mbuf_out(V_link_pfil_head, mp, ifp, NULL)) {
3476                 case PFIL_DROPPED:
3477                         return (EACCES);
3478                 case PFIL_CONSUMED:
3479                         return (0);
3480                 }
3481         }
3482
3483         /* Strip off the Ethernet header and keep a copy. */
3484         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3485         m_adj(*mp, ETHER_HDR_LEN);
3486
3487         /* Strip off snap header, if present */
3488         if (snap) {
3489                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3490                 m_adj(*mp, sizeof(struct llc));
3491         }
3492
3493         /*
3494          * Check the IP header for alignment and errors
3495          */
3496         if (dir == PFIL_IN) {
3497                 switch (ether_type) {
3498                         case ETHERTYPE_IP:
3499                                 error = bridge_ip_checkbasic(mp);
3500                                 break;
3501 #ifdef INET6
3502                         case ETHERTYPE_IPV6:
3503                                 error = bridge_ip6_checkbasic(mp);
3504                                 break;
3505 #endif /* INET6 */
3506                         default:
3507                                 error = 0;
3508                 }
3509                 if (error)
3510                         goto bad;
3511         }
3512
3513         error = 0;
3514
3515         /*
3516          * Run the packet through pfil
3517          */
3518         rv = PFIL_PASS;
3519         switch (ether_type) {
3520         case ETHERTYPE_IP:
3521                 /*
3522                  * Run pfil on the member interface and the bridge, both can
3523                  * be skipped by clearing pfil_member or pfil_bridge.
3524                  *
3525                  * Keep the order:
3526                  *   in_if -> bridge_if -> out_if
3527                  */
3528                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3529                     pfil_mbuf_out(V_inet_pfil_head, mp, bifp, NULL)) !=
3530                     PFIL_PASS)
3531                         break;
3532
3533                 if (V_pfil_member && ifp != NULL) {
3534                         rv = (dir == PFIL_OUT) ?
3535                             pfil_mbuf_out(V_inet_pfil_head, mp, ifp, NULL) :
3536                             pfil_mbuf_in(V_inet_pfil_head, mp, ifp, NULL);
3537                         if (rv != PFIL_PASS)
3538                                 break;
3539                 }
3540
3541                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3542                     pfil_mbuf_in(V_inet_pfil_head, mp, bifp, NULL)) !=
3543                     PFIL_PASS)
3544                         break;
3545
3546                 /* check if we need to fragment the packet */
3547                 /* bridge_fragment generates a mbuf chain of packets */
3548                 /* that already include eth headers */
3549                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3550                         i = (*mp)->m_pkthdr.len;
3551                         if (i > ifp->if_mtu) {
3552                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3553                                             &llc1);
3554                                 return (error);
3555                         }
3556                 }
3557
3558                 /* Recalculate the ip checksum. */
3559                 ip = mtod(*mp, struct ip *);
3560                 hlen = ip->ip_hl << 2;
3561                 if (hlen < sizeof(struct ip))
3562                         goto bad;
3563                 if (hlen > (*mp)->m_len) {
3564                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3565                                 goto bad;
3566                         ip = mtod(*mp, struct ip *);
3567                         if (ip == NULL)
3568                                 goto bad;
3569                 }
3570                 ip->ip_sum = 0;
3571                 if (hlen == sizeof(struct ip))
3572                         ip->ip_sum = in_cksum_hdr(ip);
3573                 else
3574                         ip->ip_sum = in_cksum(*mp, hlen);
3575
3576                 break;
3577 #ifdef INET6
3578         case ETHERTYPE_IPV6:
3579                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL && (rv =
3580                     pfil_mbuf_out(V_inet6_pfil_head, mp, bifp, NULL)) !=
3581                     PFIL_PASS)
3582                         break;
3583
3584                 if (V_pfil_member && ifp != NULL) {
3585                         rv = (dir == PFIL_OUT) ?
3586                             pfil_mbuf_out(V_inet6_pfil_head, mp, ifp, NULL) :
3587                             pfil_mbuf_in(V_inet6_pfil_head, mp, ifp, NULL);
3588                         if (rv != PFIL_PASS)
3589                                 break;
3590                 }
3591
3592                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL && (rv =
3593                     pfil_mbuf_in(V_inet6_pfil_head, mp, bifp, NULL)) !=
3594                     PFIL_PASS)
3595                         break;
3596                 break;
3597 #endif
3598         }
3599
3600         switch (rv) {
3601         case PFIL_CONSUMED:
3602                 return (0);
3603         case PFIL_DROPPED:
3604                 return (EACCES);
3605         default:
3606                 break;
3607         }
3608
3609         error = -1;
3610
3611         /*
3612          * Finally, put everything back the way it was and return
3613          */
3614         if (snap) {
3615                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3616                 if (*mp == NULL)
3617                         return (error);
3618                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3619         }
3620
3621         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3622         if (*mp == NULL)
3623                 return (error);
3624         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3625
3626         return (0);
3627
3628 bad:
3629         m_freem(*mp);
3630         *mp = NULL;
3631         return (error);
3632 }
3633
3634 /*
3635  * Perform basic checks on header size since
3636  * pfil assumes ip_input has already processed
3637  * it for it.  Cut-and-pasted from ip_input.c.
3638  * Given how simple the IPv6 version is,
3639  * does the IPv4 version really need to be
3640  * this complicated?
3641  *
3642  * XXX Should we update ipstat here, or not?
3643  * XXX Right now we update ipstat but not
3644  * XXX csum_counter.
3645  */
3646 static int
3647 bridge_ip_checkbasic(struct mbuf **mp)
3648 {
3649         struct mbuf *m = *mp;
3650         struct ip *ip;
3651         int len, hlen;
3652         u_short sum;
3653
3654         if (*mp == NULL)
3655                 return (-1);
3656
3657         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3658                 if ((m = m_copyup(m, sizeof(struct ip),
3659                         (max_linkhdr + 3) & ~3)) == NULL) {
3660                         /* XXXJRT new stat, please */
3661                         KMOD_IPSTAT_INC(ips_toosmall);
3662                         goto bad;
3663                 }
3664         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3665                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3666                         KMOD_IPSTAT_INC(ips_toosmall);
3667                         goto bad;
3668                 }
3669         }
3670         ip = mtod(m, struct ip *);
3671         if (ip == NULL) goto bad;
3672
3673         if (ip->ip_v != IPVERSION) {
3674                 KMOD_IPSTAT_INC(ips_badvers);
3675                 goto bad;
3676         }
3677         hlen = ip->ip_hl << 2;
3678         if (hlen < sizeof(struct ip)) { /* minimum header length */
3679                 KMOD_IPSTAT_INC(ips_badhlen);
3680                 goto bad;
3681         }
3682         if (hlen > m->m_len) {
3683                 if ((m = m_pullup(m, hlen)) == NULL) {
3684                         KMOD_IPSTAT_INC(ips_badhlen);
3685                         goto bad;
3686                 }
3687                 ip = mtod(m, struct ip *);
3688                 if (ip == NULL) goto bad;
3689         }
3690
3691         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3692                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3693         } else {
3694                 if (hlen == sizeof(struct ip)) {
3695                         sum = in_cksum_hdr(ip);
3696                 } else {
3697                         sum = in_cksum(m, hlen);
3698                 }
3699         }
3700         if (sum) {
3701                 KMOD_IPSTAT_INC(ips_badsum);
3702                 goto bad;
3703         }
3704
3705         /* Retrieve the packet length. */
3706         len = ntohs(ip->ip_len);
3707
3708         /*
3709          * Check for additional length bogosity
3710          */
3711         if (len < hlen) {
3712                 KMOD_IPSTAT_INC(ips_badlen);
3713                 goto bad;
3714         }
3715
3716         /*
3717          * Check that the amount of data in the buffers
3718          * is as at least much as the IP header would have us expect.
3719          * Drop packet if shorter than we expect.
3720          */
3721         if (m->m_pkthdr.len < len) {
3722                 KMOD_IPSTAT_INC(ips_tooshort);
3723                 goto bad;
3724         }
3725
3726         /* Checks out, proceed */
3727         *mp = m;
3728         return (0);
3729
3730 bad:
3731         *mp = m;
3732         return (-1);
3733 }
3734
3735 #ifdef INET6
3736 /*
3737  * Same as above, but for IPv6.
3738  * Cut-and-pasted from ip6_input.c.
3739  * XXX Should we update ip6stat, or not?
3740  */
3741 static int
3742 bridge_ip6_checkbasic(struct mbuf **mp)
3743 {
3744         struct mbuf *m = *mp;
3745         struct ip6_hdr *ip6;
3746
3747         /*
3748          * If the IPv6 header is not aligned, slurp it up into a new
3749          * mbuf with space for link headers, in the event we forward
3750          * it.  Otherwise, if it is aligned, make sure the entire base
3751          * IPv6 header is in the first mbuf of the chain.
3752          */
3753         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3754                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3755                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3756                             (max_linkhdr + 3) & ~3)) == NULL) {
3757                         /* XXXJRT new stat, please */
3758                         IP6STAT_INC(ip6s_toosmall);
3759                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3760                         goto bad;
3761                 }
3762         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3763                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3764                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3765                         IP6STAT_INC(ip6s_toosmall);
3766                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3767                         goto bad;
3768                 }
3769         }
3770
3771         ip6 = mtod(m, struct ip6_hdr *);
3772
3773         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3774                 IP6STAT_INC(ip6s_badvers);
3775                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3776                 goto bad;
3777         }
3778
3779         /* Checks out, proceed */
3780         *mp = m;
3781         return (0);
3782
3783 bad:
3784         *mp = m;
3785         return (-1);
3786 }
3787 #endif /* INET6 */
3788
3789 /*
3790  * bridge_fragment:
3791  *
3792  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3793  */
3794 static int
3795 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3796     int snap, struct llc *llc)
3797 {
3798         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3799         struct ip *ip;
3800         int error = -1;
3801
3802         if (m->m_len < sizeof(struct ip) &&
3803             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3804                 goto dropit;
3805         ip = mtod(m, struct ip *);
3806
3807         m->m_pkthdr.csum_flags |= CSUM_IP;
3808         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3809         if (error)
3810                 goto dropit;
3811
3812         /*
3813          * Walk the chain and re-add the Ethernet header for
3814          * each mbuf packet.
3815          */
3816         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3817                 nextpkt = mcur->m_nextpkt;
3818                 mcur->m_nextpkt = NULL;
3819                 if (snap) {
3820                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3821                         if (mcur == NULL) {
3822                                 error = ENOBUFS;
3823                                 if (mprev != NULL)
3824                                         mprev->m_nextpkt = nextpkt;
3825                                 goto dropit;
3826                         }
3827                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3828                 }
3829
3830                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3831                 if (mcur == NULL) {
3832                         error = ENOBUFS;
3833                         if (mprev != NULL)
3834                                 mprev->m_nextpkt = nextpkt;
3835                         goto dropit;
3836                 }
3837                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3838
3839                 /*
3840                  * The previous two M_PREPEND could have inserted one or two
3841                  * mbufs in front so we have to update the previous packet's
3842                  * m_nextpkt.
3843                  */
3844                 mcur->m_nextpkt = nextpkt;
3845                 if (mprev != NULL)
3846                         mprev->m_nextpkt = mcur;
3847                 else {
3848                         /* The first mbuf in the original chain needs to be
3849                          * updated. */
3850                         *mp = mcur;
3851                 }
3852                 mprev = mcur;
3853         }
3854
3855         KMOD_IPSTAT_INC(ips_fragmented);
3856         return (error);
3857
3858 dropit:
3859         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3860                 m = mcur->m_nextpkt;
3861                 m_freem(mcur);
3862         }
3863         return (error);
3864 }
3865
3866 static void
3867 bridge_linkstate(struct ifnet *ifp)
3868 {
3869         struct bridge_softc *sc = ifp->if_bridge;
3870         struct bridge_iflist *bif;
3871         struct epoch_tracker et;
3872
3873         NET_EPOCH_ENTER(et);
3874
3875         bif = bridge_lookup_member_if(sc, ifp);
3876         if (bif == NULL) {
3877                 NET_EPOCH_EXIT(et);
3878                 return;
3879         }
3880         bridge_linkcheck(sc);
3881
3882         bstp_linkstate(&bif->bif_stp);
3883
3884         NET_EPOCH_EXIT(et);
3885 }
3886
3887 static void
3888 bridge_linkcheck(struct bridge_softc *sc)
3889 {
3890         struct bridge_iflist *bif;
3891         int new_link, hasls;
3892
3893         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3894
3895         new_link = LINK_STATE_DOWN;
3896         hasls = 0;
3897         /* Our link is considered up if at least one of our ports is active */
3898         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3899                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3900                         hasls++;
3901                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3902                         new_link = LINK_STATE_UP;
3903                         break;
3904                 }
3905         }
3906         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3907                 /* If no interfaces support link-state then we default to up */
3908                 new_link = LINK_STATE_UP;
3909         }
3910         if_link_state_change(sc->sc_ifp, new_link);
3911 }