sys/net/if_bridge.c

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