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 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 191         mtx_init(&(_sc)->sc_mtx, "if_bridge", NULL, MTX_DEF);   \
 192         cv_init(&(_sc)->sc_cv, "if_bridge_cv");                 \
 193 } while (0)
 194 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 195         mtx_destroy(&(_sc)->sc_mtx);            \
 196         cv_destroy(&(_sc)->sc_cv);              \
 197 } while (0)
 198 #define BRIDGE_LOCK(_sc)                mtx_lock(&(_sc)->sc_mtx)
 199 #define BRIDGE_UNLOCK(_sc)              mtx_unlock(&(_sc)->sc_mtx)
 200 #define BRIDGE_LOCK_ASSERT(_sc)         mtx_assert(&(_sc)->sc_mtx, MA_OWNED)
 201 #define BRIDGE_UNLOCK_ASSERT(_sc)       mtx_assert(&(_sc)->sc_mtx, MA_NOTOWNED)
 202 #define BRIDGE_LOCK2REF(_sc, _err)      do {    \
 203         mtx_assert(&(_sc)->sc_mtx, MA_OWNED);   \
 204         if ((_sc)->sc_iflist_xcnt > 0)          \
 205                 (_err) = EBUSY;                 \
 206         else                                    \
 207                 (_sc)->sc_iflist_ref++;         \
 208         mtx_unlock(&(_sc)->sc_mtx);             \
 209 } while (0)
 210 #define BRIDGE_UNREF(_sc)               do {                            \
 211         mtx_lock(&(_sc)->sc_mtx);                                       \
 212         (_sc)->sc_iflist_ref--;                                         \
 213         if (((_sc)->sc_iflist_xcnt > 0) && ((_sc)->sc_iflist_ref == 0)) \
 214                 cv_broadcast(&(_sc)->sc_cv);                            \
 215         mtx_unlock(&(_sc)->sc_mtx);                                     \
 216 } while (0)
 217 #define BRIDGE_XLOCK(_sc)               do {            \
 218         mtx_assert(&(_sc)->sc_mtx, MA_OWNED);           \
 219         (_sc)->sc_iflist_xcnt++;                        \
 220         while ((_sc)->sc_iflist_ref > 0)                \
 221                 cv_wait(&(_sc)->sc_cv, &(_sc)->sc_mtx); \
 222 } while (0)
 223 #define BRIDGE_XDROP(_sc)               do {    \
 224         mtx_assert(&(_sc)->sc_mtx, MA_OWNED);   \
 225         (_sc)->sc_iflist_xcnt--;                \
 226 } while (0)
 227
 228 /*
 229  * Bridge interface list entry.
 230  */
 231 struct bridge_iflist {
 232         CK_LIST_ENTRY(bridge_iflist) bif_next;
 233         struct ifnet            *bif_ifp;       /* member if */
 234         struct bstp_port        bif_stp;        /* STP state */
 235         uint32_t                bif_flags;      /* member if flags */
 236         int                     bif_savedcaps;  /* saved capabilities */
 237         uint32_t                bif_addrmax;    /* max # of addresses */
 238         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 239         uint32_t                bif_addrexceeded;/* # of address violations */
 240
 241         struct epoch_context    bif_epoch_ctx;
 242 };
 243
 244 /*
 245  * Bridge route node.
 246  */
 247 struct bridge_rtnode {
 248         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 249         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 250         struct bridge_iflist    *brt_dst;       /* destination if */
 251         unsigned long           brt_expire;     /* expiration time */
 252         uint8_t                 brt_flags;      /* address flags */
 253         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 254         uint16_t                brt_vlan;       /* vlan id */
 255
 256         struct  vnet            *brt_vnet;
 257         struct  epoch_context   brt_epoch_ctx;
 258 };
 259 #define brt_ifp                 brt_dst->bif_ifp
 260
 261 /*
 262  * Software state for each bridge.
 263  */
 264 struct bridge_softc {
 265         struct ifnet            *sc_ifp;        /* make this an interface */
 266         LIST_ENTRY(bridge_softc) sc_list;
 267         struct mtx              sc_mtx;
 268         struct cv               sc_cv;
 269         uint32_t                sc_brtmax;      /* max # of addresses */
 270         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 271         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 272         struct callout          sc_brcallout;   /* bridge callout */
 273         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 274         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 275         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 276         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 277         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 278         uint32_t                sc_rthash_key;  /* key for hash */
 279         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 280         struct bstp_state       sc_stp;         /* STP state */
 281         uint32_t                sc_brtexceeded; /* # of cache drops */
 282         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 283         struct ether_addr       sc_defaddr;     /* Default MAC address */
 284
 285         struct epoch_context    sc_epoch_ctx;
 286 };
 287
 288 VNET_DEFINE_STATIC(struct mtx, bridge_list_mtx);
 289 #define V_bridge_list_mtx       VNET(bridge_list_mtx)
 290 static eventhandler_tag bridge_detach_cookie;
 291
 292 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 293
 294 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 295 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 296
 297 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 298 static void     bridge_clone_destroy(struct ifnet *);
 299
 300 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 301 static void     bridge_mutecaps(struct bridge_softc *);
 302 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 303                     int);
 304 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 305 static void     bridge_init(void *);
 306 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 307 static void     bridge_stop(struct ifnet *, int);
 308 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 309 static void     bridge_qflush(struct ifnet *);
 310 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 311 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 312                     struct rtentry *);
 313 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 314                     struct mbuf *);
 315 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 316
 317 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 318                     struct mbuf *m);
 319
 320 static void     bridge_timer(void *);
 321
 322 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 323                     struct mbuf *, int);
 324 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 325
 326 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 327                     uint16_t, struct bridge_iflist *, int, uint8_t);
 328 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 329                     uint16_t);
 330 static void     bridge_rttrim(struct bridge_softc *);
 331 static void     bridge_rtage(struct bridge_softc *);
 332 static void     bridge_rtflush(struct bridge_softc *, int);
 333 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 334                     uint16_t);
 335
 336 static void     bridge_rtable_init(struct bridge_softc *);
 337 static void     bridge_rtable_fini(struct bridge_softc *);
 338
 339 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 340 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 341                     const uint8_t *, uint16_t);
 342 static int      bridge_rtnode_insert(struct bridge_softc *,
 343                     struct bridge_rtnode *);
 344 static void     bridge_rtnode_destroy(struct bridge_softc *,
 345                     struct bridge_rtnode *);
 346 static void     bridge_rtable_expire(struct ifnet *, int);
 347 static void     bridge_state_change(struct ifnet *, int);
 348
 349 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 350                     const char *name);
 351 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 352                     struct ifnet *ifp);
 353 static void     bridge_delete_member(struct bridge_softc *,
 354                     struct bridge_iflist *, int);
 355 static void     bridge_delete_span(struct bridge_softc *,
 356                     struct bridge_iflist *);
 357
 358 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 381 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 382 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 383 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 384 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 385 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 386 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 387 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 388 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 389 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 390                     int);
 391 static int      bridge_ip_checkbasic(struct mbuf **mp);
 392 #ifdef INET6
 393 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 394 #endif /* INET6 */
 395 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 396                     struct ether_header *, int, struct llc *);
 397 static void     bridge_linkstate(struct ifnet *ifp);
 398 static void     bridge_linkcheck(struct bridge_softc *sc);
 399
 400
 401 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 402 #define VLANTAGOF(_m)   \
 403     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 404
 405 static struct bstp_cb_ops bridge_ops = {
 406         .bcb_state = bridge_state_change,
 407         .bcb_rtage = bridge_rtable_expire
 408 };
 409
 410 SYSCTL_DECL(_net_link);
 411 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 412
 413 /* only pass IP[46] packets when pfil is enabled */
 414 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 415 #define V_pfil_onlyip   VNET(pfil_onlyip)
 416 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 417     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 418     "Only pass IP packets when pfil is enabled");
 419
 420 /* run pfil hooks on the bridge interface */
 421 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 422 #define V_pfil_bridge   VNET(pfil_bridge)
 423 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 424     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 425     "Packet filter on the bridge interface");
 426
 427 /* layer2 filter with ipfw */
 428 VNET_DEFINE_STATIC(int, pfil_ipfw);
 429 #define V_pfil_ipfw     VNET(pfil_ipfw)
 430
 431 /* layer2 ARP filter with ipfw */
 432 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 433 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 434 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 435     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 436     "Filter ARP packets through IPFW layer2");
 437
 438 /* run pfil hooks on the member interface */
 439 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 440 #define V_pfil_member   VNET(pfil_member)
 441 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 442     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 443     "Packet filter on the member interface");
 444
 445 /* run pfil hooks on the physical interface for locally destined packets */
 446 VNET_DEFINE_STATIC(int, pfil_local_phys);
 447 #define V_pfil_local_phys       VNET(pfil_local_phys)
 448 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 449     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 450     "Packet filter on the physical interface for locally destined packets");
 451
 452 /* log STP state changes */
 453 VNET_DEFINE_STATIC(int, log_stp);
 454 #define V_log_stp       VNET(log_stp)
 455 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 456     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 457     "Log STP state changes");
 458
 459 /* share MAC with first bridge member */
 460 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 461 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 462 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 463     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 464     "Inherit MAC address from the first bridge member");
 465
 466 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 467 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 468 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 469     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 470     "Allow overlap of link-local scope "
 471     "zones of a bridge interface and the member interfaces");
 472
 473 struct bridge_control {
 474         int     (*bc_func)(struct bridge_softc *, void *);
 475         int     bc_argsize;
 476         int     bc_flags;
 477 };
 478
 479 #define BC_F_COPYIN             0x01    /* copy arguments in */
 480 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 481 #define BC_F_SUSER              0x04    /* do super-user check */
 482
 483 const struct bridge_control bridge_control_table[] = {
 484         { bridge_ioctl_add,             sizeof(struct ifbreq),
 485           BC_F_COPYIN|BC_F_SUSER },
 486         { bridge_ioctl_del,             sizeof(struct ifbreq),
 487           BC_F_COPYIN|BC_F_SUSER },
 488
 489         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 490           BC_F_COPYIN|BC_F_COPYOUT },
 491         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 492           BC_F_COPYIN|BC_F_SUSER },
 493
 494         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 495           BC_F_COPYIN|BC_F_SUSER },
 496         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 497           BC_F_COPYOUT },
 498
 499         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 500           BC_F_COPYIN|BC_F_COPYOUT },
 501         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 502           BC_F_COPYIN|BC_F_COPYOUT },
 503
 504         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 505           BC_F_COPYIN|BC_F_SUSER },
 506
 507         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 508           BC_F_COPYIN|BC_F_SUSER },
 509         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 510           BC_F_COPYOUT },
 511
 512         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 513           BC_F_COPYIN|BC_F_SUSER },
 514
 515         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 516           BC_F_COPYIN|BC_F_SUSER },
 517
 518         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 519           BC_F_COPYOUT },
 520         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 521           BC_F_COPYIN|BC_F_SUSER },
 522
 523         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 524           BC_F_COPYOUT },
 525         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 526           BC_F_COPYIN|BC_F_SUSER },
 527
 528         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 529           BC_F_COPYOUT },
 530         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 531           BC_F_COPYIN|BC_F_SUSER },
 532
 533         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 534           BC_F_COPYOUT },
 535         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 536           BC_F_COPYIN|BC_F_SUSER },
 537
 538         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 539           BC_F_COPYIN|BC_F_SUSER },
 540
 541         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 542           BC_F_COPYIN|BC_F_SUSER },
 543
 544         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 545           BC_F_COPYIN|BC_F_SUSER },
 546         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 547           BC_F_COPYIN|BC_F_SUSER },
 548
 549         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 550           BC_F_COPYOUT },
 551
 552         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 553           BC_F_COPYOUT },
 554
 555         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 556           BC_F_COPYIN|BC_F_COPYOUT },
 557
 558         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 559           BC_F_COPYIN|BC_F_SUSER },
 560
 561         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 562           BC_F_COPYIN|BC_F_SUSER },
 563
 564         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 565           BC_F_COPYIN|BC_F_SUSER },
 566
 567 };
 568 const int bridge_control_table_size = nitems(bridge_control_table);
 569
 570 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 571 #define V_bridge_list   VNET(bridge_list)
 572 #define BRIDGE_LIST_LOCK_INIT(x)        mtx_init(&V_bridge_list_mtx,    \
 573                                             "if_bridge list", NULL, MTX_DEF)
 574 #define BRIDGE_LIST_LOCK_DESTROY(x)     mtx_destroy(&V_bridge_list_mtx)
 575 #define BRIDGE_LIST_LOCK(x)             mtx_lock(&V_bridge_list_mtx)
 576 #define BRIDGE_LIST_UNLOCK(x)           mtx_unlock(&V_bridge_list_mtx)
 577
 578 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 579 #define V_bridge_cloner VNET(bridge_cloner)
 580
 581 static const char bridge_name[] = "bridge";
 582
 583 static void
 584 vnet_bridge_init(const void *unused __unused)
 585 {
 586
 587         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 588             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 589             UMA_ALIGN_PTR, 0);
 590         BRIDGE_LIST_LOCK_INIT();
 591         LIST_INIT(&V_bridge_list);
 592         V_bridge_cloner = if_clone_simple(bridge_name,
 593             bridge_clone_create, bridge_clone_destroy, 0);
 594 }
 595 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 596     vnet_bridge_init, NULL);
 597
 598 static void
 599 vnet_bridge_uninit(const void *unused __unused)
 600 {
 601
 602         if_clone_detach(V_bridge_cloner);
 603         V_bridge_cloner = NULL;
 604         BRIDGE_LIST_LOCK_DESTROY();
 605
 606         /* Before we can destroy the uma zone, because there are callbacks that
 607          * use it. */
 608         NET_EPOCH_WAIT();
 609
 610         uma_zdestroy(V_bridge_rtnode_zone);
 611 }
 612 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 613     vnet_bridge_uninit, NULL);
 614
 615 static int
 616 bridge_modevent(module_t mod, int type, void *data)
 617 {
 618
 619         switch (type) {
 620         case MOD_LOAD:
 621                 bridge_dn_p = bridge_dummynet;
 622                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 623                     ifnet_departure_event, bridge_ifdetach, NULL,
 624                     EVENTHANDLER_PRI_ANY);
 625                 break;
 626         case MOD_UNLOAD:
 627                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 628                     bridge_detach_cookie);
 629                 bridge_dn_p = NULL;
 630                 break;
 631         default:
 632                 return (EOPNOTSUPP);
 633         }
 634         return (0);
 635 }
 636
 637 static moduledata_t bridge_mod = {
 638         "if_bridge",
 639         bridge_modevent,
 640         0
 641 };
 642
 643 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 644 MODULE_VERSION(if_bridge, 1);
 645 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 646
 647 /*
 648  * handler for net.link.bridge.ipfw
 649  */
 650 static int
 651 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 652 {
 653         int enable = V_pfil_ipfw;
 654         int error;
 655
 656         error = sysctl_handle_int(oidp, &enable, 0, req);
 657         enable &= 1;
 658
 659         if (enable != V_pfil_ipfw) {
 660                 V_pfil_ipfw = enable;
 661
 662                 /*
 663                  * Disable pfil so that ipfw doesnt run twice, if the user
 664                  * really wants both then they can re-enable pfil_bridge and/or
 665                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 666                  * layer2 type.
 667                  */
 668                 if (V_pfil_ipfw) {
 669                         V_pfil_onlyip = 0;
 670                         V_pfil_bridge = 0;
 671                         V_pfil_member = 0;
 672                 }
 673         }
 674
 675         return (error);
 676 }
 677 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 678     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 679     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 680     "Layer2 filter with IPFW");
 681
 682 /*
 683  * bridge_clone_create:
 684  *
 685  *      Create a new bridge instance.
 686  */
 687 static int
 688 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 689 {
 690         struct bridge_softc *sc, *sc2;
 691         struct ifnet *bifp, *ifp;
 692         int fb, retry;
 693         unsigned long hostid;
 694
 695         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 696         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 697         if (ifp == NULL) {
 698                 free(sc, M_DEVBUF);
 699                 return (ENOSPC);
 700         }
 701
 702         BRIDGE_LOCK_INIT(sc);
 703         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 704         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 705
 706         /* Initialize our routing table. */
 707         bridge_rtable_init(sc);
 708
 709         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 710
 711         CK_LIST_INIT(&sc->sc_iflist);
 712         CK_LIST_INIT(&sc->sc_spanlist);
 713
 714         ifp->if_softc = sc;
 715         if_initname(ifp, bridge_name, unit);
 716         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 717         ifp->if_ioctl = bridge_ioctl;
 718         ifp->if_transmit = bridge_transmit;
 719         ifp->if_qflush = bridge_qflush;
 720         ifp->if_init = bridge_init;
 721         ifp->if_type = IFT_BRIDGE;
 722
 723         /*
 724          * Generate an ethernet address with a locally administered address.
 725          *
 726          * Since we are using random ethernet addresses for the bridge, it is
 727          * possible that we might have address collisions, so make sure that
 728          * this hardware address isn't already in use on another bridge.
 729          * The first try uses the hostid and falls back to arc4rand().
 730          */
 731         fb = 0;
 732         getcredhostid(curthread->td_ucred, &hostid);
 733         do {
 734                 if (fb || hostid == 0) {
 735                         ether_gen_addr(ifp, &sc->sc_defaddr);
 736                 } else {
 737                         sc->sc_defaddr.octet[0] = 0x2;
 738                         sc->sc_defaddr.octet[1] = (hostid >> 24) & 0xff;
 739                         sc->sc_defaddr.octet[2] = (hostid >> 16) & 0xff;
 740                         sc->sc_defaddr.octet[3] = (hostid >> 8 ) & 0xff;
 741                         sc->sc_defaddr.octet[4] =  hostid        & 0xff;
 742                         sc->sc_defaddr.octet[5] = ifp->if_dunit & 0xff;
 743                 }
 744
 745                 fb = 1;
 746                 retry = 0;
 747                 BRIDGE_LIST_LOCK();
 748                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 749                         bifp = sc2->sc_ifp;
 750                         if (memcmp(sc->sc_defaddr.octet,
 751                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 752                                 retry = 1;
 753                                 break;
 754                         }
 755                 }
 756                 BRIDGE_LIST_UNLOCK();
 757         } while (retry == 1);
 758
 759         bstp_attach(&sc->sc_stp, &bridge_ops);
 760         ether_ifattach(ifp, sc->sc_defaddr.octet);
 761         /* Now undo some of the damage... */
 762         ifp->if_baudrate = 0;
 763         ifp->if_type = IFT_BRIDGE;
 764
 765         BRIDGE_LIST_LOCK();
 766         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 767         BRIDGE_LIST_UNLOCK();
 768
 769         return (0);
 770 }
 771
 772 static void
 773 bridge_clone_destroy_cb(struct epoch_context *ctx)
 774 {
 775         struct bridge_softc *sc;
 776
 777         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 778
 779         BRIDGE_LOCK_DESTROY(sc);
 780         free(sc, M_DEVBUF);
 781 }
 782
 783 /*
 784  * bridge_clone_destroy:
 785  *
 786  *      Destroy a bridge instance.
 787  */
 788 static void
 789 bridge_clone_destroy(struct ifnet *ifp)
 790 {
 791         struct bridge_softc *sc = ifp->if_softc;
 792         struct bridge_iflist *bif;
 793
 794         BRIDGE_LOCK(sc);
 795
 796         bridge_stop(ifp, 1);
 797         ifp->if_flags &= ~IFF_UP;
 798
 799         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 800                 bridge_delete_member(sc, bif, 0);
 801
 802         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 803                 bridge_delete_span(sc, bif);
 804         }
 805
 806         /* Tear down the routing table. */
 807         bridge_rtable_fini(sc);
 808
 809         BRIDGE_UNLOCK(sc);
 810
 811         callout_drain(&sc->sc_brcallout);
 812
 813         BRIDGE_LIST_LOCK();
 814         LIST_REMOVE(sc, sc_list);
 815         BRIDGE_LIST_UNLOCK();
 816
 817         bstp_detach(&sc->sc_stp);
 818         ether_ifdetach(ifp);
 819         if_free(ifp);
 820
 821         epoch_call(net_epoch_preempt, &sc->sc_epoch_ctx, bridge_clone_destroy_cb);
 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         struct epoch_tracker et;
 848
 849         NET_EPOCH_ENTER_ET(et);
 850
 851         switch (cmd) {
 852
 853         case SIOCADDMULTI:
 854         case SIOCDELMULTI:
 855                 break;
 856
 857         case SIOCGDRVSPEC:
 858         case SIOCSDRVSPEC:
 859                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 860                         error = EINVAL;
 861                         break;
 862                 }
 863                 bc = &bridge_control_table[ifd->ifd_cmd];
 864
 865                 if (cmd == SIOCGDRVSPEC &&
 866                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 867                         error = EINVAL;
 868                         break;
 869                 }
 870                 else if (cmd == SIOCSDRVSPEC &&
 871                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 872                         error = EINVAL;
 873                         break;
 874                 }
 875
 876                 if (bc->bc_flags & BC_F_SUSER) {
 877                         error = priv_check(td, PRIV_NET_BRIDGE);
 878                         if (error)
 879                                 break;
 880                 }
 881
 882                 if (ifd->ifd_len != bc->bc_argsize ||
 883                     ifd->ifd_len > sizeof(args)) {
 884                         error = EINVAL;
 885                         break;
 886                 }
 887
 888                 bzero(&args, sizeof(args));
 889                 if (bc->bc_flags & BC_F_COPYIN) {
 890                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 891                         if (error)
 892                                 break;
 893                 }
 894
 895                 oldmtu = ifp->if_mtu;
 896                 BRIDGE_LOCK(sc);
 897                 error = (*bc->bc_func)(sc, &args);
 898                 BRIDGE_UNLOCK(sc);
 899                 if (error)
 900                         break;
 901
 902                 /*
 903                  * Bridge MTU may change during addition of the first port.
 904                  * If it did, do network layer specific procedure.
 905                  */
 906                 if (ifp->if_mtu != oldmtu) {
 907 #ifdef INET6
 908                         nd6_setmtu(ifp);
 909 #endif
 910                         rt_updatemtu(ifp);
 911                 }
 912
 913                 if (bc->bc_flags & BC_F_COPYOUT)
 914                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 915
 916                 break;
 917
 918         case SIOCSIFFLAGS:
 919                 if (!(ifp->if_flags & IFF_UP) &&
 920                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 921                         /*
 922                          * If interface is marked down and it is running,
 923                          * then stop and disable it.
 924                          */
 925                         BRIDGE_LOCK(sc);
 926                         bridge_stop(ifp, 1);
 927                         BRIDGE_UNLOCK(sc);
 928                 } else if ((ifp->if_flags & IFF_UP) &&
 929                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 930                         /*
 931                          * If interface is marked up and it is stopped, then
 932                          * start it.
 933                          */
 934                         (*ifp->if_init)(sc);
 935                 }
 936                 break;
 937
 938         case SIOCSIFMTU:
 939                 if (ifr->ifr_mtu < 576) {
 940                         error = EINVAL;
 941                         break;
 942                 }
 943                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 944                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 945                         break;
 946                 }
 947                 BRIDGE_LOCK(sc);
 948                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 949                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 950                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 951                                     " != %d\n", sc->sc_ifp->if_xname,
 952                                     bif->bif_ifp->if_mtu,
 953                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 954                                 error = EINVAL;
 955                                 break;
 956                         }
 957                 }
 958                 if (!error)
 959                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 960                 BRIDGE_UNLOCK(sc);
 961                 break;
 962         default:
 963                 /*
 964                  * drop the lock as ether_ioctl() will call bridge_start() and
 965                  * cause the lock to be recursed.
 966                  */
 967                 error = ether_ioctl(ifp, cmd, data);
 968                 break;
 969         }
 970
 971         NET_EPOCH_EXIT_ET(et);
 972
 973         return (error);
 974 }
 975
 976 /*
 977  * bridge_mutecaps:
 978  *
 979  *      Clear or restore unwanted capabilities on the member interface
 980  */
 981 static void
 982 bridge_mutecaps(struct bridge_softc *sc)
 983 {
 984         struct bridge_iflist *bif;
 985         int enabled, mask;
 986
 987         BRIDGE_LOCK_ASSERT(sc);
 988
 989         /* Initial bitmask of capabilities to test */
 990         mask = BRIDGE_IFCAPS_MASK;
 991
 992         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 993                 /* Every member must support it or its disabled */
 994                 mask &= bif->bif_savedcaps;
 995         }
 996
 997         BRIDGE_XLOCK(sc);
 998         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 999                 enabled = bif->bif_ifp->if_capenable;
1000                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1001                 /* strip off mask bits and enable them again if allowed */
1002                 enabled &= ~BRIDGE_IFCAPS_MASK;
1003                 enabled |= mask;
1004                 BRIDGE_UNLOCK(sc);
1005                 bridge_set_ifcap(sc, bif, enabled);
1006                 BRIDGE_LOCK(sc);
1007         }
1008         BRIDGE_XDROP(sc);
1009
1010 }
1011
1012 static void
1013 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1014 {
1015         struct ifnet *ifp = bif->bif_ifp;
1016         struct ifreq ifr;
1017         int error, mask, stuck;
1018
1019         BRIDGE_UNLOCK_ASSERT(sc);
1020
1021         bzero(&ifr, sizeof(ifr));
1022         ifr.ifr_reqcap = set;
1023
1024         if (ifp->if_capenable != set) {
1025                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1026                 if (error)
1027                         if_printf(sc->sc_ifp,
1028                             "error setting capabilities on %s: %d\n",
1029                             ifp->if_xname, error);
1030                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1031                 stuck = ifp->if_capenable & mask & ~set;
1032                 if (stuck != 0)
1033                         if_printf(sc->sc_ifp,
1034                             "can't disable some capabilities on %s: 0x%x\n",
1035                             ifp->if_xname, stuck);
1036         }
1037 }
1038
1039 /*
1040  * bridge_lookup_member:
1041  *
1042  *      Lookup a bridge member interface.
1043  */
1044 static struct bridge_iflist *
1045 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1046 {
1047         struct bridge_iflist *bif;
1048         struct ifnet *ifp;
1049
1050         MPASS(in_epoch(net_epoch_preempt));
1051
1052         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1053                 ifp = bif->bif_ifp;
1054                 if (strcmp(ifp->if_xname, name) == 0)
1055                         return (bif);
1056         }
1057
1058         return (NULL);
1059 }
1060
1061 /*
1062  * bridge_lookup_member_if:
1063  *
1064  *      Lookup a bridge member interface by ifnet*.
1065  */
1066 static struct bridge_iflist *
1067 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1068 {
1069         struct bridge_iflist *bif;
1070
1071         MPASS(in_epoch(net_epoch_preempt));
1072
1073         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1074                 if (bif->bif_ifp == member_ifp)
1075                         return (bif);
1076         }
1077
1078         return (NULL);
1079 }
1080
1081 static void
1082 bridge_delete_member_cb(struct epoch_context *ctx)
1083 {
1084         struct bridge_iflist *bif;
1085
1086         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1087
1088         free(bif, M_DEVBUF);
1089 }
1090
1091 /*
1092  * bridge_delete_member:
1093  *
1094  *      Delete the specified member interface.
1095  */
1096 static void
1097 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1098     int gone)
1099 {
1100         struct ifnet *ifs = bif->bif_ifp;
1101         struct ifnet *fif = NULL;
1102         struct bridge_iflist *bifl;
1103
1104         BRIDGE_LOCK_ASSERT(sc);
1105
1106         if (bif->bif_flags & IFBIF_STP)
1107                 bstp_disable(&bif->bif_stp);
1108
1109         ifs->if_bridge = NULL;
1110         BRIDGE_XLOCK(sc);
1111         CK_LIST_REMOVE(bif, bif_next);
1112         BRIDGE_XDROP(sc);
1113
1114         /*
1115          * If removing the interface that gave the bridge its mac address, set
1116          * the mac address of the bridge to the address of the next member, or
1117          * to its default address if no members are left.
1118          */
1119         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1120                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1121                         bcopy(&sc->sc_defaddr,
1122                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1123                         sc->sc_ifaddr = NULL;
1124                 } else {
1125                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1126                         fif = bifl->bif_ifp;
1127                         bcopy(IF_LLADDR(fif),
1128                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1129                         sc->sc_ifaddr = fif;
1130                 }
1131                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1132         }
1133
1134         bridge_linkcheck(sc);
1135         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1136         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1137         KASSERT(bif->bif_addrcnt == 0,
1138             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1139
1140         ifs->if_bridge_output = NULL;
1141         ifs->if_bridge_input = NULL;
1142         ifs->if_bridge_linkstate = NULL;
1143         BRIDGE_UNLOCK(sc);
1144         if (!gone) {
1145                 switch (ifs->if_type) {
1146                 case IFT_ETHER:
1147                 case IFT_L2VLAN:
1148                         /*
1149                          * Take the interface out of promiscuous mode, but only
1150                          * if it was promiscuous in the first place. It might
1151                          * not be if we're in the bridge_ioctl_add() error path.
1152                          */
1153                         if (ifs->if_flags & IFF_PROMISC)
1154                                 (void) ifpromisc(ifs, 0);
1155                         break;
1156
1157                 case IFT_GIF:
1158                         break;
1159
1160                 default:
1161 #ifdef DIAGNOSTIC
1162                         panic("bridge_delete_member: impossible");
1163 #endif
1164                         break;
1165                 }
1166                 /* reneable any interface capabilities */
1167                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1168         }
1169         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1170         BRIDGE_LOCK(sc);
1171
1172         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1173             bridge_delete_member_cb);
1174 }
1175
1176 /*
1177  * bridge_delete_span:
1178  *
1179  *      Delete the specified span interface.
1180  */
1181 static void
1182 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1183 {
1184         BRIDGE_LOCK_ASSERT(sc);
1185
1186         KASSERT(bif->bif_ifp->if_bridge == NULL,
1187             ("%s: not a span interface", __func__));
1188
1189         CK_LIST_REMOVE(bif, bif_next);
1190
1191         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1192             bridge_delete_member_cb);
1193 }
1194
1195 static int
1196 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1197 {
1198         struct ifbreq *req = arg;
1199         struct bridge_iflist *bif = NULL;
1200         struct ifnet *ifs;
1201         int error = 0;
1202
1203         ifs = ifunit(req->ifbr_ifsname);
1204         if (ifs == NULL)
1205                 return (ENOENT);
1206         if (ifs->if_ioctl == NULL)      /* must be supported */
1207                 return (EINVAL);
1208
1209         /* If it's in the span list, it can't be a member. */
1210         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1211                 if (ifs == bif->bif_ifp)
1212                         return (EBUSY);
1213
1214         if (ifs->if_bridge == sc)
1215                 return (EEXIST);
1216
1217         if (ifs->if_bridge != NULL)
1218                 return (EBUSY);
1219
1220         switch (ifs->if_type) {
1221         case IFT_ETHER:
1222         case IFT_L2VLAN:
1223         case IFT_GIF:
1224                 /* permitted interface types */
1225                 break;
1226         default:
1227                 return (EINVAL);
1228         }
1229
1230 #ifdef INET6
1231         /*
1232          * Two valid inet6 addresses with link-local scope must not be
1233          * on the parent interface and the member interfaces at the
1234          * same time.  This restriction is needed to prevent violation
1235          * of link-local scope zone.  Attempts to add a member
1236          * interface which has inet6 addresses when the parent has
1237          * inet6 triggers removal of all inet6 addresses on the member
1238          * interface.
1239          */
1240
1241         /* Check if the parent interface has a link-local scope addr. */
1242         if (V_allow_llz_overlap == 0 &&
1243             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1244                 /*
1245                  * If any, remove all inet6 addresses from the member
1246                  * interfaces.
1247                  */
1248                 BRIDGE_XLOCK(sc);
1249                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1250                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1251                                 BRIDGE_UNLOCK(sc);
1252                                 in6_ifdetach(bif->bif_ifp);
1253                                 BRIDGE_LOCK(sc);
1254                                 if_printf(sc->sc_ifp,
1255                                     "IPv6 addresses on %s have been removed "
1256                                     "before adding it as a member to prevent "
1257                                     "IPv6 address scope violation.\n",
1258                                     bif->bif_ifp->if_xname);
1259                         }
1260                 }
1261                 BRIDGE_XDROP(sc);
1262                 if (in6ifa_llaonifp(ifs)) {
1263                         BRIDGE_UNLOCK(sc);
1264                         in6_ifdetach(ifs);
1265                         BRIDGE_LOCK(sc);
1266                         if_printf(sc->sc_ifp,
1267                             "IPv6 addresses on %s have been removed "
1268                             "before adding it as a member to prevent "
1269                             "IPv6 address scope violation.\n",
1270                             ifs->if_xname);
1271                 }
1272         }
1273 #endif
1274         /* Allow the first Ethernet member to define the MTU */
1275         if (CK_LIST_EMPTY(&sc->sc_iflist))
1276                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1277         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1278                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1279                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1280                 return (EINVAL);
1281         }
1282
1283         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1284         if (bif == NULL)
1285                 return (ENOMEM);
1286
1287         bif->bif_ifp = ifs;
1288         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1289         bif->bif_savedcaps = ifs->if_capenable;
1290
1291         /*
1292          * Assign the interface's MAC address to the bridge if it's the first
1293          * member and the MAC address of the bridge has not been changed from
1294          * the default randomly generated one.
1295          */
1296         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1297             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1298                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1299                 sc->sc_ifaddr = ifs;
1300                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1301         }
1302
1303         ifs->if_bridge = sc;
1304         ifs->if_bridge_output = bridge_output;
1305         ifs->if_bridge_input = bridge_input;
1306         ifs->if_bridge_linkstate = bridge_linkstate;
1307         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1308         /*
1309          * XXX: XLOCK HERE!?!
1310          *
1311          * NOTE: insert_***HEAD*** should be safe for the traversals.
1312          */
1313         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1314
1315         /* Set interface capabilities to the intersection set of all members */
1316         bridge_mutecaps(sc);
1317         bridge_linkcheck(sc);
1318
1319         /* Place the interface into promiscuous mode */
1320         switch (ifs->if_type) {
1321                 case IFT_ETHER:
1322                 case IFT_L2VLAN:
1323                         BRIDGE_UNLOCK(sc);
1324                         error = ifpromisc(ifs, 1);
1325                         BRIDGE_LOCK(sc);
1326                         break;
1327         }
1328
1329         if (error)
1330                 bridge_delete_member(sc, bif, 0);
1331         return (error);
1332 }
1333
1334 static int
1335 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1336 {
1337         struct ifbreq *req = arg;
1338         struct bridge_iflist *bif;
1339
1340         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1341         if (bif == NULL)
1342                 return (ENOENT);
1343
1344         bridge_delete_member(sc, bif, 0);
1345
1346         return (0);
1347 }
1348
1349 static int
1350 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1351 {
1352         struct ifbreq *req = arg;
1353         struct bridge_iflist *bif;
1354         struct bstp_port *bp;
1355
1356         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1357         if (bif == NULL)
1358                 return (ENOENT);
1359
1360         bp = &bif->bif_stp;
1361         req->ifbr_ifsflags = bif->bif_flags;
1362         req->ifbr_state = bp->bp_state;
1363         req->ifbr_priority = bp->bp_priority;
1364         req->ifbr_path_cost = bp->bp_path_cost;
1365         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1366         req->ifbr_proto = bp->bp_protover;
1367         req->ifbr_role = bp->bp_role;
1368         req->ifbr_stpflags = bp->bp_flags;
1369         req->ifbr_addrcnt = bif->bif_addrcnt;
1370         req->ifbr_addrmax = bif->bif_addrmax;
1371         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1372
1373         /* Copy STP state options as flags */
1374         if (bp->bp_operedge)
1375                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1376         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1377                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1378         if (bp->bp_ptp_link)
1379                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1380         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1381                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1382         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1383                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1384         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1385                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1386         return (0);
1387 }
1388
1389 static int
1390 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1391 {
1392         struct ifbreq *req = arg;
1393         struct bridge_iflist *bif;
1394         struct bstp_port *bp;
1395         int error;
1396
1397         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1398         if (bif == NULL)
1399                 return (ENOENT);
1400         bp = &bif->bif_stp;
1401
1402         if (req->ifbr_ifsflags & IFBIF_SPAN)
1403                 /* SPAN is readonly */
1404                 return (EINVAL);
1405
1406         if (req->ifbr_ifsflags & IFBIF_STP) {
1407                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1408                         error = bstp_enable(&bif->bif_stp);
1409                         if (error)
1410                                 return (error);
1411                 }
1412         } else {
1413                 if ((bif->bif_flags & IFBIF_STP) != 0)
1414                         bstp_disable(&bif->bif_stp);
1415         }
1416
1417         /* Pass on STP flags */
1418         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1419         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1420         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1421         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1422
1423         /* Save the bits relating to the bridge */
1424         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1425
1426         return (0);
1427 }
1428
1429 static int
1430 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1431 {
1432         struct ifbrparam *param = arg;
1433
1434         sc->sc_brtmax = param->ifbrp_csize;
1435         bridge_rttrim(sc);
1436
1437         return (0);
1438 }
1439
1440 static int
1441 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1442 {
1443         struct ifbrparam *param = arg;
1444
1445         param->ifbrp_csize = sc->sc_brtmax;
1446
1447         return (0);
1448 }
1449
1450 static int
1451 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1452 {
1453         struct ifbifconf *bifc = arg;
1454         struct bridge_iflist *bif;
1455         struct ifbreq breq;
1456         char *buf, *outbuf;
1457         int count, buflen, len, error = 0;
1458
1459         count = 0;
1460         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1461                 count++;
1462         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1463                 count++;
1464
1465         buflen = sizeof(breq) * count;
1466         if (bifc->ifbic_len == 0) {
1467                 bifc->ifbic_len = buflen;
1468                 return (0);
1469         }
1470         BRIDGE_UNLOCK(sc);
1471         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1472         BRIDGE_LOCK(sc);
1473
1474         count = 0;
1475         buf = outbuf;
1476         len = min(bifc->ifbic_len, buflen);
1477         bzero(&breq, sizeof(breq));
1478         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1479                 if (len < sizeof(breq))
1480                         break;
1481
1482                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1483                     sizeof(breq.ifbr_ifsname));
1484                 /* Fill in the ifbreq structure */
1485                 error = bridge_ioctl_gifflags(sc, &breq);
1486                 if (error)
1487                         break;
1488                 memcpy(buf, &breq, sizeof(breq));
1489                 count++;
1490                 buf += sizeof(breq);
1491                 len -= sizeof(breq);
1492         }
1493         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1494                 if (len < sizeof(breq))
1495                         break;
1496
1497                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1498                     sizeof(breq.ifbr_ifsname));
1499                 breq.ifbr_ifsflags = bif->bif_flags;
1500                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1501                 memcpy(buf, &breq, sizeof(breq));
1502                 count++;
1503                 buf += sizeof(breq);
1504                 len -= sizeof(breq);
1505         }
1506
1507         BRIDGE_UNLOCK(sc);
1508         bifc->ifbic_len = sizeof(breq) * count;
1509         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1510         BRIDGE_LOCK(sc);
1511         free(outbuf, M_TEMP);
1512         return (error);
1513 }
1514
1515 static int
1516 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1517 {
1518         struct ifbaconf *bac = arg;
1519         struct bridge_rtnode *brt;
1520         struct ifbareq bareq;
1521         char *buf, *outbuf;
1522         int count, buflen, len, error = 0;
1523
1524         if (bac->ifbac_len == 0)
1525                 return (0);
1526
1527         count = 0;
1528         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1529                 count++;
1530         buflen = sizeof(bareq) * count;
1531
1532         BRIDGE_UNLOCK(sc);
1533         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1534         BRIDGE_LOCK(sc);
1535
1536         count = 0;
1537         buf = outbuf;
1538         len = min(bac->ifbac_len, buflen);
1539         bzero(&bareq, sizeof(bareq));
1540         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1541                 if (len < sizeof(bareq))
1542                         goto out;
1543                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1544                     sizeof(bareq.ifba_ifsname));
1545                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1546                 bareq.ifba_vlan = brt->brt_vlan;
1547                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1548                                 time_uptime < brt->brt_expire)
1549                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1550                 else
1551                         bareq.ifba_expire = 0;
1552                 bareq.ifba_flags = brt->brt_flags;
1553
1554                 memcpy(buf, &bareq, sizeof(bareq));
1555                 count++;
1556                 buf += sizeof(bareq);
1557                 len -= sizeof(bareq);
1558         }
1559 out:
1560         BRIDGE_UNLOCK(sc);
1561         bac->ifbac_len = sizeof(bareq) * count;
1562         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1563         BRIDGE_LOCK(sc);
1564         free(outbuf, M_TEMP);
1565         return (error);
1566 }
1567
1568 static int
1569 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1570 {
1571         struct ifbareq *req = arg;
1572         struct bridge_iflist *bif;
1573         int error;
1574
1575         MPASS(in_epoch(net_epoch_preempt));
1576
1577         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1578         if (bif == NULL)
1579                 return (ENOENT);
1580
1581         /* bridge_rtupdate() may acquire the lock. */
1582         BRIDGE_UNLOCK(sc);
1583         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1584             req->ifba_flags);
1585         BRIDGE_LOCK(sc);
1586
1587         return (error);
1588 }
1589
1590 static int
1591 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1592 {
1593         struct ifbrparam *param = arg;
1594
1595         sc->sc_brttimeout = param->ifbrp_ctime;
1596         return (0);
1597 }
1598
1599 static int
1600 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1601 {
1602         struct ifbrparam *param = arg;
1603
1604         param->ifbrp_ctime = sc->sc_brttimeout;
1605         return (0);
1606 }
1607
1608 static int
1609 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1610 {
1611         struct ifbareq *req = arg;
1612
1613         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1614 }
1615
1616 static int
1617 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1618 {
1619         struct ifbreq *req = arg;
1620
1621         bridge_rtflush(sc, req->ifbr_ifsflags);
1622         return (0);
1623 }
1624
1625 static int
1626 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1627 {
1628         struct ifbrparam *param = arg;
1629         struct bstp_state *bs = &sc->sc_stp;
1630
1631         param->ifbrp_prio = bs->bs_bridge_priority;
1632         return (0);
1633 }
1634
1635 static int
1636 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1637 {
1638         struct ifbrparam *param = arg;
1639
1640         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1641 }
1642
1643 static int
1644 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1645 {
1646         struct ifbrparam *param = arg;
1647         struct bstp_state *bs = &sc->sc_stp;
1648
1649         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1650         return (0);
1651 }
1652
1653 static int
1654 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1655 {
1656         struct ifbrparam *param = arg;
1657
1658         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1659 }
1660
1661 static int
1662 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1663 {
1664         struct ifbrparam *param = arg;
1665         struct bstp_state *bs = &sc->sc_stp;
1666
1667         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1668         return (0);
1669 }
1670
1671 static int
1672 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1673 {
1674         struct ifbrparam *param = arg;
1675
1676         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1677 }
1678
1679 static int
1680 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1681 {
1682         struct ifbrparam *param = arg;
1683         struct bstp_state *bs = &sc->sc_stp;
1684
1685         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1686         return (0);
1687 }
1688
1689 static int
1690 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1691 {
1692         struct ifbrparam *param = arg;
1693
1694         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1695 }
1696
1697 static int
1698 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1699 {
1700         struct ifbreq *req = arg;
1701         struct bridge_iflist *bif;
1702
1703         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1704         if (bif == NULL)
1705                 return (ENOENT);
1706
1707         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1708 }
1709
1710 static int
1711 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1712 {
1713         struct ifbreq *req = arg;
1714         struct bridge_iflist *bif;
1715
1716         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1717         if (bif == NULL)
1718                 return (ENOENT);
1719
1720         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1721 }
1722
1723 static int
1724 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1725 {
1726         struct ifbreq *req = arg;
1727         struct bridge_iflist *bif;
1728
1729         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1730         if (bif == NULL)
1731                 return (ENOENT);
1732
1733         bif->bif_addrmax = req->ifbr_addrmax;
1734         return (0);
1735 }
1736
1737 static int
1738 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1739 {
1740         struct ifbreq *req = arg;
1741         struct bridge_iflist *bif = NULL;
1742         struct ifnet *ifs;
1743
1744         ifs = ifunit(req->ifbr_ifsname);
1745         if (ifs == NULL)
1746                 return (ENOENT);
1747
1748         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1749                 if (ifs == bif->bif_ifp)
1750                         return (EBUSY);
1751
1752         if (ifs->if_bridge != NULL)
1753                 return (EBUSY);
1754
1755         switch (ifs->if_type) {
1756                 case IFT_ETHER:
1757                 case IFT_GIF:
1758                 case IFT_L2VLAN:
1759                         break;
1760                 default:
1761                         return (EINVAL);
1762         }
1763
1764         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1765         if (bif == NULL)
1766                 return (ENOMEM);
1767
1768         bif->bif_ifp = ifs;
1769         bif->bif_flags = IFBIF_SPAN;
1770
1771         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1772
1773         return (0);
1774 }
1775
1776 static int
1777 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1778 {
1779         struct ifbreq *req = arg;
1780         struct bridge_iflist *bif;
1781         struct ifnet *ifs;
1782
1783         ifs = ifunit(req->ifbr_ifsname);
1784         if (ifs == NULL)
1785                 return (ENOENT);
1786
1787         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1788                 if (ifs == bif->bif_ifp)
1789                         break;
1790
1791         if (bif == NULL)
1792                 return (ENOENT);
1793
1794         bridge_delete_span(sc, bif);
1795
1796         return (0);
1797 }
1798
1799 static int
1800 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1801 {
1802         struct ifbropreq *req = arg;
1803         struct bstp_state *bs = &sc->sc_stp;
1804         struct bstp_port *root_port;
1805
1806         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1807         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1808         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1809
1810         root_port = bs->bs_root_port;
1811         if (root_port == NULL)
1812                 req->ifbop_root_port = 0;
1813         else
1814                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1815
1816         req->ifbop_holdcount = bs->bs_txholdcount;
1817         req->ifbop_priority = bs->bs_bridge_priority;
1818         req->ifbop_protocol = bs->bs_protover;
1819         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1820         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1821         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1822         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1823         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1824         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1825
1826         return (0);
1827 }
1828
1829 static int
1830 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1831 {
1832         struct ifbrparam *param = arg;
1833
1834         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1835         return (0);
1836 }
1837
1838 static int
1839 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1840 {
1841         struct ifbpstpconf *bifstp = arg;
1842         struct bridge_iflist *bif;
1843         struct bstp_port *bp;
1844         struct ifbpstpreq bpreq;
1845         char *buf, *outbuf;
1846         int count, buflen, len, error = 0;
1847
1848         count = 0;
1849         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1850                 if ((bif->bif_flags & IFBIF_STP) != 0)
1851                         count++;
1852         }
1853
1854         buflen = sizeof(bpreq) * count;
1855         if (bifstp->ifbpstp_len == 0) {
1856                 bifstp->ifbpstp_len = buflen;
1857                 return (0);
1858         }
1859
1860         BRIDGE_UNLOCK(sc);
1861         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1862         BRIDGE_LOCK(sc);
1863
1864         count = 0;
1865         buf = outbuf;
1866         len = min(bifstp->ifbpstp_len, buflen);
1867         bzero(&bpreq, sizeof(bpreq));
1868         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1869                 if (len < sizeof(bpreq))
1870                         break;
1871
1872                 if ((bif->bif_flags & IFBIF_STP) == 0)
1873                         continue;
1874
1875                 bp = &bif->bif_stp;
1876                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1877                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1878                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1879                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1880                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1881                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1882
1883                 memcpy(buf, &bpreq, sizeof(bpreq));
1884                 count++;
1885                 buf += sizeof(bpreq);
1886                 len -= sizeof(bpreq);
1887         }
1888
1889         BRIDGE_UNLOCK(sc);
1890         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1891         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1892         BRIDGE_LOCK(sc);
1893         free(outbuf, M_TEMP);
1894         return (error);
1895 }
1896
1897 static int
1898 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1899 {
1900         struct ifbrparam *param = arg;
1901
1902         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1903 }
1904
1905 static int
1906 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1907 {
1908         struct ifbrparam *param = arg;
1909
1910         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1911 }
1912
1913 /*
1914  * bridge_ifdetach:
1915  *
1916  *      Detach an interface from a bridge.  Called when a member
1917  *      interface is detaching.
1918  */
1919 static void
1920 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1921 {
1922         struct bridge_softc *sc = ifp->if_bridge;
1923         struct bridge_iflist *bif;
1924         struct epoch_tracker et;
1925
1926         if (ifp->if_flags & IFF_RENAMING)
1927                 return;
1928         if (V_bridge_cloner == NULL) {
1929                 /*
1930                  * This detach handler can be called after
1931                  * vnet_bridge_uninit().  Just return in that case.
1932                  */
1933                 return;
1934         }
1935         NET_EPOCH_ENTER_ET(et);
1936         /* Check if the interface is a bridge member */
1937         if (sc != NULL) {
1938                 BRIDGE_LOCK(sc);
1939
1940                 bif = bridge_lookup_member_if(sc, ifp);
1941                 if (bif != NULL)
1942                         bridge_delete_member(sc, bif, 1);
1943
1944                 BRIDGE_UNLOCK(sc);
1945                 NET_EPOCH_EXIT_ET(et);
1946                 return;
1947         }
1948
1949         /* Check if the interface is a span port */
1950         BRIDGE_LIST_LOCK();
1951         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1952                 BRIDGE_LOCK(sc);
1953                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1954                         if (ifp == bif->bif_ifp) {
1955                                 bridge_delete_span(sc, bif);
1956                                 break;
1957                         }
1958
1959                 BRIDGE_UNLOCK(sc);
1960         }
1961         BRIDGE_LIST_UNLOCK();
1962         NET_EPOCH_EXIT_ET(et);
1963 }
1964
1965 /*
1966  * bridge_init:
1967  *
1968  *      Initialize a bridge interface.
1969  */
1970 static void
1971 bridge_init(void *xsc)
1972 {
1973         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1974         struct ifnet *ifp = sc->sc_ifp;
1975
1976         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1977                 return;
1978
1979         BRIDGE_LOCK(sc);
1980         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1981             bridge_timer, sc);
1982
1983         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1984         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1985
1986         BRIDGE_UNLOCK(sc);
1987 }
1988
1989 /*
1990  * bridge_stop:
1991  *
1992  *      Stop the bridge interface.
1993  */
1994 static void
1995 bridge_stop(struct ifnet *ifp, int disable)
1996 {
1997         struct bridge_softc *sc = ifp->if_softc;
1998
1999         BRIDGE_LOCK_ASSERT(sc);
2000
2001         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2002                 return;
2003
2004         callout_stop(&sc->sc_brcallout);
2005         bstp_stop(&sc->sc_stp);
2006
2007         bridge_rtflush(sc, IFBF_FLUSHDYN);
2008
2009         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2010 }
2011
2012 /*
2013  * bridge_enqueue:
2014  *
2015  *      Enqueue a packet on a bridge member interface.
2016  *
2017  */
2018 static int
2019 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2020 {
2021         int len, err = 0;
2022         short mflags;
2023         struct mbuf *m0;
2024
2025         /* We may be sending a fragment so traverse the mbuf */
2026         for (; m; m = m0) {
2027                 m0 = m->m_nextpkt;
2028                 m->m_nextpkt = NULL;
2029                 len = m->m_pkthdr.len;
2030                 mflags = m->m_flags;
2031
2032                 /*
2033                  * If underlying interface can not do VLAN tag insertion itself
2034                  * then attach a packet tag that holds it.
2035                  */
2036                 if ((m->m_flags & M_VLANTAG) &&
2037                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2038                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2039                         if (m == NULL) {
2040                                 if_printf(dst_ifp,
2041                                     "unable to prepend VLAN header\n");
2042                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2043                                 continue;
2044                         }
2045                         m->m_flags &= ~M_VLANTAG;
2046                 }
2047
2048                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2049                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2050                         m_freem(m0);
2051                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2052                         break;
2053                 }
2054
2055                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2056                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2057                 if (mflags & M_MCAST)
2058                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2059         }
2060
2061         return (err);
2062 }
2063
2064 /*
2065  * bridge_dummynet:
2066  *
2067  *      Receive a queued packet from dummynet and pass it on to the output
2068  *      interface.
2069  *
2070  *      The mbuf has the Ethernet header already attached.
2071  */
2072 static void
2073 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2074 {
2075         struct bridge_softc *sc;
2076
2077         sc = ifp->if_bridge;
2078
2079         /*
2080          * The packet didnt originate from a member interface. This should only
2081          * ever happen if a member interface is removed while packets are
2082          * queued for it.
2083          */
2084         if (sc == NULL) {
2085                 m_freem(m);
2086                 return;
2087         }
2088
2089         if (PFIL_HOOKED(&V_inet_pfil_hook)
2090 #ifdef INET6
2091             || PFIL_HOOKED(&V_inet6_pfil_hook)
2092 #endif
2093             ) {
2094                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2095                         return;
2096                 if (m == NULL)
2097                         return;
2098         }
2099
2100         bridge_enqueue(sc, ifp, m);
2101 }
2102
2103 /*
2104  * bridge_output:
2105  *
2106  *      Send output from a bridge member interface.  This
2107  *      performs the bridging function for locally originated
2108  *      packets.
2109  *
2110  *      The mbuf has the Ethernet header already attached.  We must
2111  *      enqueue or free the mbuf before returning.
2112  */
2113 static int
2114 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2115     struct rtentry *rt)
2116 {
2117         struct ether_header *eh;
2118         struct ifnet *dst_if;
2119         struct bridge_softc *sc;
2120         uint16_t vlan;
2121
2122         MPASS(in_epoch(net_epoch_preempt));
2123
2124         if (m->m_len < ETHER_HDR_LEN) {
2125                 m = m_pullup(m, ETHER_HDR_LEN);
2126                 if (m == NULL)
2127                         return (0);
2128         }
2129
2130         eh = mtod(m, struct ether_header *);
2131         sc = ifp->if_bridge;
2132         vlan = VLANTAGOF(m);
2133
2134         /*
2135          * If bridge is down, but the original output interface is up,
2136          * go ahead and send out that interface.  Otherwise, the packet
2137          * is dropped below.
2138          */
2139         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2140                 dst_if = ifp;
2141                 goto sendunicast;
2142         }
2143
2144         /*
2145          * If the packet is a multicast, or we don't know a better way to
2146          * get there, send to all interfaces.
2147          */
2148         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2149                 dst_if = NULL;
2150         else
2151                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2152         if (dst_if == NULL) {
2153                 struct bridge_iflist *bif;
2154                 struct mbuf *mc;
2155                 int used = 0;
2156
2157                 bridge_span(sc, m);
2158
2159                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2160                         dst_if = bif->bif_ifp;
2161
2162                         if (dst_if->if_type == IFT_GIF)
2163                                 continue;
2164                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2165                                 continue;
2166
2167                         /*
2168                          * If this is not the original output interface,
2169                          * and the interface is participating in spanning
2170                          * tree, make sure the port is in a state that
2171                          * allows forwarding.
2172                          */
2173                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2174                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2175                                 continue;
2176
2177                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2178                                 used = 1;
2179                                 mc = m;
2180                         } else {
2181                                 mc = m_copypacket(m, M_NOWAIT);
2182                                 if (mc == NULL) {
2183                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2184                                         continue;
2185                                 }
2186                         }
2187
2188                         bridge_enqueue(sc, dst_if, mc);
2189                 }
2190                 if (used == 0)
2191                         m_freem(m);
2192                 return (0);
2193         }
2194
2195 sendunicast:
2196         /*
2197          * XXX Spanning tree consideration here?
2198          */
2199
2200         bridge_span(sc, m);
2201         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2202                 m_freem(m);
2203                 return (0);
2204         }
2205
2206         bridge_enqueue(sc, dst_if, m);
2207         return (0);
2208 }
2209
2210 /*
2211  * bridge_transmit:
2212  *
2213  *      Do output on a bridge.
2214  *
2215  */
2216 static int
2217 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2218 {
2219         struct bridge_softc *sc;
2220         struct ether_header *eh;
2221         struct ifnet *dst_if;
2222         int error = 0;
2223
2224         sc = ifp->if_softc;
2225
2226         ETHER_BPF_MTAP(ifp, m);
2227
2228         eh = mtod(m, struct ether_header *);
2229
2230         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2231             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2232                 error = bridge_enqueue(sc, dst_if, m);
2233         } else
2234                 bridge_broadcast(sc, ifp, m, 0);
2235
2236         return (error);
2237 }
2238
2239 /*
2240  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2241  */
2242 static void
2243 bridge_qflush(struct ifnet *ifp __unused)
2244 {
2245 }
2246
2247 /*
2248  * bridge_forward:
2249  *
2250  *      The forwarding function of the bridge.
2251  *
2252  *      NOTE: Releases the lock on return.
2253  */
2254 static void
2255 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2256     struct mbuf *m)
2257 {
2258         struct bridge_iflist *dbif;
2259         struct ifnet *src_if, *dst_if, *ifp;
2260         struct ether_header *eh;
2261         uint16_t vlan;
2262         uint8_t *dst;
2263         int error;
2264
2265         MPASS(in_epoch(net_epoch_preempt));
2266
2267         src_if = m->m_pkthdr.rcvif;
2268         ifp = sc->sc_ifp;
2269
2270         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2271         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2272         vlan = VLANTAGOF(m);
2273
2274         if ((sbif->bif_flags & IFBIF_STP) &&
2275             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2276                 goto drop;
2277
2278         eh = mtod(m, struct ether_header *);
2279         dst = eh->ether_dhost;
2280
2281         /* If the interface is learning, record the address. */
2282         if (sbif->bif_flags & IFBIF_LEARNING) {
2283                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2284                     sbif, 0, IFBAF_DYNAMIC);
2285                 /*
2286                  * If the interface has addresses limits then deny any source
2287                  * that is not in the cache.
2288                  */
2289                 if (error && sbif->bif_addrmax)
2290                         goto drop;
2291         }
2292
2293         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2294             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2295                 goto drop;
2296
2297         /*
2298          * At this point, the port either doesn't participate
2299          * in spanning tree or it is in the forwarding state.
2300          */
2301
2302         /*
2303          * If the packet is unicast, destined for someone on
2304          * "this" side of the bridge, drop it.
2305          */
2306         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2307                 dst_if = bridge_rtlookup(sc, dst, vlan);
2308                 if (src_if == dst_if)
2309                         goto drop;
2310         } else {
2311                 /*
2312                  * Check if its a reserved multicast address, any address
2313                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2314                  * bridge.
2315                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2316                  */
2317                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2318                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2319                     dst[4] == 0x00 && dst[5] <= 0x0f)
2320                         goto drop;
2321
2322                 /* ...forward it to all interfaces. */
2323                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2324                 dst_if = NULL;
2325         }
2326
2327         /*
2328          * If we have a destination interface which is a member of our bridge,
2329          * OR this is a unicast packet, push it through the bpf(4) machinery.
2330          * For broadcast or multicast packets, don't bother because it will
2331          * be reinjected into ether_input. We do this before we pass the packets
2332          * through the pfil(9) framework, as it is possible that pfil(9) will
2333          * drop the packet, or possibly modify it, making it difficult to debug
2334          * firewall issues on the bridge.
2335          */
2336         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2337                 ETHER_BPF_MTAP(ifp, m);
2338
2339         /* run the packet filter */
2340         if (PFIL_HOOKED(&V_inet_pfil_hook)
2341 #ifdef INET6
2342             || PFIL_HOOKED(&V_inet6_pfil_hook)
2343 #endif
2344             ) {
2345                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2346                         return;
2347                 if (m == NULL)
2348                         return;
2349         }
2350
2351         if (dst_if == NULL) {
2352                 bridge_broadcast(sc, src_if, m, 1);
2353                 return;
2354         }
2355
2356         /*
2357          * At this point, we're dealing with a unicast frame
2358          * going to a different interface.
2359          */
2360         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2361                 goto drop;
2362
2363         dbif = bridge_lookup_member_if(sc, dst_if);
2364         if (dbif == NULL)
2365                 /* Not a member of the bridge (anymore?) */
2366                 goto drop;
2367
2368         /* Private segments can not talk to each other */
2369         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2370                 goto drop;
2371
2372         if ((dbif->bif_flags & IFBIF_STP) &&
2373             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2374                 goto drop;
2375
2376         if (PFIL_HOOKED(&V_inet_pfil_hook)
2377 #ifdef INET6
2378             || PFIL_HOOKED(&V_inet6_pfil_hook)
2379 #endif
2380             ) {
2381                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2382                         return;
2383                 if (m == NULL)
2384                         return;
2385         }
2386
2387         bridge_enqueue(sc, dst_if, m);
2388         return;
2389
2390 drop:
2391         m_freem(m);
2392 }
2393
2394 /*
2395  * bridge_input:
2396  *
2397  *      Receive input from a member interface.  Queue the packet for
2398  *      bridging if it is not for us.
2399  */
2400 static struct mbuf *
2401 bridge_input(struct ifnet *ifp, struct mbuf *m)
2402 {
2403         struct bridge_softc *sc = ifp->if_bridge;
2404         struct bridge_iflist *bif, *bif2;
2405         struct ifnet *bifp;
2406         struct ether_header *eh;
2407         struct mbuf *mc, *mc2;
2408         uint16_t vlan;
2409         int error;
2410
2411         MPASS(in_epoch(net_epoch_preempt));
2412
2413         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2414                 return (m);
2415
2416         bifp = sc->sc_ifp;
2417         vlan = VLANTAGOF(m);
2418
2419         /*
2420          * Implement support for bridge monitoring. If this flag has been
2421          * set on this interface, discard the packet once we push it through
2422          * the bpf(4) machinery, but before we do, increment the byte and
2423          * packet counters associated with this interface.
2424          */
2425         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2426                 m->m_pkthdr.rcvif  = bifp;
2427                 ETHER_BPF_MTAP(bifp, m);
2428                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2429                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2430                 m_freem(m);
2431                 return (NULL);
2432         }
2433         bif = bridge_lookup_member_if(sc, ifp);
2434         if (bif == NULL) {
2435                 return (m);
2436         }
2437
2438         eh = mtod(m, struct ether_header *);
2439
2440         bridge_span(sc, m);
2441
2442         if (m->m_flags & (M_BCAST|M_MCAST)) {
2443                 /* Tap off 802.1D packets; they do not get forwarded. */
2444                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2445                     ETHER_ADDR_LEN) == 0) {
2446                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2447                         return (NULL);
2448                 }
2449
2450                 if ((bif->bif_flags & IFBIF_STP) &&
2451                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2452                         return (m);
2453                 }
2454
2455                 /*
2456                  * Make a deep copy of the packet and enqueue the copy
2457                  * for bridge processing; return the original packet for
2458                  * local processing.
2459                  */
2460                 mc = m_dup(m, M_NOWAIT);
2461                 if (mc == NULL) {
2462                         return (m);
2463                 }
2464
2465                 /* Perform the bridge forwarding function with the copy. */
2466                 bridge_forward(sc, bif, mc);
2467
2468                 /*
2469                  * Reinject the mbuf as arriving on the bridge so we have a
2470                  * chance at claiming multicast packets. We can not loop back
2471                  * here from ether_input as a bridge is never a member of a
2472                  * bridge.
2473                  */
2474                 KASSERT(bifp->if_bridge == NULL,
2475                     ("loop created in bridge_input"));
2476                 mc2 = m_dup(m, M_NOWAIT);
2477                 if (mc2 != NULL) {
2478                         /* Keep the layer3 header aligned */
2479                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2480                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2481                 }
2482                 if (mc2 != NULL) {
2483                         mc2->m_pkthdr.rcvif = bifp;
2484                         (*bifp->if_input)(bifp, mc2);
2485                 }
2486
2487                 /* Return the original packet for local processing. */
2488                 return (m);
2489         }
2490
2491         if ((bif->bif_flags & IFBIF_STP) &&
2492             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2493                 return (m);
2494         }
2495
2496 #if (defined(INET) || defined(INET6))
2497 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2498         || ((iface)->if_carp \
2499             && (*carp_forus_p)((iface), eh->ether_dhost))
2500 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2501         || ((iface)->if_carp \
2502             && (*carp_forus_p)((iface), eh->ether_shost))
2503 #else
2504 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2505 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2506 #endif
2507
2508 #ifdef INET6
2509 #   define OR_PFIL_HOOKED_INET6 \
2510         || PFIL_HOOKED(&V_inet6_pfil_hook)
2511 #else
2512 #   define OR_PFIL_HOOKED_INET6
2513 #endif
2514
2515 #define GRAB_OUR_PACKETS(iface) \
2516         if ((iface)->if_type == IFT_GIF) \
2517                 continue; \
2518         /* It is destined for us. */ \
2519         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2520             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2521             ) {                                                         \
2522                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2523                         ETHER_BPF_MTAP(iface, m);                       \
2524                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2525                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2526                         /* Filter on the physical interface. */         \
2527                         if (V_pfil_local_phys &&                        \
2528                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2529                              OR_PFIL_HOOKED_INET6)) {                   \
2530                                 if (bridge_pfil(&m, NULL, ifp,          \
2531                                     PFIL_IN) != 0 || m == NULL) {       \
2532                                         return (NULL);                  \
2533                                 }                                       \
2534                                 eh = mtod(m, struct ether_header *);    \
2535                         }                                               \
2536                 }                                                       \
2537                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2538                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2539                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2540                         if (error && bif->bif_addrmax) {                \
2541                                 m_freem(m);                             \
2542                                 return (NULL);                          \
2543                         }                                               \
2544                 }                                                       \
2545                 m->m_pkthdr.rcvif = iface;                              \
2546                 return (m);                                             \
2547         }                                                               \
2548                                                                         \
2549         /* We just received a packet that we sent out. */               \
2550         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2551             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2552             ) {                                                         \
2553                 m_freem(m);                                             \
2554                 return (NULL);                                          \
2555         }
2556
2557         /*
2558          * Unicast.  Make sure it's not for the bridge.
2559          */
2560         do { GRAB_OUR_PACKETS(bifp) } while (0);
2561
2562         /*
2563          * Give a chance for ifp at first priority. This will help when the
2564          * packet comes through the interface like VLAN's with the same MACs
2565          * on several interfaces from the same bridge. This also will save
2566          * some CPU cycles in case the destination interface and the input
2567          * interface (eq ifp) are the same.
2568          */
2569         do { GRAB_OUR_PACKETS(ifp) } while (0);
2570
2571         /* Now check the all bridge members. */
2572         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2573                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2574         }
2575
2576 #undef OR_CARP_CHECK_WE_ARE_DST
2577 #undef OR_CARP_CHECK_WE_ARE_SRC
2578 #undef OR_PFIL_HOOKED_INET6
2579 #undef GRAB_OUR_PACKETS
2580
2581         /* Perform the bridge forwarding function. */
2582         bridge_forward(sc, bif, m);
2583
2584         return (NULL);
2585 }
2586
2587 /*
2588  * bridge_broadcast:
2589  *
2590  *      Send a frame to all interfaces that are members of
2591  *      the bridge, except for the one on which the packet
2592  *      arrived.
2593  *
2594  *      NOTE: Releases the lock on return.
2595  */
2596 static void
2597 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2598     struct mbuf *m, int runfilt)
2599 {
2600         struct bridge_iflist *dbif, *sbif;
2601         struct mbuf *mc;
2602         struct ifnet *dst_if;
2603         int used = 0, i;
2604
2605         MPASS(in_epoch(net_epoch_preempt));
2606
2607         sbif = bridge_lookup_member_if(sc, src_if);
2608
2609         /* Filter on the bridge interface before broadcasting */
2610         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2611 #ifdef INET6
2612             || PFIL_HOOKED(&V_inet6_pfil_hook)
2613 #endif
2614             )) {
2615                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2616                         return;
2617                 if (m == NULL)
2618                         return;
2619         }
2620
2621         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2622                 dst_if = dbif->bif_ifp;
2623                 if (dst_if == src_if)
2624                         continue;
2625
2626                 /* Private segments can not talk to each other */
2627                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2628                         continue;
2629
2630                 if ((dbif->bif_flags & IFBIF_STP) &&
2631                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2632                         continue;
2633
2634                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2635                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2636                         continue;
2637
2638                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2639                         continue;
2640
2641                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2642                         mc = m;
2643                         used = 1;
2644                 } else {
2645                         mc = m_dup(m, M_NOWAIT);
2646                         if (mc == NULL) {
2647                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2648                                 continue;
2649                         }
2650                 }
2651
2652                 /*
2653                  * Filter on the output interface. Pass a NULL bridge interface
2654                  * pointer so we do not redundantly filter on the bridge for
2655                  * each interface we broadcast on.
2656                  */
2657                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2658 #ifdef INET6
2659                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2660 #endif
2661                     )) {
2662                         if (used == 0) {
2663                                 /* Keep the layer3 header aligned */
2664                                 i = min(mc->m_pkthdr.len, max_protohdr);
2665                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2666                                 if (mc == NULL) {
2667                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2668                                         continue;
2669                                 }
2670                         }
2671                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2672                                 continue;
2673                         if (mc == NULL)
2674                                 continue;
2675                 }
2676
2677                 bridge_enqueue(sc, dst_if, mc);
2678         }
2679         if (used == 0)
2680                 m_freem(m);
2681 }
2682
2683 /*
2684  * bridge_span:
2685  *
2686  *      Duplicate a packet out one or more interfaces that are in span mode,
2687  *      the original mbuf is unmodified.
2688  */
2689 static void
2690 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2691 {
2692         struct bridge_iflist *bif;
2693         struct ifnet *dst_if;
2694         struct mbuf *mc;
2695
2696         MPASS(in_epoch(net_epoch_preempt));
2697
2698         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2699                 return;
2700
2701         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2702                 dst_if = bif->bif_ifp;
2703
2704                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2705                         continue;
2706
2707                 mc = m_copypacket(m, M_NOWAIT);
2708                 if (mc == NULL) {
2709                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2710                         continue;
2711                 }
2712
2713                 bridge_enqueue(sc, dst_if, mc);
2714         }
2715 }
2716
2717 /*
2718  * bridge_rtupdate:
2719  *
2720  *      Add a bridge routing entry.
2721  */
2722 static int
2723 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2724     struct bridge_iflist *bif, int setflags, uint8_t flags)
2725 {
2726         struct bridge_rtnode *brt;
2727         int error;
2728
2729         MPASS(in_epoch(net_epoch_preempt));
2730         BRIDGE_UNLOCK_ASSERT(sc);
2731
2732         /* Check the source address is valid and not multicast. */
2733         if (ETHER_IS_MULTICAST(dst) ||
2734             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2735              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2736                 return (EINVAL);
2737
2738         /* 802.1p frames map to vlan 1 */
2739         if (vlan == 0)
2740                 vlan = 1;
2741
2742         /*
2743          * A route for this destination might already exist.  If so,
2744          * update it, otherwise create a new one.
2745          */
2746         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2747                 BRIDGE_LOCK(sc);
2748
2749                 /* Check again, now that we have the lock. There could have
2750                  * been a race and we only want to insert this once. */
2751                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2752                         BRIDGE_UNLOCK(sc);
2753                         return (0);
2754                 }
2755
2756                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2757                         sc->sc_brtexceeded++;
2758                         BRIDGE_UNLOCK(sc);
2759                         return (ENOSPC);
2760                 }
2761                 /* Check per interface address limits (if enabled) */
2762                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2763                         bif->bif_addrexceeded++;
2764                         BRIDGE_UNLOCK(sc);
2765                         return (ENOSPC);
2766                 }
2767
2768                 /*
2769                  * Allocate a new bridge forwarding node, and
2770                  * initialize the expiration time and Ethernet
2771                  * address.
2772                  */
2773                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2774                 if (brt == NULL) {
2775                         BRIDGE_UNLOCK(sc);
2776                         return (ENOMEM);
2777                 }
2778                 brt->brt_vnet = curvnet;
2779
2780                 if (bif->bif_flags & IFBIF_STICKY)
2781                         brt->brt_flags = IFBAF_STICKY;
2782                 else
2783                         brt->brt_flags = IFBAF_DYNAMIC;
2784
2785                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2786                 brt->brt_vlan = vlan;
2787
2788                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2789                         uma_zfree(V_bridge_rtnode_zone, brt);
2790                         BRIDGE_UNLOCK(sc);
2791                         return (error);
2792                 }
2793                 brt->brt_dst = bif;
2794                 bif->bif_addrcnt++;
2795
2796                 BRIDGE_UNLOCK(sc);
2797         }
2798
2799         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2800             brt->brt_dst != bif) {
2801                 BRIDGE_LOCK(sc);
2802                 brt->brt_dst->bif_addrcnt--;
2803                 brt->brt_dst = bif;
2804                 brt->brt_dst->bif_addrcnt++;
2805                 BRIDGE_UNLOCK(sc);
2806         }
2807
2808         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2809                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2810         if (setflags)
2811                 brt->brt_flags = flags;
2812
2813         return (0);
2814 }
2815
2816 /*
2817  * bridge_rtlookup:
2818  *
2819  *      Lookup the destination interface for an address.
2820  */
2821 static struct ifnet *
2822 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2823 {
2824         struct bridge_rtnode *brt;
2825
2826         MPASS(in_epoch(net_epoch_preempt));
2827
2828         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2829                 return (NULL);
2830
2831         return (brt->brt_ifp);
2832 }
2833
2834 /*
2835  * bridge_rttrim:
2836  *
2837  *      Trim the routine table so that we have a number
2838  *      of routing entries less than or equal to the
2839  *      maximum number.
2840  */
2841 static void
2842 bridge_rttrim(struct bridge_softc *sc)
2843 {
2844         struct bridge_rtnode *brt, *nbrt;
2845
2846         BRIDGE_LOCK_ASSERT(sc);
2847
2848         /* Make sure we actually need to do this. */
2849         if (sc->sc_brtcnt <= sc->sc_brtmax)
2850                 return;
2851
2852         /* Force an aging cycle; this might trim enough addresses. */
2853         bridge_rtage(sc);
2854         if (sc->sc_brtcnt <= sc->sc_brtmax)
2855                 return;
2856
2857         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2858                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2859                         bridge_rtnode_destroy(sc, brt);
2860                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2861                                 return;
2862                 }
2863         }
2864 }
2865
2866 /*
2867  * bridge_timer:
2868  *
2869  *      Aging timer for the bridge.
2870  */
2871 static void
2872 bridge_timer(void *arg)
2873 {
2874         struct bridge_softc *sc = arg;
2875
2876         BRIDGE_LOCK_ASSERT(sc);
2877
2878         /* Destruction of rtnodes requires a proper vnet context */
2879         CURVNET_SET(sc->sc_ifp->if_vnet);
2880         bridge_rtage(sc);
2881
2882         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2883                 callout_reset(&sc->sc_brcallout,
2884                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2885         CURVNET_RESTORE();
2886 }
2887
2888 /*
2889  * bridge_rtage:
2890  *
2891  *      Perform an aging cycle.
2892  */
2893 static void
2894 bridge_rtage(struct bridge_softc *sc)
2895 {
2896         struct bridge_rtnode *brt, *nbrt;
2897
2898         BRIDGE_LOCK_ASSERT(sc);
2899
2900         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2901                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2902                         if (time_uptime >= brt->brt_expire)
2903                                 bridge_rtnode_destroy(sc, brt);
2904                 }
2905         }
2906 }
2907
2908 /*
2909  * bridge_rtflush:
2910  *
2911  *      Remove all dynamic addresses from the bridge.
2912  */
2913 static void
2914 bridge_rtflush(struct bridge_softc *sc, int full)
2915 {
2916         struct bridge_rtnode *brt, *nbrt;
2917
2918         BRIDGE_LOCK_ASSERT(sc);
2919
2920         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2921                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2922                         bridge_rtnode_destroy(sc, brt);
2923         }
2924 }
2925
2926 /*
2927  * bridge_rtdaddr:
2928  *
2929  *      Remove an address from the table.
2930  */
2931 static int
2932 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2933 {
2934         struct bridge_rtnode *brt;
2935         int found = 0;
2936
2937         BRIDGE_LOCK_ASSERT(sc);
2938
2939         /*
2940          * If vlan is zero then we want to delete for all vlans so the lookup
2941          * may return more than one.
2942          */
2943         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2944                 bridge_rtnode_destroy(sc, brt);
2945                 found = 1;
2946         }
2947
2948         return (found ? 0 : ENOENT);
2949 }
2950
2951 /*
2952  * bridge_rtdelete:
2953  *
2954  *      Delete routes to a speicifc member interface.
2955  */
2956 static void
2957 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2958 {
2959         struct bridge_rtnode *brt, *nbrt;
2960
2961         BRIDGE_LOCK_ASSERT(sc);
2962
2963         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2964                 if (brt->brt_ifp == ifp && (full ||
2965                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2966                         bridge_rtnode_destroy(sc, brt);
2967         }
2968 }
2969
2970 /*
2971  * bridge_rtable_init:
2972  *
2973  *      Initialize the route table for this bridge.
2974  */
2975 static void
2976 bridge_rtable_init(struct bridge_softc *sc)
2977 {
2978         int i;
2979
2980         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2981             M_DEVBUF, M_WAITOK);
2982
2983         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2984                 CK_LIST_INIT(&sc->sc_rthash[i]);
2985
2986         sc->sc_rthash_key = arc4random();
2987         CK_LIST_INIT(&sc->sc_rtlist);
2988 }
2989
2990 /*
2991  * bridge_rtable_fini:
2992  *
2993  *      Deconstruct the route table for this bridge.
2994  */
2995 static void
2996 bridge_rtable_fini(struct bridge_softc *sc)
2997 {
2998
2999         KASSERT(sc->sc_brtcnt == 0,
3000             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3001         free(sc->sc_rthash, M_DEVBUF);
3002 }
3003
3004 /*
3005  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3006  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3007  */
3008 #define mix(a, b, c)                                                    \
3009 do {                                                                    \
3010         a -= b; a -= c; a ^= (c >> 13);                                 \
3011         b -= c; b -= a; b ^= (a << 8);                                  \
3012         c -= a; c -= b; c ^= (b >> 13);                                 \
3013         a -= b; a -= c; a ^= (c >> 12);                                 \
3014         b -= c; b -= a; b ^= (a << 16);                                 \
3015         c -= a; c -= b; c ^= (b >> 5);                                  \
3016         a -= b; a -= c; a ^= (c >> 3);                                  \
3017         b -= c; b -= a; b ^= (a << 10);                                 \
3018         c -= a; c -= b; c ^= (b >> 15);                                 \
3019 } while (/*CONSTCOND*/0)
3020
3021 static __inline uint32_t
3022 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3023 {
3024         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3025
3026         b += addr[5] << 8;
3027         b += addr[4];
3028         a += addr[3] << 24;
3029         a += addr[2] << 16;
3030         a += addr[1] << 8;
3031         a += addr[0];
3032
3033         mix(a, b, c);
3034
3035         return (c & BRIDGE_RTHASH_MASK);
3036 }
3037
3038 #undef mix
3039
3040 static int
3041 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3042 {
3043         int i, d;
3044
3045         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3046                 d = ((int)a[i]) - ((int)b[i]);
3047         }
3048
3049         return (d);
3050 }
3051
3052 /*
3053  * bridge_rtnode_lookup:
3054  *
3055  *      Look up a bridge route node for the specified destination. Compare the
3056  *      vlan id or if zero then just return the first match.
3057  */
3058 static struct bridge_rtnode *
3059 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3060 {
3061         struct bridge_rtnode *brt;
3062         uint32_t hash;
3063         int dir;
3064
3065         MPASS(in_epoch(net_epoch_preempt));
3066
3067         hash = bridge_rthash(sc, addr);
3068         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3069                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3070                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3071                         return (brt);
3072                 if (dir > 0)
3073                         return (NULL);
3074         }
3075
3076         return (NULL);
3077 }
3078
3079 /*
3080  * bridge_rtnode_insert:
3081  *
3082  *      Insert the specified bridge node into the route table.  We
3083  *      assume the entry is not already in the table.
3084  */
3085 static int
3086 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3087 {
3088         struct bridge_rtnode *lbrt;
3089         uint32_t hash;
3090         int dir;
3091
3092         BRIDGE_LOCK_ASSERT(sc);
3093
3094         hash = bridge_rthash(sc, brt->brt_addr);
3095
3096         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3097         if (lbrt == NULL) {
3098                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3099                 goto out;
3100         }
3101
3102         do {
3103                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3104                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3105                         return (EEXIST);
3106                 if (dir > 0) {
3107                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3108                         goto out;
3109                 }
3110                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3111                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3112                         goto out;
3113                 }
3114                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3115         } while (lbrt != NULL);
3116
3117 #ifdef DIAGNOSTIC
3118         panic("bridge_rtnode_insert: impossible");
3119 #endif
3120
3121 out:
3122         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3123         sc->sc_brtcnt++;
3124
3125         return (0);
3126 }
3127
3128 static void
3129 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3130 {
3131         struct bridge_rtnode *brt;
3132
3133         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3134
3135         CURVNET_SET(brt->brt_vnet);
3136         uma_zfree(V_bridge_rtnode_zone, brt);
3137         CURVNET_RESTORE();
3138 }
3139
3140 /*
3141  * bridge_rtnode_destroy:
3142  *
3143  *      Destroy a bridge rtnode.
3144  */
3145 static void
3146 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3147 {
3148         BRIDGE_LOCK_ASSERT(sc);
3149
3150         CK_LIST_REMOVE(brt, brt_hash);
3151
3152         CK_LIST_REMOVE(brt, brt_list);
3153         sc->sc_brtcnt--;
3154         brt->brt_dst->bif_addrcnt--;
3155
3156         epoch_call(net_epoch_preempt, &brt->brt_epoch_ctx,
3157             bridge_rtnode_destroy_cb);
3158 }
3159
3160 /*
3161  * bridge_rtable_expire:
3162  *
3163  *      Set the expiry time for all routes on an interface.
3164  */
3165 static void
3166 bridge_rtable_expire(struct ifnet *ifp, int age)
3167 {
3168         struct bridge_softc *sc = ifp->if_bridge;
3169         struct bridge_rtnode *brt;
3170
3171         CURVNET_SET(ifp->if_vnet);
3172         BRIDGE_LOCK(sc);
3173
3174         /*
3175          * If the age is zero then flush, otherwise set all the expiry times to
3176          * age for the interface
3177          */
3178         if (age == 0)
3179                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3180         else {
3181                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3182                         /* Cap the expiry time to 'age' */
3183                         if (brt->brt_ifp == ifp &&
3184                             brt->brt_expire > time_uptime + age &&
3185                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3186                                 brt->brt_expire = time_uptime + age;
3187                 }
3188         }
3189         BRIDGE_UNLOCK(sc);
3190         CURVNET_RESTORE();
3191 }
3192
3193 /*
3194  * bridge_state_change:
3195  *
3196  *      Callback from the bridgestp code when a port changes states.
3197  */
3198 static void
3199 bridge_state_change(struct ifnet *ifp, int state)
3200 {
3201         struct bridge_softc *sc = ifp->if_bridge;
3202         static const char *stpstates[] = {
3203                 "disabled",
3204                 "listening",
3205                 "learning",
3206                 "forwarding",
3207                 "blocking",
3208                 "discarding"
3209         };
3210
3211         CURVNET_SET(ifp->if_vnet);
3212         if (V_log_stp)
3213                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3214                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3215         CURVNET_RESTORE();
3216 }
3217
3218 /*
3219  * Send bridge packets through pfil if they are one of the types pfil can deal
3220  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3221  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3222  * that interface.
3223  */
3224 static int
3225 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3226 {
3227         int snap, error, i, hlen;
3228         struct ether_header *eh1, eh2;
3229         struct ip *ip;
3230         struct llc llc1;
3231         u_int16_t ether_type;
3232
3233         snap = 0;
3234         error = -1;     /* Default error if not error == 0 */
3235
3236 #if 0
3237         /* we may return with the IP fields swapped, ensure its not shared */
3238         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3239 #endif
3240
3241         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3242                 return (0); /* filtering is disabled */
3243
3244         i = min((*mp)->m_pkthdr.len, max_protohdr);
3245         if ((*mp)->m_len < i) {
3246             *mp = m_pullup(*mp, i);
3247             if (*mp == NULL) {
3248                 printf("%s: m_pullup failed\n", __func__);
3249                 return (-1);
3250             }
3251         }
3252
3253         eh1 = mtod(*mp, struct ether_header *);
3254         ether_type = ntohs(eh1->ether_type);
3255
3256         /*
3257          * Check for SNAP/LLC.
3258          */
3259         if (ether_type < ETHERMTU) {
3260                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3261
3262                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3263                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3264                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3265                     llc2->llc_control == LLC_UI) {
3266                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3267                         snap = 1;
3268                 }
3269         }
3270
3271         /*
3272          * If we're trying to filter bridge traffic, don't look at anything
3273          * other than IP and ARP traffic.  If the filter doesn't understand
3274          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3275          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3276          * but of course we don't have an AppleTalk filter to begin with.
3277          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3278          * ARP traffic.)
3279          */
3280         switch (ether_type) {
3281                 case ETHERTYPE_ARP:
3282                 case ETHERTYPE_REVARP:
3283                         if (V_pfil_ipfw_arp == 0)
3284                                 return (0); /* Automatically pass */
3285                         break;
3286
3287                 case ETHERTYPE_IP:
3288 #ifdef INET6
3289                 case ETHERTYPE_IPV6:
3290 #endif /* INET6 */
3291                         break;
3292                 default:
3293                         /*
3294                          * Check to see if the user wants to pass non-ip
3295                          * packets, these will not be checked by pfil(9) and
3296                          * passed unconditionally so the default is to drop.
3297                          */
3298                         if (V_pfil_onlyip)
3299                                 goto bad;
3300         }
3301
3302         /* Run the packet through pfil before stripping link headers */
3303         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3304                         dir == PFIL_OUT && ifp != NULL) {
3305
3306                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, 0,
3307                     NULL);
3308
3309                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3310                         return (error);
3311         }
3312
3313         /* Strip off the Ethernet header and keep a copy. */
3314         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3315         m_adj(*mp, ETHER_HDR_LEN);
3316
3317         /* Strip off snap header, if present */
3318         if (snap) {
3319                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3320                 m_adj(*mp, sizeof(struct llc));
3321         }
3322
3323         /*
3324          * Check the IP header for alignment and errors
3325          */
3326         if (dir == PFIL_IN) {
3327                 switch (ether_type) {
3328                         case ETHERTYPE_IP:
3329                                 error = bridge_ip_checkbasic(mp);
3330                                 break;
3331 #ifdef INET6
3332                         case ETHERTYPE_IPV6:
3333                                 error = bridge_ip6_checkbasic(mp);
3334                                 break;
3335 #endif /* INET6 */
3336                         default:
3337                                 error = 0;
3338                 }
3339                 if (error)
3340                         goto bad;
3341         }
3342
3343         error = 0;
3344
3345         /*
3346          * Run the packet through pfil
3347          */
3348         switch (ether_type) {
3349         case ETHERTYPE_IP:
3350                 /*
3351                  * Run pfil on the member interface and the bridge, both can
3352                  * be skipped by clearing pfil_member or pfil_bridge.
3353                  *
3354                  * Keep the order:
3355                  *   in_if -> bridge_if -> out_if
3356                  */
3357                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3358                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3359                                         dir, 0, NULL);
3360
3361                 if (*mp == NULL || error != 0) /* filter may consume */
3362                         break;
3363
3364                 if (V_pfil_member && ifp != NULL)
3365                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3366                                         dir, 0, NULL);
3367
3368                 if (*mp == NULL || error != 0) /* filter may consume */
3369                         break;
3370
3371                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3372                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3373                                         dir, 0, NULL);
3374
3375                 if (*mp == NULL || error != 0) /* filter may consume */
3376                         break;
3377
3378                 /* check if we need to fragment the packet */
3379                 /* bridge_fragment generates a mbuf chain of packets */
3380                 /* that already include eth headers */
3381                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3382                         i = (*mp)->m_pkthdr.len;
3383                         if (i > ifp->if_mtu) {
3384                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3385                                             &llc1);
3386                                 return (error);
3387                         }
3388                 }
3389
3390                 /* Recalculate the ip checksum. */
3391                 ip = mtod(*mp, struct ip *);
3392                 hlen = ip->ip_hl << 2;
3393                 if (hlen < sizeof(struct ip))
3394                         goto bad;
3395                 if (hlen > (*mp)->m_len) {
3396                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3397                                 goto bad;
3398                         ip = mtod(*mp, struct ip *);
3399                         if (ip == NULL)
3400                                 goto bad;
3401                 }
3402                 ip->ip_sum = 0;
3403                 if (hlen == sizeof(struct ip))
3404                         ip->ip_sum = in_cksum_hdr(ip);
3405                 else
3406                         ip->ip_sum = in_cksum(*mp, hlen);
3407
3408                 break;
3409 #ifdef INET6
3410         case ETHERTYPE_IPV6:
3411                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3412                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3413                                         dir, 0, NULL);
3414
3415                 if (*mp == NULL || error != 0) /* filter may consume */
3416                         break;
3417
3418                 if (V_pfil_member && ifp != NULL)
3419                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3420                                         dir, 0, NULL);
3421
3422                 if (*mp == NULL || error != 0) /* filter may consume */
3423                         break;
3424
3425                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3426                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3427                                         dir, 0, NULL);
3428                 break;
3429 #endif
3430         default:
3431                 error = 0;
3432                 break;
3433         }
3434
3435         if (*mp == NULL)
3436                 return (error);
3437         if (error != 0)
3438                 goto bad;
3439
3440         error = -1;
3441
3442         /*
3443          * Finally, put everything back the way it was and return
3444          */
3445         if (snap) {
3446                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3447                 if (*mp == NULL)
3448                         return (error);
3449                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3450         }
3451
3452         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3453         if (*mp == NULL)
3454                 return (error);
3455         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3456
3457         return (0);
3458
3459 bad:
3460         m_freem(*mp);
3461         *mp = NULL;
3462         return (error);
3463 }
3464
3465 /*
3466  * Perform basic checks on header size since
3467  * pfil assumes ip_input has already processed
3468  * it for it.  Cut-and-pasted from ip_input.c.
3469  * Given how simple the IPv6 version is,
3470  * does the IPv4 version really need to be
3471  * this complicated?
3472  *
3473  * XXX Should we update ipstat here, or not?
3474  * XXX Right now we update ipstat but not
3475  * XXX csum_counter.
3476  */
3477 static int
3478 bridge_ip_checkbasic(struct mbuf **mp)
3479 {
3480         struct mbuf *m = *mp;
3481         struct ip *ip;
3482         int len, hlen;
3483         u_short sum;
3484
3485         if (*mp == NULL)
3486                 return (-1);
3487
3488         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3489                 if ((m = m_copyup(m, sizeof(struct ip),
3490                         (max_linkhdr + 3) & ~3)) == NULL) {
3491                         /* XXXJRT new stat, please */
3492                         KMOD_IPSTAT_INC(ips_toosmall);
3493                         goto bad;
3494                 }
3495         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3496                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3497                         KMOD_IPSTAT_INC(ips_toosmall);
3498                         goto bad;
3499                 }
3500         }
3501         ip = mtod(m, struct ip *);
3502         if (ip == NULL) goto bad;
3503
3504         if (ip->ip_v != IPVERSION) {
3505                 KMOD_IPSTAT_INC(ips_badvers);
3506                 goto bad;
3507         }
3508         hlen = ip->ip_hl << 2;
3509         if (hlen < sizeof(struct ip)) { /* minimum header length */
3510                 KMOD_IPSTAT_INC(ips_badhlen);
3511                 goto bad;
3512         }
3513         if (hlen > m->m_len) {
3514                 if ((m = m_pullup(m, hlen)) == NULL) {
3515                         KMOD_IPSTAT_INC(ips_badhlen);
3516                         goto bad;
3517                 }
3518                 ip = mtod(m, struct ip *);
3519                 if (ip == NULL) goto bad;
3520         }
3521
3522         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3523                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3524         } else {
3525                 if (hlen == sizeof(struct ip)) {
3526                         sum = in_cksum_hdr(ip);
3527                 } else {
3528                         sum = in_cksum(m, hlen);
3529                 }
3530         }
3531         if (sum) {
3532                 KMOD_IPSTAT_INC(ips_badsum);
3533                 goto bad;
3534         }
3535
3536         /* Retrieve the packet length. */
3537         len = ntohs(ip->ip_len);
3538
3539         /*
3540          * Check for additional length bogosity
3541          */
3542         if (len < hlen) {
3543                 KMOD_IPSTAT_INC(ips_badlen);
3544                 goto bad;
3545         }
3546
3547         /*
3548          * Check that the amount of data in the buffers
3549          * is as at least much as the IP header would have us expect.
3550          * Drop packet if shorter than we expect.
3551          */
3552         if (m->m_pkthdr.len < len) {
3553                 KMOD_IPSTAT_INC(ips_tooshort);
3554                 goto bad;
3555         }
3556
3557         /* Checks out, proceed */
3558         *mp = m;
3559         return (0);
3560
3561 bad:
3562         *mp = m;
3563         return (-1);
3564 }
3565
3566 #ifdef INET6
3567 /*
3568  * Same as above, but for IPv6.
3569  * Cut-and-pasted from ip6_input.c.
3570  * XXX Should we update ip6stat, or not?
3571  */
3572 static int
3573 bridge_ip6_checkbasic(struct mbuf **mp)
3574 {
3575         struct mbuf *m = *mp;
3576         struct ip6_hdr *ip6;
3577
3578         /*
3579          * If the IPv6 header is not aligned, slurp it up into a new
3580          * mbuf with space for link headers, in the event we forward
3581          * it.  Otherwise, if it is aligned, make sure the entire base
3582          * IPv6 header is in the first mbuf of the chain.
3583          */
3584         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3585                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3586                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3587                             (max_linkhdr + 3) & ~3)) == NULL) {
3588                         /* XXXJRT new stat, please */
3589                         IP6STAT_INC(ip6s_toosmall);
3590                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3591                         goto bad;
3592                 }
3593         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3594                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3595                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3596                         IP6STAT_INC(ip6s_toosmall);
3597                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3598                         goto bad;
3599                 }
3600         }
3601
3602         ip6 = mtod(m, struct ip6_hdr *);
3603
3604         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3605                 IP6STAT_INC(ip6s_badvers);
3606                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3607                 goto bad;
3608         }
3609
3610         /* Checks out, proceed */
3611         *mp = m;
3612         return (0);
3613
3614 bad:
3615         *mp = m;
3616         return (-1);
3617 }
3618 #endif /* INET6 */
3619
3620 /*
3621  * bridge_fragment:
3622  *
3623  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3624  */
3625 static int
3626 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3627     int snap, struct llc *llc)
3628 {
3629         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3630         struct ip *ip;
3631         int error = -1;
3632
3633         if (m->m_len < sizeof(struct ip) &&
3634             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3635                 goto dropit;
3636         ip = mtod(m, struct ip *);
3637
3638         m->m_pkthdr.csum_flags |= CSUM_IP;
3639         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3640         if (error)
3641                 goto dropit;
3642
3643         /*
3644          * Walk the chain and re-add the Ethernet header for
3645          * each mbuf packet.
3646          */
3647         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3648                 nextpkt = mcur->m_nextpkt;
3649                 mcur->m_nextpkt = NULL;
3650                 if (snap) {
3651                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3652                         if (mcur == NULL) {
3653                                 error = ENOBUFS;
3654                                 if (mprev != NULL)
3655                                         mprev->m_nextpkt = nextpkt;
3656                                 goto dropit;
3657                         }
3658                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3659                 }
3660
3661                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3662                 if (mcur == NULL) {
3663                         error = ENOBUFS;
3664                         if (mprev != NULL)
3665                                 mprev->m_nextpkt = nextpkt;
3666                         goto dropit;
3667                 }
3668                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3669
3670                 /*
3671                  * The previous two M_PREPEND could have inserted one or two
3672                  * mbufs in front so we have to update the previous packet's
3673                  * m_nextpkt.
3674                  */
3675                 mcur->m_nextpkt = nextpkt;
3676                 if (mprev != NULL)
3677                         mprev->m_nextpkt = mcur;
3678                 else {
3679                         /* The first mbuf in the original chain needs to be
3680                          * updated. */
3681                         *mp = mcur;
3682                 }
3683                 mprev = mcur;
3684         }
3685
3686         KMOD_IPSTAT_INC(ips_fragmented);
3687         return (error);
3688
3689 dropit:
3690         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3691                 m = mcur->m_nextpkt;
3692                 m_freem(mcur);
3693         }
3694         return (error);
3695 }
3696
3697 static void
3698 bridge_linkstate(struct ifnet *ifp)
3699 {
3700         struct bridge_softc *sc = ifp->if_bridge;
3701         struct bridge_iflist *bif;
3702         struct epoch_tracker et;
3703
3704         NET_EPOCH_ENTER_ET(et);
3705
3706         bif = bridge_lookup_member_if(sc, ifp);
3707         if (bif == NULL) {
3708                 NET_EPOCH_EXIT_ET(et);
3709                 return;
3710         }
3711         bridge_linkcheck(sc);
3712
3713         bstp_linkstate(&bif->bif_stp);
3714
3715         NET_EPOCH_EXIT_ET(et);
3716 }
3717
3718 static void
3719 bridge_linkcheck(struct bridge_softc *sc)
3720 {
3721         struct bridge_iflist *bif;
3722         int new_link, hasls;
3723
3724         MPASS(in_epoch(net_epoch_preempt));
3725
3726         new_link = LINK_STATE_DOWN;
3727         hasls = 0;
3728         /* Our link is considered up if at least one of our ports is active */
3729         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3730                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3731                         hasls++;
3732                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3733                         new_link = LINK_STATE_UP;
3734                         break;
3735                 }
3736         }
3737         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3738                 /* If no interfaces support link-state then we default to up */
3739                 new_link = LINK_STATE_UP;
3740         }
3741         if_link_state_change(sc->sc_ifp, new_link);
3742 }