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 } while (0)
 193 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 194         mtx_destroy(&(_sc)->sc_mtx);            \
 195 } while (0)
 196 #define BRIDGE_LOCK(_sc)                mtx_lock(&(_sc)->sc_mtx)
 197 #define BRIDGE_UNLOCK(_sc)              mtx_unlock(&(_sc)->sc_mtx)
 198 #define BRIDGE_LOCK_ASSERT(_sc)         mtx_assert(&(_sc)->sc_mtx, MA_OWNED)
 199 #define BRIDGE_UNLOCK_ASSERT(_sc)       mtx_assert(&(_sc)->sc_mtx, MA_NOTOWNED)
 200
 201 /*
 202  * Bridge interface list entry.
 203  */
 204 struct bridge_iflist {
 205         CK_LIST_ENTRY(bridge_iflist) bif_next;
 206         struct ifnet            *bif_ifp;       /* member if */
 207         struct bstp_port        bif_stp;        /* STP state */
 208         uint32_t                bif_flags;      /* member if flags */
 209         int                     bif_savedcaps;  /* saved capabilities */
 210         uint32_t                bif_addrmax;    /* max # of addresses */
 211         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 212         uint32_t                bif_addrexceeded;/* # of address violations */
 213
 214         struct epoch_context    bif_epoch_ctx;
 215 };
 216
 217 /*
 218  * Bridge route node.
 219  */
 220 struct bridge_rtnode {
 221         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 222         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 223         struct bridge_iflist    *brt_dst;       /* destination if */
 224         unsigned long           brt_expire;     /* expiration time */
 225         uint8_t                 brt_flags;      /* address flags */
 226         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 227         uint16_t                brt_vlan;       /* vlan id */
 228
 229         struct  vnet            *brt_vnet;
 230         struct  epoch_context   brt_epoch_ctx;
 231 };
 232 #define brt_ifp                 brt_dst->bif_ifp
 233
 234 /*
 235  * Software state for each bridge.
 236  */
 237 struct bridge_softc {
 238         struct ifnet            *sc_ifp;        /* make this an interface */
 239         LIST_ENTRY(bridge_softc) sc_list;
 240         struct mtx              sc_mtx;
 241         uint32_t                sc_brtmax;      /* max # of addresses */
 242         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 243         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 244         struct callout          sc_brcallout;   /* bridge callout */
 245         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 246         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 247         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 248         uint32_t                sc_rthash_key;  /* key for hash */
 249         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 250         struct bstp_state       sc_stp;         /* STP state */
 251         uint32_t                sc_brtexceeded; /* # of cache drops */
 252         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 253         struct ether_addr       sc_defaddr;     /* Default MAC address */
 254
 255         struct epoch_context    sc_epoch_ctx;
 256 };
 257
 258 VNET_DEFINE_STATIC(struct mtx, bridge_list_mtx);
 259 #define V_bridge_list_mtx       VNET(bridge_list_mtx)
 260 static eventhandler_tag bridge_detach_cookie;
 261
 262 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 263
 264 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 265 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 266
 267 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 268 static void     bridge_clone_destroy(struct ifnet *);
 269
 270 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 271 static void     bridge_mutecaps(struct bridge_softc *);
 272 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 273                     int);
 274 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 275 static void     bridge_init(void *);
 276 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 277 static void     bridge_stop(struct ifnet *, int);
 278 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 279 static void     bridge_qflush(struct ifnet *);
 280 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 281 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 282                     struct rtentry *);
 283 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 284                     struct mbuf *);
 285 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 286
 287 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 288                     struct mbuf *m);
 289
 290 static void     bridge_timer(void *);
 291
 292 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 293                     struct mbuf *, int);
 294 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 295
 296 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 297                     uint16_t, struct bridge_iflist *, int, uint8_t);
 298 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 299                     uint16_t);
 300 static void     bridge_rttrim(struct bridge_softc *);
 301 static void     bridge_rtage(struct bridge_softc *);
 302 static void     bridge_rtflush(struct bridge_softc *, int);
 303 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 304                     uint16_t);
 305
 306 static void     bridge_rtable_init(struct bridge_softc *);
 307 static void     bridge_rtable_fini(struct bridge_softc *);
 308
 309 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 310 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 311                     const uint8_t *, uint16_t);
 312 static int      bridge_rtnode_insert(struct bridge_softc *,
 313                     struct bridge_rtnode *);
 314 static void     bridge_rtnode_destroy(struct bridge_softc *,
 315                     struct bridge_rtnode *);
 316 static void     bridge_rtable_expire(struct ifnet *, int);
 317 static void     bridge_state_change(struct ifnet *, int);
 318
 319 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 320                     const char *name);
 321 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 322                     struct ifnet *ifp);
 323 static void     bridge_delete_member(struct bridge_softc *,
 324                     struct bridge_iflist *, int);
 325 static void     bridge_delete_span(struct bridge_softc *,
 326                     struct bridge_iflist *);
 327
 328 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 329 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 330 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 331 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 332 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 333 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 334 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 335 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 336 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 337 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 338 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 339 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 340 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 341 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 342 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 343 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 344 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 345 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 346 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 347 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 348 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 349 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 350 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 351 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 352 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 353 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 354 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 355 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 356 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 357 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 359 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 360                     int);
 361 static int      bridge_ip_checkbasic(struct mbuf **mp);
 362 #ifdef INET6
 363 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 364 #endif /* INET6 */
 365 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 366                     struct ether_header *, int, struct llc *);
 367 static void     bridge_linkstate(struct ifnet *ifp);
 368 static void     bridge_linkcheck(struct bridge_softc *sc);
 369
 370
 371 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 372 #define VLANTAGOF(_m)   \
 373     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 374
 375 static struct bstp_cb_ops bridge_ops = {
 376         .bcb_state = bridge_state_change,
 377         .bcb_rtage = bridge_rtable_expire
 378 };
 379
 380 SYSCTL_DECL(_net_link);
 381 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 382
 383 /* only pass IP[46] packets when pfil is enabled */
 384 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 385 #define V_pfil_onlyip   VNET(pfil_onlyip)
 386 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 387     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 388     "Only pass IP packets when pfil is enabled");
 389
 390 /* run pfil hooks on the bridge interface */
 391 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 392 #define V_pfil_bridge   VNET(pfil_bridge)
 393 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 394     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 395     "Packet filter on the bridge interface");
 396
 397 /* layer2 filter with ipfw */
 398 VNET_DEFINE_STATIC(int, pfil_ipfw);
 399 #define V_pfil_ipfw     VNET(pfil_ipfw)
 400
 401 /* layer2 ARP filter with ipfw */
 402 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 403 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 404 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 405     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 406     "Filter ARP packets through IPFW layer2");
 407
 408 /* run pfil hooks on the member interface */
 409 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 410 #define V_pfil_member   VNET(pfil_member)
 411 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 412     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 413     "Packet filter on the member interface");
 414
 415 /* run pfil hooks on the physical interface for locally destined packets */
 416 VNET_DEFINE_STATIC(int, pfil_local_phys);
 417 #define V_pfil_local_phys       VNET(pfil_local_phys)
 418 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 419     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 420     "Packet filter on the physical interface for locally destined packets");
 421
 422 /* log STP state changes */
 423 VNET_DEFINE_STATIC(int, log_stp);
 424 #define V_log_stp       VNET(log_stp)
 425 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 426     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 427     "Log STP state changes");
 428
 429 /* share MAC with first bridge member */
 430 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 431 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 432 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 433     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 434     "Inherit MAC address from the first bridge member");
 435
 436 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 437 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 438 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 439     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 440     "Allow overlap of link-local scope "
 441     "zones of a bridge interface and the member interfaces");
 442
 443 struct bridge_control {
 444         int     (*bc_func)(struct bridge_softc *, void *);
 445         int     bc_argsize;
 446         int     bc_flags;
 447 };
 448
 449 #define BC_F_COPYIN             0x01    /* copy arguments in */
 450 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 451 #define BC_F_SUSER              0x04    /* do super-user check */
 452
 453 const struct bridge_control bridge_control_table[] = {
 454         { bridge_ioctl_add,             sizeof(struct ifbreq),
 455           BC_F_COPYIN|BC_F_SUSER },
 456         { bridge_ioctl_del,             sizeof(struct ifbreq),
 457           BC_F_COPYIN|BC_F_SUSER },
 458
 459         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 460           BC_F_COPYIN|BC_F_COPYOUT },
 461         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 462           BC_F_COPYIN|BC_F_SUSER },
 463
 464         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 465           BC_F_COPYIN|BC_F_SUSER },
 466         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 467           BC_F_COPYOUT },
 468
 469         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 470           BC_F_COPYIN|BC_F_COPYOUT },
 471         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 472           BC_F_COPYIN|BC_F_COPYOUT },
 473
 474         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 475           BC_F_COPYIN|BC_F_SUSER },
 476
 477         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 478           BC_F_COPYIN|BC_F_SUSER },
 479         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 480           BC_F_COPYOUT },
 481
 482         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 483           BC_F_COPYIN|BC_F_SUSER },
 484
 485         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 486           BC_F_COPYIN|BC_F_SUSER },
 487
 488         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 489           BC_F_COPYOUT },
 490         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 491           BC_F_COPYIN|BC_F_SUSER },
 492
 493         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 494           BC_F_COPYOUT },
 495         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 496           BC_F_COPYIN|BC_F_SUSER },
 497
 498         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 499           BC_F_COPYOUT },
 500         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 501           BC_F_COPYIN|BC_F_SUSER },
 502
 503         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 504           BC_F_COPYOUT },
 505         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 506           BC_F_COPYIN|BC_F_SUSER },
 507
 508         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 509           BC_F_COPYIN|BC_F_SUSER },
 510
 511         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 512           BC_F_COPYIN|BC_F_SUSER },
 513
 514         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 515           BC_F_COPYIN|BC_F_SUSER },
 516         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 517           BC_F_COPYIN|BC_F_SUSER },
 518
 519         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 520           BC_F_COPYOUT },
 521
 522         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 523           BC_F_COPYOUT },
 524
 525         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 526           BC_F_COPYIN|BC_F_COPYOUT },
 527
 528         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 529           BC_F_COPYIN|BC_F_SUSER },
 530
 531         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 532           BC_F_COPYIN|BC_F_SUSER },
 533
 534         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 535           BC_F_COPYIN|BC_F_SUSER },
 536
 537 };
 538 const int bridge_control_table_size = nitems(bridge_control_table);
 539
 540 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 541 #define V_bridge_list   VNET(bridge_list)
 542 #define BRIDGE_LIST_LOCK_INIT(x)        mtx_init(&V_bridge_list_mtx,    \
 543                                             "if_bridge list", NULL, MTX_DEF)
 544 #define BRIDGE_LIST_LOCK_DESTROY(x)     mtx_destroy(&V_bridge_list_mtx)
 545 #define BRIDGE_LIST_LOCK(x)             mtx_lock(&V_bridge_list_mtx)
 546 #define BRIDGE_LIST_UNLOCK(x)           mtx_unlock(&V_bridge_list_mtx)
 547
 548 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 549 #define V_bridge_cloner VNET(bridge_cloner)
 550
 551 static const char bridge_name[] = "bridge";
 552
 553 static void
 554 vnet_bridge_init(const void *unused __unused)
 555 {
 556
 557         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 558             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 559             UMA_ALIGN_PTR, 0);
 560         BRIDGE_LIST_LOCK_INIT();
 561         LIST_INIT(&V_bridge_list);
 562         V_bridge_cloner = if_clone_simple(bridge_name,
 563             bridge_clone_create, bridge_clone_destroy, 0);
 564 }
 565 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 566     vnet_bridge_init, NULL);
 567
 568 static void
 569 vnet_bridge_uninit(const void *unused __unused)
 570 {
 571
 572         if_clone_detach(V_bridge_cloner);
 573         V_bridge_cloner = NULL;
 574         BRIDGE_LIST_LOCK_DESTROY();
 575
 576         /* Before we can destroy the uma zone, because there are callbacks that
 577          * use it. */
 578         epoch_drain_callbacks(net_epoch_preempt);
 579
 580         uma_zdestroy(V_bridge_rtnode_zone);
 581 }
 582 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 583     vnet_bridge_uninit, NULL);
 584
 585 static int
 586 bridge_modevent(module_t mod, int type, void *data)
 587 {
 588
 589         switch (type) {
 590         case MOD_LOAD:
 591                 bridge_dn_p = bridge_dummynet;
 592                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 593                     ifnet_departure_event, bridge_ifdetach, NULL,
 594                     EVENTHANDLER_PRI_ANY);
 595                 break;
 596         case MOD_UNLOAD:
 597                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 598                     bridge_detach_cookie);
 599                 bridge_dn_p = NULL;
 600                 break;
 601         default:
 602                 return (EOPNOTSUPP);
 603         }
 604         return (0);
 605 }
 606
 607 static moduledata_t bridge_mod = {
 608         "if_bridge",
 609         bridge_modevent,
 610         0
 611 };
 612
 613 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 614 MODULE_VERSION(if_bridge, 1);
 615 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 616
 617 /*
 618  * handler for net.link.bridge.ipfw
 619  */
 620 static int
 621 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 622 {
 623         int enable = V_pfil_ipfw;
 624         int error;
 625
 626         error = sysctl_handle_int(oidp, &enable, 0, req);
 627         enable &= 1;
 628
 629         if (enable != V_pfil_ipfw) {
 630                 V_pfil_ipfw = enable;
 631
 632                 /*
 633                  * Disable pfil so that ipfw doesnt run twice, if the user
 634                  * really wants both then they can re-enable pfil_bridge and/or
 635                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 636                  * layer2 type.
 637                  */
 638                 if (V_pfil_ipfw) {
 639                         V_pfil_onlyip = 0;
 640                         V_pfil_bridge = 0;
 641                         V_pfil_member = 0;
 642                 }
 643         }
 644
 645         return (error);
 646 }
 647 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 648     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 649     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 650     "Layer2 filter with IPFW");
 651
 652 /*
 653  * bridge_clone_create:
 654  *
 655  *      Create a new bridge instance.
 656  */
 657 static int
 658 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 659 {
 660         struct bridge_softc *sc, *sc2;
 661         struct ifnet *bifp, *ifp;
 662         int fb, retry;
 663         unsigned long hostid;
 664
 665         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 666         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 667         if (ifp == NULL) {
 668                 free(sc, M_DEVBUF);
 669                 return (ENOSPC);
 670         }
 671
 672         BRIDGE_LOCK_INIT(sc);
 673         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 674         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 675
 676         /* Initialize our routing table. */
 677         bridge_rtable_init(sc);
 678
 679         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 680
 681         CK_LIST_INIT(&sc->sc_iflist);
 682         CK_LIST_INIT(&sc->sc_spanlist);
 683
 684         ifp->if_softc = sc;
 685         if_initname(ifp, bridge_name, unit);
 686         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 687         ifp->if_ioctl = bridge_ioctl;
 688         ifp->if_transmit = bridge_transmit;
 689         ifp->if_qflush = bridge_qflush;
 690         ifp->if_init = bridge_init;
 691         ifp->if_type = IFT_BRIDGE;
 692
 693         /*
 694          * Generate an ethernet address with a locally administered address.
 695          *
 696          * Since we are using random ethernet addresses for the bridge, it is
 697          * possible that we might have address collisions, so make sure that
 698          * this hardware address isn't already in use on another bridge.
 699          * The first try uses the hostid and falls back to arc4rand().
 700          */
 701         fb = 0;
 702         getcredhostid(curthread->td_ucred, &hostid);
 703         do {
 704                 if (fb || hostid == 0) {
 705                         ether_gen_addr(ifp, &sc->sc_defaddr);
 706                 } else {
 707                         sc->sc_defaddr.octet[0] = 0x2;
 708                         sc->sc_defaddr.octet[1] = (hostid >> 24) & 0xff;
 709                         sc->sc_defaddr.octet[2] = (hostid >> 16) & 0xff;
 710                         sc->sc_defaddr.octet[3] = (hostid >> 8 ) & 0xff;
 711                         sc->sc_defaddr.octet[4] =  hostid        & 0xff;
 712                         sc->sc_defaddr.octet[5] = ifp->if_dunit & 0xff;
 713                 }
 714
 715                 fb = 1;
 716                 retry = 0;
 717                 BRIDGE_LIST_LOCK();
 718                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 719                         bifp = sc2->sc_ifp;
 720                         if (memcmp(sc->sc_defaddr.octet,
 721                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 722                                 retry = 1;
 723                                 break;
 724                         }
 725                 }
 726                 BRIDGE_LIST_UNLOCK();
 727         } while (retry == 1);
 728
 729         bstp_attach(&sc->sc_stp, &bridge_ops);
 730         ether_ifattach(ifp, sc->sc_defaddr.octet);
 731         /* Now undo some of the damage... */
 732         ifp->if_baudrate = 0;
 733         ifp->if_type = IFT_BRIDGE;
 734
 735         BRIDGE_LIST_LOCK();
 736         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 737         BRIDGE_LIST_UNLOCK();
 738
 739         return (0);
 740 }
 741
 742 static void
 743 bridge_clone_destroy_cb(struct epoch_context *ctx)
 744 {
 745         struct bridge_softc *sc;
 746
 747         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 748
 749         BRIDGE_LOCK_DESTROY(sc);
 750         free(sc, M_DEVBUF);
 751 }
 752
 753 /*
 754  * bridge_clone_destroy:
 755  *
 756  *      Destroy a bridge instance.
 757  */
 758 static void
 759 bridge_clone_destroy(struct ifnet *ifp)
 760 {
 761         struct bridge_softc *sc = ifp->if_softc;
 762         struct bridge_iflist *bif;
 763         struct epoch_tracker et;
 764
 765         NET_EPOCH_ENTER_ET(et);
 766         BRIDGE_LOCK(sc);
 767
 768         bridge_stop(ifp, 1);
 769         ifp->if_flags &= ~IFF_UP;
 770
 771         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 772                 bridge_delete_member(sc, bif, 0);
 773
 774         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 775                 bridge_delete_span(sc, bif);
 776         }
 777
 778         /* Tear down the routing table. */
 779         bridge_rtable_fini(sc);
 780
 781         BRIDGE_UNLOCK(sc);
 782
 783         callout_drain(&sc->sc_brcallout);
 784
 785         BRIDGE_LIST_LOCK();
 786         LIST_REMOVE(sc, sc_list);
 787         BRIDGE_LIST_UNLOCK();
 788
 789         bstp_detach(&sc->sc_stp);
 790         NET_EPOCH_EXIT_ET(et);
 791
 792         ether_ifdetach(ifp);
 793         if_free(ifp);
 794
 795         epoch_call(net_epoch_preempt, &sc->sc_epoch_ctx, bridge_clone_destroy_cb);
 796 }
 797
 798 /*
 799  * bridge_ioctl:
 800  *
 801  *      Handle a control request from the operator.
 802  */
 803 static int
 804 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 805 {
 806         struct bridge_softc *sc = ifp->if_softc;
 807         struct ifreq *ifr = (struct ifreq *)data;
 808         struct bridge_iflist *bif;
 809         struct thread *td = curthread;
 810         union {
 811                 struct ifbreq ifbreq;
 812                 struct ifbifconf ifbifconf;
 813                 struct ifbareq ifbareq;
 814                 struct ifbaconf ifbaconf;
 815                 struct ifbrparam ifbrparam;
 816                 struct ifbropreq ifbropreq;
 817         } args;
 818         struct ifdrv *ifd = (struct ifdrv *) data;
 819         const struct bridge_control *bc;
 820         int error = 0, oldmtu;
 821         struct epoch_tracker et;
 822
 823         NET_EPOCH_ENTER_ET(et);
 824
 825         switch (cmd) {
 826
 827         case SIOCADDMULTI:
 828         case SIOCDELMULTI:
 829                 break;
 830
 831         case SIOCGDRVSPEC:
 832         case SIOCSDRVSPEC:
 833                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 834                         error = EINVAL;
 835                         break;
 836                 }
 837                 bc = &bridge_control_table[ifd->ifd_cmd];
 838
 839                 if (cmd == SIOCGDRVSPEC &&
 840                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 841                         error = EINVAL;
 842                         break;
 843                 }
 844                 else if (cmd == SIOCSDRVSPEC &&
 845                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 846                         error = EINVAL;
 847                         break;
 848                 }
 849
 850                 if (bc->bc_flags & BC_F_SUSER) {
 851                         error = priv_check(td, PRIV_NET_BRIDGE);
 852                         if (error)
 853                                 break;
 854                 }
 855
 856                 if (ifd->ifd_len != bc->bc_argsize ||
 857                     ifd->ifd_len > sizeof(args)) {
 858                         error = EINVAL;
 859                         break;
 860                 }
 861
 862                 bzero(&args, sizeof(args));
 863                 if (bc->bc_flags & BC_F_COPYIN) {
 864                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 865                         if (error)
 866                                 break;
 867                 }
 868
 869                 oldmtu = ifp->if_mtu;
 870                 BRIDGE_LOCK(sc);
 871                 error = (*bc->bc_func)(sc, &args);
 872                 BRIDGE_UNLOCK(sc);
 873                 if (error)
 874                         break;
 875
 876                 /*
 877                  * Bridge MTU may change during addition of the first port.
 878                  * If it did, do network layer specific procedure.
 879                  */
 880                 if (ifp->if_mtu != oldmtu) {
 881 #ifdef INET6
 882                         nd6_setmtu(ifp);
 883 #endif
 884                         rt_updatemtu(ifp);
 885                 }
 886
 887                 if (bc->bc_flags & BC_F_COPYOUT)
 888                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 889
 890                 break;
 891
 892         case SIOCSIFFLAGS:
 893                 if (!(ifp->if_flags & IFF_UP) &&
 894                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 895                         /*
 896                          * If interface is marked down and it is running,
 897                          * then stop and disable it.
 898                          */
 899                         BRIDGE_LOCK(sc);
 900                         bridge_stop(ifp, 1);
 901                         BRIDGE_UNLOCK(sc);
 902                 } else if ((ifp->if_flags & IFF_UP) &&
 903                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 904                         /*
 905                          * If interface is marked up and it is stopped, then
 906                          * start it.
 907                          */
 908                         (*ifp->if_init)(sc);
 909                 }
 910                 break;
 911
 912         case SIOCSIFMTU:
 913                 if (ifr->ifr_mtu < 576) {
 914                         error = EINVAL;
 915                         break;
 916                 }
 917                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 918                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 919                         break;
 920                 }
 921                 BRIDGE_LOCK(sc);
 922                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 923                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 924                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 925                                     " != %d\n", sc->sc_ifp->if_xname,
 926                                     bif->bif_ifp->if_mtu,
 927                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 928                                 error = EINVAL;
 929                                 break;
 930                         }
 931                 }
 932                 if (!error)
 933                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 934                 BRIDGE_UNLOCK(sc);
 935                 break;
 936         default:
 937                 /*
 938                  * drop the lock as ether_ioctl() will call bridge_start() and
 939                  * cause the lock to be recursed.
 940                  */
 941                 error = ether_ioctl(ifp, cmd, data);
 942                 break;
 943         }
 944
 945         NET_EPOCH_EXIT_ET(et);
 946
 947         return (error);
 948 }
 949
 950 /*
 951  * bridge_mutecaps:
 952  *
 953  *      Clear or restore unwanted capabilities on the member interface
 954  */
 955 static void
 956 bridge_mutecaps(struct bridge_softc *sc)
 957 {
 958         struct bridge_iflist *bif;
 959         int enabled, mask;
 960
 961         BRIDGE_LOCK_ASSERT(sc);
 962
 963         /* Initial bitmask of capabilities to test */
 964         mask = BRIDGE_IFCAPS_MASK;
 965
 966         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 967                 /* Every member must support it or its disabled */
 968                 mask &= bif->bif_savedcaps;
 969         }
 970
 971         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 972                 enabled = bif->bif_ifp->if_capenable;
 973                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 974                 /* strip off mask bits and enable them again if allowed */
 975                 enabled &= ~BRIDGE_IFCAPS_MASK;
 976                 enabled |= mask;
 977                 BRIDGE_UNLOCK(sc);
 978                 bridge_set_ifcap(sc, bif, enabled);
 979                 BRIDGE_LOCK(sc);
 980         }
 981 }
 982
 983 static void
 984 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 985 {
 986         struct ifnet *ifp = bif->bif_ifp;
 987         struct ifreq ifr;
 988         int error, mask, stuck;
 989
 990         BRIDGE_UNLOCK_ASSERT(sc);
 991
 992         bzero(&ifr, sizeof(ifr));
 993         ifr.ifr_reqcap = set;
 994
 995         if (ifp->if_capenable != set) {
 996                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 997                 if (error)
 998                         if_printf(sc->sc_ifp,
 999                             "error setting capabilities on %s: %d\n",
1000                             ifp->if_xname, error);
1001                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1002                 stuck = ifp->if_capenable & mask & ~set;
1003                 if (stuck != 0)
1004                         if_printf(sc->sc_ifp,
1005                             "can't disable some capabilities on %s: 0x%x\n",
1006                             ifp->if_xname, stuck);
1007         }
1008 }
1009
1010 /*
1011  * bridge_lookup_member:
1012  *
1013  *      Lookup a bridge member interface.
1014  */
1015 static struct bridge_iflist *
1016 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1017 {
1018         struct bridge_iflist *bif;
1019         struct ifnet *ifp;
1020
1021         MPASS(in_epoch(net_epoch_preempt));
1022
1023         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1024                 ifp = bif->bif_ifp;
1025                 if (strcmp(ifp->if_xname, name) == 0)
1026                         return (bif);
1027         }
1028
1029         return (NULL);
1030 }
1031
1032 /*
1033  * bridge_lookup_member_if:
1034  *
1035  *      Lookup a bridge member interface by ifnet*.
1036  */
1037 static struct bridge_iflist *
1038 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1039 {
1040         struct bridge_iflist *bif;
1041
1042         MPASS(in_epoch(net_epoch_preempt));
1043
1044         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1045                 if (bif->bif_ifp == member_ifp)
1046                         return (bif);
1047         }
1048
1049         return (NULL);
1050 }
1051
1052 static void
1053 bridge_delete_member_cb(struct epoch_context *ctx)
1054 {
1055         struct bridge_iflist *bif;
1056
1057         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1058
1059         free(bif, M_DEVBUF);
1060 }
1061
1062 /*
1063  * bridge_delete_member:
1064  *
1065  *      Delete the specified member interface.
1066  */
1067 static void
1068 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1069     int gone)
1070 {
1071         struct ifnet *ifs = bif->bif_ifp;
1072         struct ifnet *fif = NULL;
1073         struct bridge_iflist *bifl;
1074
1075         BRIDGE_LOCK_ASSERT(sc);
1076
1077         if (bif->bif_flags & IFBIF_STP)
1078                 bstp_disable(&bif->bif_stp);
1079
1080         ifs->if_bridge = NULL;
1081         CK_LIST_REMOVE(bif, bif_next);
1082
1083         /*
1084          * If removing the interface that gave the bridge its mac address, set
1085          * the mac address of the bridge to the address of the next member, or
1086          * to its default address if no members are left.
1087          */
1088         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1089                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1090                         bcopy(&sc->sc_defaddr,
1091                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1092                         sc->sc_ifaddr = NULL;
1093                 } else {
1094                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1095                         fif = bifl->bif_ifp;
1096                         bcopy(IF_LLADDR(fif),
1097                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1098                         sc->sc_ifaddr = fif;
1099                 }
1100                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1101         }
1102
1103         bridge_linkcheck(sc);
1104         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1105         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1106         KASSERT(bif->bif_addrcnt == 0,
1107             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1108
1109         ifs->if_bridge_output = NULL;
1110         ifs->if_bridge_input = NULL;
1111         ifs->if_bridge_linkstate = NULL;
1112         BRIDGE_UNLOCK(sc);
1113         if (!gone) {
1114                 switch (ifs->if_type) {
1115                 case IFT_ETHER:
1116                 case IFT_L2VLAN:
1117                         /*
1118                          * Take the interface out of promiscuous mode, but only
1119                          * if it was promiscuous in the first place. It might
1120                          * not be if we're in the bridge_ioctl_add() error path.
1121                          */
1122                         if (ifs->if_flags & IFF_PROMISC)
1123                                 (void) ifpromisc(ifs, 0);
1124                         break;
1125
1126                 case IFT_GIF:
1127                         break;
1128
1129                 default:
1130 #ifdef DIAGNOSTIC
1131                         panic("bridge_delete_member: impossible");
1132 #endif
1133                         break;
1134                 }
1135                 /* reneable any interface capabilities */
1136                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1137         }
1138         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1139         BRIDGE_LOCK(sc);
1140
1141         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1142             bridge_delete_member_cb);
1143 }
1144
1145 /*
1146  * bridge_delete_span:
1147  *
1148  *      Delete the specified span interface.
1149  */
1150 static void
1151 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1152 {
1153         BRIDGE_LOCK_ASSERT(sc);
1154
1155         KASSERT(bif->bif_ifp->if_bridge == NULL,
1156             ("%s: not a span interface", __func__));
1157
1158         CK_LIST_REMOVE(bif, bif_next);
1159
1160         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1161             bridge_delete_member_cb);
1162 }
1163
1164 static int
1165 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1166 {
1167         struct ifbreq *req = arg;
1168         struct bridge_iflist *bif = NULL;
1169         struct ifnet *ifs;
1170         int error = 0;
1171
1172         ifs = ifunit(req->ifbr_ifsname);
1173         if (ifs == NULL)
1174                 return (ENOENT);
1175         if (ifs->if_ioctl == NULL)      /* must be supported */
1176                 return (EINVAL);
1177
1178         /* If it's in the span list, it can't be a member. */
1179         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1180                 if (ifs == bif->bif_ifp)
1181                         return (EBUSY);
1182
1183         if (ifs->if_bridge == sc)
1184                 return (EEXIST);
1185
1186         if (ifs->if_bridge != NULL)
1187                 return (EBUSY);
1188
1189         switch (ifs->if_type) {
1190         case IFT_ETHER:
1191         case IFT_L2VLAN:
1192         case IFT_GIF:
1193                 /* permitted interface types */
1194                 break;
1195         default:
1196                 return (EINVAL);
1197         }
1198
1199 #ifdef INET6
1200         /*
1201          * Two valid inet6 addresses with link-local scope must not be
1202          * on the parent interface and the member interfaces at the
1203          * same time.  This restriction is needed to prevent violation
1204          * of link-local scope zone.  Attempts to add a member
1205          * interface which has inet6 addresses when the parent has
1206          * inet6 triggers removal of all inet6 addresses on the member
1207          * interface.
1208          */
1209
1210         /* Check if the parent interface has a link-local scope addr. */
1211         if (V_allow_llz_overlap == 0 &&
1212             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1213                 /*
1214                  * If any, remove all inet6 addresses from the member
1215                  * interfaces.
1216                  */
1217                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1218                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1219                                 BRIDGE_UNLOCK(sc);
1220                                 in6_ifdetach(bif->bif_ifp);
1221                                 BRIDGE_LOCK(sc);
1222                                 if_printf(sc->sc_ifp,
1223                                     "IPv6 addresses on %s have been removed "
1224                                     "before adding it as a member to prevent "
1225                                     "IPv6 address scope violation.\n",
1226                                     bif->bif_ifp->if_xname);
1227                         }
1228                 }
1229                 if (in6ifa_llaonifp(ifs)) {
1230                         BRIDGE_UNLOCK(sc);
1231                         in6_ifdetach(ifs);
1232                         BRIDGE_LOCK(sc);
1233                         if_printf(sc->sc_ifp,
1234                             "IPv6 addresses on %s have been removed "
1235                             "before adding it as a member to prevent "
1236                             "IPv6 address scope violation.\n",
1237                             ifs->if_xname);
1238                 }
1239         }
1240 #endif
1241         /* Allow the first Ethernet member to define the MTU */
1242         if (CK_LIST_EMPTY(&sc->sc_iflist))
1243                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1244         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1245                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1246                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1247                 return (EINVAL);
1248         }
1249
1250         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1251         if (bif == NULL)
1252                 return (ENOMEM);
1253
1254         bif->bif_ifp = ifs;
1255         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1256         bif->bif_savedcaps = ifs->if_capenable;
1257
1258         /*
1259          * Assign the interface's MAC address to the bridge if it's the first
1260          * member and the MAC address of the bridge has not been changed from
1261          * the default randomly generated one.
1262          */
1263         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1264             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1265                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1266                 sc->sc_ifaddr = ifs;
1267                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1268         }
1269
1270         ifs->if_bridge = sc;
1271         ifs->if_bridge_output = bridge_output;
1272         ifs->if_bridge_input = bridge_input;
1273         ifs->if_bridge_linkstate = bridge_linkstate;
1274         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1275         /*
1276          * XXX: XLOCK HERE!?!
1277          *
1278          * NOTE: insert_***HEAD*** should be safe for the traversals.
1279          */
1280         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1281
1282         /* Set interface capabilities to the intersection set of all members */
1283         bridge_mutecaps(sc);
1284         bridge_linkcheck(sc);
1285
1286         /* Place the interface into promiscuous mode */
1287         switch (ifs->if_type) {
1288                 case IFT_ETHER:
1289                 case IFT_L2VLAN:
1290                         BRIDGE_UNLOCK(sc);
1291                         error = ifpromisc(ifs, 1);
1292                         BRIDGE_LOCK(sc);
1293                         break;
1294         }
1295
1296         if (error)
1297                 bridge_delete_member(sc, bif, 0);
1298         return (error);
1299 }
1300
1301 static int
1302 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1303 {
1304         struct ifbreq *req = arg;
1305         struct bridge_iflist *bif;
1306
1307         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1308         if (bif == NULL)
1309                 return (ENOENT);
1310
1311         bridge_delete_member(sc, bif, 0);
1312
1313         return (0);
1314 }
1315
1316 static int
1317 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1318 {
1319         struct ifbreq *req = arg;
1320         struct bridge_iflist *bif;
1321         struct bstp_port *bp;
1322
1323         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1324         if (bif == NULL)
1325                 return (ENOENT);
1326
1327         bp = &bif->bif_stp;
1328         req->ifbr_ifsflags = bif->bif_flags;
1329         req->ifbr_state = bp->bp_state;
1330         req->ifbr_priority = bp->bp_priority;
1331         req->ifbr_path_cost = bp->bp_path_cost;
1332         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1333         req->ifbr_proto = bp->bp_protover;
1334         req->ifbr_role = bp->bp_role;
1335         req->ifbr_stpflags = bp->bp_flags;
1336         req->ifbr_addrcnt = bif->bif_addrcnt;
1337         req->ifbr_addrmax = bif->bif_addrmax;
1338         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1339
1340         /* Copy STP state options as flags */
1341         if (bp->bp_operedge)
1342                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1343         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1344                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1345         if (bp->bp_ptp_link)
1346                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1347         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1348                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1349         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1350                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1351         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1352                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1353         return (0);
1354 }
1355
1356 static int
1357 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1358 {
1359         struct ifbreq *req = arg;
1360         struct bridge_iflist *bif;
1361         struct bstp_port *bp;
1362         int error;
1363
1364         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1365         if (bif == NULL)
1366                 return (ENOENT);
1367         bp = &bif->bif_stp;
1368
1369         if (req->ifbr_ifsflags & IFBIF_SPAN)
1370                 /* SPAN is readonly */
1371                 return (EINVAL);
1372
1373         if (req->ifbr_ifsflags & IFBIF_STP) {
1374                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1375                         error = bstp_enable(&bif->bif_stp);
1376                         if (error)
1377                                 return (error);
1378                 }
1379         } else {
1380                 if ((bif->bif_flags & IFBIF_STP) != 0)
1381                         bstp_disable(&bif->bif_stp);
1382         }
1383
1384         /* Pass on STP flags */
1385         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1386         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1387         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1388         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1389
1390         /* Save the bits relating to the bridge */
1391         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1392
1393         return (0);
1394 }
1395
1396 static int
1397 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1398 {
1399         struct ifbrparam *param = arg;
1400
1401         sc->sc_brtmax = param->ifbrp_csize;
1402         bridge_rttrim(sc);
1403
1404         return (0);
1405 }
1406
1407 static int
1408 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1409 {
1410         struct ifbrparam *param = arg;
1411
1412         param->ifbrp_csize = sc->sc_brtmax;
1413
1414         return (0);
1415 }
1416
1417 static int
1418 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1419 {
1420         struct ifbifconf *bifc = arg;
1421         struct bridge_iflist *bif;
1422         struct ifbreq breq;
1423         char *buf, *outbuf;
1424         int count, buflen, len, error = 0;
1425
1426         count = 0;
1427         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1428                 count++;
1429         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1430                 count++;
1431
1432         buflen = sizeof(breq) * count;
1433         if (bifc->ifbic_len == 0) {
1434                 bifc->ifbic_len = buflen;
1435                 return (0);
1436         }
1437         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1438         if (outbuf == NULL)
1439                 return (ENOMEM);
1440
1441         count = 0;
1442         buf = outbuf;
1443         len = min(bifc->ifbic_len, buflen);
1444         bzero(&breq, sizeof(breq));
1445         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1446                 if (len < sizeof(breq))
1447                         break;
1448
1449                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1450                     sizeof(breq.ifbr_ifsname));
1451                 /* Fill in the ifbreq structure */
1452                 error = bridge_ioctl_gifflags(sc, &breq);
1453                 if (error)
1454                         break;
1455                 memcpy(buf, &breq, sizeof(breq));
1456                 count++;
1457                 buf += sizeof(breq);
1458                 len -= sizeof(breq);
1459         }
1460         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1461                 if (len < sizeof(breq))
1462                         break;
1463
1464                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1465                     sizeof(breq.ifbr_ifsname));
1466                 breq.ifbr_ifsflags = bif->bif_flags;
1467                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1468                 memcpy(buf, &breq, sizeof(breq));
1469                 count++;
1470                 buf += sizeof(breq);
1471                 len -= sizeof(breq);
1472         }
1473
1474         BRIDGE_UNLOCK(sc);
1475         bifc->ifbic_len = sizeof(breq) * count;
1476         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1477         BRIDGE_LOCK(sc);
1478         free(outbuf, M_TEMP);
1479         return (error);
1480 }
1481
1482 static int
1483 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1484 {
1485         struct ifbaconf *bac = arg;
1486         struct bridge_rtnode *brt;
1487         struct ifbareq bareq;
1488         char *buf, *outbuf;
1489         int count, buflen, len, error = 0;
1490
1491         if (bac->ifbac_len == 0)
1492                 return (0);
1493
1494         count = 0;
1495         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1496                 count++;
1497         buflen = sizeof(bareq) * count;
1498
1499         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1500         if (outbuf == NULL)
1501                 return (ENOMEM);
1502
1503         count = 0;
1504         buf = outbuf;
1505         len = min(bac->ifbac_len, buflen);
1506         bzero(&bareq, sizeof(bareq));
1507         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1508                 if (len < sizeof(bareq))
1509                         goto out;
1510                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1511                     sizeof(bareq.ifba_ifsname));
1512                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1513                 bareq.ifba_vlan = brt->brt_vlan;
1514                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1515                                 time_uptime < brt->brt_expire)
1516                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1517                 else
1518                         bareq.ifba_expire = 0;
1519                 bareq.ifba_flags = brt->brt_flags;
1520
1521                 memcpy(buf, &bareq, sizeof(bareq));
1522                 count++;
1523                 buf += sizeof(bareq);
1524                 len -= sizeof(bareq);
1525         }
1526 out:
1527         BRIDGE_UNLOCK(sc);
1528         bac->ifbac_len = sizeof(bareq) * count;
1529         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1530         BRIDGE_LOCK(sc);
1531         free(outbuf, M_TEMP);
1532         return (error);
1533 }
1534
1535 static int
1536 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1537 {
1538         struct ifbareq *req = arg;
1539         struct bridge_iflist *bif;
1540         int error;
1541
1542         MPASS(in_epoch(net_epoch_preempt));
1543
1544         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1545         if (bif == NULL)
1546                 return (ENOENT);
1547
1548         /* bridge_rtupdate() may acquire the lock. */
1549         BRIDGE_UNLOCK(sc);
1550         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1551             req->ifba_flags);
1552         BRIDGE_LOCK(sc);
1553
1554         return (error);
1555 }
1556
1557 static int
1558 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1559 {
1560         struct ifbrparam *param = arg;
1561
1562         sc->sc_brttimeout = param->ifbrp_ctime;
1563         return (0);
1564 }
1565
1566 static int
1567 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1568 {
1569         struct ifbrparam *param = arg;
1570
1571         param->ifbrp_ctime = sc->sc_brttimeout;
1572         return (0);
1573 }
1574
1575 static int
1576 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1577 {
1578         struct ifbareq *req = arg;
1579
1580         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1581 }
1582
1583 static int
1584 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1585 {
1586         struct ifbreq *req = arg;
1587
1588         bridge_rtflush(sc, req->ifbr_ifsflags);
1589         return (0);
1590 }
1591
1592 static int
1593 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1594 {
1595         struct ifbrparam *param = arg;
1596         struct bstp_state *bs = &sc->sc_stp;
1597
1598         param->ifbrp_prio = bs->bs_bridge_priority;
1599         return (0);
1600 }
1601
1602 static int
1603 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1604 {
1605         struct ifbrparam *param = arg;
1606
1607         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1608 }
1609
1610 static int
1611 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1612 {
1613         struct ifbrparam *param = arg;
1614         struct bstp_state *bs = &sc->sc_stp;
1615
1616         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1617         return (0);
1618 }
1619
1620 static int
1621 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1622 {
1623         struct ifbrparam *param = arg;
1624
1625         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1626 }
1627
1628 static int
1629 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1630 {
1631         struct ifbrparam *param = arg;
1632         struct bstp_state *bs = &sc->sc_stp;
1633
1634         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1635         return (0);
1636 }
1637
1638 static int
1639 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1640 {
1641         struct ifbrparam *param = arg;
1642
1643         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1644 }
1645
1646 static int
1647 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1648 {
1649         struct ifbrparam *param = arg;
1650         struct bstp_state *bs = &sc->sc_stp;
1651
1652         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1653         return (0);
1654 }
1655
1656 static int
1657 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1658 {
1659         struct ifbrparam *param = arg;
1660
1661         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1662 }
1663
1664 static int
1665 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1666 {
1667         struct ifbreq *req = arg;
1668         struct bridge_iflist *bif;
1669
1670         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1671         if (bif == NULL)
1672                 return (ENOENT);
1673
1674         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1675 }
1676
1677 static int
1678 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1679 {
1680         struct ifbreq *req = arg;
1681         struct bridge_iflist *bif;
1682
1683         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1684         if (bif == NULL)
1685                 return (ENOENT);
1686
1687         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1688 }
1689
1690 static int
1691 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1692 {
1693         struct ifbreq *req = arg;
1694         struct bridge_iflist *bif;
1695
1696         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1697         if (bif == NULL)
1698                 return (ENOENT);
1699
1700         bif->bif_addrmax = req->ifbr_addrmax;
1701         return (0);
1702 }
1703
1704 static int
1705 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1706 {
1707         struct ifbreq *req = arg;
1708         struct bridge_iflist *bif = NULL;
1709         struct ifnet *ifs;
1710
1711         ifs = ifunit(req->ifbr_ifsname);
1712         if (ifs == NULL)
1713                 return (ENOENT);
1714
1715         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1716                 if (ifs == bif->bif_ifp)
1717                         return (EBUSY);
1718
1719         if (ifs->if_bridge != NULL)
1720                 return (EBUSY);
1721
1722         switch (ifs->if_type) {
1723                 case IFT_ETHER:
1724                 case IFT_GIF:
1725                 case IFT_L2VLAN:
1726                         break;
1727                 default:
1728                         return (EINVAL);
1729         }
1730
1731         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1732         if (bif == NULL)
1733                 return (ENOMEM);
1734
1735         bif->bif_ifp = ifs;
1736         bif->bif_flags = IFBIF_SPAN;
1737
1738         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1739
1740         return (0);
1741 }
1742
1743 static int
1744 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1745 {
1746         struct ifbreq *req = arg;
1747         struct bridge_iflist *bif;
1748         struct ifnet *ifs;
1749
1750         ifs = ifunit(req->ifbr_ifsname);
1751         if (ifs == NULL)
1752                 return (ENOENT);
1753
1754         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1755                 if (ifs == bif->bif_ifp)
1756                         break;
1757
1758         if (bif == NULL)
1759                 return (ENOENT);
1760
1761         bridge_delete_span(sc, bif);
1762
1763         return (0);
1764 }
1765
1766 static int
1767 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1768 {
1769         struct ifbropreq *req = arg;
1770         struct bstp_state *bs = &sc->sc_stp;
1771         struct bstp_port *root_port;
1772
1773         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1774         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1775         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1776
1777         root_port = bs->bs_root_port;
1778         if (root_port == NULL)
1779                 req->ifbop_root_port = 0;
1780         else
1781                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1782
1783         req->ifbop_holdcount = bs->bs_txholdcount;
1784         req->ifbop_priority = bs->bs_bridge_priority;
1785         req->ifbop_protocol = bs->bs_protover;
1786         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1787         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1788         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1789         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1790         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1791         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1792
1793         return (0);
1794 }
1795
1796 static int
1797 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1798 {
1799         struct ifbrparam *param = arg;
1800
1801         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1802         return (0);
1803 }
1804
1805 static int
1806 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1807 {
1808         struct ifbpstpconf *bifstp = arg;
1809         struct bridge_iflist *bif;
1810         struct bstp_port *bp;
1811         struct ifbpstpreq bpreq;
1812         char *buf, *outbuf;
1813         int count, buflen, len, error = 0;
1814
1815         count = 0;
1816         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1817                 if ((bif->bif_flags & IFBIF_STP) != 0)
1818                         count++;
1819         }
1820
1821         buflen = sizeof(bpreq) * count;
1822         if (bifstp->ifbpstp_len == 0) {
1823                 bifstp->ifbpstp_len = buflen;
1824                 return (0);
1825         }
1826
1827         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1828         if (outbuf == NULL)
1829                 return (ENOMEM);
1830
1831         count = 0;
1832         buf = outbuf;
1833         len = min(bifstp->ifbpstp_len, buflen);
1834         bzero(&bpreq, sizeof(bpreq));
1835         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1836                 if (len < sizeof(bpreq))
1837                         break;
1838
1839                 if ((bif->bif_flags & IFBIF_STP) == 0)
1840                         continue;
1841
1842                 bp = &bif->bif_stp;
1843                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1844                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1845                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1846                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1847                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1848                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1849
1850                 memcpy(buf, &bpreq, sizeof(bpreq));
1851                 count++;
1852                 buf += sizeof(bpreq);
1853                 len -= sizeof(bpreq);
1854         }
1855
1856         BRIDGE_UNLOCK(sc);
1857         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1858         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1859         BRIDGE_LOCK(sc);
1860         free(outbuf, M_TEMP);
1861         return (error);
1862 }
1863
1864 static int
1865 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1866 {
1867         struct ifbrparam *param = arg;
1868
1869         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1870 }
1871
1872 static int
1873 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1874 {
1875         struct ifbrparam *param = arg;
1876
1877         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1878 }
1879
1880 /*
1881  * bridge_ifdetach:
1882  *
1883  *      Detach an interface from a bridge.  Called when a member
1884  *      interface is detaching.
1885  */
1886 static void
1887 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1888 {
1889         struct bridge_softc *sc = ifp->if_bridge;
1890         struct bridge_iflist *bif;
1891         struct epoch_tracker et;
1892
1893         if (ifp->if_flags & IFF_RENAMING)
1894                 return;
1895         if (V_bridge_cloner == NULL) {
1896                 /*
1897                  * This detach handler can be called after
1898                  * vnet_bridge_uninit().  Just return in that case.
1899                  */
1900                 return;
1901         }
1902         NET_EPOCH_ENTER_ET(et);
1903         /* Check if the interface is a bridge member */
1904         if (sc != NULL) {
1905                 BRIDGE_LOCK(sc);
1906
1907                 bif = bridge_lookup_member_if(sc, ifp);
1908                 if (bif != NULL)
1909                         bridge_delete_member(sc, bif, 1);
1910
1911                 BRIDGE_UNLOCK(sc);
1912                 NET_EPOCH_EXIT_ET(et);
1913                 return;
1914         }
1915
1916         /* Check if the interface is a span port */
1917         BRIDGE_LIST_LOCK();
1918         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1919                 BRIDGE_LOCK(sc);
1920                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1921                         if (ifp == bif->bif_ifp) {
1922                                 bridge_delete_span(sc, bif);
1923                                 break;
1924                         }
1925
1926                 BRIDGE_UNLOCK(sc);
1927         }
1928         BRIDGE_LIST_UNLOCK();
1929         NET_EPOCH_EXIT_ET(et);
1930 }
1931
1932 /*
1933  * bridge_init:
1934  *
1935  *      Initialize a bridge interface.
1936  */
1937 static void
1938 bridge_init(void *xsc)
1939 {
1940         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1941         struct ifnet *ifp = sc->sc_ifp;
1942
1943         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1944                 return;
1945
1946         BRIDGE_LOCK(sc);
1947         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1948             bridge_timer, sc);
1949
1950         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1951         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1952
1953         BRIDGE_UNLOCK(sc);
1954 }
1955
1956 /*
1957  * bridge_stop:
1958  *
1959  *      Stop the bridge interface.
1960  */
1961 static void
1962 bridge_stop(struct ifnet *ifp, int disable)
1963 {
1964         struct bridge_softc *sc = ifp->if_softc;
1965
1966         BRIDGE_LOCK_ASSERT(sc);
1967
1968         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1969                 return;
1970
1971         callout_stop(&sc->sc_brcallout);
1972         bstp_stop(&sc->sc_stp);
1973
1974         bridge_rtflush(sc, IFBF_FLUSHDYN);
1975
1976         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1977 }
1978
1979 /*
1980  * bridge_enqueue:
1981  *
1982  *      Enqueue a packet on a bridge member interface.
1983  *
1984  */
1985 static int
1986 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1987 {
1988         int len, err = 0;
1989         short mflags;
1990         struct mbuf *m0;
1991
1992         /* We may be sending a fragment so traverse the mbuf */
1993         for (; m; m = m0) {
1994                 m0 = m->m_nextpkt;
1995                 m->m_nextpkt = NULL;
1996                 len = m->m_pkthdr.len;
1997                 mflags = m->m_flags;
1998
1999                 /*
2000                  * If underlying interface can not do VLAN tag insertion itself
2001                  * then attach a packet tag that holds it.
2002                  */
2003                 if ((m->m_flags & M_VLANTAG) &&
2004                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2005                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2006                         if (m == NULL) {
2007                                 if_printf(dst_ifp,
2008                                     "unable to prepend VLAN header\n");
2009                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2010                                 continue;
2011                         }
2012                         m->m_flags &= ~M_VLANTAG;
2013                 }
2014
2015                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2016                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2017                         m_freem(m0);
2018                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2019                         break;
2020                 }
2021
2022                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2023                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2024                 if (mflags & M_MCAST)
2025                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2026         }
2027
2028         return (err);
2029 }
2030
2031 /*
2032  * bridge_dummynet:
2033  *
2034  *      Receive a queued packet from dummynet and pass it on to the output
2035  *      interface.
2036  *
2037  *      The mbuf has the Ethernet header already attached.
2038  */
2039 static void
2040 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2041 {
2042         struct bridge_softc *sc;
2043
2044         sc = ifp->if_bridge;
2045
2046         /*
2047          * The packet didnt originate from a member interface. This should only
2048          * ever happen if a member interface is removed while packets are
2049          * queued for it.
2050          */
2051         if (sc == NULL) {
2052                 m_freem(m);
2053                 return;
2054         }
2055
2056         if (PFIL_HOOKED(&V_inet_pfil_hook)
2057 #ifdef INET6
2058             || PFIL_HOOKED(&V_inet6_pfil_hook)
2059 #endif
2060             ) {
2061                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2062                         return;
2063                 if (m == NULL)
2064                         return;
2065         }
2066
2067         bridge_enqueue(sc, ifp, m);
2068 }
2069
2070 /*
2071  * bridge_output:
2072  *
2073  *      Send output from a bridge member interface.  This
2074  *      performs the bridging function for locally originated
2075  *      packets.
2076  *
2077  *      The mbuf has the Ethernet header already attached.  We must
2078  *      enqueue or free the mbuf before returning.
2079  */
2080 static int
2081 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2082     struct rtentry *rt)
2083 {
2084         struct epoch_tracker et;
2085         struct ether_header *eh;
2086         struct ifnet *dst_if;
2087         struct bridge_softc *sc;
2088         uint16_t vlan;
2089
2090         NET_EPOCH_ENTER_ET(et);
2091
2092         if (m->m_len < ETHER_HDR_LEN) {
2093                 m = m_pullup(m, ETHER_HDR_LEN);
2094                 if (m == NULL) {
2095                         NET_EPOCH_EXIT_ET(et);
2096                         return (0);
2097                 }
2098         }
2099
2100         eh = mtod(m, struct ether_header *);
2101         sc = ifp->if_bridge;
2102         vlan = VLANTAGOF(m);
2103
2104         /*
2105          * If bridge is down, but the original output interface is up,
2106          * go ahead and send out that interface.  Otherwise, the packet
2107          * is dropped below.
2108          */
2109         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2110                 dst_if = ifp;
2111                 goto sendunicast;
2112         }
2113
2114         /*
2115          * If the packet is a multicast, or we don't know a better way to
2116          * get there, send to all interfaces.
2117          */
2118         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2119                 dst_if = NULL;
2120         else
2121                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2122         if (dst_if == NULL) {
2123                 struct bridge_iflist *bif;
2124                 struct mbuf *mc;
2125                 int used = 0;
2126
2127                 bridge_span(sc, m);
2128
2129                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2130                         dst_if = bif->bif_ifp;
2131
2132                         if (dst_if->if_type == IFT_GIF)
2133                                 continue;
2134                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2135                                 continue;
2136
2137                         /*
2138                          * If this is not the original output interface,
2139                          * and the interface is participating in spanning
2140                          * tree, make sure the port is in a state that
2141                          * allows forwarding.
2142                          */
2143                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2144                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2145                                 continue;
2146
2147                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2148                                 used = 1;
2149                                 mc = m;
2150                         } else {
2151                                 mc = m_copypacket(m, M_NOWAIT);
2152                                 if (mc == NULL) {
2153                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2154                                         continue;
2155                                 }
2156                         }
2157
2158                         bridge_enqueue(sc, dst_if, mc);
2159                 }
2160                 if (used == 0)
2161                         m_freem(m);
2162                 NET_EPOCH_EXIT_ET(et);
2163                 return (0);
2164         }
2165
2166 sendunicast:
2167         /*
2168          * XXX Spanning tree consideration here?
2169          */
2170
2171         bridge_span(sc, m);
2172         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2173                 m_freem(m);
2174                 NET_EPOCH_EXIT_ET(et);
2175                 return (0);
2176         }
2177
2178         bridge_enqueue(sc, dst_if, m);
2179         NET_EPOCH_EXIT_ET(et);
2180         return (0);
2181 }
2182
2183 /*
2184  * bridge_transmit:
2185  *
2186  *      Do output on a bridge.
2187  *
2188  */
2189 static int
2190 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2191 {
2192         struct bridge_softc *sc;
2193         struct ether_header *eh;
2194         struct ifnet *dst_if;
2195         int error = 0;
2196
2197         sc = ifp->if_softc;
2198
2199         ETHER_BPF_MTAP(ifp, m);
2200
2201         eh = mtod(m, struct ether_header *);
2202
2203         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2204             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2205                 error = bridge_enqueue(sc, dst_if, m);
2206         } else
2207                 bridge_broadcast(sc, ifp, m, 0);
2208
2209         return (error);
2210 }
2211
2212 /*
2213  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2214  */
2215 static void
2216 bridge_qflush(struct ifnet *ifp __unused)
2217 {
2218 }
2219
2220 /*
2221  * bridge_forward:
2222  *
2223  *      The forwarding function of the bridge.
2224  *
2225  *      NOTE: Releases the lock on return.
2226  */
2227 static void
2228 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2229     struct mbuf *m)
2230 {
2231         struct bridge_iflist *dbif;
2232         struct ifnet *src_if, *dst_if, *ifp;
2233         struct ether_header *eh;
2234         uint16_t vlan;
2235         uint8_t *dst;
2236         int error;
2237
2238         MPASS(in_epoch(net_epoch_preempt));
2239
2240         src_if = m->m_pkthdr.rcvif;
2241         ifp = sc->sc_ifp;
2242
2243         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2244         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2245         vlan = VLANTAGOF(m);
2246
2247         if ((sbif->bif_flags & IFBIF_STP) &&
2248             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2249                 goto drop;
2250
2251         eh = mtod(m, struct ether_header *);
2252         dst = eh->ether_dhost;
2253
2254         /* If the interface is learning, record the address. */
2255         if (sbif->bif_flags & IFBIF_LEARNING) {
2256                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2257                     sbif, 0, IFBAF_DYNAMIC);
2258                 /*
2259                  * If the interface has addresses limits then deny any source
2260                  * that is not in the cache.
2261                  */
2262                 if (error && sbif->bif_addrmax)
2263                         goto drop;
2264         }
2265
2266         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2267             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2268                 goto drop;
2269
2270         /*
2271          * At this point, the port either doesn't participate
2272          * in spanning tree or it is in the forwarding state.
2273          */
2274
2275         /*
2276          * If the packet is unicast, destined for someone on
2277          * "this" side of the bridge, drop it.
2278          */
2279         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2280                 dst_if = bridge_rtlookup(sc, dst, vlan);
2281                 if (src_if == dst_if)
2282                         goto drop;
2283         } else {
2284                 /*
2285                  * Check if its a reserved multicast address, any address
2286                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2287                  * bridge.
2288                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2289                  */
2290                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2291                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2292                     dst[4] == 0x00 && dst[5] <= 0x0f)
2293                         goto drop;
2294
2295                 /* ...forward it to all interfaces. */
2296                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2297                 dst_if = NULL;
2298         }
2299
2300         /*
2301          * If we have a destination interface which is a member of our bridge,
2302          * OR this is a unicast packet, push it through the bpf(4) machinery.
2303          * For broadcast or multicast packets, don't bother because it will
2304          * be reinjected into ether_input. We do this before we pass the packets
2305          * through the pfil(9) framework, as it is possible that pfil(9) will
2306          * drop the packet, or possibly modify it, making it difficult to debug
2307          * firewall issues on the bridge.
2308          */
2309         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2310                 ETHER_BPF_MTAP(ifp, m);
2311
2312         /* run the packet filter */
2313         if (PFIL_HOOKED(&V_inet_pfil_hook)
2314 #ifdef INET6
2315             || PFIL_HOOKED(&V_inet6_pfil_hook)
2316 #endif
2317             ) {
2318                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2319                         return;
2320                 if (m == NULL)
2321                         return;
2322         }
2323
2324         if (dst_if == NULL) {
2325                 bridge_broadcast(sc, src_if, m, 1);
2326                 return;
2327         }
2328
2329         /*
2330          * At this point, we're dealing with a unicast frame
2331          * going to a different interface.
2332          */
2333         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2334                 goto drop;
2335
2336         dbif = bridge_lookup_member_if(sc, dst_if);
2337         if (dbif == NULL)
2338                 /* Not a member of the bridge (anymore?) */
2339                 goto drop;
2340
2341         /* Private segments can not talk to each other */
2342         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2343                 goto drop;
2344
2345         if ((dbif->bif_flags & IFBIF_STP) &&
2346             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2347                 goto drop;
2348
2349         if (PFIL_HOOKED(&V_inet_pfil_hook)
2350 #ifdef INET6
2351             || PFIL_HOOKED(&V_inet6_pfil_hook)
2352 #endif
2353             ) {
2354                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2355                         return;
2356                 if (m == NULL)
2357                         return;
2358         }
2359
2360         bridge_enqueue(sc, dst_if, m);
2361         return;
2362
2363 drop:
2364         m_freem(m);
2365 }
2366
2367 /*
2368  * bridge_input:
2369  *
2370  *      Receive input from a member interface.  Queue the packet for
2371  *      bridging if it is not for us.
2372  */
2373 static struct mbuf *
2374 bridge_input(struct ifnet *ifp, struct mbuf *m)
2375 {
2376         struct epoch_tracker et;
2377         struct bridge_softc *sc = ifp->if_bridge;
2378         struct bridge_iflist *bif, *bif2;
2379         struct ifnet *bifp;
2380         struct ether_header *eh;
2381         struct mbuf *mc, *mc2;
2382         uint16_t vlan;
2383         int error;
2384
2385         NET_EPOCH_ENTER_ET(et);
2386
2387         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2388                 NET_EPOCH_EXIT_ET(et);
2389                 return (m);
2390         }
2391
2392         bifp = sc->sc_ifp;
2393         vlan = VLANTAGOF(m);
2394
2395         /*
2396          * Implement support for bridge monitoring. If this flag has been
2397          * set on this interface, discard the packet once we push it through
2398          * the bpf(4) machinery, but before we do, increment the byte and
2399          * packet counters associated with this interface.
2400          */
2401         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2402                 m->m_pkthdr.rcvif  = bifp;
2403                 ETHER_BPF_MTAP(bifp, m);
2404                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2405                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2406                 m_freem(m);
2407                 NET_EPOCH_EXIT_ET(et);
2408                 return (NULL);
2409         }
2410         bif = bridge_lookup_member_if(sc, ifp);
2411         if (bif == NULL) {
2412                 NET_EPOCH_EXIT_ET(et);
2413                 return (m);
2414         }
2415
2416         eh = mtod(m, struct ether_header *);
2417
2418         bridge_span(sc, m);
2419
2420         if (m->m_flags & (M_BCAST|M_MCAST)) {
2421                 /* Tap off 802.1D packets; they do not get forwarded. */
2422                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2423                     ETHER_ADDR_LEN) == 0) {
2424                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2425                         NET_EPOCH_EXIT_ET(et);
2426                         return (NULL);
2427                 }
2428
2429                 if ((bif->bif_flags & IFBIF_STP) &&
2430                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2431                         NET_EPOCH_EXIT_ET(et);
2432                         return (m);
2433                 }
2434
2435                 /*
2436                  * Make a deep copy of the packet and enqueue the copy
2437                  * for bridge processing; return the original packet for
2438                  * local processing.
2439                  */
2440                 mc = m_dup(m, M_NOWAIT);
2441                 if (mc == NULL) {
2442                         NET_EPOCH_EXIT_ET(et);
2443                         return (m);
2444                 }
2445
2446                 /* Perform the bridge forwarding function with the copy. */
2447                 bridge_forward(sc, bif, mc);
2448
2449                 /*
2450                  * Reinject the mbuf as arriving on the bridge so we have a
2451                  * chance at claiming multicast packets. We can not loop back
2452                  * here from ether_input as a bridge is never a member of a
2453                  * bridge.
2454                  */
2455                 KASSERT(bifp->if_bridge == NULL,
2456                     ("loop created in bridge_input"));
2457                 mc2 = m_dup(m, M_NOWAIT);
2458                 if (mc2 != NULL) {
2459                         /* Keep the layer3 header aligned */
2460                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2461                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2462                 }
2463                 if (mc2 != NULL) {
2464                         mc2->m_pkthdr.rcvif = bifp;
2465                         (*bifp->if_input)(bifp, mc2);
2466                 }
2467
2468                 /* Return the original packet for local processing. */
2469                 NET_EPOCH_EXIT_ET(et);
2470                 return (m);
2471         }
2472
2473         if ((bif->bif_flags & IFBIF_STP) &&
2474             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2475                 NET_EPOCH_EXIT_ET(et);
2476                 return (m);
2477         }
2478
2479 #if (defined(INET) || defined(INET6))
2480 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2481         || ((iface)->if_carp \
2482             && (*carp_forus_p)((iface), eh->ether_dhost))
2483 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2484         || ((iface)->if_carp \
2485             && (*carp_forus_p)((iface), eh->ether_shost))
2486 #else
2487 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2488 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2489 #endif
2490
2491 #ifdef INET6
2492 #   define OR_PFIL_HOOKED_INET6 \
2493         || PFIL_HOOKED(&V_inet6_pfil_hook)
2494 #else
2495 #   define OR_PFIL_HOOKED_INET6
2496 #endif
2497
2498 #define GRAB_OUR_PACKETS(iface) \
2499         if ((iface)->if_type == IFT_GIF) \
2500                 continue; \
2501         /* It is destined for us. */ \
2502         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2503             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2504             ) {                                                         \
2505                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2506                         ETHER_BPF_MTAP(iface, m);                       \
2507                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2508                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2509                         /* Filter on the physical interface. */         \
2510                         if (V_pfil_local_phys &&                        \
2511                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2512                              OR_PFIL_HOOKED_INET6)) {                   \
2513                                 if (bridge_pfil(&m, NULL, ifp,          \
2514                                     PFIL_IN) != 0 || m == NULL) {       \
2515                                         return (NULL);                  \
2516                                 }                                       \
2517                                 eh = mtod(m, struct ether_header *);    \
2518                         }                                               \
2519                 }                                                       \
2520                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2521                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2522                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2523                         if (error && bif->bif_addrmax) {                \
2524                                 m_freem(m);                             \
2525                                 NET_EPOCH_EXIT_ET(et);                  \
2526                                 return (NULL);                          \
2527                         }                                               \
2528                 }                                                       \
2529                 m->m_pkthdr.rcvif = iface;                              \
2530                 NET_EPOCH_EXIT_ET(et);                                  \
2531                 return (m);                                             \
2532         }                                                               \
2533                                                                         \
2534         /* We just received a packet that we sent out. */               \
2535         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2536             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2537             ) {                                                         \
2538                 m_freem(m);                                             \
2539                 NET_EPOCH_EXIT_ET(et);                                  \
2540                 return (NULL);                                          \
2541         }
2542
2543         /*
2544          * Unicast.  Make sure it's not for the bridge.
2545          */
2546         do { GRAB_OUR_PACKETS(bifp) } while (0);
2547
2548         /*
2549          * Give a chance for ifp at first priority. This will help when the
2550          * packet comes through the interface like VLAN's with the same MACs
2551          * on several interfaces from the same bridge. This also will save
2552          * some CPU cycles in case the destination interface and the input
2553          * interface (eq ifp) are the same.
2554          */
2555         do { GRAB_OUR_PACKETS(ifp) } while (0);
2556
2557         /* Now check the all bridge members. */
2558         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2559                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2560         }
2561
2562 #undef OR_CARP_CHECK_WE_ARE_DST
2563 #undef OR_CARP_CHECK_WE_ARE_SRC
2564 #undef OR_PFIL_HOOKED_INET6
2565 #undef GRAB_OUR_PACKETS
2566
2567         /* Perform the bridge forwarding function. */
2568         bridge_forward(sc, bif, m);
2569
2570         NET_EPOCH_EXIT_ET(et);
2571         return (NULL);
2572 }
2573
2574 /*
2575  * bridge_broadcast:
2576  *
2577  *      Send a frame to all interfaces that are members of
2578  *      the bridge, except for the one on which the packet
2579  *      arrived.
2580  *
2581  *      NOTE: Releases the lock on return.
2582  */
2583 static void
2584 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2585     struct mbuf *m, int runfilt)
2586 {
2587         struct bridge_iflist *dbif, *sbif;
2588         struct mbuf *mc;
2589         struct ifnet *dst_if;
2590         int used = 0, i;
2591
2592         MPASS(in_epoch(net_epoch_preempt));
2593
2594         sbif = bridge_lookup_member_if(sc, src_if);
2595
2596         /* Filter on the bridge interface before broadcasting */
2597         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2598 #ifdef INET6
2599             || PFIL_HOOKED(&V_inet6_pfil_hook)
2600 #endif
2601             )) {
2602                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2603                         return;
2604                 if (m == NULL)
2605                         return;
2606         }
2607
2608         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2609                 dst_if = dbif->bif_ifp;
2610                 if (dst_if == src_if)
2611                         continue;
2612
2613                 /* Private segments can not talk to each other */
2614                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2615                         continue;
2616
2617                 if ((dbif->bif_flags & IFBIF_STP) &&
2618                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2619                         continue;
2620
2621                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2622                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2623                         continue;
2624
2625                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2626                         continue;
2627
2628                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2629                         mc = m;
2630                         used = 1;
2631                 } else {
2632                         mc = m_dup(m, M_NOWAIT);
2633                         if (mc == NULL) {
2634                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2635                                 continue;
2636                         }
2637                 }
2638
2639                 /*
2640                  * Filter on the output interface. Pass a NULL bridge interface
2641                  * pointer so we do not redundantly filter on the bridge for
2642                  * each interface we broadcast on.
2643                  */
2644                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2645 #ifdef INET6
2646                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2647 #endif
2648                     )) {
2649                         if (used == 0) {
2650                                 /* Keep the layer3 header aligned */
2651                                 i = min(mc->m_pkthdr.len, max_protohdr);
2652                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2653                                 if (mc == NULL) {
2654                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2655                                         continue;
2656                                 }
2657                         }
2658                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2659                                 continue;
2660                         if (mc == NULL)
2661                                 continue;
2662                 }
2663
2664                 bridge_enqueue(sc, dst_if, mc);
2665         }
2666         if (used == 0)
2667                 m_freem(m);
2668 }
2669
2670 /*
2671  * bridge_span:
2672  *
2673  *      Duplicate a packet out one or more interfaces that are in span mode,
2674  *      the original mbuf is unmodified.
2675  */
2676 static void
2677 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2678 {
2679         struct bridge_iflist *bif;
2680         struct ifnet *dst_if;
2681         struct mbuf *mc;
2682
2683         MPASS(in_epoch(net_epoch_preempt));
2684
2685         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2686                 return;
2687
2688         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2689                 dst_if = bif->bif_ifp;
2690
2691                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2692                         continue;
2693
2694                 mc = m_copypacket(m, M_NOWAIT);
2695                 if (mc == NULL) {
2696                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2697                         continue;
2698                 }
2699
2700                 bridge_enqueue(sc, dst_if, mc);
2701         }
2702 }
2703
2704 /*
2705  * bridge_rtupdate:
2706  *
2707  *      Add a bridge routing entry.
2708  */
2709 static int
2710 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2711     struct bridge_iflist *bif, int setflags, uint8_t flags)
2712 {
2713         struct bridge_rtnode *brt;
2714         int error;
2715
2716         MPASS(in_epoch(net_epoch_preempt));
2717         BRIDGE_UNLOCK_ASSERT(sc);
2718
2719         /* Check the source address is valid and not multicast. */
2720         if (ETHER_IS_MULTICAST(dst) ||
2721             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2722              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2723                 return (EINVAL);
2724
2725         /* 802.1p frames map to vlan 1 */
2726         if (vlan == 0)
2727                 vlan = 1;
2728
2729         /*
2730          * A route for this destination might already exist.  If so,
2731          * update it, otherwise create a new one.
2732          */
2733         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2734                 BRIDGE_LOCK(sc);
2735
2736                 /* Check again, now that we have the lock. There could have
2737                  * been a race and we only want to insert this once. */
2738                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2739                         BRIDGE_UNLOCK(sc);
2740                         return (0);
2741                 }
2742
2743                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2744                         sc->sc_brtexceeded++;
2745                         BRIDGE_UNLOCK(sc);
2746                         return (ENOSPC);
2747                 }
2748                 /* Check per interface address limits (if enabled) */
2749                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2750                         bif->bif_addrexceeded++;
2751                         BRIDGE_UNLOCK(sc);
2752                         return (ENOSPC);
2753                 }
2754
2755                 /*
2756                  * Allocate a new bridge forwarding node, and
2757                  * initialize the expiration time and Ethernet
2758                  * address.
2759                  */
2760                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2761                 if (brt == NULL) {
2762                         BRIDGE_UNLOCK(sc);
2763                         return (ENOMEM);
2764                 }
2765                 brt->brt_vnet = curvnet;
2766
2767                 if (bif->bif_flags & IFBIF_STICKY)
2768                         brt->brt_flags = IFBAF_STICKY;
2769                 else
2770                         brt->brt_flags = IFBAF_DYNAMIC;
2771
2772                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2773                 brt->brt_vlan = vlan;
2774
2775                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2776                         uma_zfree(V_bridge_rtnode_zone, brt);
2777                         BRIDGE_UNLOCK(sc);
2778                         return (error);
2779                 }
2780                 brt->brt_dst = bif;
2781                 bif->bif_addrcnt++;
2782
2783                 BRIDGE_UNLOCK(sc);
2784         }
2785
2786         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2787             brt->brt_dst != bif) {
2788                 BRIDGE_LOCK(sc);
2789                 brt->brt_dst->bif_addrcnt--;
2790                 brt->brt_dst = bif;
2791                 brt->brt_dst->bif_addrcnt++;
2792                 BRIDGE_UNLOCK(sc);
2793         }
2794
2795         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2796                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2797         if (setflags)
2798                 brt->brt_flags = flags;
2799
2800         return (0);
2801 }
2802
2803 /*
2804  * bridge_rtlookup:
2805  *
2806  *      Lookup the destination interface for an address.
2807  */
2808 static struct ifnet *
2809 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2810 {
2811         struct bridge_rtnode *brt;
2812
2813         MPASS(in_epoch(net_epoch_preempt));
2814
2815         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2816                 return (NULL);
2817
2818         return (brt->brt_ifp);
2819 }
2820
2821 /*
2822  * bridge_rttrim:
2823  *
2824  *      Trim the routine table so that we have a number
2825  *      of routing entries less than or equal to the
2826  *      maximum number.
2827  */
2828 static void
2829 bridge_rttrim(struct bridge_softc *sc)
2830 {
2831         struct bridge_rtnode *brt, *nbrt;
2832
2833         BRIDGE_LOCK_ASSERT(sc);
2834
2835         /* Make sure we actually need to do this. */
2836         if (sc->sc_brtcnt <= sc->sc_brtmax)
2837                 return;
2838
2839         /* Force an aging cycle; this might trim enough addresses. */
2840         bridge_rtage(sc);
2841         if (sc->sc_brtcnt <= sc->sc_brtmax)
2842                 return;
2843
2844         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2845                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2846                         bridge_rtnode_destroy(sc, brt);
2847                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2848                                 return;
2849                 }
2850         }
2851 }
2852
2853 /*
2854  * bridge_timer:
2855  *
2856  *      Aging timer for the bridge.
2857  */
2858 static void
2859 bridge_timer(void *arg)
2860 {
2861         struct bridge_softc *sc = arg;
2862
2863         BRIDGE_LOCK_ASSERT(sc);
2864
2865         /* Destruction of rtnodes requires a proper vnet context */
2866         CURVNET_SET(sc->sc_ifp->if_vnet);
2867         bridge_rtage(sc);
2868
2869         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2870                 callout_reset(&sc->sc_brcallout,
2871                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2872         CURVNET_RESTORE();
2873 }
2874
2875 /*
2876  * bridge_rtage:
2877  *
2878  *      Perform an aging cycle.
2879  */
2880 static void
2881 bridge_rtage(struct bridge_softc *sc)
2882 {
2883         struct bridge_rtnode *brt, *nbrt;
2884
2885         BRIDGE_LOCK_ASSERT(sc);
2886
2887         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2888                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2889                         if (time_uptime >= brt->brt_expire)
2890                                 bridge_rtnode_destroy(sc, brt);
2891                 }
2892         }
2893 }
2894
2895 /*
2896  * bridge_rtflush:
2897  *
2898  *      Remove all dynamic addresses from the bridge.
2899  */
2900 static void
2901 bridge_rtflush(struct bridge_softc *sc, int full)
2902 {
2903         struct bridge_rtnode *brt, *nbrt;
2904
2905         BRIDGE_LOCK_ASSERT(sc);
2906
2907         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2908                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2909                         bridge_rtnode_destroy(sc, brt);
2910         }
2911 }
2912
2913 /*
2914  * bridge_rtdaddr:
2915  *
2916  *      Remove an address from the table.
2917  */
2918 static int
2919 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2920 {
2921         struct bridge_rtnode *brt;
2922         int found = 0;
2923
2924         BRIDGE_LOCK_ASSERT(sc);
2925
2926         /*
2927          * If vlan is zero then we want to delete for all vlans so the lookup
2928          * may return more than one.
2929          */
2930         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2931                 bridge_rtnode_destroy(sc, brt);
2932                 found = 1;
2933         }
2934
2935         return (found ? 0 : ENOENT);
2936 }
2937
2938 /*
2939  * bridge_rtdelete:
2940  *
2941  *      Delete routes to a speicifc member interface.
2942  */
2943 static void
2944 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2945 {
2946         struct bridge_rtnode *brt, *nbrt;
2947
2948         BRIDGE_LOCK_ASSERT(sc);
2949
2950         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2951                 if (brt->brt_ifp == ifp && (full ||
2952                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2953                         bridge_rtnode_destroy(sc, brt);
2954         }
2955 }
2956
2957 /*
2958  * bridge_rtable_init:
2959  *
2960  *      Initialize the route table for this bridge.
2961  */
2962 static void
2963 bridge_rtable_init(struct bridge_softc *sc)
2964 {
2965         int i;
2966
2967         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2968             M_DEVBUF, M_WAITOK);
2969
2970         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2971                 CK_LIST_INIT(&sc->sc_rthash[i]);
2972
2973         sc->sc_rthash_key = arc4random();
2974         CK_LIST_INIT(&sc->sc_rtlist);
2975 }
2976
2977 /*
2978  * bridge_rtable_fini:
2979  *
2980  *      Deconstruct the route table for this bridge.
2981  */
2982 static void
2983 bridge_rtable_fini(struct bridge_softc *sc)
2984 {
2985
2986         KASSERT(sc->sc_brtcnt == 0,
2987             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2988         free(sc->sc_rthash, M_DEVBUF);
2989 }
2990
2991 /*
2992  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2993  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2994  */
2995 #define mix(a, b, c)                                                    \
2996 do {                                                                    \
2997         a -= b; a -= c; a ^= (c >> 13);                                 \
2998         b -= c; b -= a; b ^= (a << 8);                                  \
2999         c -= a; c -= b; c ^= (b >> 13);                                 \
3000         a -= b; a -= c; a ^= (c >> 12);                                 \
3001         b -= c; b -= a; b ^= (a << 16);                                 \
3002         c -= a; c -= b; c ^= (b >> 5);                                  \
3003         a -= b; a -= c; a ^= (c >> 3);                                  \
3004         b -= c; b -= a; b ^= (a << 10);                                 \
3005         c -= a; c -= b; c ^= (b >> 15);                                 \
3006 } while (/*CONSTCOND*/0)
3007
3008 static __inline uint32_t
3009 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3010 {
3011         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3012
3013         b += addr[5] << 8;
3014         b += addr[4];
3015         a += addr[3] << 24;
3016         a += addr[2] << 16;
3017         a += addr[1] << 8;
3018         a += addr[0];
3019
3020         mix(a, b, c);
3021
3022         return (c & BRIDGE_RTHASH_MASK);
3023 }
3024
3025 #undef mix
3026
3027 static int
3028 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3029 {
3030         int i, d;
3031
3032         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3033                 d = ((int)a[i]) - ((int)b[i]);
3034         }
3035
3036         return (d);
3037 }
3038
3039 /*
3040  * bridge_rtnode_lookup:
3041  *
3042  *      Look up a bridge route node for the specified destination. Compare the
3043  *      vlan id or if zero then just return the first match.
3044  */
3045 static struct bridge_rtnode *
3046 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3047 {
3048         struct bridge_rtnode *brt;
3049         uint32_t hash;
3050         int dir;
3051
3052         MPASS(in_epoch(net_epoch_preempt));
3053
3054         hash = bridge_rthash(sc, addr);
3055         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3056                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3057                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3058                         return (brt);
3059                 if (dir > 0)
3060                         return (NULL);
3061         }
3062
3063         return (NULL);
3064 }
3065
3066 /*
3067  * bridge_rtnode_insert:
3068  *
3069  *      Insert the specified bridge node into the route table.  We
3070  *      assume the entry is not already in the table.
3071  */
3072 static int
3073 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3074 {
3075         struct bridge_rtnode *lbrt;
3076         uint32_t hash;
3077         int dir;
3078
3079         BRIDGE_LOCK_ASSERT(sc);
3080
3081         hash = bridge_rthash(sc, brt->brt_addr);
3082
3083         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3084         if (lbrt == NULL) {
3085                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3086                 goto out;
3087         }
3088
3089         do {
3090                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3091                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3092                         return (EEXIST);
3093                 if (dir > 0) {
3094                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3095                         goto out;
3096                 }
3097                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3098                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3099                         goto out;
3100                 }
3101                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3102         } while (lbrt != NULL);
3103
3104 #ifdef DIAGNOSTIC
3105         panic("bridge_rtnode_insert: impossible");
3106 #endif
3107
3108 out:
3109         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3110         sc->sc_brtcnt++;
3111
3112         return (0);
3113 }
3114
3115 static void
3116 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3117 {
3118         struct bridge_rtnode *brt;
3119
3120         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3121
3122         CURVNET_SET(brt->brt_vnet);
3123         uma_zfree(V_bridge_rtnode_zone, brt);
3124         CURVNET_RESTORE();
3125 }
3126
3127 /*
3128  * bridge_rtnode_destroy:
3129  *
3130  *      Destroy a bridge rtnode.
3131  */
3132 static void
3133 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3134 {
3135         BRIDGE_LOCK_ASSERT(sc);
3136
3137         CK_LIST_REMOVE(brt, brt_hash);
3138
3139         CK_LIST_REMOVE(brt, brt_list);
3140         sc->sc_brtcnt--;
3141         brt->brt_dst->bif_addrcnt--;
3142
3143         epoch_call(net_epoch_preempt, &brt->brt_epoch_ctx,
3144             bridge_rtnode_destroy_cb);
3145 }
3146
3147 /*
3148  * bridge_rtable_expire:
3149  *
3150  *      Set the expiry time for all routes on an interface.
3151  */
3152 static void
3153 bridge_rtable_expire(struct ifnet *ifp, int age)
3154 {
3155         struct bridge_softc *sc = ifp->if_bridge;
3156         struct bridge_rtnode *brt;
3157
3158         CURVNET_SET(ifp->if_vnet);
3159         BRIDGE_LOCK(sc);
3160
3161         /*
3162          * If the age is zero then flush, otherwise set all the expiry times to
3163          * age for the interface
3164          */
3165         if (age == 0)
3166                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3167         else {
3168                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3169                         /* Cap the expiry time to 'age' */
3170                         if (brt->brt_ifp == ifp &&
3171                             brt->brt_expire > time_uptime + age &&
3172                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3173                                 brt->brt_expire = time_uptime + age;
3174                 }
3175         }
3176         BRIDGE_UNLOCK(sc);
3177         CURVNET_RESTORE();
3178 }
3179
3180 /*
3181  * bridge_state_change:
3182  *
3183  *      Callback from the bridgestp code when a port changes states.
3184  */
3185 static void
3186 bridge_state_change(struct ifnet *ifp, int state)
3187 {
3188         struct bridge_softc *sc = ifp->if_bridge;
3189         static const char *stpstates[] = {
3190                 "disabled",
3191                 "listening",
3192                 "learning",
3193                 "forwarding",
3194                 "blocking",
3195                 "discarding"
3196         };
3197
3198         CURVNET_SET(ifp->if_vnet);
3199         if (V_log_stp)
3200                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3201                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3202         CURVNET_RESTORE();
3203 }
3204
3205 /*
3206  * Send bridge packets through pfil if they are one of the types pfil can deal
3207  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3208  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3209  * that interface.
3210  */
3211 static int
3212 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3213 {
3214         int snap, error, i, hlen;
3215         struct ether_header *eh1, eh2;
3216         struct ip *ip;
3217         struct llc llc1;
3218         u_int16_t ether_type;
3219
3220         snap = 0;
3221         error = -1;     /* Default error if not error == 0 */
3222
3223 #if 0
3224         /* we may return with the IP fields swapped, ensure its not shared */
3225         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3226 #endif
3227
3228         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3229                 return (0); /* filtering is disabled */
3230
3231         i = min((*mp)->m_pkthdr.len, max_protohdr);
3232         if ((*mp)->m_len < i) {
3233             *mp = m_pullup(*mp, i);
3234             if (*mp == NULL) {
3235                 printf("%s: m_pullup failed\n", __func__);
3236                 return (-1);
3237             }
3238         }
3239
3240         eh1 = mtod(*mp, struct ether_header *);
3241         ether_type = ntohs(eh1->ether_type);
3242
3243         /*
3244          * Check for SNAP/LLC.
3245          */
3246         if (ether_type < ETHERMTU) {
3247                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3248
3249                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3250                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3251                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3252                     llc2->llc_control == LLC_UI) {
3253                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3254                         snap = 1;
3255                 }
3256         }
3257
3258         /*
3259          * If we're trying to filter bridge traffic, don't look at anything
3260          * other than IP and ARP traffic.  If the filter doesn't understand
3261          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3262          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3263          * but of course we don't have an AppleTalk filter to begin with.
3264          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3265          * ARP traffic.)
3266          */
3267         switch (ether_type) {
3268                 case ETHERTYPE_ARP:
3269                 case ETHERTYPE_REVARP:
3270                         if (V_pfil_ipfw_arp == 0)
3271                                 return (0); /* Automatically pass */
3272                         break;
3273
3274                 case ETHERTYPE_IP:
3275 #ifdef INET6
3276                 case ETHERTYPE_IPV6:
3277 #endif /* INET6 */
3278                         break;
3279                 default:
3280                         /*
3281                          * Check to see if the user wants to pass non-ip
3282                          * packets, these will not be checked by pfil(9) and
3283                          * passed unconditionally so the default is to drop.
3284                          */
3285                         if (V_pfil_onlyip)
3286                                 goto bad;
3287         }
3288
3289         /* Run the packet through pfil before stripping link headers */
3290         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3291                         dir == PFIL_OUT && ifp != NULL) {
3292
3293                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, 0,
3294                     NULL);
3295
3296                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3297                         return (error);
3298         }
3299
3300         /* Strip off the Ethernet header and keep a copy. */
3301         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3302         m_adj(*mp, ETHER_HDR_LEN);
3303
3304         /* Strip off snap header, if present */
3305         if (snap) {
3306                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3307                 m_adj(*mp, sizeof(struct llc));
3308         }
3309
3310         /*
3311          * Check the IP header for alignment and errors
3312          */
3313         if (dir == PFIL_IN) {
3314                 switch (ether_type) {
3315                         case ETHERTYPE_IP:
3316                                 error = bridge_ip_checkbasic(mp);
3317                                 break;
3318 #ifdef INET6
3319                         case ETHERTYPE_IPV6:
3320                                 error = bridge_ip6_checkbasic(mp);
3321                                 break;
3322 #endif /* INET6 */
3323                         default:
3324                                 error = 0;
3325                 }
3326                 if (error)
3327                         goto bad;
3328         }
3329
3330         error = 0;
3331
3332         /*
3333          * Run the packet through pfil
3334          */
3335         switch (ether_type) {
3336         case ETHERTYPE_IP:
3337                 /*
3338                  * Run pfil on the member interface and the bridge, both can
3339                  * be skipped by clearing pfil_member or pfil_bridge.
3340                  *
3341                  * Keep the order:
3342                  *   in_if -> bridge_if -> out_if
3343                  */
3344                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3345                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3346                                         dir, 0, NULL);
3347
3348                 if (*mp == NULL || error != 0) /* filter may consume */
3349                         break;
3350
3351                 if (V_pfil_member && ifp != NULL)
3352                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3353                                         dir, 0, NULL);
3354
3355                 if (*mp == NULL || error != 0) /* filter may consume */
3356                         break;
3357
3358                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3359                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3360                                         dir, 0, NULL);
3361
3362                 if (*mp == NULL || error != 0) /* filter may consume */
3363                         break;
3364
3365                 /* check if we need to fragment the packet */
3366                 /* bridge_fragment generates a mbuf chain of packets */
3367                 /* that already include eth headers */
3368                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3369                         i = (*mp)->m_pkthdr.len;
3370                         if (i > ifp->if_mtu) {
3371                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3372                                             &llc1);
3373                                 return (error);
3374                         }
3375                 }
3376
3377                 /* Recalculate the ip checksum. */
3378                 ip = mtod(*mp, struct ip *);
3379                 hlen = ip->ip_hl << 2;
3380                 if (hlen < sizeof(struct ip))
3381                         goto bad;
3382                 if (hlen > (*mp)->m_len) {
3383                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3384                                 goto bad;
3385                         ip = mtod(*mp, struct ip *);
3386                         if (ip == NULL)
3387                                 goto bad;
3388                 }
3389                 ip->ip_sum = 0;
3390                 if (hlen == sizeof(struct ip))
3391                         ip->ip_sum = in_cksum_hdr(ip);
3392                 else
3393                         ip->ip_sum = in_cksum(*mp, hlen);
3394
3395                 break;
3396 #ifdef INET6
3397         case ETHERTYPE_IPV6:
3398                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3399                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3400                                         dir, 0, NULL);
3401
3402                 if (*mp == NULL || error != 0) /* filter may consume */
3403                         break;
3404
3405                 if (V_pfil_member && ifp != NULL)
3406                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3407                                         dir, 0, NULL);
3408
3409                 if (*mp == NULL || error != 0) /* filter may consume */
3410                         break;
3411
3412                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3413                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3414                                         dir, 0, NULL);
3415                 break;
3416 #endif
3417         default:
3418                 error = 0;
3419                 break;
3420         }
3421
3422         if (*mp == NULL)
3423                 return (error);
3424         if (error != 0)
3425                 goto bad;
3426
3427         error = -1;
3428
3429         /*
3430          * Finally, put everything back the way it was and return
3431          */
3432         if (snap) {
3433                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3434                 if (*mp == NULL)
3435                         return (error);
3436                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3437         }
3438
3439         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3440         if (*mp == NULL)
3441                 return (error);
3442         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3443
3444         return (0);
3445
3446 bad:
3447         m_freem(*mp);
3448         *mp = NULL;
3449         return (error);
3450 }
3451
3452 /*
3453  * Perform basic checks on header size since
3454  * pfil assumes ip_input has already processed
3455  * it for it.  Cut-and-pasted from ip_input.c.
3456  * Given how simple the IPv6 version is,
3457  * does the IPv4 version really need to be
3458  * this complicated?
3459  *
3460  * XXX Should we update ipstat here, or not?
3461  * XXX Right now we update ipstat but not
3462  * XXX csum_counter.
3463  */
3464 static int
3465 bridge_ip_checkbasic(struct mbuf **mp)
3466 {
3467         struct mbuf *m = *mp;
3468         struct ip *ip;
3469         int len, hlen;
3470         u_short sum;
3471
3472         if (*mp == NULL)
3473                 return (-1);
3474
3475         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3476                 if ((m = m_copyup(m, sizeof(struct ip),
3477                         (max_linkhdr + 3) & ~3)) == NULL) {
3478                         /* XXXJRT new stat, please */
3479                         KMOD_IPSTAT_INC(ips_toosmall);
3480                         goto bad;
3481                 }
3482         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3483                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3484                         KMOD_IPSTAT_INC(ips_toosmall);
3485                         goto bad;
3486                 }
3487         }
3488         ip = mtod(m, struct ip *);
3489         if (ip == NULL) goto bad;
3490
3491         if (ip->ip_v != IPVERSION) {
3492                 KMOD_IPSTAT_INC(ips_badvers);
3493                 goto bad;
3494         }
3495         hlen = ip->ip_hl << 2;
3496         if (hlen < sizeof(struct ip)) { /* minimum header length */
3497                 KMOD_IPSTAT_INC(ips_badhlen);
3498                 goto bad;
3499         }
3500         if (hlen > m->m_len) {
3501                 if ((m = m_pullup(m, hlen)) == NULL) {
3502                         KMOD_IPSTAT_INC(ips_badhlen);
3503                         goto bad;
3504                 }
3505                 ip = mtod(m, struct ip *);
3506                 if (ip == NULL) goto bad;
3507         }
3508
3509         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3510                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3511         } else {
3512                 if (hlen == sizeof(struct ip)) {
3513                         sum = in_cksum_hdr(ip);
3514                 } else {
3515                         sum = in_cksum(m, hlen);
3516                 }
3517         }
3518         if (sum) {
3519                 KMOD_IPSTAT_INC(ips_badsum);
3520                 goto bad;
3521         }
3522
3523         /* Retrieve the packet length. */
3524         len = ntohs(ip->ip_len);
3525
3526         /*
3527          * Check for additional length bogosity
3528          */
3529         if (len < hlen) {
3530                 KMOD_IPSTAT_INC(ips_badlen);
3531                 goto bad;
3532         }
3533
3534         /*
3535          * Check that the amount of data in the buffers
3536          * is as at least much as the IP header would have us expect.
3537          * Drop packet if shorter than we expect.
3538          */
3539         if (m->m_pkthdr.len < len) {
3540                 KMOD_IPSTAT_INC(ips_tooshort);
3541                 goto bad;
3542         }
3543
3544         /* Checks out, proceed */
3545         *mp = m;
3546         return (0);
3547
3548 bad:
3549         *mp = m;
3550         return (-1);
3551 }
3552
3553 #ifdef INET6
3554 /*
3555  * Same as above, but for IPv6.
3556  * Cut-and-pasted from ip6_input.c.
3557  * XXX Should we update ip6stat, or not?
3558  */
3559 static int
3560 bridge_ip6_checkbasic(struct mbuf **mp)
3561 {
3562         struct mbuf *m = *mp;
3563         struct ip6_hdr *ip6;
3564
3565         /*
3566          * If the IPv6 header is not aligned, slurp it up into a new
3567          * mbuf with space for link headers, in the event we forward
3568          * it.  Otherwise, if it is aligned, make sure the entire base
3569          * IPv6 header is in the first mbuf of the chain.
3570          */
3571         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3572                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3573                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3574                             (max_linkhdr + 3) & ~3)) == NULL) {
3575                         /* XXXJRT new stat, please */
3576                         IP6STAT_INC(ip6s_toosmall);
3577                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3578                         goto bad;
3579                 }
3580         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3581                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3582                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3583                         IP6STAT_INC(ip6s_toosmall);
3584                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3585                         goto bad;
3586                 }
3587         }
3588
3589         ip6 = mtod(m, struct ip6_hdr *);
3590
3591         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3592                 IP6STAT_INC(ip6s_badvers);
3593                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3594                 goto bad;
3595         }
3596
3597         /* Checks out, proceed */
3598         *mp = m;
3599         return (0);
3600
3601 bad:
3602         *mp = m;
3603         return (-1);
3604 }
3605 #endif /* INET6 */
3606
3607 /*
3608  * bridge_fragment:
3609  *
3610  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3611  */
3612 static int
3613 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3614     int snap, struct llc *llc)
3615 {
3616         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3617         struct ip *ip;
3618         int error = -1;
3619
3620         if (m->m_len < sizeof(struct ip) &&
3621             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3622                 goto dropit;
3623         ip = mtod(m, struct ip *);
3624
3625         m->m_pkthdr.csum_flags |= CSUM_IP;
3626         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3627         if (error)
3628                 goto dropit;
3629
3630         /*
3631          * Walk the chain and re-add the Ethernet header for
3632          * each mbuf packet.
3633          */
3634         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3635                 nextpkt = mcur->m_nextpkt;
3636                 mcur->m_nextpkt = NULL;
3637                 if (snap) {
3638                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3639                         if (mcur == NULL) {
3640                                 error = ENOBUFS;
3641                                 if (mprev != NULL)
3642                                         mprev->m_nextpkt = nextpkt;
3643                                 goto dropit;
3644                         }
3645                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3646                 }
3647
3648                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3649                 if (mcur == NULL) {
3650                         error = ENOBUFS;
3651                         if (mprev != NULL)
3652                                 mprev->m_nextpkt = nextpkt;
3653                         goto dropit;
3654                 }
3655                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3656
3657                 /*
3658                  * The previous two M_PREPEND could have inserted one or two
3659                  * mbufs in front so we have to update the previous packet's
3660                  * m_nextpkt.
3661                  */
3662                 mcur->m_nextpkt = nextpkt;
3663                 if (mprev != NULL)
3664                         mprev->m_nextpkt = mcur;
3665                 else {
3666                         /* The first mbuf in the original chain needs to be
3667                          * updated. */
3668                         *mp = mcur;
3669                 }
3670                 mprev = mcur;
3671         }
3672
3673         KMOD_IPSTAT_INC(ips_fragmented);
3674         return (error);
3675
3676 dropit:
3677         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3678                 m = mcur->m_nextpkt;
3679                 m_freem(mcur);
3680         }
3681         return (error);
3682 }
3683
3684 static void
3685 bridge_linkstate(struct ifnet *ifp)
3686 {
3687         struct bridge_softc *sc = ifp->if_bridge;
3688         struct bridge_iflist *bif;
3689         struct epoch_tracker et;
3690
3691         NET_EPOCH_ENTER_ET(et);
3692
3693         bif = bridge_lookup_member_if(sc, ifp);
3694         if (bif == NULL) {
3695                 NET_EPOCH_EXIT_ET(et);
3696                 return;
3697         }
3698         bridge_linkcheck(sc);
3699
3700         bstp_linkstate(&bif->bif_stp);
3701
3702         NET_EPOCH_EXIT_ET(et);
3703 }
3704
3705 static void
3706 bridge_linkcheck(struct bridge_softc *sc)
3707 {
3708         struct bridge_iflist *bif;
3709         int new_link, hasls;
3710
3711         MPASS(in_epoch(net_epoch_preempt));
3712
3713         new_link = LINK_STATE_DOWN;
3714         hasls = 0;
3715         /* Our link is considered up if at least one of our ports is active */
3716         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3717                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3718                         hasls++;
3719                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3720                         new_link = LINK_STATE_UP;
3721                         break;
3722                 }
3723         }
3724         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3725                 /* If no interfaces support link-state then we default to up */
3726                 new_link = LINK_STATE_UP;
3727         }
3728         if_link_state_change(sc->sc_ifp, new_link);
3729 }