sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*-
   4  * SPDX-License-Identifier: BSD-4-Clause
   5  *
   6  * Copyright 2001 Wasabi Systems, Inc.
   7  * All rights reserved.
   8  *
   9  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
  10  *
  11  * Redistribution and use in source and binary forms, with or without
  12  * modification, are permitted provided that the following conditions
  13  * are met:
  14  * 1. Redistributions of source code must retain the above copyright
  15  *    notice, this list of conditions and the following disclaimer.
  16  * 2. Redistributions in binary form must reproduce the above copyright
  17  *    notice, this list of conditions and the following disclaimer in the
  18  *    documentation and/or other materials provided with the distribution.
  19  * 3. All advertising materials mentioning features or use of this software
  20  *    must display the following acknowledgement:
  21  *      This product includes software developed for the NetBSD Project by
  22  *      Wasabi Systems, Inc.
  23  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  24  *    or promote products derived from this software without specific prior
  25  *    written permission.
  26  *
  27  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  28  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  29  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  30  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  31  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  32  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  33  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  34  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  35  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  36  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  37  * POSSIBILITY OF SUCH DAMAGE.
  38  */
  39
  40 /*
  41  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  42  * All rights reserved.
  43  *
  44  * Redistribution and use in source and binary forms, with or without
  45  * modification, are permitted provided that the following conditions
  46  * are met:
  47  * 1. Redistributions of source code must retain the above copyright
  48  *    notice, this list of conditions and the following disclaimer.
  49  * 2. Redistributions in binary form must reproduce the above copyright
  50  *    notice, this list of conditions and the following disclaimer in the
  51  *    documentation and/or other materials provided with the distribution.
  52  *
  53  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  54  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  55  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  56  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  57  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  58  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  59  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  60  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  61  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  62  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  63  * POSSIBILITY OF SUCH DAMAGE.
  64  *
  65  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  66  */
  67
  68 /*
  69  * Network interface bridge support.
  70  *
  71  * TODO:
  72  *
  73  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  74  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  75  *        to bridge other types of interfaces (maybe consider
  76  *        heterogeneous bridges).
  77  */
  78
  79 #include <sys/cdefs.h>
  80 __FBSDID("$FreeBSD$");
  81
  82 #include "opt_inet.h"
  83 #include "opt_inet6.h"
  84
  85 #include <sys/param.h>
  86 #include <sys/eventhandler.h>
  87 #include <sys/mbuf.h>
  88 #include <sys/malloc.h>
  89 #include <sys/protosw.h>
  90 #include <sys/systm.h>
  91 #include <sys/jail.h>
  92 #include <sys/time.h>
  93 #include <sys/socket.h> /* for net/if.h */
  94 #include <sys/sockio.h>
  95 #include <sys/ctype.h>  /* string functions */
  96 #include <sys/kernel.h>
  97 #include <sys/random.h>
  98 #include <sys/syslog.h>
  99 #include <sys/sysctl.h>
 100 #include <vm/uma.h>
 101 #include <sys/module.h>
 102 #include <sys/priv.h>
 103 #include <sys/proc.h>
 104 #include <sys/lock.h>
 105 #include <sys/mutex.h>
 106
 107 #include <net/bpf.h>
 108 #include <net/if.h>
 109 #include <net/if_clone.h>
 110 #include <net/if_dl.h>
 111 #include <net/if_types.h>
 112 #include <net/if_var.h>
 113 #include <net/pfil.h>
 114 #include <net/vnet.h>
 115
 116 #include <netinet/in.h>
 117 #include <netinet/in_systm.h>
 118 #include <netinet/in_var.h>
 119 #include <netinet/ip.h>
 120 #include <netinet/ip_var.h>
 121 #ifdef INET6
 122 #include <netinet/ip6.h>
 123 #include <netinet6/ip6_var.h>
 124 #include <netinet6/in6_ifattach.h>
 125 #endif
 126 #if defined(INET) || defined(INET6)
 127 #include <netinet/ip_carp.h>
 128 #endif
 129 #include <machine/in_cksum.h>
 130 #include <netinet/if_ether.h>
 131 #include <net/bridgestp.h>
 132 #include <net/if_bridgevar.h>
 133 #include <net/if_llc.h>
 134 #include <net/if_vlan_var.h>
 135
 136 #include <net/route.h>
 137
 138 #ifdef INET6
 139 /*
 140  * XXX: declare here to avoid to include many inet6 related files..
 141  * should be more generalized?
 142  */
 143 extern void     nd6_setmtu(struct ifnet *);
 144 #endif
 145
 146 /*
 147  * Size of the route hash table.  Must be a power of two.
 148  */
 149 #ifndef BRIDGE_RTHASH_SIZE
 150 #define BRIDGE_RTHASH_SIZE              1024
 151 #endif
 152
 153 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 154
 155 /*
 156  * Default maximum number of addresses to cache.
 157  */
 158 #ifndef BRIDGE_RTABLE_MAX
 159 #define BRIDGE_RTABLE_MAX               2000
 160 #endif
 161
 162 /*
 163  * Timeout (in seconds) for entries learned dynamically.
 164  */
 165 #ifndef BRIDGE_RTABLE_TIMEOUT
 166 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 167 #endif
 168
 169 /*
 170  * Number of seconds between walks of the route list.
 171  */
 172 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 173 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 174 #endif
 175
 176 /*
 177  * List of capabilities to possibly mask on the member interface.
 178  */
 179 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM|\
 180                                          IFCAP_TXCSUM_IPV6)
 181
 182 /*
 183  * List of capabilities to strip
 184  */
 185 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 186
 187 /*
 188  * Bridge locking
 189  *
 190  * The bridge relies heavily on the epoch(9) system to protect its data
 191  * structures. This means we can safely use CK_LISTs while in NET_EPOCH, but we
 192  * must ensure there is only one writer at a time.
 193  *
 194  * That is: for read accesses we only need to be in NET_EPOCH, but for write
 195  * accesses we must hold:
 196  *
 197  *  - BRIDGE_RT_LOCK, for any change to bridge_rtnodes
 198  *  - BRIDGE_LOCK, for any other change
 199  *
 200  * The BRIDGE_LOCK is a sleepable lock, because it is held accross ioctl()
 201  * calls to bridge member interfaces and these ioctl()s can sleep.
 202  * The BRIDGE_RT_LOCK is a non-sleepable mutex, because it is sometimes
 203  * required while we're in NET_EPOCH and then we're not allowed to sleep.
 204  */
 205 #define BRIDGE_LOCK_INIT(_sc)           do {                    \
 206         sx_init(&(_sc)->sc_sx, "if_bridge");                    \
 207         mtx_init(&(_sc)->sc_rt_mtx, "if_bridge rt", NULL, MTX_DEF);     \
 208 } while (0)
 209 #define BRIDGE_LOCK_DESTROY(_sc)        do {    \
 210         sx_destroy(&(_sc)->sc_sx);              \
 211         mtx_destroy(&(_sc)->sc_rt_mtx);         \
 212 } while (0)
 213 #define BRIDGE_LOCK(_sc)                sx_xlock(&(_sc)->sc_sx)
 214 #define BRIDGE_UNLOCK(_sc)              sx_xunlock(&(_sc)->sc_sx)
 215 #define BRIDGE_LOCK_ASSERT(_sc)         sx_assert(&(_sc)->sc_sx, SX_XLOCKED)
 216 #define BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(_sc)    \
 217             MPASS(in_epoch(net_epoch_preempt) || sx_xlocked(&(_sc)->sc_sx))
 218 #define BRIDGE_UNLOCK_ASSERT(_sc)       sx_assert(&(_sc)->sc_sx, SX_UNLOCKED)
 219 #define BRIDGE_RT_LOCK(_sc)             mtx_lock(&(_sc)->sc_rt_mtx)
 220 #define BRIDGE_RT_UNLOCK(_sc)           mtx_unlock(&(_sc)->sc_rt_mtx)
 221 #define BRIDGE_RT_LOCK_ASSERT(_sc)      mtx_assert(&(_sc)->sc_rt_mtx, MA_OWNED)
 222 #define BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(_sc) \
 223             MPASS(in_epoch(net_epoch_preempt) || mtx_owned(&(_sc)->sc_rt_mtx))
 224
 225 /*
 226  * Bridge interface list entry.
 227  */
 228 struct bridge_iflist {
 229         CK_LIST_ENTRY(bridge_iflist) bif_next;
 230         struct ifnet            *bif_ifp;       /* member if */
 231         struct bstp_port        bif_stp;        /* STP state */
 232         uint32_t                bif_flags;      /* member if flags */
 233         int                     bif_savedcaps;  /* saved capabilities */
 234         uint32_t                bif_addrmax;    /* max # of addresses */
 235         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 236         uint32_t                bif_addrexceeded;/* # of address violations */
 237
 238         struct epoch_context    bif_epoch_ctx;
 239 };
 240
 241 /*
 242  * Bridge route node.
 243  */
 244 struct bridge_rtnode {
 245         CK_LIST_ENTRY(bridge_rtnode) brt_hash;  /* hash table linkage */
 246         CK_LIST_ENTRY(bridge_rtnode) brt_list;  /* list linkage */
 247         struct bridge_iflist    *brt_dst;       /* destination if */
 248         unsigned long           brt_expire;     /* expiration time */
 249         uint8_t                 brt_flags;      /* address flags */
 250         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 251         uint16_t                brt_vlan;       /* vlan id */
 252
 253         struct  vnet            *brt_vnet;
 254         struct  epoch_context   brt_epoch_ctx;
 255 };
 256 #define brt_ifp                 brt_dst->bif_ifp
 257
 258 /*
 259  * Software state for each bridge.
 260  */
 261 struct bridge_softc {
 262         struct ifnet            *sc_ifp;        /* make this an interface */
 263         LIST_ENTRY(bridge_softc) sc_list;
 264         struct sx               sc_sx;
 265         struct mtx              sc_rt_mtx;
 266         uint32_t                sc_brtmax;      /* max # of addresses */
 267         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 268         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 269         struct callout          sc_brcallout;   /* bridge callout */
 270         CK_LIST_HEAD(, bridge_iflist) sc_iflist;        /* member interface list */
 271         CK_LIST_HEAD(, bridge_rtnode) *sc_rthash;       /* our forwarding table */
 272         CK_LIST_HEAD(, bridge_rtnode) sc_rtlist;        /* list version of above */
 273         uint32_t                sc_rthash_key;  /* key for hash */
 274         CK_LIST_HEAD(, bridge_iflist) sc_spanlist;      /* span ports list */
 275         struct bstp_state       sc_stp;         /* STP state */
 276         uint32_t                sc_brtexceeded; /* # of cache drops */
 277         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 278         struct ether_addr       sc_defaddr;     /* Default MAC address */
 279
 280         struct epoch_context    sc_epoch_ctx;
 281 };
 282
 283 VNET_DEFINE_STATIC(struct sx, bridge_list_sx);
 284 #define V_bridge_list_sx        VNET(bridge_list_sx)
 285 static eventhandler_tag bridge_detach_cookie;
 286
 287 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 288
 289 VNET_DEFINE_STATIC(uma_zone_t, bridge_rtnode_zone);
 290 #define V_bridge_rtnode_zone    VNET(bridge_rtnode_zone)
 291
 292 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 293 static void     bridge_clone_destroy(struct ifnet *);
 294
 295 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 296 static void     bridge_mutecaps(struct bridge_softc *);
 297 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 298                     int);
 299 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 300 static void     bridge_init(void *);
 301 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 302 static void     bridge_stop(struct ifnet *, int);
 303 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 304 #ifdef ALTQ
 305 static void     bridge_altq_start(if_t);
 306 static int      bridge_altq_transmit(if_t, struct mbuf *);
 307 #endif
 308 static void     bridge_qflush(struct ifnet *);
 309 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 310 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 311                     struct rtentry *);
 312 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 313                     struct mbuf *);
 314 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 315
 316 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 317                     struct mbuf *m);
 318
 319 static void     bridge_timer(void *);
 320
 321 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 322                     struct mbuf *, int);
 323 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 324
 325 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 326                     uint16_t, struct bridge_iflist *, int, uint8_t);
 327 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 328                     uint16_t);
 329 static void     bridge_rttrim(struct bridge_softc *);
 330 static void     bridge_rtage(struct bridge_softc *);
 331 static void     bridge_rtflush(struct bridge_softc *, int);
 332 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 333                     uint16_t);
 334
 335 static void     bridge_rtable_init(struct bridge_softc *);
 336 static void     bridge_rtable_fini(struct bridge_softc *);
 337
 338 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 339 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 340                     const uint8_t *, uint16_t);
 341 static int      bridge_rtnode_insert(struct bridge_softc *,
 342                     struct bridge_rtnode *);
 343 static void     bridge_rtnode_destroy(struct bridge_softc *,
 344                     struct bridge_rtnode *);
 345 static void     bridge_rtable_expire(struct ifnet *, int);
 346 static void     bridge_state_change(struct ifnet *, int);
 347
 348 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 349                     const char *name);
 350 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 351                     struct ifnet *ifp);
 352 static void     bridge_delete_member(struct bridge_softc *,
 353                     struct bridge_iflist *, int);
 354 static void     bridge_delete_span(struct bridge_softc *,
 355                     struct bridge_iflist *);
 356
 357 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 358 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 359 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 360 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 361 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 362 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 363 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 364 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 365 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 366 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 367 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 368 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 369 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 370 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 371 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 372 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 373 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 374 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 375 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 376 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 377 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 378 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 379 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 380 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 381 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 382 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 383 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 384 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 385 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 386 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 387 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 388 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 389                     int);
 390 static int      bridge_ip_checkbasic(struct mbuf **mp);
 391 #ifdef INET6
 392 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 393 #endif /* INET6 */
 394 static int      bridge_fragment(struct ifnet *, struct mbuf **mp,
 395                     struct ether_header *, int, struct llc *);
 396 static void     bridge_linkstate(struct ifnet *ifp);
 397 static void     bridge_linkcheck(struct bridge_softc *sc);
 398
 399
 400 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 401 #define VLANTAGOF(_m)   \
 402     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 403
 404 static struct bstp_cb_ops bridge_ops = {
 405         .bcb_state = bridge_state_change,
 406         .bcb_rtage = bridge_rtable_expire
 407 };
 408
 409 SYSCTL_DECL(_net_link);
 410 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 411
 412 /* only pass IP[46] packets when pfil is enabled */
 413 VNET_DEFINE_STATIC(int, pfil_onlyip) = 1;
 414 #define V_pfil_onlyip   VNET(pfil_onlyip)
 415 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 416     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 417     "Only pass IP packets when pfil is enabled");
 418
 419 /* run pfil hooks on the bridge interface */
 420 VNET_DEFINE_STATIC(int, pfil_bridge) = 1;
 421 #define V_pfil_bridge   VNET(pfil_bridge)
 422 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 423     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 424     "Packet filter on the bridge interface");
 425
 426 /* layer2 filter with ipfw */
 427 VNET_DEFINE_STATIC(int, pfil_ipfw);
 428 #define V_pfil_ipfw     VNET(pfil_ipfw)
 429
 430 /* layer2 ARP filter with ipfw */
 431 VNET_DEFINE_STATIC(int, pfil_ipfw_arp);
 432 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 433 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 434     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 435     "Filter ARP packets through IPFW layer2");
 436
 437 /* run pfil hooks on the member interface */
 438 VNET_DEFINE_STATIC(int, pfil_member) = 1;
 439 #define V_pfil_member   VNET(pfil_member)
 440 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 441     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 442     "Packet filter on the member interface");
 443
 444 /* run pfil hooks on the physical interface for locally destined packets */
 445 VNET_DEFINE_STATIC(int, pfil_local_phys);
 446 #define V_pfil_local_phys       VNET(pfil_local_phys)
 447 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 448     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 449     "Packet filter on the physical interface for locally destined packets");
 450
 451 /* log STP state changes */
 452 VNET_DEFINE_STATIC(int, log_stp);
 453 #define V_log_stp       VNET(log_stp)
 454 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 455     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 456     "Log STP state changes");
 457
 458 /* share MAC with first bridge member */
 459 VNET_DEFINE_STATIC(int, bridge_inherit_mac);
 460 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 461 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 462     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 463     "Inherit MAC address from the first bridge member");
 464
 465 VNET_DEFINE_STATIC(int, allow_llz_overlap) = 0;
 466 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 467 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 468     CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 469     "Allow overlap of link-local scope "
 470     "zones of a bridge interface and the member interfaces");
 471
 472 struct bridge_control {
 473         int     (*bc_func)(struct bridge_softc *, void *);
 474         int     bc_argsize;
 475         int     bc_flags;
 476 };
 477
 478 #define BC_F_COPYIN             0x01    /* copy arguments in */
 479 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 480 #define BC_F_SUSER              0x04    /* do super-user check */
 481
 482 const struct bridge_control bridge_control_table[] = {
 483         { bridge_ioctl_add,             sizeof(struct ifbreq),
 484           BC_F_COPYIN|BC_F_SUSER },
 485         { bridge_ioctl_del,             sizeof(struct ifbreq),
 486           BC_F_COPYIN|BC_F_SUSER },
 487
 488         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 489           BC_F_COPYIN|BC_F_COPYOUT },
 490         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 491           BC_F_COPYIN|BC_F_SUSER },
 492
 493         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 494           BC_F_COPYIN|BC_F_SUSER },
 495         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 496           BC_F_COPYOUT },
 497
 498         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 499           BC_F_COPYIN|BC_F_COPYOUT },
 500         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 501           BC_F_COPYIN|BC_F_COPYOUT },
 502
 503         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 504           BC_F_COPYIN|BC_F_SUSER },
 505
 506         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 507           BC_F_COPYIN|BC_F_SUSER },
 508         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 509           BC_F_COPYOUT },
 510
 511         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 512           BC_F_COPYIN|BC_F_SUSER },
 513
 514         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 515           BC_F_COPYIN|BC_F_SUSER },
 516
 517         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 518           BC_F_COPYOUT },
 519         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 520           BC_F_COPYIN|BC_F_SUSER },
 521
 522         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 523           BC_F_COPYOUT },
 524         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 525           BC_F_COPYIN|BC_F_SUSER },
 526
 527         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 528           BC_F_COPYOUT },
 529         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 530           BC_F_COPYIN|BC_F_SUSER },
 531
 532         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 533           BC_F_COPYOUT },
 534         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 535           BC_F_COPYIN|BC_F_SUSER },
 536
 537         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 538           BC_F_COPYIN|BC_F_SUSER },
 539
 540         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 541           BC_F_COPYIN|BC_F_SUSER },
 542
 543         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 544           BC_F_COPYIN|BC_F_SUSER },
 545         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 546           BC_F_COPYIN|BC_F_SUSER },
 547
 548         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 549           BC_F_COPYOUT },
 550
 551         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 552           BC_F_COPYOUT },
 553
 554         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 555           BC_F_COPYIN|BC_F_COPYOUT },
 556
 557         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 558           BC_F_COPYIN|BC_F_SUSER },
 559
 560         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 561           BC_F_COPYIN|BC_F_SUSER },
 562
 563         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 564           BC_F_COPYIN|BC_F_SUSER },
 565
 566 };
 567 const int bridge_control_table_size = nitems(bridge_control_table);
 568
 569 VNET_DEFINE_STATIC(LIST_HEAD(, bridge_softc), bridge_list);
 570 #define V_bridge_list   VNET(bridge_list)
 571 #define BRIDGE_LIST_LOCK_INIT(x)        sx_init(&V_bridge_list_sx,      \
 572                                             "if_bridge list")
 573 #define BRIDGE_LIST_LOCK_DESTROY(x)     sx_destroy(&V_bridge_list_sx)
 574 #define BRIDGE_LIST_LOCK(x)             sx_xlock(&V_bridge_list_sx)
 575 #define BRIDGE_LIST_UNLOCK(x)           sx_xunlock(&V_bridge_list_sx)
 576
 577 VNET_DEFINE_STATIC(struct if_clone *, bridge_cloner);
 578 #define V_bridge_cloner VNET(bridge_cloner)
 579
 580 static const char bridge_name[] = "bridge";
 581
 582 static void
 583 vnet_bridge_init(const void *unused __unused)
 584 {
 585
 586         V_bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 587             sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 588             UMA_ALIGN_PTR, 0);
 589         BRIDGE_LIST_LOCK_INIT();
 590         LIST_INIT(&V_bridge_list);
 591         V_bridge_cloner = if_clone_simple(bridge_name,
 592             bridge_clone_create, bridge_clone_destroy, 0);
 593 }
 594 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 595     vnet_bridge_init, NULL);
 596
 597 static void
 598 vnet_bridge_uninit(const void *unused __unused)
 599 {
 600
 601         if_clone_detach(V_bridge_cloner);
 602         V_bridge_cloner = NULL;
 603         BRIDGE_LIST_LOCK_DESTROY();
 604
 605         /* Before we can destroy the uma zone, because there are callbacks that
 606          * use it. */
 607         epoch_drain_callbacks(net_epoch_preempt);
 608
 609         uma_zdestroy(V_bridge_rtnode_zone);
 610 }
 611 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PSEUDO, SI_ORDER_ANY,
 612     vnet_bridge_uninit, NULL);
 613
 614 static int
 615 bridge_modevent(module_t mod, int type, void *data)
 616 {
 617
 618         switch (type) {
 619         case MOD_LOAD:
 620                 bridge_dn_p = bridge_dummynet;
 621                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 622                     ifnet_departure_event, bridge_ifdetach, NULL,
 623                     EVENTHANDLER_PRI_ANY);
 624                 break;
 625         case MOD_UNLOAD:
 626                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 627                     bridge_detach_cookie);
 628                 bridge_dn_p = NULL;
 629                 break;
 630         default:
 631                 return (EOPNOTSUPP);
 632         }
 633         return (0);
 634 }
 635
 636 static moduledata_t bridge_mod = {
 637         "if_bridge",
 638         bridge_modevent,
 639         0
 640 };
 641
 642 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 643 MODULE_VERSION(if_bridge, 1);
 644 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 645
 646 /*
 647  * handler for net.link.bridge.ipfw
 648  */
 649 static int
 650 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 651 {
 652         int enable = V_pfil_ipfw;
 653         int error;
 654
 655         error = sysctl_handle_int(oidp, &enable, 0, req);
 656         enable &= 1;
 657
 658         if (enable != V_pfil_ipfw) {
 659                 V_pfil_ipfw = enable;
 660
 661                 /*
 662                  * Disable pfil so that ipfw doesnt run twice, if the user
 663                  * really wants both then they can re-enable pfil_bridge and/or
 664                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 665                  * layer2 type.
 666                  */
 667                 if (V_pfil_ipfw) {
 668                         V_pfil_onlyip = 0;
 669                         V_pfil_bridge = 0;
 670                         V_pfil_member = 0;
 671                 }
 672         }
 673
 674         return (error);
 675 }
 676 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 677     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 678     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 679     "Layer2 filter with IPFW");
 680
 681 #ifdef VIMAGE
 682 static void
 683 bridge_reassign(struct ifnet *ifp, struct vnet *newvnet, char *arg)
 684 {
 685         struct bridge_softc *sc = ifp->if_softc;
 686         struct bridge_iflist *bif;
 687
 688         BRIDGE_LOCK(sc);
 689
 690         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 691                 bridge_delete_member(sc, bif, 0);
 692
 693         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 694                 bridge_delete_span(sc, bif);
 695         }
 696
 697         BRIDGE_UNLOCK(sc);
 698
 699         ether_reassign(ifp, newvnet, arg);
 700 }
 701 #endif
 702
 703 /*
 704  * bridge_clone_create:
 705  *
 706  *      Create a new bridge instance.
 707  */
 708 static int
 709 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 710 {
 711         struct bridge_softc *sc, *sc2;
 712         struct ifnet *bifp, *ifp;
 713         int fb, retry;
 714         unsigned long hostid;
 715
 716         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 717         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 718         if (ifp == NULL) {
 719                 free(sc, M_DEVBUF);
 720                 return (ENOSPC);
 721         }
 722
 723         BRIDGE_LOCK_INIT(sc);
 724         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 725         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 726
 727         /* Initialize our routing table. */
 728         bridge_rtable_init(sc);
 729
 730         callout_init_mtx(&sc->sc_brcallout, &sc->sc_rt_mtx, 0);
 731
 732         CK_LIST_INIT(&sc->sc_iflist);
 733         CK_LIST_INIT(&sc->sc_spanlist);
 734
 735         ifp->if_softc = sc;
 736         if_initname(ifp, bridge_name, unit);
 737         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 738         ifp->if_ioctl = bridge_ioctl;
 739 #ifdef ALTQ
 740         ifp->if_start = bridge_altq_start;
 741         ifp->if_transmit = bridge_altq_transmit;
 742         IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
 743         ifp->if_snd.ifq_drv_maxlen = 0;
 744         IFQ_SET_READY(&ifp->if_snd);
 745 #else
 746         ifp->if_transmit = bridge_transmit;
 747 #endif
 748         ifp->if_qflush = bridge_qflush;
 749         ifp->if_init = bridge_init;
 750         ifp->if_type = IFT_BRIDGE;
 751
 752         /*
 753          * Generate an ethernet address with a locally administered address.
 754          *
 755          * Since we are using random ethernet addresses for the bridge, it is
 756          * possible that we might have address collisions, so make sure that
 757          * this hardware address isn't already in use on another bridge.
 758          * The first try uses the hostid and falls back to arc4rand().
 759          */
 760         fb = 0;
 761         getcredhostid(curthread->td_ucred, &hostid);
 762         do {
 763                 if (fb || hostid == 0) {
 764                         ether_gen_addr(ifp, &sc->sc_defaddr);
 765                 } else {
 766                         sc->sc_defaddr.octet[0] = 0x2;
 767                         sc->sc_defaddr.octet[1] = (hostid >> 24) & 0xff;
 768                         sc->sc_defaddr.octet[2] = (hostid >> 16) & 0xff;
 769                         sc->sc_defaddr.octet[3] = (hostid >> 8 ) & 0xff;
 770                         sc->sc_defaddr.octet[4] =  hostid        & 0xff;
 771                         sc->sc_defaddr.octet[5] = ifp->if_dunit & 0xff;
 772                 }
 773
 774                 fb = 1;
 775                 retry = 0;
 776                 BRIDGE_LIST_LOCK();
 777                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 778                         bifp = sc2->sc_ifp;
 779                         if (memcmp(sc->sc_defaddr.octet,
 780                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 781                                 retry = 1;
 782                                 break;
 783                         }
 784                 }
 785                 BRIDGE_LIST_UNLOCK();
 786         } while (retry == 1);
 787
 788         bstp_attach(&sc->sc_stp, &bridge_ops);
 789         ether_ifattach(ifp, sc->sc_defaddr.octet);
 790         /* Now undo some of the damage... */
 791         ifp->if_baudrate = 0;
 792         ifp->if_type = IFT_BRIDGE;
 793 #ifdef VIMAGE
 794         ifp->if_reassign = bridge_reassign;
 795 #endif
 796
 797         BRIDGE_LIST_LOCK();
 798         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 799         BRIDGE_LIST_UNLOCK();
 800
 801         return (0);
 802 }
 803
 804 static void
 805 bridge_clone_destroy_cb(struct epoch_context *ctx)
 806 {
 807         struct bridge_softc *sc;
 808
 809         sc = __containerof(ctx, struct bridge_softc, sc_epoch_ctx);
 810
 811         BRIDGE_LOCK_DESTROY(sc);
 812         free(sc, M_DEVBUF);
 813 }
 814
 815 /*
 816  * bridge_clone_destroy:
 817  *
 818  *      Destroy a bridge instance.
 819  */
 820 static void
 821 bridge_clone_destroy(struct ifnet *ifp)
 822 {
 823         struct bridge_softc *sc = ifp->if_softc;
 824         struct bridge_iflist *bif;
 825         struct epoch_tracker et;
 826
 827         BRIDGE_LOCK(sc);
 828
 829         bridge_stop(ifp, 1);
 830         ifp->if_flags &= ~IFF_UP;
 831
 832         while ((bif = CK_LIST_FIRST(&sc->sc_iflist)) != NULL)
 833                 bridge_delete_member(sc, bif, 0);
 834
 835         while ((bif = CK_LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 836                 bridge_delete_span(sc, bif);
 837         }
 838
 839         /* Tear down the routing table. */
 840         bridge_rtable_fini(sc);
 841
 842         BRIDGE_UNLOCK(sc);
 843
 844         NET_EPOCH_ENTER_ET(et);
 845
 846         callout_drain(&sc->sc_brcallout);
 847
 848         BRIDGE_LIST_LOCK();
 849         LIST_REMOVE(sc, sc_list);
 850         BRIDGE_LIST_UNLOCK();
 851
 852         bstp_detach(&sc->sc_stp);
 853 #ifdef ALTQ
 854         IFQ_PURGE(&ifp->if_snd);
 855 #endif
 856         NET_EPOCH_EXIT_ET(et);
 857
 858         ether_ifdetach(ifp);
 859         if_free(ifp);
 860
 861         epoch_call(net_epoch_preempt, &sc->sc_epoch_ctx, bridge_clone_destroy_cb);
 862 }
 863
 864 /*
 865  * bridge_ioctl:
 866  *
 867  *      Handle a control request from the operator.
 868  */
 869 static int
 870 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 871 {
 872         struct bridge_softc *sc = ifp->if_softc;
 873         struct ifreq *ifr = (struct ifreq *)data;
 874         struct bridge_iflist *bif;
 875         struct thread *td = curthread;
 876         union {
 877                 struct ifbreq ifbreq;
 878                 struct ifbifconf ifbifconf;
 879                 struct ifbareq ifbareq;
 880                 struct ifbaconf ifbaconf;
 881                 struct ifbrparam ifbrparam;
 882                 struct ifbropreq ifbropreq;
 883         } args;
 884         struct ifdrv *ifd = (struct ifdrv *) data;
 885         const struct bridge_control *bc;
 886         int error = 0, oldmtu;
 887
 888         BRIDGE_LOCK(sc);
 889
 890         switch (cmd) {
 891
 892         case SIOCADDMULTI:
 893         case SIOCDELMULTI:
 894                 break;
 895
 896         case SIOCGDRVSPEC:
 897         case SIOCSDRVSPEC:
 898                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 899                         error = EINVAL;
 900                         break;
 901                 }
 902                 bc = &bridge_control_table[ifd->ifd_cmd];
 903
 904                 if (cmd == SIOCGDRVSPEC &&
 905                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 906                         error = EINVAL;
 907                         break;
 908                 }
 909                 else if (cmd == SIOCSDRVSPEC &&
 910                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 911                         error = EINVAL;
 912                         break;
 913                 }
 914
 915                 if (bc->bc_flags & BC_F_SUSER) {
 916                         error = priv_check(td, PRIV_NET_BRIDGE);
 917                         if (error)
 918                                 break;
 919                 }
 920
 921                 if (ifd->ifd_len != bc->bc_argsize ||
 922                     ifd->ifd_len > sizeof(args)) {
 923                         error = EINVAL;
 924                         break;
 925                 }
 926
 927                 bzero(&args, sizeof(args));
 928                 if (bc->bc_flags & BC_F_COPYIN) {
 929                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 930                         if (error)
 931                                 break;
 932                 }
 933
 934                 oldmtu = ifp->if_mtu;
 935                 error = (*bc->bc_func)(sc, &args);
 936                 if (error)
 937                         break;
 938
 939                 /*
 940                  * Bridge MTU may change during addition of the first port.
 941                  * If it did, do network layer specific procedure.
 942                  */
 943                 if (ifp->if_mtu != oldmtu) {
 944 #ifdef INET6
 945                         nd6_setmtu(ifp);
 946 #endif
 947                         rt_updatemtu(ifp);
 948                 }
 949
 950                 if (bc->bc_flags & BC_F_COPYOUT)
 951                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 952
 953                 break;
 954
 955         case SIOCSIFFLAGS:
 956                 if (!(ifp->if_flags & IFF_UP) &&
 957                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 958                         /*
 959                          * If interface is marked down and it is running,
 960                          * then stop and disable it.
 961                          */
 962                         bridge_stop(ifp, 1);
 963                 } else if ((ifp->if_flags & IFF_UP) &&
 964                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 965                         /*
 966                          * If interface is marked up and it is stopped, then
 967                          * start it.
 968                          */
 969                         BRIDGE_UNLOCK(sc);
 970                         (*ifp->if_init)(sc);
 971                         BRIDGE_LOCK(sc);
 972                 }
 973                 break;
 974
 975         case SIOCSIFMTU:
 976                 if (ifr->ifr_mtu < 576) {
 977                         error = EINVAL;
 978                         break;
 979                 }
 980                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
 981                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 982                         break;
 983                 }
 984                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 985                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 986                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 987                                     " != %d\n", sc->sc_ifp->if_xname,
 988                                     bif->bif_ifp->if_mtu,
 989                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 990                                 error = EINVAL;
 991                                 break;
 992                         }
 993                 }
 994                 if (!error)
 995                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 996                 break;
 997         default:
 998                 /*
 999                  * drop the lock as ether_ioctl() will call bridge_start() and
1000                  * cause the lock to be recursed.
1001                  */
1002                 BRIDGE_UNLOCK(sc);
1003                 error = ether_ioctl(ifp, cmd, data);
1004                 BRIDGE_LOCK(sc);
1005                 break;
1006         }
1007
1008         BRIDGE_UNLOCK(sc);
1009
1010         return (error);
1011 }
1012
1013 /*
1014  * bridge_mutecaps:
1015  *
1016  *      Clear or restore unwanted capabilities on the member interface
1017  */
1018 static void
1019 bridge_mutecaps(struct bridge_softc *sc)
1020 {
1021         struct bridge_iflist *bif;
1022         int enabled, mask;
1023
1024         BRIDGE_LOCK_ASSERT(sc);
1025
1026         /* Initial bitmask of capabilities to test */
1027         mask = BRIDGE_IFCAPS_MASK;
1028
1029         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1030                 /* Every member must support it or its disabled */
1031                 mask &= bif->bif_savedcaps;
1032         }
1033
1034         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1035                 enabled = bif->bif_ifp->if_capenable;
1036                 enabled &= ~BRIDGE_IFCAPS_STRIP;
1037                 /* strip off mask bits and enable them again if allowed */
1038                 enabled &= ~BRIDGE_IFCAPS_MASK;
1039                 enabled |= mask;
1040                 bridge_set_ifcap(sc, bif, enabled);
1041         }
1042 }
1043
1044 static void
1045 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
1046 {
1047         struct ifnet *ifp = bif->bif_ifp;
1048         struct ifreq ifr;
1049         int error, mask, stuck;
1050
1051         bzero(&ifr, sizeof(ifr));
1052         ifr.ifr_reqcap = set;
1053
1054         if (ifp->if_capenable != set) {
1055                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
1056                 if (error)
1057                         if_printf(sc->sc_ifp,
1058                             "error setting capabilities on %s: %d\n",
1059                             ifp->if_xname, error);
1060                 mask = BRIDGE_IFCAPS_MASK | BRIDGE_IFCAPS_STRIP;
1061                 stuck = ifp->if_capenable & mask & ~set;
1062                 if (stuck != 0)
1063                         if_printf(sc->sc_ifp,
1064                             "can't disable some capabilities on %s: 0x%x\n",
1065                             ifp->if_xname, stuck);
1066         }
1067 }
1068
1069 /*
1070  * bridge_lookup_member:
1071  *
1072  *      Lookup a bridge member interface.
1073  */
1074 static struct bridge_iflist *
1075 bridge_lookup_member(struct bridge_softc *sc, const char *name)
1076 {
1077         struct bridge_iflist *bif;
1078         struct ifnet *ifp;
1079
1080         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1081
1082         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1083                 ifp = bif->bif_ifp;
1084                 if (strcmp(ifp->if_xname, name) == 0)
1085                         return (bif);
1086         }
1087
1088         return (NULL);
1089 }
1090
1091 /*
1092  * bridge_lookup_member_if:
1093  *
1094  *      Lookup a bridge member interface by ifnet*.
1095  */
1096 static struct bridge_iflist *
1097 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
1098 {
1099         struct bridge_iflist *bif;
1100
1101         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
1102
1103         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1104                 if (bif->bif_ifp == member_ifp)
1105                         return (bif);
1106         }
1107
1108         return (NULL);
1109 }
1110
1111 static void
1112 bridge_delete_member_cb(struct epoch_context *ctx)
1113 {
1114         struct bridge_iflist *bif;
1115
1116         bif = __containerof(ctx, struct bridge_iflist, bif_epoch_ctx);
1117
1118         free(bif, M_DEVBUF);
1119 }
1120
1121 /*
1122  * bridge_delete_member:
1123  *
1124  *      Delete the specified member interface.
1125  */
1126 static void
1127 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
1128     int gone)
1129 {
1130         struct ifnet *ifs = bif->bif_ifp;
1131         struct ifnet *fif = NULL;
1132         struct bridge_iflist *bifl;
1133
1134         BRIDGE_LOCK_ASSERT(sc);
1135
1136         if (bif->bif_flags & IFBIF_STP)
1137                 bstp_disable(&bif->bif_stp);
1138
1139         ifs->if_bridge = NULL;
1140         CK_LIST_REMOVE(bif, bif_next);
1141
1142         /*
1143          * If removing the interface that gave the bridge its mac address, set
1144          * the mac address of the bridge to the address of the next member, or
1145          * to its default address if no members are left.
1146          */
1147         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1148                 if (CK_LIST_EMPTY(&sc->sc_iflist)) {
1149                         bcopy(&sc->sc_defaddr,
1150                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1151                         sc->sc_ifaddr = NULL;
1152                 } else {
1153                         bifl = CK_LIST_FIRST(&sc->sc_iflist);
1154                         fif = bifl->bif_ifp;
1155                         bcopy(IF_LLADDR(fif),
1156                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1157                         sc->sc_ifaddr = fif;
1158                 }
1159                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1160         }
1161
1162         bridge_linkcheck(sc);
1163         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1164         BRIDGE_RT_LOCK(sc);
1165         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1166         BRIDGE_RT_UNLOCK(sc);
1167         KASSERT(bif->bif_addrcnt == 0,
1168             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1169
1170         ifs->if_bridge_output = NULL;
1171         ifs->if_bridge_input = NULL;
1172         ifs->if_bridge_linkstate = NULL;
1173         if (!gone) {
1174                 switch (ifs->if_type) {
1175                 case IFT_ETHER:
1176                 case IFT_L2VLAN:
1177                         /*
1178                          * Take the interface out of promiscuous mode, but only
1179                          * if it was promiscuous in the first place. It might
1180                          * not be if we're in the bridge_ioctl_add() error path.
1181                          */
1182                         if (ifs->if_flags & IFF_PROMISC)
1183                                 (void) ifpromisc(ifs, 0);
1184                         break;
1185
1186                 case IFT_GIF:
1187                         break;
1188
1189                 default:
1190 #ifdef DIAGNOSTIC
1191                         panic("bridge_delete_member: impossible");
1192 #endif
1193                         break;
1194                 }
1195                 /* reneable any interface capabilities */
1196                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1197         }
1198         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1199
1200         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1201             bridge_delete_member_cb);
1202 }
1203
1204 /*
1205  * bridge_delete_span:
1206  *
1207  *      Delete the specified span interface.
1208  */
1209 static void
1210 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1211 {
1212         BRIDGE_LOCK_ASSERT(sc);
1213
1214         KASSERT(bif->bif_ifp->if_bridge == NULL,
1215             ("%s: not a span interface", __func__));
1216
1217         CK_LIST_REMOVE(bif, bif_next);
1218
1219         epoch_call(net_epoch_preempt, &bif->bif_epoch_ctx,
1220             bridge_delete_member_cb);
1221 }
1222
1223 static int
1224 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1225 {
1226         struct ifbreq *req = arg;
1227         struct bridge_iflist *bif = NULL;
1228         struct ifnet *ifs;
1229         int error = 0;
1230
1231         ifs = ifunit(req->ifbr_ifsname);
1232         if (ifs == NULL)
1233                 return (ENOENT);
1234         if (ifs->if_ioctl == NULL)      /* must be supported */
1235                 return (EINVAL);
1236
1237         /* If it's in the span list, it can't be a member. */
1238         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1239                 if (ifs == bif->bif_ifp)
1240                         return (EBUSY);
1241
1242         if (ifs->if_bridge == sc)
1243                 return (EEXIST);
1244
1245         if (ifs->if_bridge != NULL)
1246                 return (EBUSY);
1247
1248         switch (ifs->if_type) {
1249         case IFT_ETHER:
1250         case IFT_L2VLAN:
1251         case IFT_GIF:
1252                 /* permitted interface types */
1253                 break;
1254         default:
1255                 return (EINVAL);
1256         }
1257
1258 #ifdef INET6
1259         /*
1260          * Two valid inet6 addresses with link-local scope must not be
1261          * on the parent interface and the member interfaces at the
1262          * same time.  This restriction is needed to prevent violation
1263          * of link-local scope zone.  Attempts to add a member
1264          * interface which has inet6 addresses when the parent has
1265          * inet6 triggers removal of all inet6 addresses on the member
1266          * interface.
1267          */
1268
1269         /* Check if the parent interface has a link-local scope addr. */
1270         if (V_allow_llz_overlap == 0 &&
1271             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1272                 /*
1273                  * If any, remove all inet6 addresses from the member
1274                  * interfaces.
1275                  */
1276                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1277                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1278                                 in6_ifdetach(bif->bif_ifp);
1279                                 if_printf(sc->sc_ifp,
1280                                     "IPv6 addresses on %s have been removed "
1281                                     "before adding it as a member to prevent "
1282                                     "IPv6 address scope violation.\n",
1283                                     bif->bif_ifp->if_xname);
1284                         }
1285                 }
1286                 if (in6ifa_llaonifp(ifs)) {
1287                         in6_ifdetach(ifs);
1288                         if_printf(sc->sc_ifp,
1289                             "IPv6 addresses on %s have been removed "
1290                             "before adding it as a member to prevent "
1291                             "IPv6 address scope violation.\n",
1292                             ifs->if_xname);
1293                 }
1294         }
1295 #endif
1296         /* Allow the first Ethernet member to define the MTU */
1297         if (CK_LIST_EMPTY(&sc->sc_iflist))
1298                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1299         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1300                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1301                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1302                 return (EINVAL);
1303         }
1304
1305         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1306         if (bif == NULL)
1307                 return (ENOMEM);
1308
1309         bif->bif_ifp = ifs;
1310         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1311         bif->bif_savedcaps = ifs->if_capenable;
1312
1313         /*
1314          * Assign the interface's MAC address to the bridge if it's the first
1315          * member and the MAC address of the bridge has not been changed from
1316          * the default randomly generated one.
1317          */
1318         if (V_bridge_inherit_mac && CK_LIST_EMPTY(&sc->sc_iflist) &&
1319             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr.octet, ETHER_ADDR_LEN)) {
1320                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1321                 sc->sc_ifaddr = ifs;
1322                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1323         }
1324
1325         ifs->if_bridge = sc;
1326         ifs->if_bridge_output = bridge_output;
1327         ifs->if_bridge_input = bridge_input;
1328         ifs->if_bridge_linkstate = bridge_linkstate;
1329         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1330         /*
1331          * XXX: XLOCK HERE!?!
1332          *
1333          * NOTE: insert_***HEAD*** should be safe for the traversals.
1334          */
1335         CK_LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1336
1337         /* Set interface capabilities to the intersection set of all members */
1338         bridge_mutecaps(sc);
1339         bridge_linkcheck(sc);
1340
1341         /* Place the interface into promiscuous mode */
1342         switch (ifs->if_type) {
1343                 case IFT_ETHER:
1344                 case IFT_L2VLAN:
1345                         error = ifpromisc(ifs, 1);
1346                         break;
1347         }
1348
1349         if (error)
1350                 bridge_delete_member(sc, bif, 0);
1351         return (error);
1352 }
1353
1354 static int
1355 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1356 {
1357         struct ifbreq *req = arg;
1358         struct bridge_iflist *bif;
1359
1360         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1361         if (bif == NULL)
1362                 return (ENOENT);
1363
1364         bridge_delete_member(sc, bif, 0);
1365
1366         return (0);
1367 }
1368
1369 static int
1370 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1371 {
1372         struct ifbreq *req = arg;
1373         struct bridge_iflist *bif;
1374         struct bstp_port *bp;
1375
1376         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1377         if (bif == NULL)
1378                 return (ENOENT);
1379
1380         bp = &bif->bif_stp;
1381         req->ifbr_ifsflags = bif->bif_flags;
1382         req->ifbr_state = bp->bp_state;
1383         req->ifbr_priority = bp->bp_priority;
1384         req->ifbr_path_cost = bp->bp_path_cost;
1385         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1386         req->ifbr_proto = bp->bp_protover;
1387         req->ifbr_role = bp->bp_role;
1388         req->ifbr_stpflags = bp->bp_flags;
1389         req->ifbr_addrcnt = bif->bif_addrcnt;
1390         req->ifbr_addrmax = bif->bif_addrmax;
1391         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1392
1393         /* Copy STP state options as flags */
1394         if (bp->bp_operedge)
1395                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1396         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1397                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1398         if (bp->bp_ptp_link)
1399                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1400         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1401                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1402         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1403                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1404         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1405                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1406         return (0);
1407 }
1408
1409 static int
1410 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1411 {
1412         struct ifbreq *req = arg;
1413         struct bridge_iflist *bif;
1414         struct bstp_port *bp;
1415         int error;
1416
1417         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1418         if (bif == NULL)
1419                 return (ENOENT);
1420         bp = &bif->bif_stp;
1421
1422         if (req->ifbr_ifsflags & IFBIF_SPAN)
1423                 /* SPAN is readonly */
1424                 return (EINVAL);
1425
1426         NET_EPOCH_ENTER();
1427
1428         if (req->ifbr_ifsflags & IFBIF_STP) {
1429                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1430                         error = bstp_enable(&bif->bif_stp);
1431                         if (error) {
1432                                 NET_EPOCH_EXIT();
1433                                 return (error);
1434                         }
1435                 }
1436         } else {
1437                 if ((bif->bif_flags & IFBIF_STP) != 0)
1438                         bstp_disable(&bif->bif_stp);
1439         }
1440
1441         /* Pass on STP flags */
1442         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1443         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1444         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1445         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1446
1447         /* Save the bits relating to the bridge */
1448         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1449
1450         NET_EPOCH_EXIT();
1451
1452         return (0);
1453 }
1454
1455 static int
1456 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1457 {
1458         struct ifbrparam *param = arg;
1459
1460         sc->sc_brtmax = param->ifbrp_csize;
1461         bridge_rttrim(sc);
1462
1463         return (0);
1464 }
1465
1466 static int
1467 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1468 {
1469         struct ifbrparam *param = arg;
1470
1471         param->ifbrp_csize = sc->sc_brtmax;
1472
1473         return (0);
1474 }
1475
1476 static int
1477 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1478 {
1479         struct ifbifconf *bifc = arg;
1480         struct bridge_iflist *bif;
1481         struct ifbreq breq;
1482         char *buf, *outbuf;
1483         int count, buflen, len, error = 0;
1484
1485         count = 0;
1486         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1487                 count++;
1488         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1489                 count++;
1490
1491         buflen = sizeof(breq) * count;
1492         if (bifc->ifbic_len == 0) {
1493                 bifc->ifbic_len = buflen;
1494                 return (0);
1495         }
1496         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1497         if (outbuf == NULL)
1498                 return (ENOMEM);
1499
1500         count = 0;
1501         buf = outbuf;
1502         len = min(bifc->ifbic_len, buflen);
1503         bzero(&breq, sizeof(breq));
1504         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1505                 if (len < sizeof(breq))
1506                         break;
1507
1508                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1509                     sizeof(breq.ifbr_ifsname));
1510                 /* Fill in the ifbreq structure */
1511                 error = bridge_ioctl_gifflags(sc, &breq);
1512                 if (error)
1513                         break;
1514                 memcpy(buf, &breq, sizeof(breq));
1515                 count++;
1516                 buf += sizeof(breq);
1517                 len -= sizeof(breq);
1518         }
1519         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1520                 if (len < sizeof(breq))
1521                         break;
1522
1523                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1524                     sizeof(breq.ifbr_ifsname));
1525                 breq.ifbr_ifsflags = bif->bif_flags;
1526                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1527                 memcpy(buf, &breq, sizeof(breq));
1528                 count++;
1529                 buf += sizeof(breq);
1530                 len -= sizeof(breq);
1531         }
1532
1533         bifc->ifbic_len = sizeof(breq) * count;
1534         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1535         free(outbuf, M_TEMP);
1536         return (error);
1537 }
1538
1539 static int
1540 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1541 {
1542         struct ifbaconf *bac = arg;
1543         struct bridge_rtnode *brt;
1544         struct ifbareq bareq;
1545         char *buf, *outbuf;
1546         int count, buflen, len, error = 0;
1547
1548         if (bac->ifbac_len == 0)
1549                 return (0);
1550
1551         count = 0;
1552         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1553                 count++;
1554         buflen = sizeof(bareq) * count;
1555
1556         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1557         if (outbuf == NULL)
1558                 return (ENOMEM);
1559
1560         count = 0;
1561         buf = outbuf;
1562         len = min(bac->ifbac_len, buflen);
1563         bzero(&bareq, sizeof(bareq));
1564         CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1565                 if (len < sizeof(bareq))
1566                         goto out;
1567                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1568                     sizeof(bareq.ifba_ifsname));
1569                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1570                 bareq.ifba_vlan = brt->brt_vlan;
1571                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1572                                 time_uptime < brt->brt_expire)
1573                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1574                 else
1575                         bareq.ifba_expire = 0;
1576                 bareq.ifba_flags = brt->brt_flags;
1577
1578                 memcpy(buf, &bareq, sizeof(bareq));
1579                 count++;
1580                 buf += sizeof(bareq);
1581                 len -= sizeof(bareq);
1582         }
1583 out:
1584         bac->ifbac_len = sizeof(bareq) * count;
1585         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1586         free(outbuf, M_TEMP);
1587         return (error);
1588 }
1589
1590 static int
1591 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1592 {
1593         struct ifbareq *req = arg;
1594         struct bridge_iflist *bif;
1595         struct epoch_tracker et;
1596         int error;
1597
1598         NET_EPOCH_ENTER_ET(et);
1599
1600         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1601         if (bif == NULL) {
1602                 NET_EPOCH_EXIT_ET(et);
1603                 return (ENOENT);
1604         }
1605
1606         /* bridge_rtupdate() may acquire the lock. */
1607         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1608             req->ifba_flags);
1609         NET_EPOCH_EXIT_ET(et);
1610
1611         return (error);
1612 }
1613
1614 static int
1615 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1616 {
1617         struct ifbrparam *param = arg;
1618
1619         sc->sc_brttimeout = param->ifbrp_ctime;
1620         return (0);
1621 }
1622
1623 static int
1624 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1625 {
1626         struct ifbrparam *param = arg;
1627
1628         param->ifbrp_ctime = sc->sc_brttimeout;
1629         return (0);
1630 }
1631
1632 static int
1633 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1634 {
1635         struct ifbareq *req = arg;
1636
1637         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1638 }
1639
1640 static int
1641 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1642 {
1643         struct ifbreq *req = arg;
1644
1645         BRIDGE_RT_LOCK(sc);
1646         bridge_rtflush(sc, req->ifbr_ifsflags);
1647         BRIDGE_RT_UNLOCK(sc);
1648
1649         return (0);
1650 }
1651
1652 static int
1653 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1654 {
1655         struct ifbrparam *param = arg;
1656         struct bstp_state *bs = &sc->sc_stp;
1657
1658         param->ifbrp_prio = bs->bs_bridge_priority;
1659         return (0);
1660 }
1661
1662 static int
1663 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1664 {
1665         struct ifbrparam *param = arg;
1666
1667         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1668 }
1669
1670 static int
1671 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1672 {
1673         struct ifbrparam *param = arg;
1674         struct bstp_state *bs = &sc->sc_stp;
1675
1676         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1677         return (0);
1678 }
1679
1680 static int
1681 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1682 {
1683         struct ifbrparam *param = arg;
1684
1685         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1686 }
1687
1688 static int
1689 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1690 {
1691         struct ifbrparam *param = arg;
1692         struct bstp_state *bs = &sc->sc_stp;
1693
1694         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1695         return (0);
1696 }
1697
1698 static int
1699 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1700 {
1701         struct ifbrparam *param = arg;
1702
1703         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1704 }
1705
1706 static int
1707 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1708 {
1709         struct ifbrparam *param = arg;
1710         struct bstp_state *bs = &sc->sc_stp;
1711
1712         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1713         return (0);
1714 }
1715
1716 static int
1717 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1718 {
1719         struct ifbrparam *param = arg;
1720
1721         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1722 }
1723
1724 static int
1725 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1726 {
1727         struct ifbreq *req = arg;
1728         struct bridge_iflist *bif;
1729
1730         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1731         if (bif == NULL)
1732                 return (ENOENT);
1733
1734         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1735 }
1736
1737 static int
1738 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1739 {
1740         struct ifbreq *req = arg;
1741         struct bridge_iflist *bif;
1742
1743         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1744         if (bif == NULL)
1745                 return (ENOENT);
1746
1747         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1748 }
1749
1750 static int
1751 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1752 {
1753         struct ifbreq *req = arg;
1754         struct bridge_iflist *bif;
1755
1756         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1757         if (bif == NULL)
1758                 return (ENOENT);
1759
1760         bif->bif_addrmax = req->ifbr_addrmax;
1761         return (0);
1762 }
1763
1764 static int
1765 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1766 {
1767         struct ifbreq *req = arg;
1768         struct bridge_iflist *bif = NULL;
1769         struct ifnet *ifs;
1770
1771         ifs = ifunit(req->ifbr_ifsname);
1772         if (ifs == NULL)
1773                 return (ENOENT);
1774
1775         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1776                 if (ifs == bif->bif_ifp)
1777                         return (EBUSY);
1778
1779         if (ifs->if_bridge != NULL)
1780                 return (EBUSY);
1781
1782         switch (ifs->if_type) {
1783                 case IFT_ETHER:
1784                 case IFT_GIF:
1785                 case IFT_L2VLAN:
1786                         break;
1787                 default:
1788                         return (EINVAL);
1789         }
1790
1791         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1792         if (bif == NULL)
1793                 return (ENOMEM);
1794
1795         bif->bif_ifp = ifs;
1796         bif->bif_flags = IFBIF_SPAN;
1797
1798         CK_LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1799
1800         return (0);
1801 }
1802
1803 static int
1804 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1805 {
1806         struct ifbreq *req = arg;
1807         struct bridge_iflist *bif;
1808         struct ifnet *ifs;
1809
1810         ifs = ifunit(req->ifbr_ifsname);
1811         if (ifs == NULL)
1812                 return (ENOENT);
1813
1814         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1815                 if (ifs == bif->bif_ifp)
1816                         break;
1817
1818         if (bif == NULL)
1819                 return (ENOENT);
1820
1821         bridge_delete_span(sc, bif);
1822
1823         return (0);
1824 }
1825
1826 static int
1827 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1828 {
1829         struct ifbropreq *req = arg;
1830         struct bstp_state *bs = &sc->sc_stp;
1831         struct bstp_port *root_port;
1832
1833         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1834         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1835         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1836
1837         root_port = bs->bs_root_port;
1838         if (root_port == NULL)
1839                 req->ifbop_root_port = 0;
1840         else
1841                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1842
1843         req->ifbop_holdcount = bs->bs_txholdcount;
1844         req->ifbop_priority = bs->bs_bridge_priority;
1845         req->ifbop_protocol = bs->bs_protover;
1846         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1847         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1848         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1849         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1850         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1851         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1852
1853         return (0);
1854 }
1855
1856 static int
1857 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1858 {
1859         struct ifbrparam *param = arg;
1860
1861         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1862         return (0);
1863 }
1864
1865 static int
1866 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1867 {
1868         struct ifbpstpconf *bifstp = arg;
1869         struct bridge_iflist *bif;
1870         struct bstp_port *bp;
1871         struct ifbpstpreq bpreq;
1872         char *buf, *outbuf;
1873         int count, buflen, len, error = 0;
1874
1875         count = 0;
1876         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1877                 if ((bif->bif_flags & IFBIF_STP) != 0)
1878                         count++;
1879         }
1880
1881         buflen = sizeof(bpreq) * count;
1882         if (bifstp->ifbpstp_len == 0) {
1883                 bifstp->ifbpstp_len = buflen;
1884                 return (0);
1885         }
1886
1887         outbuf = malloc(buflen, M_TEMP, M_NOWAIT | M_ZERO);
1888         if (outbuf == NULL)
1889                 return (ENOMEM);
1890
1891         count = 0;
1892         buf = outbuf;
1893         len = min(bifstp->ifbpstp_len, buflen);
1894         bzero(&bpreq, sizeof(bpreq));
1895         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1896                 if (len < sizeof(bpreq))
1897                         break;
1898
1899                 if ((bif->bif_flags & IFBIF_STP) == 0)
1900                         continue;
1901
1902                 bp = &bif->bif_stp;
1903                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1904                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1905                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1906                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1907                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1908                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1909
1910                 memcpy(buf, &bpreq, sizeof(bpreq));
1911                 count++;
1912                 buf += sizeof(bpreq);
1913                 len -= sizeof(bpreq);
1914         }
1915
1916         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1917         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1918         free(outbuf, M_TEMP);
1919         return (error);
1920 }
1921
1922 static int
1923 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1924 {
1925         struct ifbrparam *param = arg;
1926
1927         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1928 }
1929
1930 static int
1931 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1932 {
1933         struct ifbrparam *param = arg;
1934
1935         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1936 }
1937
1938 /*
1939  * bridge_ifdetach:
1940  *
1941  *      Detach an interface from a bridge.  Called when a member
1942  *      interface is detaching.
1943  */
1944 static void
1945 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1946 {
1947         struct bridge_softc *sc = ifp->if_bridge;
1948         struct bridge_iflist *bif;
1949
1950         if (ifp->if_flags & IFF_RENAMING)
1951                 return;
1952         if (V_bridge_cloner == NULL) {
1953                 /*
1954                  * This detach handler can be called after
1955                  * vnet_bridge_uninit().  Just return in that case.
1956                  */
1957                 return;
1958         }
1959         /* Check if the interface is a bridge member */
1960         if (sc != NULL) {
1961                 BRIDGE_LOCK(sc);
1962
1963                 bif = bridge_lookup_member_if(sc, ifp);
1964                 if (bif != NULL)
1965                         bridge_delete_member(sc, bif, 1);
1966
1967                 BRIDGE_UNLOCK(sc);
1968                 return;
1969         }
1970
1971         /* Check if the interface is a span port */
1972         BRIDGE_LIST_LOCK();
1973         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1974                 BRIDGE_LOCK(sc);
1975                 CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1976                         if (ifp == bif->bif_ifp) {
1977                                 bridge_delete_span(sc, bif);
1978                                 break;
1979                         }
1980
1981                 BRIDGE_UNLOCK(sc);
1982         }
1983         BRIDGE_LIST_UNLOCK();
1984 }
1985
1986 /*
1987  * bridge_init:
1988  *
1989  *      Initialize a bridge interface.
1990  */
1991 static void
1992 bridge_init(void *xsc)
1993 {
1994         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1995         struct ifnet *ifp = sc->sc_ifp;
1996
1997         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1998                 return;
1999
2000         BRIDGE_LOCK(sc);
2001         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
2002             bridge_timer, sc);
2003
2004         ifp->if_drv_flags |= IFF_DRV_RUNNING;
2005         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
2006
2007         BRIDGE_UNLOCK(sc);
2008 }
2009
2010 /*
2011  * bridge_stop:
2012  *
2013  *      Stop the bridge interface.
2014  */
2015 static void
2016 bridge_stop(struct ifnet *ifp, int disable)
2017 {
2018         struct bridge_softc *sc = ifp->if_softc;
2019
2020         BRIDGE_LOCK_ASSERT(sc);
2021
2022         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2023                 return;
2024
2025         BRIDGE_RT_LOCK(sc);
2026         callout_stop(&sc->sc_brcallout);
2027
2028         bstp_stop(&sc->sc_stp);
2029
2030         bridge_rtflush(sc, IFBF_FLUSHDYN);
2031         BRIDGE_RT_UNLOCK(sc);
2032
2033         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
2034 }
2035
2036 /*
2037  * bridge_enqueue:
2038  *
2039  *      Enqueue a packet on a bridge member interface.
2040  *
2041  */
2042 static int
2043 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
2044 {
2045         int len, err = 0;
2046         short mflags;
2047         struct mbuf *m0;
2048
2049         /* We may be sending a fragment so traverse the mbuf */
2050         for (; m; m = m0) {
2051                 m0 = m->m_nextpkt;
2052                 m->m_nextpkt = NULL;
2053                 len = m->m_pkthdr.len;
2054                 mflags = m->m_flags;
2055
2056                 /*
2057                  * If underlying interface can not do VLAN tag insertion itself
2058                  * then attach a packet tag that holds it.
2059                  */
2060                 if ((m->m_flags & M_VLANTAG) &&
2061                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
2062                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
2063                         if (m == NULL) {
2064                                 if_printf(dst_ifp,
2065                                     "unable to prepend VLAN header\n");
2066                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
2067                                 continue;
2068                         }
2069                         m->m_flags &= ~M_VLANTAG;
2070                 }
2071
2072                 M_ASSERTPKTHDR(m); /* We shouldn't transmit mbuf without pkthdr */
2073                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
2074                         m_freem(m0);
2075                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2076                         break;
2077                 }
2078
2079                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
2080                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
2081                 if (mflags & M_MCAST)
2082                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
2083         }
2084
2085         return (err);
2086 }
2087
2088 /*
2089  * bridge_dummynet:
2090  *
2091  *      Receive a queued packet from dummynet and pass it on to the output
2092  *      interface.
2093  *
2094  *      The mbuf has the Ethernet header already attached.
2095  */
2096 static void
2097 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
2098 {
2099         struct bridge_softc *sc;
2100
2101         sc = ifp->if_bridge;
2102
2103         /*
2104          * The packet didnt originate from a member interface. This should only
2105          * ever happen if a member interface is removed while packets are
2106          * queued for it.
2107          */
2108         if (sc == NULL) {
2109                 m_freem(m);
2110                 return;
2111         }
2112
2113         if (PFIL_HOOKED(&V_inet_pfil_hook)
2114 #ifdef INET6
2115             || PFIL_HOOKED(&V_inet6_pfil_hook)
2116 #endif
2117             ) {
2118                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
2119                         return;
2120                 if (m == NULL)
2121                         return;
2122         }
2123
2124         bridge_enqueue(sc, ifp, m);
2125 }
2126
2127 /*
2128  * bridge_output:
2129  *
2130  *      Send output from a bridge member interface.  This
2131  *      performs the bridging function for locally originated
2132  *      packets.
2133  *
2134  *      The mbuf has the Ethernet header already attached.  We must
2135  *      enqueue or free the mbuf before returning.
2136  */
2137 static int
2138 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
2139     struct rtentry *rt)
2140 {
2141         struct epoch_tracker et;
2142         struct ether_header *eh;
2143         struct ifnet *dst_if;
2144         struct bridge_softc *sc;
2145         uint16_t vlan;
2146
2147         NET_EPOCH_ENTER_ET(et);
2148
2149         if (m->m_len < ETHER_HDR_LEN) {
2150                 m = m_pullup(m, ETHER_HDR_LEN);
2151                 if (m == NULL) {
2152                         NET_EPOCH_EXIT_ET(et);
2153                         return (0);
2154                 }
2155         }
2156
2157         eh = mtod(m, struct ether_header *);
2158         sc = ifp->if_bridge;
2159         vlan = VLANTAGOF(m);
2160
2161         /*
2162          * If bridge is down, but the original output interface is up,
2163          * go ahead and send out that interface.  Otherwise, the packet
2164          * is dropped below.
2165          */
2166         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2167                 dst_if = ifp;
2168                 goto sendunicast;
2169         }
2170
2171         /*
2172          * If the packet is a multicast, or we don't know a better way to
2173          * get there, send to all interfaces.
2174          */
2175         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2176                 dst_if = NULL;
2177         else
2178                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2179         if (dst_if == NULL) {
2180                 struct bridge_iflist *bif;
2181                 struct mbuf *mc;
2182                 int used = 0;
2183
2184                 bridge_span(sc, m);
2185
2186                 CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2187                         dst_if = bif->bif_ifp;
2188
2189                         if (dst_if->if_type == IFT_GIF)
2190                                 continue;
2191                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2192                                 continue;
2193
2194                         /*
2195                          * If this is not the original output interface,
2196                          * and the interface is participating in spanning
2197                          * tree, make sure the port is in a state that
2198                          * allows forwarding.
2199                          */
2200                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2201                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2202                                 continue;
2203
2204                         if (CK_LIST_NEXT(bif, bif_next) == NULL) {
2205                                 used = 1;
2206                                 mc = m;
2207                         } else {
2208                                 mc = m_copypacket(m, M_NOWAIT);
2209                                 if (mc == NULL) {
2210                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2211                                         continue;
2212                                 }
2213                         }
2214
2215                         bridge_enqueue(sc, dst_if, mc);
2216                 }
2217                 if (used == 0)
2218                         m_freem(m);
2219                 NET_EPOCH_EXIT_ET(et);
2220                 return (0);
2221         }
2222
2223 sendunicast:
2224         /*
2225          * XXX Spanning tree consideration here?
2226          */
2227
2228         bridge_span(sc, m);
2229         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2230                 m_freem(m);
2231                 NET_EPOCH_EXIT_ET(et);
2232                 return (0);
2233         }
2234
2235         bridge_enqueue(sc, dst_if, m);
2236         NET_EPOCH_EXIT_ET(et);
2237         return (0);
2238 }
2239
2240 /*
2241  * bridge_transmit:
2242  *
2243  *      Do output on a bridge.
2244  *
2245  */
2246 static int
2247 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2248 {
2249         struct epoch_tracker et;
2250         struct bridge_softc *sc;
2251         struct ether_header *eh;
2252         struct ifnet *dst_if;
2253         int error = 0;
2254
2255         NET_EPOCH_ENTER_ET(et);
2256
2257         sc = ifp->if_softc;
2258
2259         ETHER_BPF_MTAP(ifp, m);
2260
2261         eh = mtod(m, struct ether_header *);
2262
2263         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2264             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2265                 error = bridge_enqueue(sc, dst_if, m);
2266         } else
2267                 bridge_broadcast(sc, ifp, m, 0);
2268
2269         NET_EPOCH_EXIT_ET(et);
2270
2271         return (error);
2272 }
2273
2274 #ifdef ALTQ
2275 static void
2276 bridge_altq_start(if_t ifp)
2277 {
2278         struct ifaltq *ifq = &ifp->if_snd;
2279         struct mbuf *m;
2280
2281         IFQ_LOCK(ifq);
2282         IFQ_DEQUEUE_NOLOCK(ifq, m);
2283         while (m != NULL) {
2284                 bridge_transmit(ifp, m);
2285                 IFQ_DEQUEUE_NOLOCK(ifq, m);
2286         }
2287         IFQ_UNLOCK(ifq);
2288 }
2289
2290 static int
2291 bridge_altq_transmit(if_t ifp, struct mbuf *m)
2292 {
2293         int err;
2294
2295         if (ALTQ_IS_ENABLED(&ifp->if_snd)) {
2296                 IFQ_ENQUEUE(&ifp->if_snd, m, err);
2297                 if (err == 0)
2298                         bridge_altq_start(ifp);
2299         } else
2300                 err = bridge_transmit(ifp, m);
2301
2302         return (err);
2303 }
2304 #endif  /* ALTQ */
2305
2306 /*
2307  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2308  */
2309 static void
2310 bridge_qflush(struct ifnet *ifp __unused)
2311 {
2312 }
2313
2314 /*
2315  * bridge_forward:
2316  *
2317  *      The forwarding function of the bridge.
2318  *
2319  *      NOTE: Releases the lock on return.
2320  */
2321 static void
2322 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2323     struct mbuf *m)
2324 {
2325         struct bridge_iflist *dbif;
2326         struct ifnet *src_if, *dst_if, *ifp;
2327         struct ether_header *eh;
2328         uint16_t vlan;
2329         uint8_t *dst;
2330         int error;
2331
2332         MPASS(in_epoch(net_epoch_preempt));
2333
2334         src_if = m->m_pkthdr.rcvif;
2335         ifp = sc->sc_ifp;
2336
2337         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2338         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2339         vlan = VLANTAGOF(m);
2340
2341         if ((sbif->bif_flags & IFBIF_STP) &&
2342             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2343                 goto drop;
2344
2345         eh = mtod(m, struct ether_header *);
2346         dst = eh->ether_dhost;
2347
2348         /* If the interface is learning, record the address. */
2349         if (sbif->bif_flags & IFBIF_LEARNING) {
2350                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2351                     sbif, 0, IFBAF_DYNAMIC);
2352                 /*
2353                  * If the interface has addresses limits then deny any source
2354                  * that is not in the cache.
2355                  */
2356                 if (error && sbif->bif_addrmax)
2357                         goto drop;
2358         }
2359
2360         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2361             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2362                 goto drop;
2363
2364         /*
2365          * At this point, the port either doesn't participate
2366          * in spanning tree or it is in the forwarding state.
2367          */
2368
2369         /*
2370          * If the packet is unicast, destined for someone on
2371          * "this" side of the bridge, drop it.
2372          */
2373         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2374                 dst_if = bridge_rtlookup(sc, dst, vlan);
2375                 if (src_if == dst_if)
2376                         goto drop;
2377         } else {
2378                 /*
2379                  * Check if its a reserved multicast address, any address
2380                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2381                  * bridge.
2382                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2383                  */
2384                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2385                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2386                     dst[4] == 0x00 && dst[5] <= 0x0f)
2387                         goto drop;
2388
2389                 /* ...forward it to all interfaces. */
2390                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2391                 dst_if = NULL;
2392         }
2393
2394         /*
2395          * If we have a destination interface which is a member of our bridge,
2396          * OR this is a unicast packet, push it through the bpf(4) machinery.
2397          * For broadcast or multicast packets, don't bother because it will
2398          * be reinjected into ether_input. We do this before we pass the packets
2399          * through the pfil(9) framework, as it is possible that pfil(9) will
2400          * drop the packet, or possibly modify it, making it difficult to debug
2401          * firewall issues on the bridge.
2402          */
2403         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2404                 ETHER_BPF_MTAP(ifp, m);
2405
2406         /* run the packet filter */
2407         if (PFIL_HOOKED(&V_inet_pfil_hook)
2408 #ifdef INET6
2409             || PFIL_HOOKED(&V_inet6_pfil_hook)
2410 #endif
2411             ) {
2412                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2413                         return;
2414                 if (m == NULL)
2415                         return;
2416         }
2417
2418         if (dst_if == NULL) {
2419                 bridge_broadcast(sc, src_if, m, 1);
2420                 return;
2421         }
2422
2423         /*
2424          * At this point, we're dealing with a unicast frame
2425          * going to a different interface.
2426          */
2427         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2428                 goto drop;
2429
2430         dbif = bridge_lookup_member_if(sc, dst_if);
2431         if (dbif == NULL)
2432                 /* Not a member of the bridge (anymore?) */
2433                 goto drop;
2434
2435         /* Private segments can not talk to each other */
2436         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2437                 goto drop;
2438
2439         if ((dbif->bif_flags & IFBIF_STP) &&
2440             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2441                 goto drop;
2442
2443         if (PFIL_HOOKED(&V_inet_pfil_hook)
2444 #ifdef INET6
2445             || PFIL_HOOKED(&V_inet6_pfil_hook)
2446 #endif
2447             ) {
2448                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2449                         return;
2450                 if (m == NULL)
2451                         return;
2452         }
2453
2454         bridge_enqueue(sc, dst_if, m);
2455         return;
2456
2457 drop:
2458         m_freem(m);
2459 }
2460
2461 /*
2462  * bridge_input:
2463  *
2464  *      Receive input from a member interface.  Queue the packet for
2465  *      bridging if it is not for us.
2466  */
2467 static struct mbuf *
2468 bridge_input(struct ifnet *ifp, struct mbuf *m)
2469 {
2470         struct epoch_tracker et;
2471         struct bridge_softc *sc = ifp->if_bridge;
2472         struct bridge_iflist *bif, *bif2;
2473         struct ifnet *bifp;
2474         struct ether_header *eh;
2475         struct mbuf *mc, *mc2;
2476         uint16_t vlan;
2477         int error;
2478
2479         NET_EPOCH_ENTER_ET(et);
2480
2481         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2482                 NET_EPOCH_EXIT_ET(et);
2483                 return (m);
2484         }
2485
2486         bifp = sc->sc_ifp;
2487         vlan = VLANTAGOF(m);
2488
2489         /*
2490          * Implement support for bridge monitoring. If this flag has been
2491          * set on this interface, discard the packet once we push it through
2492          * the bpf(4) machinery, but before we do, increment the byte and
2493          * packet counters associated with this interface.
2494          */
2495         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2496                 m->m_pkthdr.rcvif  = bifp;
2497                 ETHER_BPF_MTAP(bifp, m);
2498                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2499                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2500                 m_freem(m);
2501                 NET_EPOCH_EXIT_ET(et);
2502                 return (NULL);
2503         }
2504         bif = bridge_lookup_member_if(sc, ifp);
2505         if (bif == NULL) {
2506                 NET_EPOCH_EXIT_ET(et);
2507                 return (m);
2508         }
2509
2510         eh = mtod(m, struct ether_header *);
2511
2512         bridge_span(sc, m);
2513
2514         if (m->m_flags & (M_BCAST|M_MCAST)) {
2515                 /* Tap off 802.1D packets; they do not get forwarded. */
2516                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2517                     ETHER_ADDR_LEN) == 0) {
2518                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2519                         NET_EPOCH_EXIT_ET(et);
2520                         return (NULL);
2521                 }
2522
2523                 if ((bif->bif_flags & IFBIF_STP) &&
2524                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2525                         NET_EPOCH_EXIT_ET(et);
2526                         return (m);
2527                 }
2528
2529                 /*
2530                  * Make a deep copy of the packet and enqueue the copy
2531                  * for bridge processing; return the original packet for
2532                  * local processing.
2533                  */
2534                 mc = m_dup(m, M_NOWAIT);
2535                 if (mc == NULL) {
2536                         NET_EPOCH_EXIT_ET(et);
2537                         return (m);
2538                 }
2539
2540                 /* Perform the bridge forwarding function with the copy. */
2541                 bridge_forward(sc, bif, mc);
2542
2543                 /*
2544                  * Reinject the mbuf as arriving on the bridge so we have a
2545                  * chance at claiming multicast packets. We can not loop back
2546                  * here from ether_input as a bridge is never a member of a
2547                  * bridge.
2548                  */
2549                 KASSERT(bifp->if_bridge == NULL,
2550                     ("loop created in bridge_input"));
2551                 mc2 = m_dup(m, M_NOWAIT);
2552                 if (mc2 != NULL) {
2553                         /* Keep the layer3 header aligned */
2554                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2555                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2556                 }
2557                 if (mc2 != NULL) {
2558                         mc2->m_pkthdr.rcvif = bifp;
2559                         (*bifp->if_input)(bifp, mc2);
2560                 }
2561
2562                 /* Return the original packet for local processing. */
2563                 NET_EPOCH_EXIT_ET(et);
2564                 return (m);
2565         }
2566
2567         if ((bif->bif_flags & IFBIF_STP) &&
2568             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2569                 NET_EPOCH_EXIT_ET(et);
2570                 return (m);
2571         }
2572
2573 #if (defined(INET) || defined(INET6))
2574 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2575         || ((iface)->if_carp \
2576             && (*carp_forus_p)((iface), eh->ether_dhost))
2577 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2578         || ((iface)->if_carp \
2579             && (*carp_forus_p)((iface), eh->ether_shost))
2580 #else
2581 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2582 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2583 #endif
2584
2585 #ifdef INET6
2586 #   define OR_PFIL_HOOKED_INET6 \
2587         || PFIL_HOOKED(&V_inet6_pfil_hook)
2588 #else
2589 #   define OR_PFIL_HOOKED_INET6
2590 #endif
2591
2592 #define GRAB_OUR_PACKETS(iface) \
2593         if ((iface)->if_type == IFT_GIF) \
2594                 continue; \
2595         /* It is destined for us. */ \
2596         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2597             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2598             ) {                                                         \
2599                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2600                         ETHER_BPF_MTAP(iface, m);                       \
2601                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2602                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2603                         /* Filter on the physical interface. */         \
2604                         if (V_pfil_local_phys &&                        \
2605                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2606                              OR_PFIL_HOOKED_INET6)) {                   \
2607                                 if (bridge_pfil(&m, NULL, ifp,          \
2608                                     PFIL_IN) != 0 || m == NULL) {       \
2609                                         NET_EPOCH_EXIT_ET(et);  \
2610                                         return (NULL);                  \
2611                                 }                                       \
2612                                 eh = mtod(m, struct ether_header *);    \
2613                         }                                               \
2614                 }                                                       \
2615                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2616                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2617                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2618                         if (error && bif->bif_addrmax) {                \
2619                                 m_freem(m);                             \
2620                                 NET_EPOCH_EXIT_ET(et);                  \
2621                                 return (NULL);                          \
2622                         }                                               \
2623                 }                                                       \
2624                 m->m_pkthdr.rcvif = iface;                              \
2625                 NET_EPOCH_EXIT_ET(et);                                  \
2626                 return (m);                                             \
2627         }                                                               \
2628                                                                         \
2629         /* We just received a packet that we sent out. */               \
2630         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2631             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2632             ) {                                                         \
2633                 m_freem(m);                                             \
2634                 NET_EPOCH_EXIT_ET(et);                                  \
2635                 return (NULL);                                          \
2636         }
2637
2638         /*
2639          * Unicast.  Make sure it's not for the bridge.
2640          */
2641         do { GRAB_OUR_PACKETS(bifp) } while (0);
2642
2643         /*
2644          * Give a chance for ifp at first priority. This will help when the
2645          * packet comes through the interface like VLAN's with the same MACs
2646          * on several interfaces from the same bridge. This also will save
2647          * some CPU cycles in case the destination interface and the input
2648          * interface (eq ifp) are the same.
2649          */
2650         do { GRAB_OUR_PACKETS(ifp) } while (0);
2651
2652         /* Now check the all bridge members. */
2653         CK_LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2654                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2655         }
2656
2657 #undef OR_CARP_CHECK_WE_ARE_DST
2658 #undef OR_CARP_CHECK_WE_ARE_SRC
2659 #undef OR_PFIL_HOOKED_INET6
2660 #undef GRAB_OUR_PACKETS
2661
2662         /* Perform the bridge forwarding function. */
2663         bridge_forward(sc, bif, m);
2664
2665         NET_EPOCH_EXIT_ET(et);
2666         return (NULL);
2667 }
2668
2669 /*
2670  * bridge_broadcast:
2671  *
2672  *      Send a frame to all interfaces that are members of
2673  *      the bridge, except for the one on which the packet
2674  *      arrived.
2675  *
2676  *      NOTE: Releases the lock on return.
2677  */
2678 static void
2679 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2680     struct mbuf *m, int runfilt)
2681 {
2682         struct bridge_iflist *dbif, *sbif;
2683         struct mbuf *mc;
2684         struct ifnet *dst_if;
2685         int used = 0, i;
2686
2687         MPASS(in_epoch(net_epoch_preempt));
2688
2689         sbif = bridge_lookup_member_if(sc, src_if);
2690
2691         /* Filter on the bridge interface before broadcasting */
2692         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2693 #ifdef INET6
2694             || PFIL_HOOKED(&V_inet6_pfil_hook)
2695 #endif
2696             )) {
2697                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2698                         return;
2699                 if (m == NULL)
2700                         return;
2701         }
2702
2703         CK_LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2704                 dst_if = dbif->bif_ifp;
2705                 if (dst_if == src_if)
2706                         continue;
2707
2708                 /* Private segments can not talk to each other */
2709                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2710                         continue;
2711
2712                 if ((dbif->bif_flags & IFBIF_STP) &&
2713                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2714                         continue;
2715
2716                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2717                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2718                         continue;
2719
2720                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2721                         continue;
2722
2723                 if (CK_LIST_NEXT(dbif, bif_next) == NULL) {
2724                         mc = m;
2725                         used = 1;
2726                 } else {
2727                         mc = m_dup(m, M_NOWAIT);
2728                         if (mc == NULL) {
2729                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2730                                 continue;
2731                         }
2732                 }
2733
2734                 /*
2735                  * Filter on the output interface. Pass a NULL bridge interface
2736                  * pointer so we do not redundantly filter on the bridge for
2737                  * each interface we broadcast on.
2738                  */
2739                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2740 #ifdef INET6
2741                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2742 #endif
2743                     )) {
2744                         if (used == 0) {
2745                                 /* Keep the layer3 header aligned */
2746                                 i = min(mc->m_pkthdr.len, max_protohdr);
2747                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2748                                 if (mc == NULL) {
2749                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2750                                         continue;
2751                                 }
2752                         }
2753                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2754                                 continue;
2755                         if (mc == NULL)
2756                                 continue;
2757                 }
2758
2759                 bridge_enqueue(sc, dst_if, mc);
2760         }
2761         if (used == 0)
2762                 m_freem(m);
2763 }
2764
2765 /*
2766  * bridge_span:
2767  *
2768  *      Duplicate a packet out one or more interfaces that are in span mode,
2769  *      the original mbuf is unmodified.
2770  */
2771 static void
2772 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2773 {
2774         struct bridge_iflist *bif;
2775         struct ifnet *dst_if;
2776         struct mbuf *mc;
2777
2778         MPASS(in_epoch(net_epoch_preempt));
2779
2780         if (CK_LIST_EMPTY(&sc->sc_spanlist))
2781                 return;
2782
2783         CK_LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2784                 dst_if = bif->bif_ifp;
2785
2786                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2787                         continue;
2788
2789                 mc = m_copypacket(m, M_NOWAIT);
2790                 if (mc == NULL) {
2791                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2792                         continue;
2793                 }
2794
2795                 bridge_enqueue(sc, dst_if, mc);
2796         }
2797 }
2798
2799 /*
2800  * bridge_rtupdate:
2801  *
2802  *      Add a bridge routing entry.
2803  */
2804 static int
2805 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2806     struct bridge_iflist *bif, int setflags, uint8_t flags)
2807 {
2808         struct bridge_rtnode *brt;
2809         int error;
2810
2811         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
2812
2813         /* Check the source address is valid and not multicast. */
2814         if (ETHER_IS_MULTICAST(dst) ||
2815             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2816              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2817                 return (EINVAL);
2818
2819         /* 802.1p frames map to vlan 1 */
2820         if (vlan == 0)
2821                 vlan = 1;
2822
2823         /*
2824          * A route for this destination might already exist.  If so,
2825          * update it, otherwise create a new one.
2826          */
2827         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2828                 BRIDGE_RT_LOCK(sc);
2829
2830                 /* Check again, now that we have the lock. There could have
2831                  * been a race and we only want to insert this once. */
2832                 if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) != NULL) {
2833                         BRIDGE_RT_UNLOCK(sc);
2834                         return (0);
2835                 }
2836
2837                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2838                         sc->sc_brtexceeded++;
2839                         BRIDGE_RT_UNLOCK(sc);
2840                         return (ENOSPC);
2841                 }
2842                 /* Check per interface address limits (if enabled) */
2843                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2844                         bif->bif_addrexceeded++;
2845                         BRIDGE_RT_UNLOCK(sc);
2846                         return (ENOSPC);
2847                 }
2848
2849                 /*
2850                  * Allocate a new bridge forwarding node, and
2851                  * initialize the expiration time and Ethernet
2852                  * address.
2853                  */
2854                 brt = uma_zalloc(V_bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2855                 if (brt == NULL) {
2856                         BRIDGE_RT_UNLOCK(sc);
2857                         return (ENOMEM);
2858                 }
2859                 brt->brt_vnet = curvnet;
2860
2861                 if (bif->bif_flags & IFBIF_STICKY)
2862                         brt->brt_flags = IFBAF_STICKY;
2863                 else
2864                         brt->brt_flags = IFBAF_DYNAMIC;
2865
2866                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2867                 brt->brt_vlan = vlan;
2868
2869                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2870                         uma_zfree(V_bridge_rtnode_zone, brt);
2871                         BRIDGE_RT_UNLOCK(sc);
2872                         return (error);
2873                 }
2874                 brt->brt_dst = bif;
2875                 bif->bif_addrcnt++;
2876
2877                 BRIDGE_RT_UNLOCK(sc);
2878         }
2879
2880         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2881             brt->brt_dst != bif) {
2882                 BRIDGE_RT_LOCK(sc);
2883                 brt->brt_dst->bif_addrcnt--;
2884                 brt->brt_dst = bif;
2885                 brt->brt_dst->bif_addrcnt++;
2886                 BRIDGE_RT_UNLOCK(sc);
2887         }
2888
2889         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2890                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2891         if (setflags)
2892                 brt->brt_flags = flags;
2893
2894         return (0);
2895 }
2896
2897 /*
2898  * bridge_rtlookup:
2899  *
2900  *      Lookup the destination interface for an address.
2901  */
2902 static struct ifnet *
2903 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2904 {
2905         struct bridge_rtnode *brt;
2906
2907         MPASS(in_epoch(net_epoch_preempt));
2908
2909         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2910                 return (NULL);
2911
2912         return (brt->brt_ifp);
2913 }
2914
2915 /*
2916  * bridge_rttrim:
2917  *
2918  *      Trim the routine table so that we have a number
2919  *      of routing entries less than or equal to the
2920  *      maximum number.
2921  */
2922 static void
2923 bridge_rttrim(struct bridge_softc *sc)
2924 {
2925         struct bridge_rtnode *brt, *nbrt;
2926
2927         MPASS(in_epoch(net_epoch_preempt));
2928         BRIDGE_RT_LOCK_ASSERT(sc);
2929
2930         /* Make sure we actually need to do this. */
2931         if (sc->sc_brtcnt <= sc->sc_brtmax)
2932                 return;
2933
2934         /* Force an aging cycle; this might trim enough addresses. */
2935         bridge_rtage(sc);
2936         if (sc->sc_brtcnt <= sc->sc_brtmax)
2937                 return;
2938
2939         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2940                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2941                         bridge_rtnode_destroy(sc, brt);
2942                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2943                                 return;
2944                 }
2945         }
2946 }
2947
2948 /*
2949  * bridge_timer:
2950  *
2951  *      Aging timer for the bridge.
2952  */
2953 static void
2954 bridge_timer(void *arg)
2955 {
2956         struct bridge_softc *sc = arg;
2957
2958         BRIDGE_RT_LOCK_ASSERT(sc);
2959
2960         /* Destruction of rtnodes requires a proper vnet context */
2961         CURVNET_SET(sc->sc_ifp->if_vnet);
2962         bridge_rtage(sc);
2963
2964         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2965                 callout_reset(&sc->sc_brcallout,
2966                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2967         CURVNET_RESTORE();
2968 }
2969
2970 /*
2971  * bridge_rtage:
2972  *
2973  *      Perform an aging cycle.
2974  */
2975 static void
2976 bridge_rtage(struct bridge_softc *sc)
2977 {
2978         struct bridge_rtnode *brt, *nbrt;
2979
2980         BRIDGE_RT_LOCK_ASSERT(sc);
2981
2982         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2983                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2984                         if (time_uptime >= brt->brt_expire)
2985                                 bridge_rtnode_destroy(sc, brt);
2986                 }
2987         }
2988 }
2989
2990 /*
2991  * bridge_rtflush:
2992  *
2993  *      Remove all dynamic addresses from the bridge.
2994  */
2995 static void
2996 bridge_rtflush(struct bridge_softc *sc, int full)
2997 {
2998         struct bridge_rtnode *brt, *nbrt;
2999
3000         BRIDGE_RT_LOCK_ASSERT(sc);
3001
3002         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3003                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3004                         bridge_rtnode_destroy(sc, brt);
3005         }
3006 }
3007
3008 /*
3009  * bridge_rtdaddr:
3010  *
3011  *      Remove an address from the table.
3012  */
3013 static int
3014 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3015 {
3016         struct bridge_rtnode *brt;
3017         int found = 0;
3018
3019         BRIDGE_RT_LOCK(sc);
3020
3021         /*
3022          * If vlan is zero then we want to delete for all vlans so the lookup
3023          * may return more than one.
3024          */
3025         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
3026                 bridge_rtnode_destroy(sc, brt);
3027                 found = 1;
3028         }
3029
3030         BRIDGE_RT_UNLOCK(sc);
3031
3032         return (found ? 0 : ENOENT);
3033 }
3034
3035 /*
3036  * bridge_rtdelete:
3037  *
3038  *      Delete routes to a speicifc member interface.
3039  */
3040 static void
3041 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
3042 {
3043         struct bridge_rtnode *brt, *nbrt;
3044
3045         BRIDGE_RT_LOCK_ASSERT(sc);
3046
3047         CK_LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
3048                 if (brt->brt_ifp == ifp && (full ||
3049                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
3050                         bridge_rtnode_destroy(sc, brt);
3051         }
3052 }
3053
3054 /*
3055  * bridge_rtable_init:
3056  *
3057  *      Initialize the route table for this bridge.
3058  */
3059 static void
3060 bridge_rtable_init(struct bridge_softc *sc)
3061 {
3062         int i;
3063
3064         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
3065             M_DEVBUF, M_WAITOK);
3066
3067         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
3068                 CK_LIST_INIT(&sc->sc_rthash[i]);
3069
3070         sc->sc_rthash_key = arc4random();
3071         CK_LIST_INIT(&sc->sc_rtlist);
3072 }
3073
3074 /*
3075  * bridge_rtable_fini:
3076  *
3077  *      Deconstruct the route table for this bridge.
3078  */
3079 static void
3080 bridge_rtable_fini(struct bridge_softc *sc)
3081 {
3082
3083         KASSERT(sc->sc_brtcnt == 0,
3084             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
3085         free(sc->sc_rthash, M_DEVBUF);
3086 }
3087
3088 /*
3089  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
3090  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
3091  */
3092 #define mix(a, b, c)                                                    \
3093 do {                                                                    \
3094         a -= b; a -= c; a ^= (c >> 13);                                 \
3095         b -= c; b -= a; b ^= (a << 8);                                  \
3096         c -= a; c -= b; c ^= (b >> 13);                                 \
3097         a -= b; a -= c; a ^= (c >> 12);                                 \
3098         b -= c; b -= a; b ^= (a << 16);                                 \
3099         c -= a; c -= b; c ^= (b >> 5);                                  \
3100         a -= b; a -= c; a ^= (c >> 3);                                  \
3101         b -= c; b -= a; b ^= (a << 10);                                 \
3102         c -= a; c -= b; c ^= (b >> 15);                                 \
3103 } while (/*CONSTCOND*/0)
3104
3105 static __inline uint32_t
3106 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
3107 {
3108         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
3109
3110         b += addr[5] << 8;
3111         b += addr[4];
3112         a += addr[3] << 24;
3113         a += addr[2] << 16;
3114         a += addr[1] << 8;
3115         a += addr[0];
3116
3117         mix(a, b, c);
3118
3119         return (c & BRIDGE_RTHASH_MASK);
3120 }
3121
3122 #undef mix
3123
3124 static int
3125 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
3126 {
3127         int i, d;
3128
3129         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
3130                 d = ((int)a[i]) - ((int)b[i]);
3131         }
3132
3133         return (d);
3134 }
3135
3136 /*
3137  * bridge_rtnode_lookup:
3138  *
3139  *      Look up a bridge route node for the specified destination. Compare the
3140  *      vlan id or if zero then just return the first match.
3141  */
3142 static struct bridge_rtnode *
3143 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
3144 {
3145         struct bridge_rtnode *brt;
3146         uint32_t hash;
3147         int dir;
3148
3149         BRIDGE_RT_LOCK_OR_NET_EPOCH_ASSERT(sc);
3150
3151         hash = bridge_rthash(sc, addr);
3152         CK_LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
3153                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
3154                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
3155                         return (brt);
3156                 if (dir > 0)
3157                         return (NULL);
3158         }
3159
3160         return (NULL);
3161 }
3162
3163 /*
3164  * bridge_rtnode_insert:
3165  *
3166  *      Insert the specified bridge node into the route table.  We
3167  *      assume the entry is not already in the table.
3168  */
3169 static int
3170 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
3171 {
3172         struct bridge_rtnode *lbrt;
3173         uint32_t hash;
3174         int dir;
3175
3176         BRIDGE_RT_LOCK_ASSERT(sc);
3177
3178         hash = bridge_rthash(sc, brt->brt_addr);
3179
3180         lbrt = CK_LIST_FIRST(&sc->sc_rthash[hash]);
3181         if (lbrt == NULL) {
3182                 CK_LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
3183                 goto out;
3184         }
3185
3186         do {
3187                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
3188                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
3189                         return (EEXIST);
3190                 if (dir > 0) {
3191                         CK_LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
3192                         goto out;
3193                 }
3194                 if (CK_LIST_NEXT(lbrt, brt_hash) == NULL) {
3195                         CK_LIST_INSERT_AFTER(lbrt, brt, brt_hash);
3196                         goto out;
3197                 }
3198                 lbrt = CK_LIST_NEXT(lbrt, brt_hash);
3199         } while (lbrt != NULL);
3200
3201 #ifdef DIAGNOSTIC
3202         panic("bridge_rtnode_insert: impossible");
3203 #endif
3204
3205 out:
3206         CK_LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
3207         sc->sc_brtcnt++;
3208
3209         return (0);
3210 }
3211
3212 static void
3213 bridge_rtnode_destroy_cb(struct epoch_context *ctx)
3214 {
3215         struct bridge_rtnode *brt;
3216
3217         brt = __containerof(ctx, struct bridge_rtnode, brt_epoch_ctx);
3218
3219         CURVNET_SET(brt->brt_vnet);
3220         uma_zfree(V_bridge_rtnode_zone, brt);
3221         CURVNET_RESTORE();
3222 }
3223
3224 /*
3225  * bridge_rtnode_destroy:
3226  *
3227  *      Destroy a bridge rtnode.
3228  */
3229 static void
3230 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3231 {
3232
3233         BRIDGE_RT_LOCK_ASSERT(sc);
3234
3235         CK_LIST_REMOVE(brt, brt_hash);
3236
3237         CK_LIST_REMOVE(brt, brt_list);
3238         sc->sc_brtcnt--;
3239         brt->brt_dst->bif_addrcnt--;
3240
3241         epoch_call(net_epoch_preempt, &brt->brt_epoch_ctx,
3242             bridge_rtnode_destroy_cb);
3243 }
3244
3245 /*
3246  * bridge_rtable_expire:
3247  *
3248  *      Set the expiry time for all routes on an interface.
3249  */
3250 static void
3251 bridge_rtable_expire(struct ifnet *ifp, int age)
3252 {
3253         struct bridge_softc *sc = ifp->if_bridge;
3254         struct bridge_rtnode *brt;
3255
3256         CURVNET_SET(ifp->if_vnet);
3257         BRIDGE_RT_LOCK(sc);
3258
3259         /*
3260          * If the age is zero then flush, otherwise set all the expiry times to
3261          * age for the interface
3262          */
3263         if (age == 0)
3264                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3265         else {
3266                 CK_LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3267                         /* Cap the expiry time to 'age' */
3268                         if (brt->brt_ifp == ifp &&
3269                             brt->brt_expire > time_uptime + age &&
3270                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3271                                 brt->brt_expire = time_uptime + age;
3272                 }
3273         }
3274         BRIDGE_RT_UNLOCK(sc);
3275         CURVNET_RESTORE();
3276 }
3277
3278 /*
3279  * bridge_state_change:
3280  *
3281  *      Callback from the bridgestp code when a port changes states.
3282  */
3283 static void
3284 bridge_state_change(struct ifnet *ifp, int state)
3285 {
3286         struct bridge_softc *sc = ifp->if_bridge;
3287         static const char *stpstates[] = {
3288                 "disabled",
3289                 "listening",
3290                 "learning",
3291                 "forwarding",
3292                 "blocking",
3293                 "discarding"
3294         };
3295
3296         CURVNET_SET(ifp->if_vnet);
3297         if (V_log_stp)
3298                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3299                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3300         CURVNET_RESTORE();
3301 }
3302
3303 /*
3304  * Send bridge packets through pfil if they are one of the types pfil can deal
3305  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3306  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3307  * that interface.
3308  */
3309 static int
3310 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3311 {
3312         int snap, error, i, hlen;
3313         struct ether_header *eh1, eh2;
3314         struct ip *ip;
3315         struct llc llc1;
3316         u_int16_t ether_type;
3317
3318         snap = 0;
3319         error = -1;     /* Default error if not error == 0 */
3320
3321 #if 0
3322         /* we may return with the IP fields swapped, ensure its not shared */
3323         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3324 #endif
3325
3326         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3327                 return (0); /* filtering is disabled */
3328
3329         i = min((*mp)->m_pkthdr.len, max_protohdr);
3330         if ((*mp)->m_len < i) {
3331             *mp = m_pullup(*mp, i);
3332             if (*mp == NULL) {
3333                 printf("%s: m_pullup failed\n", __func__);
3334                 return (-1);
3335             }
3336         }
3337
3338         eh1 = mtod(*mp, struct ether_header *);
3339         ether_type = ntohs(eh1->ether_type);
3340
3341         /*
3342          * Check for SNAP/LLC.
3343          */
3344         if (ether_type < ETHERMTU) {
3345                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3346
3347                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3348                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3349                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3350                     llc2->llc_control == LLC_UI) {
3351                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3352                         snap = 1;
3353                 }
3354         }
3355
3356         /*
3357          * If we're trying to filter bridge traffic, don't look at anything
3358          * other than IP and ARP traffic.  If the filter doesn't understand
3359          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3360          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3361          * but of course we don't have an AppleTalk filter to begin with.
3362          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3363          * ARP traffic.)
3364          */
3365         switch (ether_type) {
3366                 case ETHERTYPE_ARP:
3367                 case ETHERTYPE_REVARP:
3368                         if (V_pfil_ipfw_arp == 0)
3369                                 return (0); /* Automatically pass */
3370                         break;
3371
3372                 case ETHERTYPE_IP:
3373 #ifdef INET6
3374                 case ETHERTYPE_IPV6:
3375 #endif /* INET6 */
3376                         break;
3377                 default:
3378                         /*
3379                          * Check to see if the user wants to pass non-ip
3380                          * packets, these will not be checked by pfil(9) and
3381                          * passed unconditionally so the default is to drop.
3382                          */
3383                         if (V_pfil_onlyip)
3384                                 goto bad;
3385         }
3386
3387         /* Run the packet through pfil before stripping link headers */
3388         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3389                         dir == PFIL_OUT && ifp != NULL) {
3390
3391                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, 0,
3392                     NULL);
3393
3394                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3395                         return (error);
3396         }
3397
3398         /* Strip off the Ethernet header and keep a copy. */
3399         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3400         m_adj(*mp, ETHER_HDR_LEN);
3401
3402         /* Strip off snap header, if present */
3403         if (snap) {
3404                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3405                 m_adj(*mp, sizeof(struct llc));
3406         }
3407
3408         /*
3409          * Check the IP header for alignment and errors
3410          */
3411         if (dir == PFIL_IN) {
3412                 switch (ether_type) {
3413                         case ETHERTYPE_IP:
3414                                 error = bridge_ip_checkbasic(mp);
3415                                 break;
3416 #ifdef INET6
3417                         case ETHERTYPE_IPV6:
3418                                 error = bridge_ip6_checkbasic(mp);
3419                                 break;
3420 #endif /* INET6 */
3421                         default:
3422                                 error = 0;
3423                 }
3424                 if (error)
3425                         goto bad;
3426         }
3427
3428         error = 0;
3429
3430         /*
3431          * Run the packet through pfil
3432          */
3433         switch (ether_type) {
3434         case ETHERTYPE_IP:
3435                 /*
3436                  * Run pfil on the member interface and the bridge, both can
3437                  * be skipped by clearing pfil_member or pfil_bridge.
3438                  *
3439                  * Keep the order:
3440                  *   in_if -> bridge_if -> out_if
3441                  */
3442                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3443                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3444                                         dir, 0, NULL);
3445
3446                 if (*mp == NULL || error != 0) /* filter may consume */
3447                         break;
3448
3449                 if (V_pfil_member && ifp != NULL)
3450                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3451                                         dir, 0, NULL);
3452
3453                 if (*mp == NULL || error != 0) /* filter may consume */
3454                         break;
3455
3456                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3457                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3458                                         dir, 0, NULL);
3459
3460                 if (*mp == NULL || error != 0) /* filter may consume */
3461                         break;
3462
3463                 /* check if we need to fragment the packet */
3464                 /* bridge_fragment generates a mbuf chain of packets */
3465                 /* that already include eth headers */
3466                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3467                         i = (*mp)->m_pkthdr.len;
3468                         if (i > ifp->if_mtu) {
3469                                 error = bridge_fragment(ifp, mp, &eh2, snap,
3470                                             &llc1);
3471                                 return (error);
3472                         }
3473                 }
3474
3475                 /* Recalculate the ip checksum. */
3476                 ip = mtod(*mp, struct ip *);
3477                 hlen = ip->ip_hl << 2;
3478                 if (hlen < sizeof(struct ip))
3479                         goto bad;
3480                 if (hlen > (*mp)->m_len) {
3481                         if ((*mp = m_pullup(*mp, hlen)) == NULL)
3482                                 goto bad;
3483                         ip = mtod(*mp, struct ip *);
3484                         if (ip == NULL)
3485                                 goto bad;
3486                 }
3487                 ip->ip_sum = 0;
3488                 if (hlen == sizeof(struct ip))
3489                         ip->ip_sum = in_cksum_hdr(ip);
3490                 else
3491                         ip->ip_sum = in_cksum(*mp, hlen);
3492
3493                 break;
3494 #ifdef INET6
3495         case ETHERTYPE_IPV6:
3496                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3497                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3498                                         dir, 0, NULL);
3499
3500                 if (*mp == NULL || error != 0) /* filter may consume */
3501                         break;
3502
3503                 if (V_pfil_member && ifp != NULL)
3504                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3505                                         dir, 0, NULL);
3506
3507                 if (*mp == NULL || error != 0) /* filter may consume */
3508                         break;
3509
3510                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3511                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3512                                         dir, 0, NULL);
3513                 break;
3514 #endif
3515         default:
3516                 error = 0;
3517                 break;
3518         }
3519
3520         if (*mp == NULL)
3521                 return (error);
3522         if (error != 0)
3523                 goto bad;
3524
3525         error = -1;
3526
3527         /*
3528          * Finally, put everything back the way it was and return
3529          */
3530         if (snap) {
3531                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3532                 if (*mp == NULL)
3533                         return (error);
3534                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3535         }
3536
3537         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3538         if (*mp == NULL)
3539                 return (error);
3540         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3541
3542         return (0);
3543
3544 bad:
3545         m_freem(*mp);
3546         *mp = NULL;
3547         return (error);
3548 }
3549
3550 /*
3551  * Perform basic checks on header size since
3552  * pfil assumes ip_input has already processed
3553  * it for it.  Cut-and-pasted from ip_input.c.
3554  * Given how simple the IPv6 version is,
3555  * does the IPv4 version really need to be
3556  * this complicated?
3557  *
3558  * XXX Should we update ipstat here, or not?
3559  * XXX Right now we update ipstat but not
3560  * XXX csum_counter.
3561  */
3562 static int
3563 bridge_ip_checkbasic(struct mbuf **mp)
3564 {
3565         struct mbuf *m = *mp;
3566         struct ip *ip;
3567         int len, hlen;
3568         u_short sum;
3569
3570         if (*mp == NULL)
3571                 return (-1);
3572
3573         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3574                 if ((m = m_copyup(m, sizeof(struct ip),
3575                         (max_linkhdr + 3) & ~3)) == NULL) {
3576                         /* XXXJRT new stat, please */
3577                         KMOD_IPSTAT_INC(ips_toosmall);
3578                         goto bad;
3579                 }
3580         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3581                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3582                         KMOD_IPSTAT_INC(ips_toosmall);
3583                         goto bad;
3584                 }
3585         }
3586         ip = mtod(m, struct ip *);
3587         if (ip == NULL) goto bad;
3588
3589         if (ip->ip_v != IPVERSION) {
3590                 KMOD_IPSTAT_INC(ips_badvers);
3591                 goto bad;
3592         }
3593         hlen = ip->ip_hl << 2;
3594         if (hlen < sizeof(struct ip)) { /* minimum header length */
3595                 KMOD_IPSTAT_INC(ips_badhlen);
3596                 goto bad;
3597         }
3598         if (hlen > m->m_len) {
3599                 if ((m = m_pullup(m, hlen)) == NULL) {
3600                         KMOD_IPSTAT_INC(ips_badhlen);
3601                         goto bad;
3602                 }
3603                 ip = mtod(m, struct ip *);
3604                 if (ip == NULL) goto bad;
3605         }
3606
3607         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3608                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3609         } else {
3610                 if (hlen == sizeof(struct ip)) {
3611                         sum = in_cksum_hdr(ip);
3612                 } else {
3613                         sum = in_cksum(m, hlen);
3614                 }
3615         }
3616         if (sum) {
3617                 KMOD_IPSTAT_INC(ips_badsum);
3618                 goto bad;
3619         }
3620
3621         /* Retrieve the packet length. */
3622         len = ntohs(ip->ip_len);
3623
3624         /*
3625          * Check for additional length bogosity
3626          */
3627         if (len < hlen) {
3628                 KMOD_IPSTAT_INC(ips_badlen);
3629                 goto bad;
3630         }
3631
3632         /*
3633          * Check that the amount of data in the buffers
3634          * is as at least much as the IP header would have us expect.
3635          * Drop packet if shorter than we expect.
3636          */
3637         if (m->m_pkthdr.len < len) {
3638                 KMOD_IPSTAT_INC(ips_tooshort);
3639                 goto bad;
3640         }
3641
3642         /* Checks out, proceed */
3643         *mp = m;
3644         return (0);
3645
3646 bad:
3647         *mp = m;
3648         return (-1);
3649 }
3650
3651 #ifdef INET6
3652 /*
3653  * Same as above, but for IPv6.
3654  * Cut-and-pasted from ip6_input.c.
3655  * XXX Should we update ip6stat, or not?
3656  */
3657 static int
3658 bridge_ip6_checkbasic(struct mbuf **mp)
3659 {
3660         struct mbuf *m = *mp;
3661         struct ip6_hdr *ip6;
3662
3663         /*
3664          * If the IPv6 header is not aligned, slurp it up into a new
3665          * mbuf with space for link headers, in the event we forward
3666          * it.  Otherwise, if it is aligned, make sure the entire base
3667          * IPv6 header is in the first mbuf of the chain.
3668          */
3669         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3670                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3671                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3672                             (max_linkhdr + 3) & ~3)) == NULL) {
3673                         /* XXXJRT new stat, please */
3674                         IP6STAT_INC(ip6s_toosmall);
3675                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3676                         goto bad;
3677                 }
3678         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3679                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3680                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3681                         IP6STAT_INC(ip6s_toosmall);
3682                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3683                         goto bad;
3684                 }
3685         }
3686
3687         ip6 = mtod(m, struct ip6_hdr *);
3688
3689         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3690                 IP6STAT_INC(ip6s_badvers);
3691                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3692                 goto bad;
3693         }
3694
3695         /* Checks out, proceed */
3696         *mp = m;
3697         return (0);
3698
3699 bad:
3700         *mp = m;
3701         return (-1);
3702 }
3703 #endif /* INET6 */
3704
3705 /*
3706  * bridge_fragment:
3707  *
3708  *      Fragment mbuf chain in multiple packets and prepend ethernet header.
3709  */
3710 static int
3711 bridge_fragment(struct ifnet *ifp, struct mbuf **mp, struct ether_header *eh,
3712     int snap, struct llc *llc)
3713 {
3714         struct mbuf *m = *mp, *nextpkt = NULL, *mprev = NULL, *mcur = NULL;
3715         struct ip *ip;
3716         int error = -1;
3717
3718         if (m->m_len < sizeof(struct ip) &&
3719             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3720                 goto dropit;
3721         ip = mtod(m, struct ip *);
3722
3723         m->m_pkthdr.csum_flags |= CSUM_IP;
3724         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3725         if (error)
3726                 goto dropit;
3727
3728         /*
3729          * Walk the chain and re-add the Ethernet header for
3730          * each mbuf packet.
3731          */
3732         for (mcur = m; mcur; mcur = mcur->m_nextpkt) {
3733                 nextpkt = mcur->m_nextpkt;
3734                 mcur->m_nextpkt = NULL;
3735                 if (snap) {
3736                         M_PREPEND(mcur, sizeof(struct llc), M_NOWAIT);
3737                         if (mcur == NULL) {
3738                                 error = ENOBUFS;
3739                                 if (mprev != NULL)
3740                                         mprev->m_nextpkt = nextpkt;
3741                                 goto dropit;
3742                         }
3743                         bcopy(llc, mtod(mcur, caddr_t),sizeof(struct llc));
3744                 }
3745
3746                 M_PREPEND(mcur, ETHER_HDR_LEN, M_NOWAIT);
3747                 if (mcur == NULL) {
3748                         error = ENOBUFS;
3749                         if (mprev != NULL)
3750                                 mprev->m_nextpkt = nextpkt;
3751                         goto dropit;
3752                 }
3753                 bcopy(eh, mtod(mcur, caddr_t), ETHER_HDR_LEN);
3754
3755                 /*
3756                  * The previous two M_PREPEND could have inserted one or two
3757                  * mbufs in front so we have to update the previous packet's
3758                  * m_nextpkt.
3759                  */
3760                 mcur->m_nextpkt = nextpkt;
3761                 if (mprev != NULL)
3762                         mprev->m_nextpkt = mcur;
3763                 else {
3764                         /* The first mbuf in the original chain needs to be
3765                          * updated. */
3766                         *mp = mcur;
3767                 }
3768                 mprev = mcur;
3769         }
3770
3771         KMOD_IPSTAT_INC(ips_fragmented);
3772         return (error);
3773
3774 dropit:
3775         for (mcur = *mp; mcur; mcur = m) { /* droping the full packet chain */
3776                 m = mcur->m_nextpkt;
3777                 m_freem(mcur);
3778         }
3779         return (error);
3780 }
3781
3782 static void
3783 bridge_linkstate(struct ifnet *ifp)
3784 {
3785         struct bridge_softc *sc = ifp->if_bridge;
3786         struct bridge_iflist *bif;
3787         struct epoch_tracker et;
3788
3789         NET_EPOCH_ENTER_ET(et);
3790
3791         bif = bridge_lookup_member_if(sc, ifp);
3792         if (bif == NULL) {
3793                 NET_EPOCH_EXIT_ET(et);
3794                 return;
3795         }
3796         bridge_linkcheck(sc);
3797
3798         bstp_linkstate(&bif->bif_stp);
3799
3800         NET_EPOCH_EXIT_ET(et);
3801 }
3802
3803 static void
3804 bridge_linkcheck(struct bridge_softc *sc)
3805 {
3806         struct bridge_iflist *bif;
3807         int new_link, hasls;
3808
3809         BRIDGE_LOCK_OR_NET_EPOCH_ASSERT(sc);
3810
3811         new_link = LINK_STATE_DOWN;
3812         hasls = 0;
3813         /* Our link is considered up if at least one of our ports is active */
3814         CK_LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3815                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3816                         hasls++;
3817                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3818                         new_link = LINK_STATE_UP;
3819                         break;
3820                 }
3821         }
3822         if (!CK_LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3823                 /* If no interfaces support link-state then we default to up */
3824                 new_link = LINK_STATE_UP;
3825         }
3826         if_link_state_change(sc->sc_ifp, new_link);
3827 }