sys/net/if_bridge.c

   1 /*      $NetBSD: if_bridge.c,v 1.31 2005/06/01 19:45:34 jdc Exp $       */
   2
   3 /*
   4  * Copyright 2001 Wasabi Systems, Inc.
   5  * All rights reserved.
   6  *
   7  * Written by Jason R. Thorpe for Wasabi Systems, Inc.
   8  *
   9  * Redistribution and use in source and binary forms, with or without
  10  * modification, are permitted provided that the following conditions
  11  * are met:
  12  * 1. Redistributions of source code must retain the above copyright
  13  *    notice, this list of conditions and the following disclaimer.
  14  * 2. Redistributions in binary form must reproduce the above copyright
  15  *    notice, this list of conditions and the following disclaimer in the
  16  *    documentation and/or other materials provided with the distribution.
  17  * 3. All advertising materials mentioning features or use of this software
  18  *    must display the following acknowledgement:
  19  *      This product includes software developed for the NetBSD Project by
  20  *      Wasabi Systems, Inc.
  21  * 4. The name of Wasabi Systems, Inc. may not be used to endorse
  22  *    or promote products derived from this software without specific prior
  23  *    written permission.
  24  *
  25  * THIS SOFTWARE IS PROVIDED BY WASABI SYSTEMS, INC. ``AS IS'' AND
  26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  27  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  28  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL WASABI SYSTEMS, INC
  29  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  30  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  31  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  32  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  33  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  34  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  35  * POSSIBILITY OF SUCH DAMAGE.
  36  */
  37
  38 /*
  39  * Copyright (c) 1999, 2000 Jason L. Wright (jason@thought.net)
  40  * All rights reserved.
  41  *
  42  * Redistribution and use in source and binary forms, with or without
  43  * modification, are permitted provided that the following conditions
  44  * are met:
  45  * 1. Redistributions of source code must retain the above copyright
  46  *    notice, this list of conditions and the following disclaimer.
  47  * 2. Redistributions in binary form must reproduce the above copyright
  48  *    notice, this list of conditions and the following disclaimer in the
  49  *    documentation and/or other materials provided with the distribution.
  50  *
  51  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
  52  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
  53  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
  54  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
  55  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
  56  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
  57  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  58  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  59  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
  60  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  61  * POSSIBILITY OF SUCH DAMAGE.
  62  *
  63  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
  64  */
  65
  66 /*
  67  * Network interface bridge support.
  68  *
  69  * TODO:
  70  *
  71  *      - Currently only supports Ethernet-like interfaces (Ethernet,
  72  *        802.11, VLANs on Ethernet, etc.)  Figure out a nice way
  73  *        to bridge other types of interfaces (FDDI-FDDI, and maybe
  74  *        consider heterogenous bridges).
  75  */
  76
  77 #include <sys/cdefs.h>
  78 __FBSDID("$FreeBSD$");
  79
  80 #include "opt_inet.h"
  81 #include "opt_inet6.h"
  82
  83 #include <sys/param.h>
  84 #include <sys/eventhandler.h>
  85 #include <sys/mbuf.h>
  86 #include <sys/malloc.h>
  87 #include <sys/protosw.h>
  88 #include <sys/systm.h>
  89 #include <sys/jail.h>
  90 #include <sys/time.h>
  91 #include <sys/socket.h> /* for net/if.h */
  92 #include <sys/sockio.h>
  93 #include <sys/ctype.h>  /* string functions */
  94 #include <sys/kernel.h>
  95 #include <sys/random.h>
  96 #include <sys/syslog.h>
  97 #include <sys/sysctl.h>
  98 #include <vm/uma.h>
  99 #include <sys/module.h>
 100 #include <sys/priv.h>
 101 #include <sys/proc.h>
 102 #include <sys/lock.h>
 103 #include <sys/mutex.h>
 104
 105 #include <net/bpf.h>
 106 #include <net/if.h>
 107 #include <net/if_clone.h>
 108 #include <net/if_dl.h>
 109 #include <net/if_types.h>
 110 #include <net/if_var.h>
 111 #include <net/pfil.h>
 112 #include <net/vnet.h>
 113
 114 #include <netinet/in.h>
 115 #include <netinet/in_systm.h>
 116 #include <netinet/in_var.h>
 117 #include <netinet/ip.h>
 118 #include <netinet/ip_var.h>
 119 #ifdef INET6
 120 #include <netinet/ip6.h>
 121 #include <netinet6/ip6_var.h>
 122 #include <netinet6/in6_ifattach.h>
 123 #endif
 124 #if defined(INET) || defined(INET6)
 125 #include <netinet/ip_carp.h>
 126 #endif
 127 #include <machine/in_cksum.h>
 128 #include <netinet/if_ether.h>
 129 #include <net/bridgestp.h>
 130 #include <net/if_bridgevar.h>
 131 #include <net/if_llc.h>
 132 #include <net/if_vlan_var.h>
 133
 134 #include <net/route.h>
 135
 136 /*
 137  * Size of the route hash table.  Must be a power of two.
 138  */
 139 #ifndef BRIDGE_RTHASH_SIZE
 140 #define BRIDGE_RTHASH_SIZE              1024
 141 #endif
 142
 143 #define BRIDGE_RTHASH_MASK              (BRIDGE_RTHASH_SIZE - 1)
 144
 145 /*
 146  * Default maximum number of addresses to cache.
 147  */
 148 #ifndef BRIDGE_RTABLE_MAX
 149 #define BRIDGE_RTABLE_MAX               2000
 150 #endif
 151
 152 /*
 153  * Timeout (in seconds) for entries learned dynamically.
 154  */
 155 #ifndef BRIDGE_RTABLE_TIMEOUT
 156 #define BRIDGE_RTABLE_TIMEOUT           (20 * 60)       /* same as ARP */
 157 #endif
 158
 159 /*
 160  * Number of seconds between walks of the route list.
 161  */
 162 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
 163 #define BRIDGE_RTABLE_PRUNE_PERIOD      (5 * 60)
 164 #endif
 165
 166 /*
 167  * List of capabilities to possibly mask on the member interface.
 168  */
 169 #define BRIDGE_IFCAPS_MASK              (IFCAP_TOE|IFCAP_TSO|IFCAP_TXCSUM)
 170
 171 /*
 172  * List of capabilities to strip
 173  */
 174 #define BRIDGE_IFCAPS_STRIP             IFCAP_LRO
 175
 176 /*
 177  * Bridge interface list entry.
 178  */
 179 struct bridge_iflist {
 180         LIST_ENTRY(bridge_iflist) bif_next;
 181         struct ifnet            *bif_ifp;       /* member if */
 182         struct bstp_port        bif_stp;        /* STP state */
 183         uint32_t                bif_flags;      /* member if flags */
 184         int                     bif_savedcaps;  /* saved capabilities */
 185         uint32_t                bif_addrmax;    /* max # of addresses */
 186         uint32_t                bif_addrcnt;    /* cur. # of addresses */
 187         uint32_t                bif_addrexceeded;/* # of address violations */
 188 };
 189
 190 /*
 191  * Bridge route node.
 192  */
 193 struct bridge_rtnode {
 194         LIST_ENTRY(bridge_rtnode) brt_hash;     /* hash table linkage */
 195         LIST_ENTRY(bridge_rtnode) brt_list;     /* list linkage */
 196         struct bridge_iflist    *brt_dst;       /* destination if */
 197         unsigned long           brt_expire;     /* expiration time */
 198         uint8_t                 brt_flags;      /* address flags */
 199         uint8_t                 brt_addr[ETHER_ADDR_LEN];
 200         uint16_t                brt_vlan;       /* vlan id */
 201 };
 202 #define brt_ifp                 brt_dst->bif_ifp
 203
 204 /*
 205  * Software state for each bridge.
 206  */
 207 struct bridge_softc {
 208         struct ifnet            *sc_ifp;        /* make this an interface */
 209         LIST_ENTRY(bridge_softc) sc_list;
 210         struct mtx              sc_mtx;
 211         struct cv               sc_cv;
 212         uint32_t                sc_brtmax;      /* max # of addresses */
 213         uint32_t                sc_brtcnt;      /* cur. # of addresses */
 214         uint32_t                sc_brttimeout;  /* rt timeout in seconds */
 215         struct callout          sc_brcallout;   /* bridge callout */
 216         uint32_t                sc_iflist_ref;  /* refcount for sc_iflist */
 217         uint32_t                sc_iflist_xcnt; /* refcount for sc_iflist */
 218         LIST_HEAD(, bridge_iflist) sc_iflist;   /* member interface list */
 219         LIST_HEAD(, bridge_rtnode) *sc_rthash;  /* our forwarding table */
 220         LIST_HEAD(, bridge_rtnode) sc_rtlist;   /* list version of above */
 221         uint32_t                sc_rthash_key;  /* key for hash */
 222         LIST_HEAD(, bridge_iflist) sc_spanlist; /* span ports list */
 223         struct bstp_state       sc_stp;         /* STP state */
 224         uint32_t                sc_brtexceeded; /* # of cache drops */
 225         struct ifnet            *sc_ifaddr;     /* member mac copied from */
 226         u_char                  sc_defaddr[6];  /* Default MAC address */
 227 };
 228
 229 static VNET_DEFINE(struct mtx, bridge_list_mtx);
 230 #define V_bridge_list_mtx       VNET(bridge_list_mtx)
 231 static eventhandler_tag bridge_detach_cookie;
 232
 233 int     bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
 234
 235 uma_zone_t bridge_rtnode_zone;
 236
 237 static int      bridge_clone_create(struct if_clone *, int, caddr_t);
 238 static void     bridge_clone_destroy(struct ifnet *);
 239
 240 static int      bridge_ioctl(struct ifnet *, u_long, caddr_t);
 241 static void     bridge_mutecaps(struct bridge_softc *);
 242 static void     bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
 243                     int);
 244 static void     bridge_ifdetach(void *arg __unused, struct ifnet *);
 245 static void     bridge_init(void *);
 246 static void     bridge_dummynet(struct mbuf *, struct ifnet *);
 247 static void     bridge_stop(struct ifnet *, int);
 248 static int      bridge_transmit(struct ifnet *, struct mbuf *);
 249 static void     bridge_qflush(struct ifnet *);
 250 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
 251 static int      bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
 252                     struct rtentry *);
 253 static int      bridge_enqueue(struct bridge_softc *, struct ifnet *,
 254                     struct mbuf *);
 255 static void     bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
 256
 257 static void     bridge_forward(struct bridge_softc *, struct bridge_iflist *,
 258                     struct mbuf *m);
 259
 260 static void     bridge_timer(void *);
 261
 262 static void     bridge_broadcast(struct bridge_softc *, struct ifnet *,
 263                     struct mbuf *, int);
 264 static void     bridge_span(struct bridge_softc *, struct mbuf *);
 265
 266 static int      bridge_rtupdate(struct bridge_softc *, const uint8_t *,
 267                     uint16_t, struct bridge_iflist *, int, uint8_t);
 268 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
 269                     uint16_t);
 270 static void     bridge_rttrim(struct bridge_softc *);
 271 static void     bridge_rtage(struct bridge_softc *);
 272 static void     bridge_rtflush(struct bridge_softc *, int);
 273 static int      bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
 274                     uint16_t);
 275
 276 static void     bridge_rtable_init(struct bridge_softc *);
 277 static void     bridge_rtable_fini(struct bridge_softc *);
 278
 279 static int      bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
 280 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
 281                     const uint8_t *, uint16_t);
 282 static int      bridge_rtnode_insert(struct bridge_softc *,
 283                     struct bridge_rtnode *);
 284 static void     bridge_rtnode_destroy(struct bridge_softc *,
 285                     struct bridge_rtnode *);
 286 static void     bridge_rtable_expire(struct ifnet *, int);
 287 static void     bridge_state_change(struct ifnet *, int);
 288
 289 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
 290                     const char *name);
 291 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
 292                     struct ifnet *ifp);
 293 static void     bridge_delete_member(struct bridge_softc *,
 294                     struct bridge_iflist *, int);
 295 static void     bridge_delete_span(struct bridge_softc *,
 296                     struct bridge_iflist *);
 297
 298 static int      bridge_ioctl_add(struct bridge_softc *, void *);
 299 static int      bridge_ioctl_del(struct bridge_softc *, void *);
 300 static int      bridge_ioctl_gifflags(struct bridge_softc *, void *);
 301 static int      bridge_ioctl_sifflags(struct bridge_softc *, void *);
 302 static int      bridge_ioctl_scache(struct bridge_softc *, void *);
 303 static int      bridge_ioctl_gcache(struct bridge_softc *, void *);
 304 static int      bridge_ioctl_gifs(struct bridge_softc *, void *);
 305 static int      bridge_ioctl_rts(struct bridge_softc *, void *);
 306 static int      bridge_ioctl_saddr(struct bridge_softc *, void *);
 307 static int      bridge_ioctl_sto(struct bridge_softc *, void *);
 308 static int      bridge_ioctl_gto(struct bridge_softc *, void *);
 309 static int      bridge_ioctl_daddr(struct bridge_softc *, void *);
 310 static int      bridge_ioctl_flush(struct bridge_softc *, void *);
 311 static int      bridge_ioctl_gpri(struct bridge_softc *, void *);
 312 static int      bridge_ioctl_spri(struct bridge_softc *, void *);
 313 static int      bridge_ioctl_ght(struct bridge_softc *, void *);
 314 static int      bridge_ioctl_sht(struct bridge_softc *, void *);
 315 static int      bridge_ioctl_gfd(struct bridge_softc *, void *);
 316 static int      bridge_ioctl_sfd(struct bridge_softc *, void *);
 317 static int      bridge_ioctl_gma(struct bridge_softc *, void *);
 318 static int      bridge_ioctl_sma(struct bridge_softc *, void *);
 319 static int      bridge_ioctl_sifprio(struct bridge_softc *, void *);
 320 static int      bridge_ioctl_sifcost(struct bridge_softc *, void *);
 321 static int      bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
 322 static int      bridge_ioctl_addspan(struct bridge_softc *, void *);
 323 static int      bridge_ioctl_delspan(struct bridge_softc *, void *);
 324 static int      bridge_ioctl_gbparam(struct bridge_softc *, void *);
 325 static int      bridge_ioctl_grte(struct bridge_softc *, void *);
 326 static int      bridge_ioctl_gifsstp(struct bridge_softc *, void *);
 327 static int      bridge_ioctl_sproto(struct bridge_softc *, void *);
 328 static int      bridge_ioctl_stxhc(struct bridge_softc *, void *);
 329 static int      bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
 330                     int);
 331 static int      bridge_ip_checkbasic(struct mbuf **mp);
 332 #ifdef INET6
 333 static int      bridge_ip6_checkbasic(struct mbuf **mp);
 334 #endif /* INET6 */
 335 static int      bridge_fragment(struct ifnet *, struct mbuf *,
 336                     struct ether_header *, int, struct llc *);
 337 static void     bridge_linkstate(struct ifnet *ifp);
 338 static void     bridge_linkcheck(struct bridge_softc *sc);
 339
 340 extern void (*bridge_linkstate_p)(struct ifnet *ifp);
 341
 342 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
 343 #define VLANTAGOF(_m)   \
 344     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
 345
 346 static struct bstp_cb_ops bridge_ops = {
 347         .bcb_state = bridge_state_change,
 348         .bcb_rtage = bridge_rtable_expire
 349 };
 350
 351 SYSCTL_DECL(_net_link);
 352 static SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
 353
 354 /* only pass IP[46] packets when pfil is enabled */
 355 static VNET_DEFINE(int, pfil_onlyip) = 1;
 356 #define V_pfil_onlyip   VNET(pfil_onlyip)
 357 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip,
 358     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_onlyip), 0,
 359     "Only pass IP packets when pfil is enabled");
 360
 361 /* run pfil hooks on the bridge interface */
 362 static VNET_DEFINE(int, pfil_bridge) = 1;
 363 #define V_pfil_bridge   VNET(pfil_bridge)
 364 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge,
 365     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_bridge), 0,
 366     "Packet filter on the bridge interface");
 367
 368 /* layer2 filter with ipfw */
 369 static VNET_DEFINE(int, pfil_ipfw);
 370 #define V_pfil_ipfw     VNET(pfil_ipfw)
 371
 372 /* layer2 ARP filter with ipfw */
 373 static VNET_DEFINE(int, pfil_ipfw_arp);
 374 #define V_pfil_ipfw_arp VNET(pfil_ipfw_arp)
 375 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp,
 376     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_ipfw_arp), 0,
 377     "Filter ARP packets through IPFW layer2");
 378
 379 /* run pfil hooks on the member interface */
 380 static VNET_DEFINE(int, pfil_member) = 1;
 381 #define V_pfil_member   VNET(pfil_member)
 382 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member,
 383     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_member), 0,
 384     "Packet filter on the member interface");
 385
 386 /* run pfil hooks on the physical interface for locally destined packets */
 387 static VNET_DEFINE(int, pfil_local_phys);
 388 #define V_pfil_local_phys       VNET(pfil_local_phys)
 389 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys,
 390     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(pfil_local_phys), 0,
 391     "Packet filter on the physical interface for locally destined packets");
 392
 393 /* log STP state changes */
 394 static VNET_DEFINE(int, log_stp);
 395 #define V_log_stp       VNET(log_stp)
 396 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp,
 397     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(log_stp), 0,
 398     "Log STP state changes");
 399
 400 /* share MAC with first bridge member */
 401 static VNET_DEFINE(int, bridge_inherit_mac);
 402 #define V_bridge_inherit_mac    VNET(bridge_inherit_mac)
 403 SYSCTL_INT(_net_link_bridge, OID_AUTO, inherit_mac,
 404     CTLFLAG_RWTUN | CTLFLAG_VNET, &VNET_NAME(bridge_inherit_mac), 0,
 405     "Inherit MAC address from the first bridge member");
 406
 407 static VNET_DEFINE(int, allow_llz_overlap) = 0;
 408 #define V_allow_llz_overlap     VNET(allow_llz_overlap)
 409 SYSCTL_INT(_net_link_bridge, OID_AUTO, allow_llz_overlap,
 410     CTLFLAG_VNET | CTLFLAG_RW | CTLFLAG_VNET, &VNET_NAME(allow_llz_overlap), 0,
 411     "Allow overlap of link-local scope "
 412     "zones of a bridge interface and the member interfaces");
 413
 414 struct bridge_control {
 415         int     (*bc_func)(struct bridge_softc *, void *);
 416         int     bc_argsize;
 417         int     bc_flags;
 418 };
 419
 420 #define BC_F_COPYIN             0x01    /* copy arguments in */
 421 #define BC_F_COPYOUT            0x02    /* copy arguments out */
 422 #define BC_F_SUSER              0x04    /* do super-user check */
 423
 424 const struct bridge_control bridge_control_table[] = {
 425         { bridge_ioctl_add,             sizeof(struct ifbreq),
 426           BC_F_COPYIN|BC_F_SUSER },
 427         { bridge_ioctl_del,             sizeof(struct ifbreq),
 428           BC_F_COPYIN|BC_F_SUSER },
 429
 430         { bridge_ioctl_gifflags,        sizeof(struct ifbreq),
 431           BC_F_COPYIN|BC_F_COPYOUT },
 432         { bridge_ioctl_sifflags,        sizeof(struct ifbreq),
 433           BC_F_COPYIN|BC_F_SUSER },
 434
 435         { bridge_ioctl_scache,          sizeof(struct ifbrparam),
 436           BC_F_COPYIN|BC_F_SUSER },
 437         { bridge_ioctl_gcache,          sizeof(struct ifbrparam),
 438           BC_F_COPYOUT },
 439
 440         { bridge_ioctl_gifs,            sizeof(struct ifbifconf),
 441           BC_F_COPYIN|BC_F_COPYOUT },
 442         { bridge_ioctl_rts,             sizeof(struct ifbaconf),
 443           BC_F_COPYIN|BC_F_COPYOUT },
 444
 445         { bridge_ioctl_saddr,           sizeof(struct ifbareq),
 446           BC_F_COPYIN|BC_F_SUSER },
 447
 448         { bridge_ioctl_sto,             sizeof(struct ifbrparam),
 449           BC_F_COPYIN|BC_F_SUSER },
 450         { bridge_ioctl_gto,             sizeof(struct ifbrparam),
 451           BC_F_COPYOUT },
 452
 453         { bridge_ioctl_daddr,           sizeof(struct ifbareq),
 454           BC_F_COPYIN|BC_F_SUSER },
 455
 456         { bridge_ioctl_flush,           sizeof(struct ifbreq),
 457           BC_F_COPYIN|BC_F_SUSER },
 458
 459         { bridge_ioctl_gpri,            sizeof(struct ifbrparam),
 460           BC_F_COPYOUT },
 461         { bridge_ioctl_spri,            sizeof(struct ifbrparam),
 462           BC_F_COPYIN|BC_F_SUSER },
 463
 464         { bridge_ioctl_ght,             sizeof(struct ifbrparam),
 465           BC_F_COPYOUT },
 466         { bridge_ioctl_sht,             sizeof(struct ifbrparam),
 467           BC_F_COPYIN|BC_F_SUSER },
 468
 469         { bridge_ioctl_gfd,             sizeof(struct ifbrparam),
 470           BC_F_COPYOUT },
 471         { bridge_ioctl_sfd,             sizeof(struct ifbrparam),
 472           BC_F_COPYIN|BC_F_SUSER },
 473
 474         { bridge_ioctl_gma,             sizeof(struct ifbrparam),
 475           BC_F_COPYOUT },
 476         { bridge_ioctl_sma,             sizeof(struct ifbrparam),
 477           BC_F_COPYIN|BC_F_SUSER },
 478
 479         { bridge_ioctl_sifprio,         sizeof(struct ifbreq),
 480           BC_F_COPYIN|BC_F_SUSER },
 481
 482         { bridge_ioctl_sifcost,         sizeof(struct ifbreq),
 483           BC_F_COPYIN|BC_F_SUSER },
 484
 485         { bridge_ioctl_addspan,         sizeof(struct ifbreq),
 486           BC_F_COPYIN|BC_F_SUSER },
 487         { bridge_ioctl_delspan,         sizeof(struct ifbreq),
 488           BC_F_COPYIN|BC_F_SUSER },
 489
 490         { bridge_ioctl_gbparam,         sizeof(struct ifbropreq),
 491           BC_F_COPYOUT },
 492
 493         { bridge_ioctl_grte,            sizeof(struct ifbrparam),
 494           BC_F_COPYOUT },
 495
 496         { bridge_ioctl_gifsstp,         sizeof(struct ifbpstpconf),
 497           BC_F_COPYIN|BC_F_COPYOUT },
 498
 499         { bridge_ioctl_sproto,          sizeof(struct ifbrparam),
 500           BC_F_COPYIN|BC_F_SUSER },
 501
 502         { bridge_ioctl_stxhc,           sizeof(struct ifbrparam),
 503           BC_F_COPYIN|BC_F_SUSER },
 504
 505         { bridge_ioctl_sifmaxaddr,      sizeof(struct ifbreq),
 506           BC_F_COPYIN|BC_F_SUSER },
 507
 508 };
 509 const int bridge_control_table_size = nitems(bridge_control_table);
 510
 511 static VNET_DEFINE(LIST_HEAD(, bridge_softc), bridge_list);
 512 #define V_bridge_list   VNET(bridge_list)
 513 #define BRIDGE_LIST_LOCK_INIT(x)        mtx_init(&V_bridge_list_mtx,    \
 514                                             "if_bridge list", NULL, MTX_DEF)
 515 #define BRIDGE_LIST_LOCK_DESTROY(x)     mtx_destroy(&V_bridge_list_mtx)
 516 #define BRIDGE_LIST_LOCK(x)             mtx_lock(&V_bridge_list_mtx)
 517 #define BRIDGE_LIST_UNLOCK(x)           mtx_unlock(&V_bridge_list_mtx)
 518
 519 static VNET_DEFINE(struct if_clone *, bridge_cloner);
 520 #define V_bridge_cloner VNET(bridge_cloner)
 521
 522 static const char bridge_name[] = "bridge";
 523
 524 static void
 525 vnet_bridge_init(const void *unused __unused)
 526 {
 527
 528         BRIDGE_LIST_LOCK_INIT();
 529         LIST_INIT(&V_bridge_list);
 530         V_bridge_cloner = if_clone_simple(bridge_name,
 531             bridge_clone_create, bridge_clone_destroy, 0);
 532 }
 533 VNET_SYSINIT(vnet_bridge_init, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 534     vnet_bridge_init, NULL);
 535
 536 static void
 537 vnet_bridge_uninit(const void *unused __unused)
 538 {
 539
 540         if_clone_detach(V_bridge_cloner);
 541         V_bridge_cloner = NULL;
 542         BRIDGE_LIST_LOCK_DESTROY();
 543 }
 544 VNET_SYSUNINIT(vnet_bridge_uninit, SI_SUB_PROTO_IFATTACHDOMAIN, SI_ORDER_ANY,
 545     vnet_bridge_uninit, NULL);
 546
 547 static int
 548 bridge_modevent(module_t mod, int type, void *data)
 549 {
 550
 551         switch (type) {
 552         case MOD_LOAD:
 553                 bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
 554                     sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
 555                     UMA_ALIGN_PTR, 0);
 556                 bridge_input_p = bridge_input;
 557                 bridge_output_p = bridge_output;
 558                 bridge_dn_p = bridge_dummynet;
 559                 bridge_linkstate_p = bridge_linkstate;
 560                 bridge_detach_cookie = EVENTHANDLER_REGISTER(
 561                     ifnet_departure_event, bridge_ifdetach, NULL,
 562                     EVENTHANDLER_PRI_ANY);
 563                 break;
 564         case MOD_UNLOAD:
 565                 EVENTHANDLER_DEREGISTER(ifnet_departure_event,
 566                     bridge_detach_cookie);
 567                 uma_zdestroy(bridge_rtnode_zone);
 568                 bridge_input_p = NULL;
 569                 bridge_output_p = NULL;
 570                 bridge_dn_p = NULL;
 571                 bridge_linkstate_p = NULL;
 572                 break;
 573         default:
 574                 return (EOPNOTSUPP);
 575         }
 576         return (0);
 577 }
 578
 579 static moduledata_t bridge_mod = {
 580         "if_bridge",
 581         bridge_modevent,
 582         0
 583 };
 584
 585 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
 586 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
 587
 588 /*
 589  * handler for net.link.bridge.ipfw
 590  */
 591 static int
 592 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
 593 {
 594         int enable = V_pfil_ipfw;
 595         int error;
 596
 597         error = sysctl_handle_int(oidp, &enable, 0, req);
 598         enable &= 1;
 599
 600         if (enable != V_pfil_ipfw) {
 601                 V_pfil_ipfw = enable;
 602
 603                 /*
 604                  * Disable pfil so that ipfw doesnt run twice, if the user
 605                  * really wants both then they can re-enable pfil_bridge and/or
 606                  * pfil_member. Also allow non-ip packets as ipfw can filter by
 607                  * layer2 type.
 608                  */
 609                 if (V_pfil_ipfw) {
 610                         V_pfil_onlyip = 0;
 611                         V_pfil_bridge = 0;
 612                         V_pfil_member = 0;
 613                 }
 614         }
 615
 616         return (error);
 617 }
 618 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw,
 619     CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_VNET,
 620     &VNET_NAME(pfil_ipfw), 0, &sysctl_pfil_ipfw, "I",
 621     "Layer2 filter with IPFW");
 622
 623 /*
 624  * bridge_clone_create:
 625  *
 626  *      Create a new bridge instance.
 627  */
 628 static int
 629 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
 630 {
 631         struct bridge_softc *sc, *sc2;
 632         struct ifnet *bifp, *ifp;
 633         int fb, retry;
 634         unsigned long hostid;
 635
 636         sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
 637         ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
 638         if (ifp == NULL) {
 639                 free(sc, M_DEVBUF);
 640                 return (ENOSPC);
 641         }
 642
 643         BRIDGE_LOCK_INIT(sc);
 644         sc->sc_brtmax = BRIDGE_RTABLE_MAX;
 645         sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
 646
 647         /* Initialize our routing table. */
 648         bridge_rtable_init(sc);
 649
 650         callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
 651
 652         LIST_INIT(&sc->sc_iflist);
 653         LIST_INIT(&sc->sc_spanlist);
 654
 655         ifp->if_softc = sc;
 656         if_initname(ifp, bridge_name, unit);
 657         ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
 658         ifp->if_ioctl = bridge_ioctl;
 659         ifp->if_transmit = bridge_transmit;
 660         ifp->if_qflush = bridge_qflush;
 661         ifp->if_init = bridge_init;
 662         ifp->if_type = IFT_BRIDGE;
 663
 664         /*
 665          * Generate an ethernet address with a locally administered address.
 666          *
 667          * Since we are using random ethernet addresses for the bridge, it is
 668          * possible that we might have address collisions, so make sure that
 669          * this hardware address isn't already in use on another bridge.
 670          * The first try uses the hostid and falls back to arc4rand().
 671          */
 672         fb = 0;
 673         getcredhostid(curthread->td_ucred, &hostid);
 674         do {
 675                 if (fb || hostid == 0) {
 676                         arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
 677                         sc->sc_defaddr[0] &= ~1;/* clear multicast bit */
 678                         sc->sc_defaddr[0] |= 2; /* set the LAA bit */
 679                 } else {
 680                         sc->sc_defaddr[0] = 0x2;
 681                         sc->sc_defaddr[1] = (hostid >> 24) & 0xff;
 682                         sc->sc_defaddr[2] = (hostid >> 16) & 0xff;
 683                         sc->sc_defaddr[3] = (hostid >> 8 ) & 0xff;
 684                         sc->sc_defaddr[4] =  hostid        & 0xff;
 685                         sc->sc_defaddr[5] = ifp->if_dunit & 0xff;
 686                 }
 687
 688                 fb = 1;
 689                 retry = 0;
 690                 BRIDGE_LIST_LOCK();
 691                 LIST_FOREACH(sc2, &V_bridge_list, sc_list) {
 692                         bifp = sc2->sc_ifp;
 693                         if (memcmp(sc->sc_defaddr,
 694                             IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0) {
 695                                 retry = 1;
 696                                 break;
 697                         }
 698                 }
 699                 BRIDGE_LIST_UNLOCK();
 700         } while (retry == 1);
 701
 702         bstp_attach(&sc->sc_stp, &bridge_ops);
 703         ether_ifattach(ifp, sc->sc_defaddr);
 704         /* Now undo some of the damage... */
 705         ifp->if_baudrate = 0;
 706         ifp->if_type = IFT_BRIDGE;
 707
 708         BRIDGE_LIST_LOCK();
 709         LIST_INSERT_HEAD(&V_bridge_list, sc, sc_list);
 710         BRIDGE_LIST_UNLOCK();
 711
 712         return (0);
 713 }
 714
 715 /*
 716  * bridge_clone_destroy:
 717  *
 718  *      Destroy a bridge instance.
 719  */
 720 static void
 721 bridge_clone_destroy(struct ifnet *ifp)
 722 {
 723         struct bridge_softc *sc = ifp->if_softc;
 724         struct bridge_iflist *bif;
 725
 726         BRIDGE_LOCK(sc);
 727
 728         bridge_stop(ifp, 1);
 729         ifp->if_flags &= ~IFF_UP;
 730
 731         while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
 732                 bridge_delete_member(sc, bif, 0);
 733
 734         while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
 735                 bridge_delete_span(sc, bif);
 736         }
 737
 738         BRIDGE_UNLOCK(sc);
 739
 740         callout_drain(&sc->sc_brcallout);
 741
 742         BRIDGE_LIST_LOCK();
 743         LIST_REMOVE(sc, sc_list);
 744         BRIDGE_LIST_UNLOCK();
 745
 746         bstp_detach(&sc->sc_stp);
 747         ether_ifdetach(ifp);
 748         if_free(ifp);
 749
 750         /* Tear down the routing table. */
 751         bridge_rtable_fini(sc);
 752
 753         BRIDGE_LOCK_DESTROY(sc);
 754         free(sc, M_DEVBUF);
 755 }
 756
 757 /*
 758  * bridge_ioctl:
 759  *
 760  *      Handle a control request from the operator.
 761  */
 762 static int
 763 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
 764 {
 765         struct bridge_softc *sc = ifp->if_softc;
 766         struct ifreq *ifr = (struct ifreq *)data;
 767         struct bridge_iflist *bif;
 768         struct thread *td = curthread;
 769         union {
 770                 struct ifbreq ifbreq;
 771                 struct ifbifconf ifbifconf;
 772                 struct ifbareq ifbareq;
 773                 struct ifbaconf ifbaconf;
 774                 struct ifbrparam ifbrparam;
 775                 struct ifbropreq ifbropreq;
 776         } args;
 777         struct ifdrv *ifd = (struct ifdrv *) data;
 778         const struct bridge_control *bc;
 779         int error = 0;
 780
 781         switch (cmd) {
 782
 783         case SIOCADDMULTI:
 784         case SIOCDELMULTI:
 785                 break;
 786
 787         case SIOCGDRVSPEC:
 788         case SIOCSDRVSPEC:
 789                 if (ifd->ifd_cmd >= bridge_control_table_size) {
 790                         error = EINVAL;
 791                         break;
 792                 }
 793                 bc = &bridge_control_table[ifd->ifd_cmd];
 794
 795                 if (cmd == SIOCGDRVSPEC &&
 796                     (bc->bc_flags & BC_F_COPYOUT) == 0) {
 797                         error = EINVAL;
 798                         break;
 799                 }
 800                 else if (cmd == SIOCSDRVSPEC &&
 801                     (bc->bc_flags & BC_F_COPYOUT) != 0) {
 802                         error = EINVAL;
 803                         break;
 804                 }
 805
 806                 if (bc->bc_flags & BC_F_SUSER) {
 807                         error = priv_check(td, PRIV_NET_BRIDGE);
 808                         if (error)
 809                                 break;
 810                 }
 811
 812                 if (ifd->ifd_len != bc->bc_argsize ||
 813                     ifd->ifd_len > sizeof(args)) {
 814                         error = EINVAL;
 815                         break;
 816                 }
 817
 818                 bzero(&args, sizeof(args));
 819                 if (bc->bc_flags & BC_F_COPYIN) {
 820                         error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
 821                         if (error)
 822                                 break;
 823                 }
 824
 825                 BRIDGE_LOCK(sc);
 826                 error = (*bc->bc_func)(sc, &args);
 827                 BRIDGE_UNLOCK(sc);
 828                 if (error)
 829                         break;
 830
 831                 if (bc->bc_flags & BC_F_COPYOUT)
 832                         error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
 833
 834                 break;
 835
 836         case SIOCSIFFLAGS:
 837                 if (!(ifp->if_flags & IFF_UP) &&
 838                     (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 839                         /*
 840                          * If interface is marked down and it is running,
 841                          * then stop and disable it.
 842                          */
 843                         BRIDGE_LOCK(sc);
 844                         bridge_stop(ifp, 1);
 845                         BRIDGE_UNLOCK(sc);
 846                 } else if ((ifp->if_flags & IFF_UP) &&
 847                     !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
 848                         /*
 849                          * If interface is marked up and it is stopped, then
 850                          * start it.
 851                          */
 852                         (*ifp->if_init)(sc);
 853                 }
 854                 break;
 855
 856         case SIOCSIFMTU:
 857                 if (ifr->ifr_mtu < 576) {
 858                         error = EINVAL;
 859                         break;
 860                 }
 861                 if (LIST_EMPTY(&sc->sc_iflist)) {
 862                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 863                         break;
 864                 }
 865                 BRIDGE_LOCK(sc);
 866                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 867                         if (bif->bif_ifp->if_mtu != ifr->ifr_mtu) {
 868                                 log(LOG_NOTICE, "%s: invalid MTU: %u(%s)"
 869                                     " != %d\n", sc->sc_ifp->if_xname,
 870                                     bif->bif_ifp->if_mtu,
 871                                     bif->bif_ifp->if_xname, ifr->ifr_mtu);
 872                                 error = EINVAL;
 873                                 break;
 874                         }
 875                 }
 876                 if (!error)
 877                         sc->sc_ifp->if_mtu = ifr->ifr_mtu;
 878                 BRIDGE_UNLOCK(sc);
 879                 break;
 880         default:
 881                 /*
 882                  * drop the lock as ether_ioctl() will call bridge_start() and
 883                  * cause the lock to be recursed.
 884                  */
 885                 error = ether_ioctl(ifp, cmd, data);
 886                 break;
 887         }
 888
 889         return (error);
 890 }
 891
 892 /*
 893  * bridge_mutecaps:
 894  *
 895  *      Clear or restore unwanted capabilities on the member interface
 896  */
 897 static void
 898 bridge_mutecaps(struct bridge_softc *sc)
 899 {
 900         struct bridge_iflist *bif;
 901         int enabled, mask;
 902
 903         /* Initial bitmask of capabilities to test */
 904         mask = BRIDGE_IFCAPS_MASK;
 905
 906         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 907                 /* Every member must support it or its disabled */
 908                 mask &= bif->bif_savedcaps;
 909         }
 910
 911         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 912                 enabled = bif->bif_ifp->if_capenable;
 913                 enabled &= ~BRIDGE_IFCAPS_STRIP;
 914                 /* strip off mask bits and enable them again if allowed */
 915                 enabled &= ~BRIDGE_IFCAPS_MASK;
 916                 enabled |= mask;
 917                 bridge_set_ifcap(sc, bif, enabled);
 918         }
 919
 920 }
 921
 922 static void
 923 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
 924 {
 925         struct ifnet *ifp = bif->bif_ifp;
 926         struct ifreq ifr;
 927         int error;
 928
 929         bzero(&ifr, sizeof(ifr));
 930         ifr.ifr_reqcap = set;
 931
 932         if (ifp->if_capenable != set) {
 933                 error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
 934                 if (error)
 935                         if_printf(sc->sc_ifp,
 936                             "error setting interface capabilities on %s\n",
 937                             ifp->if_xname);
 938         }
 939 }
 940
 941 /*
 942  * bridge_lookup_member:
 943  *
 944  *      Lookup a bridge member interface.
 945  */
 946 static struct bridge_iflist *
 947 bridge_lookup_member(struct bridge_softc *sc, const char *name)
 948 {
 949         struct bridge_iflist *bif;
 950         struct ifnet *ifp;
 951
 952         BRIDGE_LOCK_ASSERT(sc);
 953
 954         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 955                 ifp = bif->bif_ifp;
 956                 if (strcmp(ifp->if_xname, name) == 0)
 957                         return (bif);
 958         }
 959
 960         return (NULL);
 961 }
 962
 963 /*
 964  * bridge_lookup_member_if:
 965  *
 966  *      Lookup a bridge member interface by ifnet*.
 967  */
 968 static struct bridge_iflist *
 969 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
 970 {
 971         struct bridge_iflist *bif;
 972
 973         BRIDGE_LOCK_ASSERT(sc);
 974
 975         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
 976                 if (bif->bif_ifp == member_ifp)
 977                         return (bif);
 978         }
 979
 980         return (NULL);
 981 }
 982
 983 /*
 984  * bridge_delete_member:
 985  *
 986  *      Delete the specified member interface.
 987  */
 988 static void
 989 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
 990     int gone)
 991 {
 992         struct ifnet *ifs = bif->bif_ifp;
 993         struct ifnet *fif = NULL;
 994
 995         BRIDGE_LOCK_ASSERT(sc);
 996
 997         if (bif->bif_flags & IFBIF_STP)
 998                 bstp_disable(&bif->bif_stp);
 999
1000         ifs->if_bridge = NULL;
1001         BRIDGE_XLOCK(sc);
1002         LIST_REMOVE(bif, bif_next);
1003         BRIDGE_XDROP(sc);
1004
1005         /*
1006          * If removing the interface that gave the bridge its mac address, set
1007          * the mac address of the bridge to the address of the next member, or
1008          * to its default address if no members are left.
1009          */
1010         if (V_bridge_inherit_mac && sc->sc_ifaddr == ifs) {
1011                 if (LIST_EMPTY(&sc->sc_iflist)) {
1012                         bcopy(sc->sc_defaddr,
1013                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1014                         sc->sc_ifaddr = NULL;
1015                 } else {
1016                         fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
1017                         bcopy(IF_LLADDR(fif),
1018                             IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1019                         sc->sc_ifaddr = fif;
1020                 }
1021                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1022         }
1023
1024         bridge_linkcheck(sc);
1025         bridge_mutecaps(sc);    /* recalcuate now this interface is removed */
1026         bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
1027         KASSERT(bif->bif_addrcnt == 0,
1028             ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
1029
1030         BRIDGE_UNLOCK(sc);
1031         if (!gone) {
1032                 switch (ifs->if_type) {
1033                 case IFT_ETHER:
1034                 case IFT_L2VLAN:
1035                         /*
1036                          * Take the interface out of promiscuous mode, but only
1037                          * if it was promiscuous in the first place. It might
1038                          * not be if we're in the bridge_ioctl_add() error path.
1039                          */
1040                         if (ifs->if_flags & IFF_PROMISC)
1041                                 (void) ifpromisc(ifs, 0);
1042                         break;
1043
1044                 case IFT_GIF:
1045                         break;
1046
1047                 default:
1048 #ifdef DIAGNOSTIC
1049                         panic("bridge_delete_member: impossible");
1050 #endif
1051                         break;
1052                 }
1053                 /* reneable any interface capabilities */
1054                 bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
1055         }
1056         bstp_destroy(&bif->bif_stp);    /* prepare to free */
1057         BRIDGE_LOCK(sc);
1058         free(bif, M_DEVBUF);
1059 }
1060
1061 /*
1062  * bridge_delete_span:
1063  *
1064  *      Delete the specified span interface.
1065  */
1066 static void
1067 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
1068 {
1069         BRIDGE_LOCK_ASSERT(sc);
1070
1071         KASSERT(bif->bif_ifp->if_bridge == NULL,
1072             ("%s: not a span interface", __func__));
1073
1074         LIST_REMOVE(bif, bif_next);
1075         free(bif, M_DEVBUF);
1076 }
1077
1078 static int
1079 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
1080 {
1081         struct ifbreq *req = arg;
1082         struct bridge_iflist *bif = NULL;
1083         struct ifnet *ifs;
1084         int error = 0;
1085
1086         ifs = ifunit(req->ifbr_ifsname);
1087         if (ifs == NULL)
1088                 return (ENOENT);
1089         if (ifs->if_ioctl == NULL)      /* must be supported */
1090                 return (EINVAL);
1091
1092         /* If it's in the span list, it can't be a member. */
1093         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1094                 if (ifs == bif->bif_ifp)
1095                         return (EBUSY);
1096
1097         if (ifs->if_bridge == sc)
1098                 return (EEXIST);
1099
1100         if (ifs->if_bridge != NULL)
1101                 return (EBUSY);
1102
1103         switch (ifs->if_type) {
1104         case IFT_ETHER:
1105         case IFT_L2VLAN:
1106         case IFT_GIF:
1107                 /* permitted interface types */
1108                 break;
1109         default:
1110                 return (EINVAL);
1111         }
1112
1113 #ifdef INET6
1114         /*
1115          * Two valid inet6 addresses with link-local scope must not be
1116          * on the parent interface and the member interfaces at the
1117          * same time.  This restriction is needed to prevent violation
1118          * of link-local scope zone.  Attempts to add a member
1119          * interface which has inet6 addresses when the parent has
1120          * inet6 triggers removal of all inet6 addresses on the member
1121          * interface.
1122          */
1123
1124         /* Check if the parent interface has a link-local scope addr. */
1125         if (V_allow_llz_overlap == 0 &&
1126             in6ifa_llaonifp(sc->sc_ifp) != NULL) {
1127                 /*
1128                  * If any, remove all inet6 addresses from the member
1129                  * interfaces.
1130                  */
1131                 BRIDGE_XLOCK(sc);
1132                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1133                         if (in6ifa_llaonifp(bif->bif_ifp)) {
1134                                 BRIDGE_UNLOCK(sc);
1135                                 in6_ifdetach(bif->bif_ifp);
1136                                 BRIDGE_LOCK(sc);
1137                                 if_printf(sc->sc_ifp,
1138                                     "IPv6 addresses on %s have been removed "
1139                                     "before adding it as a member to prevent "
1140                                     "IPv6 address scope violation.\n",
1141                                     bif->bif_ifp->if_xname);
1142                         }
1143                 }
1144                 BRIDGE_XDROP(sc);
1145                 if (in6ifa_llaonifp(ifs)) {
1146                         BRIDGE_UNLOCK(sc);
1147                         in6_ifdetach(ifs);
1148                         BRIDGE_LOCK(sc);
1149                         if_printf(sc->sc_ifp,
1150                             "IPv6 addresses on %s have been removed "
1151                             "before adding it as a member to prevent "
1152                             "IPv6 address scope violation.\n",
1153                             ifs->if_xname);
1154                 }
1155         }
1156 #endif
1157         /* Allow the first Ethernet member to define the MTU */
1158         if (LIST_EMPTY(&sc->sc_iflist))
1159                 sc->sc_ifp->if_mtu = ifs->if_mtu;
1160         else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
1161                 if_printf(sc->sc_ifp, "invalid MTU: %u(%s) != %u\n",
1162                     ifs->if_mtu, ifs->if_xname, sc->sc_ifp->if_mtu);
1163                 return (EINVAL);
1164         }
1165
1166         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1167         if (bif == NULL)
1168                 return (ENOMEM);
1169
1170         bif->bif_ifp = ifs;
1171         bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1172         bif->bif_savedcaps = ifs->if_capenable;
1173
1174         /*
1175          * Assign the interface's MAC address to the bridge if it's the first
1176          * member and the MAC address of the bridge has not been changed from
1177          * the default randomly generated one.
1178          */
1179         if (V_bridge_inherit_mac && LIST_EMPTY(&sc->sc_iflist) &&
1180             !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN)) {
1181                 bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1182                 sc->sc_ifaddr = ifs;
1183                 EVENTHANDLER_INVOKE(iflladdr_event, sc->sc_ifp);
1184         }
1185
1186         ifs->if_bridge = sc;
1187         bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1188         /*
1189          * XXX: XLOCK HERE!?!
1190          *
1191          * NOTE: insert_***HEAD*** should be safe for the traversals.
1192          */
1193         LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1194
1195         /* Set interface capabilities to the intersection set of all members */
1196         bridge_mutecaps(sc);
1197         bridge_linkcheck(sc);
1198
1199         /* Place the interface into promiscuous mode */
1200         switch (ifs->if_type) {
1201                 case IFT_ETHER:
1202                 case IFT_L2VLAN:
1203                         BRIDGE_UNLOCK(sc);
1204                         error = ifpromisc(ifs, 1);
1205                         BRIDGE_LOCK(sc);
1206                         break;
1207         }
1208
1209         if (error)
1210                 bridge_delete_member(sc, bif, 0);
1211         return (error);
1212 }
1213
1214 static int
1215 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1216 {
1217         struct ifbreq *req = arg;
1218         struct bridge_iflist *bif;
1219
1220         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1221         if (bif == NULL)
1222                 return (ENOENT);
1223
1224         bridge_delete_member(sc, bif, 0);
1225
1226         return (0);
1227 }
1228
1229 static int
1230 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1231 {
1232         struct ifbreq *req = arg;
1233         struct bridge_iflist *bif;
1234         struct bstp_port *bp;
1235
1236         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1237         if (bif == NULL)
1238                 return (ENOENT);
1239
1240         bp = &bif->bif_stp;
1241         req->ifbr_ifsflags = bif->bif_flags;
1242         req->ifbr_state = bp->bp_state;
1243         req->ifbr_priority = bp->bp_priority;
1244         req->ifbr_path_cost = bp->bp_path_cost;
1245         req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1246         req->ifbr_proto = bp->bp_protover;
1247         req->ifbr_role = bp->bp_role;
1248         req->ifbr_stpflags = bp->bp_flags;
1249         req->ifbr_addrcnt = bif->bif_addrcnt;
1250         req->ifbr_addrmax = bif->bif_addrmax;
1251         req->ifbr_addrexceeded = bif->bif_addrexceeded;
1252
1253         /* Copy STP state options as flags */
1254         if (bp->bp_operedge)
1255                 req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1256         if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1257                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1258         if (bp->bp_ptp_link)
1259                 req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1260         if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1261                 req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1262         if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1263                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1264         if (bp->bp_flags & BSTP_PORT_ADMCOST)
1265                 req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1266         return (0);
1267 }
1268
1269 static int
1270 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1271 {
1272         struct ifbreq *req = arg;
1273         struct bridge_iflist *bif;
1274         struct bstp_port *bp;
1275         int error;
1276
1277         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1278         if (bif == NULL)
1279                 return (ENOENT);
1280         bp = &bif->bif_stp;
1281
1282         if (req->ifbr_ifsflags & IFBIF_SPAN)
1283                 /* SPAN is readonly */
1284                 return (EINVAL);
1285
1286         if (req->ifbr_ifsflags & IFBIF_STP) {
1287                 if ((bif->bif_flags & IFBIF_STP) == 0) {
1288                         error = bstp_enable(&bif->bif_stp);
1289                         if (error)
1290                                 return (error);
1291                 }
1292         } else {
1293                 if ((bif->bif_flags & IFBIF_STP) != 0)
1294                         bstp_disable(&bif->bif_stp);
1295         }
1296
1297         /* Pass on STP flags */
1298         bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1299         bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1300         bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1301         bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1302
1303         /* Save the bits relating to the bridge */
1304         bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1305
1306         return (0);
1307 }
1308
1309 static int
1310 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1311 {
1312         struct ifbrparam *param = arg;
1313
1314         sc->sc_brtmax = param->ifbrp_csize;
1315         bridge_rttrim(sc);
1316
1317         return (0);
1318 }
1319
1320 static int
1321 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1322 {
1323         struct ifbrparam *param = arg;
1324
1325         param->ifbrp_csize = sc->sc_brtmax;
1326
1327         return (0);
1328 }
1329
1330 static int
1331 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1332 {
1333         struct ifbifconf *bifc = arg;
1334         struct bridge_iflist *bif;
1335         struct ifbreq breq;
1336         char *buf, *outbuf;
1337         int count, buflen, len, error = 0;
1338
1339         count = 0;
1340         LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1341                 count++;
1342         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1343                 count++;
1344
1345         buflen = sizeof(breq) * count;
1346         if (bifc->ifbic_len == 0) {
1347                 bifc->ifbic_len = buflen;
1348                 return (0);
1349         }
1350         BRIDGE_UNLOCK(sc);
1351         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1352         BRIDGE_LOCK(sc);
1353
1354         count = 0;
1355         buf = outbuf;
1356         len = min(bifc->ifbic_len, buflen);
1357         bzero(&breq, sizeof(breq));
1358         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1359                 if (len < sizeof(breq))
1360                         break;
1361
1362                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1363                     sizeof(breq.ifbr_ifsname));
1364                 /* Fill in the ifbreq structure */
1365                 error = bridge_ioctl_gifflags(sc, &breq);
1366                 if (error)
1367                         break;
1368                 memcpy(buf, &breq, sizeof(breq));
1369                 count++;
1370                 buf += sizeof(breq);
1371                 len -= sizeof(breq);
1372         }
1373         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1374                 if (len < sizeof(breq))
1375                         break;
1376
1377                 strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1378                     sizeof(breq.ifbr_ifsname));
1379                 breq.ifbr_ifsflags = bif->bif_flags;
1380                 breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1381                 memcpy(buf, &breq, sizeof(breq));
1382                 count++;
1383                 buf += sizeof(breq);
1384                 len -= sizeof(breq);
1385         }
1386
1387         BRIDGE_UNLOCK(sc);
1388         bifc->ifbic_len = sizeof(breq) * count;
1389         error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1390         BRIDGE_LOCK(sc);
1391         free(outbuf, M_TEMP);
1392         return (error);
1393 }
1394
1395 static int
1396 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1397 {
1398         struct ifbaconf *bac = arg;
1399         struct bridge_rtnode *brt;
1400         struct ifbareq bareq;
1401         char *buf, *outbuf;
1402         int count, buflen, len, error = 0;
1403
1404         if (bac->ifbac_len == 0)
1405                 return (0);
1406
1407         count = 0;
1408         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1409                 count++;
1410         buflen = sizeof(bareq) * count;
1411
1412         BRIDGE_UNLOCK(sc);
1413         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1414         BRIDGE_LOCK(sc);
1415
1416         count = 0;
1417         buf = outbuf;
1418         len = min(bac->ifbac_len, buflen);
1419         bzero(&bareq, sizeof(bareq));
1420         LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1421                 if (len < sizeof(bareq))
1422                         goto out;
1423                 strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1424                     sizeof(bareq.ifba_ifsname));
1425                 memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1426                 bareq.ifba_vlan = brt->brt_vlan;
1427                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1428                                 time_uptime < brt->brt_expire)
1429                         bareq.ifba_expire = brt->brt_expire - time_uptime;
1430                 else
1431                         bareq.ifba_expire = 0;
1432                 bareq.ifba_flags = brt->brt_flags;
1433
1434                 memcpy(buf, &bareq, sizeof(bareq));
1435                 count++;
1436                 buf += sizeof(bareq);
1437                 len -= sizeof(bareq);
1438         }
1439 out:
1440         BRIDGE_UNLOCK(sc);
1441         bac->ifbac_len = sizeof(bareq) * count;
1442         error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1443         BRIDGE_LOCK(sc);
1444         free(outbuf, M_TEMP);
1445         return (error);
1446 }
1447
1448 static int
1449 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1450 {
1451         struct ifbareq *req = arg;
1452         struct bridge_iflist *bif;
1453         int error;
1454
1455         bif = bridge_lookup_member(sc, req->ifba_ifsname);
1456         if (bif == NULL)
1457                 return (ENOENT);
1458
1459         error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1460             req->ifba_flags);
1461
1462         return (error);
1463 }
1464
1465 static int
1466 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1467 {
1468         struct ifbrparam *param = arg;
1469
1470         sc->sc_brttimeout = param->ifbrp_ctime;
1471         return (0);
1472 }
1473
1474 static int
1475 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1476 {
1477         struct ifbrparam *param = arg;
1478
1479         param->ifbrp_ctime = sc->sc_brttimeout;
1480         return (0);
1481 }
1482
1483 static int
1484 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1485 {
1486         struct ifbareq *req = arg;
1487
1488         return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1489 }
1490
1491 static int
1492 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1493 {
1494         struct ifbreq *req = arg;
1495
1496         bridge_rtflush(sc, req->ifbr_ifsflags);
1497         return (0);
1498 }
1499
1500 static int
1501 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1502 {
1503         struct ifbrparam *param = arg;
1504         struct bstp_state *bs = &sc->sc_stp;
1505
1506         param->ifbrp_prio = bs->bs_bridge_priority;
1507         return (0);
1508 }
1509
1510 static int
1511 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1512 {
1513         struct ifbrparam *param = arg;
1514
1515         return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1516 }
1517
1518 static int
1519 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1520 {
1521         struct ifbrparam *param = arg;
1522         struct bstp_state *bs = &sc->sc_stp;
1523
1524         param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1525         return (0);
1526 }
1527
1528 static int
1529 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1530 {
1531         struct ifbrparam *param = arg;
1532
1533         return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1534 }
1535
1536 static int
1537 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1538 {
1539         struct ifbrparam *param = arg;
1540         struct bstp_state *bs = &sc->sc_stp;
1541
1542         param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1543         return (0);
1544 }
1545
1546 static int
1547 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1548 {
1549         struct ifbrparam *param = arg;
1550
1551         return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1552 }
1553
1554 static int
1555 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1556 {
1557         struct ifbrparam *param = arg;
1558         struct bstp_state *bs = &sc->sc_stp;
1559
1560         param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1561         return (0);
1562 }
1563
1564 static int
1565 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1566 {
1567         struct ifbrparam *param = arg;
1568
1569         return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1570 }
1571
1572 static int
1573 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1574 {
1575         struct ifbreq *req = arg;
1576         struct bridge_iflist *bif;
1577
1578         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1579         if (bif == NULL)
1580                 return (ENOENT);
1581
1582         return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1583 }
1584
1585 static int
1586 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1587 {
1588         struct ifbreq *req = arg;
1589         struct bridge_iflist *bif;
1590
1591         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1592         if (bif == NULL)
1593                 return (ENOENT);
1594
1595         return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1596 }
1597
1598 static int
1599 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1600 {
1601         struct ifbreq *req = arg;
1602         struct bridge_iflist *bif;
1603
1604         bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1605         if (bif == NULL)
1606                 return (ENOENT);
1607
1608         bif->bif_addrmax = req->ifbr_addrmax;
1609         return (0);
1610 }
1611
1612 static int
1613 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1614 {
1615         struct ifbreq *req = arg;
1616         struct bridge_iflist *bif = NULL;
1617         struct ifnet *ifs;
1618
1619         ifs = ifunit(req->ifbr_ifsname);
1620         if (ifs == NULL)
1621                 return (ENOENT);
1622
1623         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1624                 if (ifs == bif->bif_ifp)
1625                         return (EBUSY);
1626
1627         if (ifs->if_bridge != NULL)
1628                 return (EBUSY);
1629
1630         switch (ifs->if_type) {
1631                 case IFT_ETHER:
1632                 case IFT_GIF:
1633                 case IFT_L2VLAN:
1634                         break;
1635                 default:
1636                         return (EINVAL);
1637         }
1638
1639         bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1640         if (bif == NULL)
1641                 return (ENOMEM);
1642
1643         bif->bif_ifp = ifs;
1644         bif->bif_flags = IFBIF_SPAN;
1645
1646         LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1647
1648         return (0);
1649 }
1650
1651 static int
1652 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1653 {
1654         struct ifbreq *req = arg;
1655         struct bridge_iflist *bif;
1656         struct ifnet *ifs;
1657
1658         ifs = ifunit(req->ifbr_ifsname);
1659         if (ifs == NULL)
1660                 return (ENOENT);
1661
1662         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1663                 if (ifs == bif->bif_ifp)
1664                         break;
1665
1666         if (bif == NULL)
1667                 return (ENOENT);
1668
1669         bridge_delete_span(sc, bif);
1670
1671         return (0);
1672 }
1673
1674 static int
1675 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1676 {
1677         struct ifbropreq *req = arg;
1678         struct bstp_state *bs = &sc->sc_stp;
1679         struct bstp_port *root_port;
1680
1681         req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1682         req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1683         req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1684
1685         root_port = bs->bs_root_port;
1686         if (root_port == NULL)
1687                 req->ifbop_root_port = 0;
1688         else
1689                 req->ifbop_root_port = root_port->bp_ifp->if_index;
1690
1691         req->ifbop_holdcount = bs->bs_txholdcount;
1692         req->ifbop_priority = bs->bs_bridge_priority;
1693         req->ifbop_protocol = bs->bs_protover;
1694         req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1695         req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1696         req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1697         req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1698         req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1699         req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1700
1701         return (0);
1702 }
1703
1704 static int
1705 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1706 {
1707         struct ifbrparam *param = arg;
1708
1709         param->ifbrp_cexceeded = sc->sc_brtexceeded;
1710         return (0);
1711 }
1712
1713 static int
1714 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1715 {
1716         struct ifbpstpconf *bifstp = arg;
1717         struct bridge_iflist *bif;
1718         struct bstp_port *bp;
1719         struct ifbpstpreq bpreq;
1720         char *buf, *outbuf;
1721         int count, buflen, len, error = 0;
1722
1723         count = 0;
1724         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1725                 if ((bif->bif_flags & IFBIF_STP) != 0)
1726                         count++;
1727         }
1728
1729         buflen = sizeof(bpreq) * count;
1730         if (bifstp->ifbpstp_len == 0) {
1731                 bifstp->ifbpstp_len = buflen;
1732                 return (0);
1733         }
1734
1735         BRIDGE_UNLOCK(sc);
1736         outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1737         BRIDGE_LOCK(sc);
1738
1739         count = 0;
1740         buf = outbuf;
1741         len = min(bifstp->ifbpstp_len, buflen);
1742         bzero(&bpreq, sizeof(bpreq));
1743         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1744                 if (len < sizeof(bpreq))
1745                         break;
1746
1747                 if ((bif->bif_flags & IFBIF_STP) == 0)
1748                         continue;
1749
1750                 bp = &bif->bif_stp;
1751                 bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1752                 bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1753                 bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1754                 bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1755                 bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1756                 bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1757
1758                 memcpy(buf, &bpreq, sizeof(bpreq));
1759                 count++;
1760                 buf += sizeof(bpreq);
1761                 len -= sizeof(bpreq);
1762         }
1763
1764         BRIDGE_UNLOCK(sc);
1765         bifstp->ifbpstp_len = sizeof(bpreq) * count;
1766         error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1767         BRIDGE_LOCK(sc);
1768         free(outbuf, M_TEMP);
1769         return (error);
1770 }
1771
1772 static int
1773 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1774 {
1775         struct ifbrparam *param = arg;
1776
1777         return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1778 }
1779
1780 static int
1781 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1782 {
1783         struct ifbrparam *param = arg;
1784
1785         return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1786 }
1787
1788 /*
1789  * bridge_ifdetach:
1790  *
1791  *      Detach an interface from a bridge.  Called when a member
1792  *      interface is detaching.
1793  */
1794 static void
1795 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1796 {
1797         struct bridge_softc *sc = ifp->if_bridge;
1798         struct bridge_iflist *bif;
1799
1800         if (ifp->if_flags & IFF_RENAMING)
1801                 return;
1802         if (V_bridge_cloner == NULL) {
1803                 /*
1804                  * This detach handler can be called after
1805                  * vnet_bridge_uninit().  Just return in that case.
1806                  */
1807                 return;
1808         }
1809         /* Check if the interface is a bridge member */
1810         if (sc != NULL) {
1811                 BRIDGE_LOCK(sc);
1812
1813                 bif = bridge_lookup_member_if(sc, ifp);
1814                 if (bif != NULL)
1815                         bridge_delete_member(sc, bif, 1);
1816
1817                 BRIDGE_UNLOCK(sc);
1818                 return;
1819         }
1820
1821         /* Check if the interface is a span port */
1822         BRIDGE_LIST_LOCK();
1823         LIST_FOREACH(sc, &V_bridge_list, sc_list) {
1824                 BRIDGE_LOCK(sc);
1825                 LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1826                         if (ifp == bif->bif_ifp) {
1827                                 bridge_delete_span(sc, bif);
1828                                 break;
1829                         }
1830
1831                 BRIDGE_UNLOCK(sc);
1832         }
1833         BRIDGE_LIST_UNLOCK();
1834 }
1835
1836 /*
1837  * bridge_init:
1838  *
1839  *      Initialize a bridge interface.
1840  */
1841 static void
1842 bridge_init(void *xsc)
1843 {
1844         struct bridge_softc *sc = (struct bridge_softc *)xsc;
1845         struct ifnet *ifp = sc->sc_ifp;
1846
1847         if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1848                 return;
1849
1850         BRIDGE_LOCK(sc);
1851         callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1852             bridge_timer, sc);
1853
1854         ifp->if_drv_flags |= IFF_DRV_RUNNING;
1855         bstp_init(&sc->sc_stp);         /* Initialize Spanning Tree */
1856
1857         BRIDGE_UNLOCK(sc);
1858 }
1859
1860 /*
1861  * bridge_stop:
1862  *
1863  *      Stop the bridge interface.
1864  */
1865 static void
1866 bridge_stop(struct ifnet *ifp, int disable)
1867 {
1868         struct bridge_softc *sc = ifp->if_softc;
1869
1870         BRIDGE_LOCK_ASSERT(sc);
1871
1872         if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1873                 return;
1874
1875         callout_stop(&sc->sc_brcallout);
1876         bstp_stop(&sc->sc_stp);
1877
1878         bridge_rtflush(sc, IFBF_FLUSHDYN);
1879
1880         ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1881 }
1882
1883 /*
1884  * bridge_enqueue:
1885  *
1886  *      Enqueue a packet on a bridge member interface.
1887  *
1888  */
1889 static int
1890 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1891 {
1892         int len, err = 0;
1893         short mflags;
1894         struct mbuf *m0;
1895
1896         /* We may be sending a fragment so traverse the mbuf */
1897         for (; m; m = m0) {
1898                 m0 = m->m_nextpkt;
1899                 m->m_nextpkt = NULL;
1900                 len = m->m_pkthdr.len;
1901                 mflags = m->m_flags;
1902
1903                 /*
1904                  * If underlying interface can not do VLAN tag insertion itself
1905                  * then attach a packet tag that holds it.
1906                  */
1907                 if ((m->m_flags & M_VLANTAG) &&
1908                     (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1909                         m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1910                         if (m == NULL) {
1911                                 if_printf(dst_ifp,
1912                                     "unable to prepend VLAN header\n");
1913                                 if_inc_counter(dst_ifp, IFCOUNTER_OERRORS, 1);
1914                                 continue;
1915                         }
1916                         m->m_flags &= ~M_VLANTAG;
1917                 }
1918
1919                 if ((err = dst_ifp->if_transmit(dst_ifp, m))) {
1920                         m_freem(m0);
1921                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
1922                         break;
1923                 }
1924
1925                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OPACKETS, 1);
1926                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OBYTES, len);
1927                 if (mflags & M_MCAST)
1928                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OMCASTS, 1);
1929         }
1930
1931         return (err);
1932 }
1933
1934 /*
1935  * bridge_dummynet:
1936  *
1937  *      Receive a queued packet from dummynet and pass it on to the output
1938  *      interface.
1939  *
1940  *      The mbuf has the Ethernet header already attached.
1941  */
1942 static void
1943 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1944 {
1945         struct bridge_softc *sc;
1946
1947         sc = ifp->if_bridge;
1948
1949         /*
1950          * The packet didnt originate from a member interface. This should only
1951          * ever happen if a member interface is removed while packets are
1952          * queued for it.
1953          */
1954         if (sc == NULL) {
1955                 m_freem(m);
1956                 return;
1957         }
1958
1959         if (PFIL_HOOKED(&V_inet_pfil_hook)
1960 #ifdef INET6
1961             || PFIL_HOOKED(&V_inet6_pfil_hook)
1962 #endif
1963             ) {
1964                 if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1965                         return;
1966                 if (m == NULL)
1967                         return;
1968         }
1969
1970         bridge_enqueue(sc, ifp, m);
1971 }
1972
1973 /*
1974  * bridge_output:
1975  *
1976  *      Send output from a bridge member interface.  This
1977  *      performs the bridging function for locally originated
1978  *      packets.
1979  *
1980  *      The mbuf has the Ethernet header already attached.  We must
1981  *      enqueue or free the mbuf before returning.
1982  */
1983 static int
1984 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1985     struct rtentry *rt)
1986 {
1987         struct ether_header *eh;
1988         struct ifnet *dst_if;
1989         struct bridge_softc *sc;
1990         uint16_t vlan;
1991
1992         if (m->m_len < ETHER_HDR_LEN) {
1993                 m = m_pullup(m, ETHER_HDR_LEN);
1994                 if (m == NULL)
1995                         return (0);
1996         }
1997
1998         eh = mtod(m, struct ether_header *);
1999         sc = ifp->if_bridge;
2000         vlan = VLANTAGOF(m);
2001
2002         BRIDGE_LOCK(sc);
2003
2004         /*
2005          * If bridge is down, but the original output interface is up,
2006          * go ahead and send out that interface.  Otherwise, the packet
2007          * is dropped below.
2008          */
2009         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2010                 dst_if = ifp;
2011                 goto sendunicast;
2012         }
2013
2014         /*
2015          * If the packet is a multicast, or we don't know a better way to
2016          * get there, send to all interfaces.
2017          */
2018         if (ETHER_IS_MULTICAST(eh->ether_dhost))
2019                 dst_if = NULL;
2020         else
2021                 dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
2022         if (dst_if == NULL) {
2023                 struct bridge_iflist *bif;
2024                 struct mbuf *mc;
2025                 int error = 0, used = 0;
2026
2027                 bridge_span(sc, m);
2028
2029                 BRIDGE_LOCK2REF(sc, error);
2030                 if (error) {
2031                         m_freem(m);
2032                         return (0);
2033                 }
2034
2035                 LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2036                         dst_if = bif->bif_ifp;
2037
2038                         if (dst_if->if_type == IFT_GIF)
2039                                 continue;
2040                         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2041                                 continue;
2042
2043                         /*
2044                          * If this is not the original output interface,
2045                          * and the interface is participating in spanning
2046                          * tree, make sure the port is in a state that
2047                          * allows forwarding.
2048                          */
2049                         if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
2050                             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2051                                 continue;
2052
2053                         if (LIST_NEXT(bif, bif_next) == NULL) {
2054                                 used = 1;
2055                                 mc = m;
2056                         } else {
2057                                 mc = m_copypacket(m, M_NOWAIT);
2058                                 if (mc == NULL) {
2059                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2060                                         continue;
2061                                 }
2062                         }
2063
2064                         bridge_enqueue(sc, dst_if, mc);
2065                 }
2066                 if (used == 0)
2067                         m_freem(m);
2068                 BRIDGE_UNREF(sc);
2069                 return (0);
2070         }
2071
2072 sendunicast:
2073         /*
2074          * XXX Spanning tree consideration here?
2075          */
2076
2077         bridge_span(sc, m);
2078         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
2079                 m_freem(m);
2080                 BRIDGE_UNLOCK(sc);
2081                 return (0);
2082         }
2083
2084         BRIDGE_UNLOCK(sc);
2085         bridge_enqueue(sc, dst_if, m);
2086         return (0);
2087 }
2088
2089 /*
2090  * bridge_transmit:
2091  *
2092  *      Do output on a bridge.
2093  *
2094  */
2095 static int
2096 bridge_transmit(struct ifnet *ifp, struct mbuf *m)
2097 {
2098         struct bridge_softc *sc;
2099         struct ether_header *eh;
2100         struct ifnet *dst_if;
2101         int error = 0;
2102
2103         sc = ifp->if_softc;
2104
2105         ETHER_BPF_MTAP(ifp, m);
2106
2107         eh = mtod(m, struct ether_header *);
2108
2109         BRIDGE_LOCK(sc);
2110         if (((m->m_flags & (M_BCAST|M_MCAST)) == 0) &&
2111             (dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1)) != NULL) {
2112                 BRIDGE_UNLOCK(sc);
2113                 error = bridge_enqueue(sc, dst_if, m);
2114         } else
2115                 bridge_broadcast(sc, ifp, m, 0);
2116
2117         return (error);
2118 }
2119
2120 /*
2121  * The ifp->if_qflush entry point for if_bridge(4) is no-op.
2122  */
2123 static void
2124 bridge_qflush(struct ifnet *ifp __unused)
2125 {
2126 }
2127
2128 /*
2129  * bridge_forward:
2130  *
2131  *      The forwarding function of the bridge.
2132  *
2133  *      NOTE: Releases the lock on return.
2134  */
2135 static void
2136 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
2137     struct mbuf *m)
2138 {
2139         struct bridge_iflist *dbif;
2140         struct ifnet *src_if, *dst_if, *ifp;
2141         struct ether_header *eh;
2142         uint16_t vlan;
2143         uint8_t *dst;
2144         int error;
2145
2146         src_if = m->m_pkthdr.rcvif;
2147         ifp = sc->sc_ifp;
2148
2149         if_inc_counter(ifp, IFCOUNTER_IPACKETS, 1);
2150         if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2151         vlan = VLANTAGOF(m);
2152
2153         if ((sbif->bif_flags & IFBIF_STP) &&
2154             sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2155                 goto drop;
2156
2157         eh = mtod(m, struct ether_header *);
2158         dst = eh->ether_dhost;
2159
2160         /* If the interface is learning, record the address. */
2161         if (sbif->bif_flags & IFBIF_LEARNING) {
2162                 error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2163                     sbif, 0, IFBAF_DYNAMIC);
2164                 /*
2165                  * If the interface has addresses limits then deny any source
2166                  * that is not in the cache.
2167                  */
2168                 if (error && sbif->bif_addrmax)
2169                         goto drop;
2170         }
2171
2172         if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2173             sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2174                 goto drop;
2175
2176         /*
2177          * At this point, the port either doesn't participate
2178          * in spanning tree or it is in the forwarding state.
2179          */
2180
2181         /*
2182          * If the packet is unicast, destined for someone on
2183          * "this" side of the bridge, drop it.
2184          */
2185         if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2186                 dst_if = bridge_rtlookup(sc, dst, vlan);
2187                 if (src_if == dst_if)
2188                         goto drop;
2189         } else {
2190                 /*
2191                  * Check if its a reserved multicast address, any address
2192                  * listed in 802.1D section 7.12.6 may not be forwarded by the
2193                  * bridge.
2194                  * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2195                  */
2196                 if (dst[0] == 0x01 && dst[1] == 0x80 &&
2197                     dst[2] == 0xc2 && dst[3] == 0x00 &&
2198                     dst[4] == 0x00 && dst[5] <= 0x0f)
2199                         goto drop;
2200
2201                 /* ...forward it to all interfaces. */
2202                 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
2203                 dst_if = NULL;
2204         }
2205
2206         /*
2207          * If we have a destination interface which is a member of our bridge,
2208          * OR this is a unicast packet, push it through the bpf(4) machinery.
2209          * For broadcast or multicast packets, don't bother because it will
2210          * be reinjected into ether_input. We do this before we pass the packets
2211          * through the pfil(9) framework, as it is possible that pfil(9) will
2212          * drop the packet, or possibly modify it, making it difficult to debug
2213          * firewall issues on the bridge.
2214          */
2215         if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2216                 ETHER_BPF_MTAP(ifp, m);
2217
2218         /* run the packet filter */
2219         if (PFIL_HOOKED(&V_inet_pfil_hook)
2220 #ifdef INET6
2221             || PFIL_HOOKED(&V_inet6_pfil_hook)
2222 #endif
2223             ) {
2224                 BRIDGE_UNLOCK(sc);
2225                 if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2226                         return;
2227                 if (m == NULL)
2228                         return;
2229                 BRIDGE_LOCK(sc);
2230         }
2231
2232         if (dst_if == NULL) {
2233                 bridge_broadcast(sc, src_if, m, 1);
2234                 return;
2235         }
2236
2237         /*
2238          * At this point, we're dealing with a unicast frame
2239          * going to a different interface.
2240          */
2241         if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2242                 goto drop;
2243
2244         dbif = bridge_lookup_member_if(sc, dst_if);
2245         if (dbif == NULL)
2246                 /* Not a member of the bridge (anymore?) */
2247                 goto drop;
2248
2249         /* Private segments can not talk to each other */
2250         if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2251                 goto drop;
2252
2253         if ((dbif->bif_flags & IFBIF_STP) &&
2254             dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2255                 goto drop;
2256
2257         BRIDGE_UNLOCK(sc);
2258
2259         if (PFIL_HOOKED(&V_inet_pfil_hook)
2260 #ifdef INET6
2261             || PFIL_HOOKED(&V_inet6_pfil_hook)
2262 #endif
2263             ) {
2264                 if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2265                         return;
2266                 if (m == NULL)
2267                         return;
2268         }
2269
2270         bridge_enqueue(sc, dst_if, m);
2271         return;
2272
2273 drop:
2274         BRIDGE_UNLOCK(sc);
2275         m_freem(m);
2276 }
2277
2278 /*
2279  * bridge_input:
2280  *
2281  *      Receive input from a member interface.  Queue the packet for
2282  *      bridging if it is not for us.
2283  */
2284 static struct mbuf *
2285 bridge_input(struct ifnet *ifp, struct mbuf *m)
2286 {
2287         struct bridge_softc *sc = ifp->if_bridge;
2288         struct bridge_iflist *bif, *bif2;
2289         struct ifnet *bifp;
2290         struct ether_header *eh;
2291         struct mbuf *mc, *mc2;
2292         uint16_t vlan;
2293         int error;
2294
2295         if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2296                 return (m);
2297
2298         bifp = sc->sc_ifp;
2299         vlan = VLANTAGOF(m);
2300
2301         /*
2302          * Implement support for bridge monitoring. If this flag has been
2303          * set on this interface, discard the packet once we push it through
2304          * the bpf(4) machinery, but before we do, increment the byte and
2305          * packet counters associated with this interface.
2306          */
2307         if ((bifp->if_flags & IFF_MONITOR) != 0) {
2308                 m->m_pkthdr.rcvif  = bifp;
2309                 ETHER_BPF_MTAP(bifp, m);
2310                 if_inc_counter(bifp, IFCOUNTER_IPACKETS, 1);
2311                 if_inc_counter(bifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
2312                 m_freem(m);
2313                 return (NULL);
2314         }
2315         BRIDGE_LOCK(sc);
2316         bif = bridge_lookup_member_if(sc, ifp);
2317         if (bif == NULL) {
2318                 BRIDGE_UNLOCK(sc);
2319                 return (m);
2320         }
2321
2322         eh = mtod(m, struct ether_header *);
2323
2324         bridge_span(sc, m);
2325
2326         if (m->m_flags & (M_BCAST|M_MCAST)) {
2327                 /* Tap off 802.1D packets; they do not get forwarded. */
2328                 if (memcmp(eh->ether_dhost, bstp_etheraddr,
2329                     ETHER_ADDR_LEN) == 0) {
2330                         bstp_input(&bif->bif_stp, ifp, m); /* consumes mbuf */
2331                         BRIDGE_UNLOCK(sc);
2332                         return (NULL);
2333                 }
2334
2335                 if ((bif->bif_flags & IFBIF_STP) &&
2336                     bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2337                         BRIDGE_UNLOCK(sc);
2338                         return (m);
2339                 }
2340
2341                 /*
2342                  * Make a deep copy of the packet and enqueue the copy
2343                  * for bridge processing; return the original packet for
2344                  * local processing.
2345                  */
2346                 mc = m_dup(m, M_NOWAIT);
2347                 if (mc == NULL) {
2348                         BRIDGE_UNLOCK(sc);
2349                         return (m);
2350                 }
2351
2352                 /* Perform the bridge forwarding function with the copy. */
2353                 bridge_forward(sc, bif, mc);
2354
2355                 /*
2356                  * Reinject the mbuf as arriving on the bridge so we have a
2357                  * chance at claiming multicast packets. We can not loop back
2358                  * here from ether_input as a bridge is never a member of a
2359                  * bridge.
2360                  */
2361                 KASSERT(bifp->if_bridge == NULL,
2362                     ("loop created in bridge_input"));
2363                 mc2 = m_dup(m, M_NOWAIT);
2364                 if (mc2 != NULL) {
2365                         /* Keep the layer3 header aligned */
2366                         int i = min(mc2->m_pkthdr.len, max_protohdr);
2367                         mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2368                 }
2369                 if (mc2 != NULL) {
2370                         mc2->m_pkthdr.rcvif = bifp;
2371                         (*bifp->if_input)(bifp, mc2);
2372                 }
2373
2374                 /* Return the original packet for local processing. */
2375                 return (m);
2376         }
2377
2378         if ((bif->bif_flags & IFBIF_STP) &&
2379             bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2380                 BRIDGE_UNLOCK(sc);
2381                 return (m);
2382         }
2383
2384 #if (defined(INET) || defined(INET6))
2385 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2386         || ((iface)->if_carp \
2387             && (*carp_forus_p)((iface), eh->ether_dhost))
2388 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2389         || ((iface)->if_carp \
2390             && (*carp_forus_p)((iface), eh->ether_shost))
2391 #else
2392 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2393 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2394 #endif
2395
2396 #ifdef INET6
2397 #   define OR_PFIL_HOOKED_INET6 \
2398         || PFIL_HOOKED(&V_inet6_pfil_hook)
2399 #else
2400 #   define OR_PFIL_HOOKED_INET6
2401 #endif
2402
2403 #define GRAB_OUR_PACKETS(iface) \
2404         if ((iface)->if_type == IFT_GIF) \
2405                 continue; \
2406         /* It is destined for us. */ \
2407         if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2408             OR_CARP_CHECK_WE_ARE_DST((iface))                           \
2409             ) {                                                         \
2410                 if ((iface)->if_type == IFT_BRIDGE) {                   \
2411                         ETHER_BPF_MTAP(iface, m);                       \
2412                         if_inc_counter(iface, IFCOUNTER_IPACKETS, 1);                           \
2413                         if_inc_counter(iface, IFCOUNTER_IBYTES, m->m_pkthdr.len);               \
2414                         /* Filter on the physical interface. */         \
2415                         if (V_pfil_local_phys &&                        \
2416                             (PFIL_HOOKED(&V_inet_pfil_hook)             \
2417                              OR_PFIL_HOOKED_INET6)) {                   \
2418                                 if (bridge_pfil(&m, NULL, ifp,          \
2419                                     PFIL_IN) != 0 || m == NULL) {       \
2420                                         BRIDGE_UNLOCK(sc);              \
2421                                         return (NULL);                  \
2422                                 }                                       \
2423                                 eh = mtod(m, struct ether_header *);    \
2424                         }                                               \
2425                 }                                                       \
2426                 if (bif->bif_flags & IFBIF_LEARNING) {                  \
2427                         error = bridge_rtupdate(sc, eh->ether_shost,    \
2428                             vlan, bif, 0, IFBAF_DYNAMIC);               \
2429                         if (error && bif->bif_addrmax) {                \
2430                                 BRIDGE_UNLOCK(sc);                      \
2431                                 m_freem(m);                             \
2432                                 return (NULL);                          \
2433                         }                                               \
2434                 }                                                       \
2435                 m->m_pkthdr.rcvif = iface;                              \
2436                 BRIDGE_UNLOCK(sc);                                      \
2437                 return (m);                                             \
2438         }                                                               \
2439                                                                         \
2440         /* We just received a packet that we sent out. */               \
2441         if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2442             OR_CARP_CHECK_WE_ARE_SRC((iface))                   \
2443             ) {                                                         \
2444                 BRIDGE_UNLOCK(sc);                                      \
2445                 m_freem(m);                                             \
2446                 return (NULL);                                          \
2447         }
2448
2449         /*
2450          * Unicast.  Make sure it's not for the bridge.
2451          */
2452         do { GRAB_OUR_PACKETS(bifp) } while (0);
2453
2454         /*
2455          * Give a chance for ifp at first priority. This will help when the
2456          * packet comes through the interface like VLAN's with the same MACs
2457          * on several interfaces from the same bridge. This also will save
2458          * some CPU cycles in case the destination interface and the input
2459          * interface (eq ifp) are the same.
2460          */
2461         do { GRAB_OUR_PACKETS(ifp) } while (0);
2462
2463         /* Now check the all bridge members. */
2464         LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2465                 GRAB_OUR_PACKETS(bif2->bif_ifp)
2466         }
2467
2468 #undef OR_CARP_CHECK_WE_ARE_DST
2469 #undef OR_CARP_CHECK_WE_ARE_SRC
2470 #undef OR_PFIL_HOOKED_INET6
2471 #undef GRAB_OUR_PACKETS
2472
2473         /* Perform the bridge forwarding function. */
2474         bridge_forward(sc, bif, m);
2475
2476         return (NULL);
2477 }
2478
2479 /*
2480  * bridge_broadcast:
2481  *
2482  *      Send a frame to all interfaces that are members of
2483  *      the bridge, except for the one on which the packet
2484  *      arrived.
2485  *
2486  *      NOTE: Releases the lock on return.
2487  */
2488 static void
2489 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2490     struct mbuf *m, int runfilt)
2491 {
2492         struct bridge_iflist *dbif, *sbif;
2493         struct mbuf *mc;
2494         struct ifnet *dst_if;
2495         int error = 0, used = 0, i;
2496
2497         sbif = bridge_lookup_member_if(sc, src_if);
2498
2499         BRIDGE_LOCK2REF(sc, error);
2500         if (error) {
2501                 m_freem(m);
2502                 return;
2503         }
2504
2505         /* Filter on the bridge interface before broadcasting */
2506         if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2507 #ifdef INET6
2508             || PFIL_HOOKED(&V_inet6_pfil_hook)
2509 #endif
2510             )) {
2511                 if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2512                         goto out;
2513                 if (m == NULL)
2514                         goto out;
2515         }
2516
2517         LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2518                 dst_if = dbif->bif_ifp;
2519                 if (dst_if == src_if)
2520                         continue;
2521
2522                 /* Private segments can not talk to each other */
2523                 if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2524                         continue;
2525
2526                 if ((dbif->bif_flags & IFBIF_STP) &&
2527                     dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2528                         continue;
2529
2530                 if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2531                     (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2532                         continue;
2533
2534                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2535                         continue;
2536
2537                 if (LIST_NEXT(dbif, bif_next) == NULL) {
2538                         mc = m;
2539                         used = 1;
2540                 } else {
2541                         mc = m_dup(m, M_NOWAIT);
2542                         if (mc == NULL) {
2543                                 if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2544                                 continue;
2545                         }
2546                 }
2547
2548                 /*
2549                  * Filter on the output interface. Pass a NULL bridge interface
2550                  * pointer so we do not redundantly filter on the bridge for
2551                  * each interface we broadcast on.
2552                  */
2553                 if (runfilt && (PFIL_HOOKED(&V_inet_pfil_hook)
2554 #ifdef INET6
2555                     || PFIL_HOOKED(&V_inet6_pfil_hook)
2556 #endif
2557                     )) {
2558                         if (used == 0) {
2559                                 /* Keep the layer3 header aligned */
2560                                 i = min(mc->m_pkthdr.len, max_protohdr);
2561                                 mc = m_copyup(mc, i, ETHER_ALIGN);
2562                                 if (mc == NULL) {
2563                                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2564                                         continue;
2565                                 }
2566                         }
2567                         if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2568                                 continue;
2569                         if (mc == NULL)
2570                                 continue;
2571                 }
2572
2573                 bridge_enqueue(sc, dst_if, mc);
2574         }
2575         if (used == 0)
2576                 m_freem(m);
2577
2578 out:
2579         BRIDGE_UNREF(sc);
2580 }
2581
2582 /*
2583  * bridge_span:
2584  *
2585  *      Duplicate a packet out one or more interfaces that are in span mode,
2586  *      the original mbuf is unmodified.
2587  */
2588 static void
2589 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2590 {
2591         struct bridge_iflist *bif;
2592         struct ifnet *dst_if;
2593         struct mbuf *mc;
2594
2595         if (LIST_EMPTY(&sc->sc_spanlist))
2596                 return;
2597
2598         LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2599                 dst_if = bif->bif_ifp;
2600
2601                 if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2602                         continue;
2603
2604                 mc = m_copypacket(m, M_NOWAIT);
2605                 if (mc == NULL) {
2606                         if_inc_counter(sc->sc_ifp, IFCOUNTER_OERRORS, 1);
2607                         continue;
2608                 }
2609
2610                 bridge_enqueue(sc, dst_if, mc);
2611         }
2612 }
2613
2614 /*
2615  * bridge_rtupdate:
2616  *
2617  *      Add a bridge routing entry.
2618  */
2619 static int
2620 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2621     struct bridge_iflist *bif, int setflags, uint8_t flags)
2622 {
2623         struct bridge_rtnode *brt;
2624         int error;
2625
2626         BRIDGE_LOCK_ASSERT(sc);
2627
2628         /* Check the source address is valid and not multicast. */
2629         if (ETHER_IS_MULTICAST(dst) ||
2630             (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2631              dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2632                 return (EINVAL);
2633
2634         /* 802.1p frames map to vlan 1 */
2635         if (vlan == 0)
2636                 vlan = 1;
2637
2638         /*
2639          * A route for this destination might already exist.  If so,
2640          * update it, otherwise create a new one.
2641          */
2642         if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2643                 if (sc->sc_brtcnt >= sc->sc_brtmax) {
2644                         sc->sc_brtexceeded++;
2645                         return (ENOSPC);
2646                 }
2647                 /* Check per interface address limits (if enabled) */
2648                 if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2649                         bif->bif_addrexceeded++;
2650                         return (ENOSPC);
2651                 }
2652
2653                 /*
2654                  * Allocate a new bridge forwarding node, and
2655                  * initialize the expiration time and Ethernet
2656                  * address.
2657                  */
2658                 brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2659                 if (brt == NULL)
2660                         return (ENOMEM);
2661
2662                 if (bif->bif_flags & IFBIF_STICKY)
2663                         brt->brt_flags = IFBAF_STICKY;
2664                 else
2665                         brt->brt_flags = IFBAF_DYNAMIC;
2666
2667                 memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2668                 brt->brt_vlan = vlan;
2669
2670                 if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2671                         uma_zfree(bridge_rtnode_zone, brt);
2672                         return (error);
2673                 }
2674                 brt->brt_dst = bif;
2675                 bif->bif_addrcnt++;
2676         }
2677
2678         if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2679             brt->brt_dst != bif) {
2680                 brt->brt_dst->bif_addrcnt--;
2681                 brt->brt_dst = bif;
2682                 brt->brt_dst->bif_addrcnt++;
2683         }
2684
2685         if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2686                 brt->brt_expire = time_uptime + sc->sc_brttimeout;
2687         if (setflags)
2688                 brt->brt_flags = flags;
2689
2690         return (0);
2691 }
2692
2693 /*
2694  * bridge_rtlookup:
2695  *
2696  *      Lookup the destination interface for an address.
2697  */
2698 static struct ifnet *
2699 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2700 {
2701         struct bridge_rtnode *brt;
2702
2703         BRIDGE_LOCK_ASSERT(sc);
2704
2705         if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2706                 return (NULL);
2707
2708         return (brt->brt_ifp);
2709 }
2710
2711 /*
2712  * bridge_rttrim:
2713  *
2714  *      Trim the routine table so that we have a number
2715  *      of routing entries less than or equal to the
2716  *      maximum number.
2717  */
2718 static void
2719 bridge_rttrim(struct bridge_softc *sc)
2720 {
2721         struct bridge_rtnode *brt, *nbrt;
2722
2723         BRIDGE_LOCK_ASSERT(sc);
2724
2725         /* Make sure we actually need to do this. */
2726         if (sc->sc_brtcnt <= sc->sc_brtmax)
2727                 return;
2728
2729         /* Force an aging cycle; this might trim enough addresses. */
2730         bridge_rtage(sc);
2731         if (sc->sc_brtcnt <= sc->sc_brtmax)
2732                 return;
2733
2734         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2735                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2736                         bridge_rtnode_destroy(sc, brt);
2737                         if (sc->sc_brtcnt <= sc->sc_brtmax)
2738                                 return;
2739                 }
2740         }
2741 }
2742
2743 /*
2744  * bridge_timer:
2745  *
2746  *      Aging timer for the bridge.
2747  */
2748 static void
2749 bridge_timer(void *arg)
2750 {
2751         struct bridge_softc *sc = arg;
2752
2753         BRIDGE_LOCK_ASSERT(sc);
2754
2755         bridge_rtage(sc);
2756
2757         if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2758                 callout_reset(&sc->sc_brcallout,
2759                     bridge_rtable_prune_period * hz, bridge_timer, sc);
2760 }
2761
2762 /*
2763  * bridge_rtage:
2764  *
2765  *      Perform an aging cycle.
2766  */
2767 static void
2768 bridge_rtage(struct bridge_softc *sc)
2769 {
2770         struct bridge_rtnode *brt, *nbrt;
2771
2772         BRIDGE_LOCK_ASSERT(sc);
2773
2774         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2775                 if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2776                         if (time_uptime >= brt->brt_expire)
2777                                 bridge_rtnode_destroy(sc, brt);
2778                 }
2779         }
2780 }
2781
2782 /*
2783  * bridge_rtflush:
2784  *
2785  *      Remove all dynamic addresses from the bridge.
2786  */
2787 static void
2788 bridge_rtflush(struct bridge_softc *sc, int full)
2789 {
2790         struct bridge_rtnode *brt, *nbrt;
2791
2792         BRIDGE_LOCK_ASSERT(sc);
2793
2794         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2795                 if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2796                         bridge_rtnode_destroy(sc, brt);
2797         }
2798 }
2799
2800 /*
2801  * bridge_rtdaddr:
2802  *
2803  *      Remove an address from the table.
2804  */
2805 static int
2806 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2807 {
2808         struct bridge_rtnode *brt;
2809         int found = 0;
2810
2811         BRIDGE_LOCK_ASSERT(sc);
2812
2813         /*
2814          * If vlan is zero then we want to delete for all vlans so the lookup
2815          * may return more than one.
2816          */
2817         while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2818                 bridge_rtnode_destroy(sc, brt);
2819                 found = 1;
2820         }
2821
2822         return (found ? 0 : ENOENT);
2823 }
2824
2825 /*
2826  * bridge_rtdelete:
2827  *
2828  *      Delete routes to a speicifc member interface.
2829  */
2830 static void
2831 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2832 {
2833         struct bridge_rtnode *brt, *nbrt;
2834
2835         BRIDGE_LOCK_ASSERT(sc);
2836
2837         LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2838                 if (brt->brt_ifp == ifp && (full ||
2839                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2840                         bridge_rtnode_destroy(sc, brt);
2841         }
2842 }
2843
2844 /*
2845  * bridge_rtable_init:
2846  *
2847  *      Initialize the route table for this bridge.
2848  */
2849 static void
2850 bridge_rtable_init(struct bridge_softc *sc)
2851 {
2852         int i;
2853
2854         sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2855             M_DEVBUF, M_WAITOK);
2856
2857         for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2858                 LIST_INIT(&sc->sc_rthash[i]);
2859
2860         sc->sc_rthash_key = arc4random();
2861         LIST_INIT(&sc->sc_rtlist);
2862 }
2863
2864 /*
2865  * bridge_rtable_fini:
2866  *
2867  *      Deconstruct the route table for this bridge.
2868  */
2869 static void
2870 bridge_rtable_fini(struct bridge_softc *sc)
2871 {
2872
2873         KASSERT(sc->sc_brtcnt == 0,
2874             ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2875         free(sc->sc_rthash, M_DEVBUF);
2876 }
2877
2878 /*
2879  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2880  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2881  */
2882 #define mix(a, b, c)                                                    \
2883 do {                                                                    \
2884         a -= b; a -= c; a ^= (c >> 13);                                 \
2885         b -= c; b -= a; b ^= (a << 8);                                  \
2886         c -= a; c -= b; c ^= (b >> 13);                                 \
2887         a -= b; a -= c; a ^= (c >> 12);                                 \
2888         b -= c; b -= a; b ^= (a << 16);                                 \
2889         c -= a; c -= b; c ^= (b >> 5);                                  \
2890         a -= b; a -= c; a ^= (c >> 3);                                  \
2891         b -= c; b -= a; b ^= (a << 10);                                 \
2892         c -= a; c -= b; c ^= (b >> 15);                                 \
2893 } while (/*CONSTCOND*/0)
2894
2895 static __inline uint32_t
2896 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2897 {
2898         uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2899
2900         b += addr[5] << 8;
2901         b += addr[4];
2902         a += addr[3] << 24;
2903         a += addr[2] << 16;
2904         a += addr[1] << 8;
2905         a += addr[0];
2906
2907         mix(a, b, c);
2908
2909         return (c & BRIDGE_RTHASH_MASK);
2910 }
2911
2912 #undef mix
2913
2914 static int
2915 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2916 {
2917         int i, d;
2918
2919         for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2920                 d = ((int)a[i]) - ((int)b[i]);
2921         }
2922
2923         return (d);
2924 }
2925
2926 /*
2927  * bridge_rtnode_lookup:
2928  *
2929  *      Look up a bridge route node for the specified destination. Compare the
2930  *      vlan id or if zero then just return the first match.
2931  */
2932 static struct bridge_rtnode *
2933 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2934 {
2935         struct bridge_rtnode *brt;
2936         uint32_t hash;
2937         int dir;
2938
2939         BRIDGE_LOCK_ASSERT(sc);
2940
2941         hash = bridge_rthash(sc, addr);
2942         LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2943                 dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2944                 if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2945                         return (brt);
2946                 if (dir > 0)
2947                         return (NULL);
2948         }
2949
2950         return (NULL);
2951 }
2952
2953 /*
2954  * bridge_rtnode_insert:
2955  *
2956  *      Insert the specified bridge node into the route table.  We
2957  *      assume the entry is not already in the table.
2958  */
2959 static int
2960 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2961 {
2962         struct bridge_rtnode *lbrt;
2963         uint32_t hash;
2964         int dir;
2965
2966         BRIDGE_LOCK_ASSERT(sc);
2967
2968         hash = bridge_rthash(sc, brt->brt_addr);
2969
2970         lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2971         if (lbrt == NULL) {
2972                 LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2973                 goto out;
2974         }
2975
2976         do {
2977                 dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2978                 if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2979                         return (EEXIST);
2980                 if (dir > 0) {
2981                         LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2982                         goto out;
2983                 }
2984                 if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2985                         LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2986                         goto out;
2987                 }
2988                 lbrt = LIST_NEXT(lbrt, brt_hash);
2989         } while (lbrt != NULL);
2990
2991 #ifdef DIAGNOSTIC
2992         panic("bridge_rtnode_insert: impossible");
2993 #endif
2994
2995 out:
2996         LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2997         sc->sc_brtcnt++;
2998
2999         return (0);
3000 }
3001
3002 /*
3003  * bridge_rtnode_destroy:
3004  *
3005  *      Destroy a bridge rtnode.
3006  */
3007 static void
3008 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
3009 {
3010         BRIDGE_LOCK_ASSERT(sc);
3011
3012         LIST_REMOVE(brt, brt_hash);
3013
3014         LIST_REMOVE(brt, brt_list);
3015         sc->sc_brtcnt--;
3016         brt->brt_dst->bif_addrcnt--;
3017         uma_zfree(bridge_rtnode_zone, brt);
3018 }
3019
3020 /*
3021  * bridge_rtable_expire:
3022  *
3023  *      Set the expiry time for all routes on an interface.
3024  */
3025 static void
3026 bridge_rtable_expire(struct ifnet *ifp, int age)
3027 {
3028         struct bridge_softc *sc = ifp->if_bridge;
3029         struct bridge_rtnode *brt;
3030
3031         BRIDGE_LOCK(sc);
3032
3033         /*
3034          * If the age is zero then flush, otherwise set all the expiry times to
3035          * age for the interface
3036          */
3037         if (age == 0)
3038                 bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
3039         else {
3040                 LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
3041                         /* Cap the expiry time to 'age' */
3042                         if (brt->brt_ifp == ifp &&
3043                             brt->brt_expire > time_uptime + age &&
3044                             (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
3045                                 brt->brt_expire = time_uptime + age;
3046                 }
3047         }
3048         BRIDGE_UNLOCK(sc);
3049 }
3050
3051 /*
3052  * bridge_state_change:
3053  *
3054  *      Callback from the bridgestp code when a port changes states.
3055  */
3056 static void
3057 bridge_state_change(struct ifnet *ifp, int state)
3058 {
3059         struct bridge_softc *sc = ifp->if_bridge;
3060         static const char *stpstates[] = {
3061                 "disabled",
3062                 "listening",
3063                 "learning",
3064                 "forwarding",
3065                 "blocking",
3066                 "discarding"
3067         };
3068
3069         CURVNET_SET(ifp->if_vnet);
3070         if (V_log_stp)
3071                 log(LOG_NOTICE, "%s: state changed to %s on %s\n",
3072                     sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
3073         CURVNET_RESTORE();
3074 }
3075
3076 /*
3077  * Send bridge packets through pfil if they are one of the types pfil can deal
3078  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
3079  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
3080  * that interface.
3081  */
3082 static int
3083 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
3084 {
3085         int snap, error, i, hlen;
3086         struct ether_header *eh1, eh2;
3087         struct ip *ip;
3088         struct llc llc1;
3089         u_int16_t ether_type;
3090
3091         snap = 0;
3092         error = -1;     /* Default error if not error == 0 */
3093
3094 #if 0
3095         /* we may return with the IP fields swapped, ensure its not shared */
3096         KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
3097 #endif
3098
3099         if (V_pfil_bridge == 0 && V_pfil_member == 0 && V_pfil_ipfw == 0)
3100                 return (0); /* filtering is disabled */
3101
3102         i = min((*mp)->m_pkthdr.len, max_protohdr);
3103         if ((*mp)->m_len < i) {
3104             *mp = m_pullup(*mp, i);
3105             if (*mp == NULL) {
3106                 printf("%s: m_pullup failed\n", __func__);
3107                 return (-1);
3108             }
3109         }
3110
3111         eh1 = mtod(*mp, struct ether_header *);
3112         ether_type = ntohs(eh1->ether_type);
3113
3114         /*
3115          * Check for SNAP/LLC.
3116          */
3117         if (ether_type < ETHERMTU) {
3118                 struct llc *llc2 = (struct llc *)(eh1 + 1);
3119
3120                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
3121                     llc2->llc_dsap == LLC_SNAP_LSAP &&
3122                     llc2->llc_ssap == LLC_SNAP_LSAP &&
3123                     llc2->llc_control == LLC_UI) {
3124                         ether_type = htons(llc2->llc_un.type_snap.ether_type);
3125                         snap = 1;
3126                 }
3127         }
3128
3129         /*
3130          * If we're trying to filter bridge traffic, don't look at anything
3131          * other than IP and ARP traffic.  If the filter doesn't understand
3132          * IPv6, don't allow IPv6 through the bridge either.  This is lame
3133          * since if we really wanted, say, an AppleTalk filter, we are hosed,
3134          * but of course we don't have an AppleTalk filter to begin with.
3135          * (Note that since pfil doesn't understand ARP it will pass *ALL*
3136          * ARP traffic.)
3137          */
3138         switch (ether_type) {
3139                 case ETHERTYPE_ARP:
3140                 case ETHERTYPE_REVARP:
3141                         if (V_pfil_ipfw_arp == 0)
3142                                 return (0); /* Automatically pass */
3143                         break;
3144
3145                 case ETHERTYPE_IP:
3146 #ifdef INET6
3147                 case ETHERTYPE_IPV6:
3148 #endif /* INET6 */
3149                         break;
3150                 default:
3151                         /*
3152                          * Check to see if the user wants to pass non-ip
3153                          * packets, these will not be checked by pfil(9) and
3154                          * passed unconditionally so the default is to drop.
3155                          */
3156                         if (V_pfil_onlyip)
3157                                 goto bad;
3158         }
3159
3160         /* Run the packet through pfil before stripping link headers */
3161         if (PFIL_HOOKED(&V_link_pfil_hook) && V_pfil_ipfw != 0 &&
3162                         dir == PFIL_OUT && ifp != NULL) {
3163
3164                 error = pfil_run_hooks(&V_link_pfil_hook, mp, ifp, dir, NULL);
3165
3166                 if (*mp == NULL || error != 0) /* packet consumed by filter */
3167                         return (error);
3168         }
3169
3170         /* Strip off the Ethernet header and keep a copy. */
3171         m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3172         m_adj(*mp, ETHER_HDR_LEN);
3173
3174         /* Strip off snap header, if present */
3175         if (snap) {
3176                 m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3177                 m_adj(*mp, sizeof(struct llc));
3178         }
3179
3180         /*
3181          * Check the IP header for alignment and errors
3182          */
3183         if (dir == PFIL_IN) {
3184                 switch (ether_type) {
3185                         case ETHERTYPE_IP:
3186                                 error = bridge_ip_checkbasic(mp);
3187                                 break;
3188 #ifdef INET6
3189                         case ETHERTYPE_IPV6:
3190                                 error = bridge_ip6_checkbasic(mp);
3191                                 break;
3192 #endif /* INET6 */
3193                         default:
3194                                 error = 0;
3195                 }
3196                 if (error)
3197                         goto bad;
3198         }
3199
3200         error = 0;
3201
3202         /*
3203          * Run the packet through pfil
3204          */
3205         switch (ether_type) {
3206         case ETHERTYPE_IP:
3207                 /*
3208                  * Run pfil on the member interface and the bridge, both can
3209                  * be skipped by clearing pfil_member or pfil_bridge.
3210                  *
3211                  * Keep the order:
3212                  *   in_if -> bridge_if -> out_if
3213                  */
3214                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3215                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3216                                         dir, NULL);
3217
3218                 if (*mp == NULL || error != 0) /* filter may consume */
3219                         break;
3220
3221                 if (V_pfil_member && ifp != NULL)
3222                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, ifp,
3223                                         dir, NULL);
3224
3225                 if (*mp == NULL || error != 0) /* filter may consume */
3226                         break;
3227
3228                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3229                         error = pfil_run_hooks(&V_inet_pfil_hook, mp, bifp,
3230                                         dir, NULL);
3231
3232                 if (*mp == NULL || error != 0) /* filter may consume */
3233                         break;
3234
3235                 /* check if we need to fragment the packet */
3236                 if (V_pfil_member && ifp != NULL && dir == PFIL_OUT) {
3237                         i = (*mp)->m_pkthdr.len;
3238                         if (i > ifp->if_mtu) {
3239                                 error = bridge_fragment(ifp, *mp, &eh2, snap,
3240                                             &llc1);
3241                                 return (error);
3242                         }
3243                 }
3244
3245                 /* Recalculate the ip checksum. */
3246                 ip = mtod(*mp, struct ip *);
3247                 hlen = ip->ip_hl << 2;
3248                 if (hlen < sizeof(struct ip))
3249                         goto bad;
3250                 if (hlen > (*mp)->m_len) {
3251                         if ((*mp = m_pullup(*mp, hlen)) == 0)
3252                                 goto bad;
3253                         ip = mtod(*mp, struct ip *);
3254                         if (ip == NULL)
3255                                 goto bad;
3256                 }
3257                 ip->ip_sum = 0;
3258                 if (hlen == sizeof(struct ip))
3259                         ip->ip_sum = in_cksum_hdr(ip);
3260                 else
3261                         ip->ip_sum = in_cksum(*mp, hlen);
3262
3263                 break;
3264 #ifdef INET6
3265         case ETHERTYPE_IPV6:
3266                 if (V_pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3267                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3268                                         dir, NULL);
3269
3270                 if (*mp == NULL || error != 0) /* filter may consume */
3271                         break;
3272
3273                 if (V_pfil_member && ifp != NULL)
3274                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, ifp,
3275                                         dir, NULL);
3276
3277                 if (*mp == NULL || error != 0) /* filter may consume */
3278                         break;
3279
3280                 if (V_pfil_bridge && dir == PFIL_IN && bifp != NULL)
3281                         error = pfil_run_hooks(&V_inet6_pfil_hook, mp, bifp,
3282                                         dir, NULL);
3283                 break;
3284 #endif
3285         default:
3286                 error = 0;
3287                 break;
3288         }
3289
3290         if (*mp == NULL)
3291                 return (error);
3292         if (error != 0)
3293                 goto bad;
3294
3295         error = -1;
3296
3297         /*
3298          * Finally, put everything back the way it was and return
3299          */
3300         if (snap) {
3301                 M_PREPEND(*mp, sizeof(struct llc), M_NOWAIT);
3302                 if (*mp == NULL)
3303                         return (error);
3304                 bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3305         }
3306
3307         M_PREPEND(*mp, ETHER_HDR_LEN, M_NOWAIT);
3308         if (*mp == NULL)
3309                 return (error);
3310         bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3311
3312         return (0);
3313
3314 bad:
3315         m_freem(*mp);
3316         *mp = NULL;
3317         return (error);
3318 }
3319
3320 /*
3321  * Perform basic checks on header size since
3322  * pfil assumes ip_input has already processed
3323  * it for it.  Cut-and-pasted from ip_input.c.
3324  * Given how simple the IPv6 version is,
3325  * does the IPv4 version really need to be
3326  * this complicated?
3327  *
3328  * XXX Should we update ipstat here, or not?
3329  * XXX Right now we update ipstat but not
3330  * XXX csum_counter.
3331  */
3332 static int
3333 bridge_ip_checkbasic(struct mbuf **mp)
3334 {
3335         struct mbuf *m = *mp;
3336         struct ip *ip;
3337         int len, hlen;
3338         u_short sum;
3339
3340         if (*mp == NULL)
3341                 return (-1);
3342
3343         if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3344                 if ((m = m_copyup(m, sizeof(struct ip),
3345                         (max_linkhdr + 3) & ~3)) == NULL) {
3346                         /* XXXJRT new stat, please */
3347                         KMOD_IPSTAT_INC(ips_toosmall);
3348                         goto bad;
3349                 }
3350         } else if (__predict_false(m->m_len < sizeof (struct ip))) {
3351                 if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3352                         KMOD_IPSTAT_INC(ips_toosmall);
3353                         goto bad;
3354                 }
3355         }
3356         ip = mtod(m, struct ip *);
3357         if (ip == NULL) goto bad;
3358
3359         if (ip->ip_v != IPVERSION) {
3360                 KMOD_IPSTAT_INC(ips_badvers);
3361                 goto bad;
3362         }
3363         hlen = ip->ip_hl << 2;
3364         if (hlen < sizeof(struct ip)) { /* minimum header length */
3365                 KMOD_IPSTAT_INC(ips_badhlen);
3366                 goto bad;
3367         }
3368         if (hlen > m->m_len) {
3369                 if ((m = m_pullup(m, hlen)) == 0) {
3370                         KMOD_IPSTAT_INC(ips_badhlen);
3371                         goto bad;
3372                 }
3373                 ip = mtod(m, struct ip *);
3374                 if (ip == NULL) goto bad;
3375         }
3376
3377         if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3378                 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3379         } else {
3380                 if (hlen == sizeof(struct ip)) {
3381                         sum = in_cksum_hdr(ip);
3382                 } else {
3383                         sum = in_cksum(m, hlen);
3384                 }
3385         }
3386         if (sum) {
3387                 KMOD_IPSTAT_INC(ips_badsum);
3388                 goto bad;
3389         }
3390
3391         /* Retrieve the packet length. */
3392         len = ntohs(ip->ip_len);
3393
3394         /*
3395          * Check for additional length bogosity
3396          */
3397         if (len < hlen) {
3398                 KMOD_IPSTAT_INC(ips_badlen);
3399                 goto bad;
3400         }
3401
3402         /*
3403          * Check that the amount of data in the buffers
3404          * is as at least much as the IP header would have us expect.
3405          * Drop packet if shorter than we expect.
3406          */
3407         if (m->m_pkthdr.len < len) {
3408                 KMOD_IPSTAT_INC(ips_tooshort);
3409                 goto bad;
3410         }
3411
3412         /* Checks out, proceed */
3413         *mp = m;
3414         return (0);
3415
3416 bad:
3417         *mp = m;
3418         return (-1);
3419 }
3420
3421 #ifdef INET6
3422 /*
3423  * Same as above, but for IPv6.
3424  * Cut-and-pasted from ip6_input.c.
3425  * XXX Should we update ip6stat, or not?
3426  */
3427 static int
3428 bridge_ip6_checkbasic(struct mbuf **mp)
3429 {
3430         struct mbuf *m = *mp;
3431         struct ip6_hdr *ip6;
3432
3433         /*
3434          * If the IPv6 header is not aligned, slurp it up into a new
3435          * mbuf with space for link headers, in the event we forward
3436          * it.  Otherwise, if it is aligned, make sure the entire base
3437          * IPv6 header is in the first mbuf of the chain.
3438          */
3439         if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3440                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3441                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3442                             (max_linkhdr + 3) & ~3)) == NULL) {
3443                         /* XXXJRT new stat, please */
3444                         IP6STAT_INC(ip6s_toosmall);
3445                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3446                         goto bad;
3447                 }
3448         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3449                 struct ifnet *inifp = m->m_pkthdr.rcvif;
3450                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3451                         IP6STAT_INC(ip6s_toosmall);
3452                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3453                         goto bad;
3454                 }
3455         }
3456
3457         ip6 = mtod(m, struct ip6_hdr *);
3458
3459         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3460                 IP6STAT_INC(ip6s_badvers);
3461                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3462                 goto bad;
3463         }
3464
3465         /* Checks out, proceed */
3466         *mp = m;
3467         return (0);
3468
3469 bad:
3470         *mp = m;
3471         return (-1);
3472 }
3473 #endif /* INET6 */
3474
3475 /*
3476  * bridge_fragment:
3477  *
3478  *      Return a fragmented mbuf chain.
3479  */
3480 static int
3481 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3482     int snap, struct llc *llc)
3483 {
3484         struct mbuf *m0;
3485         struct ip *ip;
3486         int error = -1;
3487
3488         if (m->m_len < sizeof(struct ip) &&
3489             (m = m_pullup(m, sizeof(struct ip))) == NULL)
3490                 goto out;
3491         ip = mtod(m, struct ip *);
3492
3493         m->m_pkthdr.csum_flags |= CSUM_IP;
3494         error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist);
3495         if (error)
3496                 goto out;
3497
3498         /* walk the chain and re-add the Ethernet header */
3499         for (m0 = m; m0; m0 = m0->m_nextpkt) {
3500                 if (error == 0) {
3501                         if (snap) {
3502                                 M_PREPEND(m0, sizeof(struct llc), M_NOWAIT);
3503                                 if (m0 == NULL) {
3504                                         error = ENOBUFS;
3505                                         continue;
3506                                 }
3507                                 bcopy(llc, mtod(m0, caddr_t),
3508                                     sizeof(struct llc));
3509                         }
3510                         M_PREPEND(m0, ETHER_HDR_LEN, M_NOWAIT);
3511                         if (m0 == NULL) {
3512                                 error = ENOBUFS;
3513                                 continue;
3514                         }
3515                         bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3516                 } else
3517                         m_freem(m);
3518         }
3519
3520         if (error == 0)
3521                 KMOD_IPSTAT_INC(ips_fragmented);
3522
3523         return (error);
3524
3525 out:
3526         if (m != NULL)
3527                 m_freem(m);
3528         return (error);
3529 }
3530
3531 static void
3532 bridge_linkstate(struct ifnet *ifp)
3533 {
3534         struct bridge_softc *sc = ifp->if_bridge;
3535         struct bridge_iflist *bif;
3536
3537         BRIDGE_LOCK(sc);
3538         bif = bridge_lookup_member_if(sc, ifp);
3539         if (bif == NULL) {
3540                 BRIDGE_UNLOCK(sc);
3541                 return;
3542         }
3543         bridge_linkcheck(sc);
3544         BRIDGE_UNLOCK(sc);
3545
3546         bstp_linkstate(&bif->bif_stp);
3547 }
3548
3549 static void
3550 bridge_linkcheck(struct bridge_softc *sc)
3551 {
3552         struct bridge_iflist *bif;
3553         int new_link, hasls;
3554
3555         BRIDGE_LOCK_ASSERT(sc);
3556         new_link = LINK_STATE_DOWN;
3557         hasls = 0;
3558         /* Our link is considered up if at least one of our ports is active */
3559         LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
3560                 if (bif->bif_ifp->if_capabilities & IFCAP_LINKSTATE)
3561                         hasls++;
3562                 if (bif->bif_ifp->if_link_state == LINK_STATE_UP) {
3563                         new_link = LINK_STATE_UP;
3564                         break;
3565                 }
3566         }
3567         if (!LIST_EMPTY(&sc->sc_iflist) && !hasls) {
3568                 /* If no interfaces support link-state then we default to up */
3569                 new_link = LINK_STATE_UP;
3570         }
3571         if_link_state_change(sc->sc_ifp, new_link);
3572 }