Home | History | Annotate | Line # | Download | only in net
if_bridge.c revision 1.83
      1 /*	$NetBSD: if_bridge.c,v 1.83 2014/06/18 10:51:03 ozaki-r 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 (at) 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  * 3. All advertising materials mentioning features or use of this software
     51  *    must display the following acknowledgement:
     52  *	This product includes software developed by Jason L. Wright
     53  * 4. The name of the author may not be used to endorse or promote products
     54  *    derived from this software without specific prior written permission.
     55  *
     56  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
     57  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
     58  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
     59  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
     60  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
     61  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
     62  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
     64  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
     65  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     66  * POSSIBILITY OF SUCH DAMAGE.
     67  *
     68  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
     69  */
     70 
     71 /*
     72  * Network interface bridge support.
     73  *
     74  * TODO:
     75  *
     76  *	- Currently only supports Ethernet-like interfaces (Ethernet,
     77  *	  802.11, VLANs on Ethernet, etc.)  Figure out a nice way
     78  *	  to bridge other types of interfaces (FDDI-FDDI, and maybe
     79  *	  consider heterogenous bridges).
     80  */
     81 
     82 #include <sys/cdefs.h>
     83 __KERNEL_RCSID(0, "$NetBSD: if_bridge.c,v 1.83 2014/06/18 10:51:03 ozaki-r Exp $");
     84 
     85 #ifdef _KERNEL_OPT
     86 #include "opt_bridge_ipf.h"
     87 #include "opt_inet.h"
     88 #endif /* _KERNEL_OPT */
     89 
     90 #include <sys/param.h>
     91 #include <sys/kernel.h>
     92 #include <sys/mbuf.h>
     93 #include <sys/queue.h>
     94 #include <sys/socket.h>
     95 #include <sys/socketvar.h> /* for softnet_lock */
     96 #include <sys/sockio.h>
     97 #include <sys/systm.h>
     98 #include <sys/proc.h>
     99 #include <sys/pool.h>
    100 #include <sys/kauth.h>
    101 #include <sys/cpu.h>
    102 #include <sys/cprng.h>
    103 #include <sys/xcall.h>
    104 
    105 #include <net/bpf.h>
    106 #include <net/if.h>
    107 #include <net/if_dl.h>
    108 #include <net/if_types.h>
    109 #include <net/if_llc.h>
    110 #include <net/pktqueue.h>
    111 
    112 #include <net/if_ether.h>
    113 #include <net/if_bridgevar.h>
    114 
    115 #if defined(BRIDGE_IPF)
    116 /* Used for bridge_ip[6]_checkbasic */
    117 #include <netinet/in.h>
    118 #include <netinet/in_systm.h>
    119 #include <netinet/ip.h>
    120 #include <netinet/ip_var.h>
    121 #include <netinet/ip_private.h>		/* XXX */
    122 
    123 #include <netinet/ip6.h>
    124 #include <netinet6/in6_var.h>
    125 #include <netinet6/ip6_var.h>
    126 #include <netinet6/ip6_private.h>	/* XXX */
    127 #endif /* BRIDGE_IPF */
    128 
    129 /*
    130  * Size of the route hash table.  Must be a power of two.
    131  */
    132 #ifndef BRIDGE_RTHASH_SIZE
    133 #define	BRIDGE_RTHASH_SIZE		1024
    134 #endif
    135 
    136 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
    137 
    138 #include "carp.h"
    139 #if NCARP > 0
    140 #include <netinet/in.h>
    141 #include <netinet/in_var.h>
    142 #include <netinet/ip_carp.h>
    143 #endif
    144 
    145 /*
    146  * Maximum number of addresses to cache.
    147  */
    148 #ifndef BRIDGE_RTABLE_MAX
    149 #define	BRIDGE_RTABLE_MAX		100
    150 #endif
    151 
    152 /*
    153  * Spanning tree defaults.
    154  */
    155 #define	BSTP_DEFAULT_MAX_AGE		(20 * 256)
    156 #define	BSTP_DEFAULT_HELLO_TIME		(2 * 256)
    157 #define	BSTP_DEFAULT_FORWARD_DELAY	(15 * 256)
    158 #define	BSTP_DEFAULT_HOLD_TIME		(1 * 256)
    159 #define	BSTP_DEFAULT_BRIDGE_PRIORITY	0x8000
    160 #define	BSTP_DEFAULT_PORT_PRIORITY	0x80
    161 #define	BSTP_DEFAULT_PATH_COST		55
    162 
    163 /*
    164  * Timeout (in seconds) for entries learned dynamically.
    165  */
    166 #ifndef BRIDGE_RTABLE_TIMEOUT
    167 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
    168 #endif
    169 
    170 /*
    171  * Number of seconds between walks of the route list.
    172  */
    173 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
    174 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
    175 #endif
    176 
    177 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
    178 
    179 static struct pool bridge_rtnode_pool;
    180 
    181 void	bridgeattach(int);
    182 
    183 static int	bridge_clone_create(struct if_clone *, int);
    184 static int	bridge_clone_destroy(struct ifnet *);
    185 
    186 static int	bridge_ioctl(struct ifnet *, u_long, void *);
    187 static int	bridge_init(struct ifnet *);
    188 static void	bridge_stop(struct ifnet *, int);
    189 static void	bridge_start(struct ifnet *);
    190 
    191 static void	bridge_input(struct ifnet *, struct mbuf *);
    192 static void	bridge_forward(void *);
    193 
    194 static void	bridge_timer(void *);
    195 
    196 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
    197 				 struct mbuf *);
    198 
    199 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
    200 				struct ifnet *, int, uint8_t);
    201 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
    202 static void	bridge_rttrim(struct bridge_softc *);
    203 static void	bridge_rtage(struct bridge_softc *);
    204 static void	bridge_rtflush(struct bridge_softc *, int);
    205 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
    206 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp);
    207 
    208 static int	bridge_rtable_init(struct bridge_softc *);
    209 static void	bridge_rtable_fini(struct bridge_softc *);
    210 
    211 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
    212 						  const uint8_t *);
    213 static int	bridge_rtnode_insert(struct bridge_softc *,
    214 				     struct bridge_rtnode *);
    215 static void	bridge_rtnode_destroy(struct bridge_softc *,
    216 				      struct bridge_rtnode *);
    217 
    218 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
    219 						  const char *name);
    220 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
    221 						     struct ifnet *ifp);
    222 static void	bridge_delete_member(struct bridge_softc *,
    223 				     struct bridge_iflist *);
    224 
    225 static int	bridge_ioctl_add(struct bridge_softc *, void *);
    226 static int	bridge_ioctl_del(struct bridge_softc *, void *);
    227 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
    228 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
    229 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
    230 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
    231 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
    232 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
    233 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
    234 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
    235 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
    236 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
    237 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
    238 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
    239 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
    240 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
    241 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
    242 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
    243 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
    244 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
    245 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
    246 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
    247 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
    248 #if defined(BRIDGE_IPF)
    249 static int	bridge_ioctl_gfilt(struct bridge_softc *, void *);
    250 static int	bridge_ioctl_sfilt(struct bridge_softc *, void *);
    251 static int	bridge_ipf(void *, struct mbuf **, struct ifnet *, int);
    252 static int	bridge_ip_checkbasic(struct mbuf **mp);
    253 # ifdef INET6
    254 static int	bridge_ip6_checkbasic(struct mbuf **mp);
    255 # endif /* INET6 */
    256 #endif /* BRIDGE_IPF */
    257 
    258 static void bridge_sysctl_fwdq_setup(struct sysctllog **clog,
    259     struct bridge_softc *sc);
    260 
    261 struct bridge_control {
    262 	int	(*bc_func)(struct bridge_softc *, void *);
    263 	int	bc_argsize;
    264 	int	bc_flags;
    265 };
    266 
    267 #define	BC_F_COPYIN		0x01	/* copy arguments in */
    268 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
    269 #define	BC_F_SUSER		0x04	/* do super-user check */
    270 
    271 static const struct bridge_control bridge_control_table[] = {
    272 [BRDGADD] = {bridge_ioctl_add, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    273 [BRDGDEL] = {bridge_ioctl_del, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    274 
    275 [BRDGGIFFLGS] = {bridge_ioctl_gifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_COPYOUT},
    276 [BRDGSIFFLGS] = {bridge_ioctl_sifflags, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    277 
    278 [BRDGSCACHE] = {bridge_ioctl_scache, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    279 [BRDGGCACHE] = {bridge_ioctl_gcache, sizeof(struct ifbrparam), BC_F_COPYOUT},
    280 
    281 [BRDGGIFS] = {bridge_ioctl_gifs, sizeof(struct ifbifconf), BC_F_COPYIN|BC_F_COPYOUT},
    282 [BRDGRTS] = {bridge_ioctl_rts, sizeof(struct ifbaconf), BC_F_COPYIN|BC_F_COPYOUT},
    283 
    284 [BRDGSADDR] = {bridge_ioctl_saddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    285 
    286 [BRDGSTO] = {bridge_ioctl_sto, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    287 [BRDGGTO] = {bridge_ioctl_gto, sizeof(struct ifbrparam), BC_F_COPYOUT},
    288 
    289 [BRDGDADDR] = {bridge_ioctl_daddr, sizeof(struct ifbareq), BC_F_COPYIN|BC_F_SUSER},
    290 
    291 [BRDGFLUSH] = {bridge_ioctl_flush, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    292 
    293 [BRDGGPRI] = {bridge_ioctl_gpri, sizeof(struct ifbrparam), BC_F_COPYOUT},
    294 [BRDGSPRI] = {bridge_ioctl_spri, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    295 
    296 [BRDGGHT] = {bridge_ioctl_ght, sizeof(struct ifbrparam), BC_F_COPYOUT},
    297 [BRDGSHT] = {bridge_ioctl_sht, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    298 
    299 [BRDGGFD] = {bridge_ioctl_gfd, sizeof(struct ifbrparam), BC_F_COPYOUT},
    300 [BRDGSFD] = {bridge_ioctl_sfd, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    301 
    302 [BRDGGMA] = {bridge_ioctl_gma, sizeof(struct ifbrparam), BC_F_COPYOUT},
    303 [BRDGSMA] = {bridge_ioctl_sma, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    304 
    305 [BRDGSIFPRIO] = {bridge_ioctl_sifprio, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    306 
    307 [BRDGSIFCOST] = {bridge_ioctl_sifcost, sizeof(struct ifbreq), BC_F_COPYIN|BC_F_SUSER},
    308 #if defined(BRIDGE_IPF)
    309 [BRDGGFILT] = {bridge_ioctl_gfilt, sizeof(struct ifbrparam), BC_F_COPYOUT},
    310 [BRDGSFILT] = {bridge_ioctl_sfilt, sizeof(struct ifbrparam), BC_F_COPYIN|BC_F_SUSER},
    311 #endif /* BRIDGE_IPF */
    312 };
    313 static const int bridge_control_table_size = __arraycount(bridge_control_table);
    314 
    315 static LIST_HEAD(, bridge_softc) bridge_list;
    316 
    317 static struct if_clone bridge_cloner =
    318     IF_CLONE_INITIALIZER("bridge", bridge_clone_create, bridge_clone_destroy);
    319 
    320 /*
    321  * bridgeattach:
    322  *
    323  *	Pseudo-device attach routine.
    324  */
    325 void
    326 bridgeattach(int n)
    327 {
    328 
    329 	pool_init(&bridge_rtnode_pool, sizeof(struct bridge_rtnode),
    330 	    0, 0, 0, "brtpl", NULL, IPL_NET);
    331 
    332 	LIST_INIT(&bridge_list);
    333 	if_clone_attach(&bridge_cloner);
    334 }
    335 
    336 /*
    337  * bridge_clone_create:
    338  *
    339  *	Create a new bridge instance.
    340  */
    341 static int
    342 bridge_clone_create(struct if_clone *ifc, int unit)
    343 {
    344 	struct bridge_softc *sc;
    345 	struct ifnet *ifp;
    346 	int s;
    347 
    348 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
    349 	ifp = &sc->sc_if;
    350 
    351 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
    352 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
    353 	sc->sc_bridge_max_age = BSTP_DEFAULT_MAX_AGE;
    354 	sc->sc_bridge_hello_time = BSTP_DEFAULT_HELLO_TIME;
    355 	sc->sc_bridge_forward_delay = BSTP_DEFAULT_FORWARD_DELAY;
    356 	sc->sc_bridge_priority = BSTP_DEFAULT_BRIDGE_PRIORITY;
    357 	sc->sc_hold_time = BSTP_DEFAULT_HOLD_TIME;
    358 	sc->sc_filter_flags = 0;
    359 
    360 	/* Initialize our routing table. */
    361 	bridge_rtable_init(sc);
    362 
    363 	callout_init(&sc->sc_brcallout, 0);
    364 	callout_init(&sc->sc_bstpcallout, 0);
    365 
    366 	LIST_INIT(&sc->sc_iflist);
    367 
    368 	if_initname(ifp, ifc->ifc_name, unit);
    369 	ifp->if_softc = sc;
    370 	ifp->if_mtu = ETHERMTU;
    371 	ifp->if_ioctl = bridge_ioctl;
    372 	ifp->if_output = bridge_output;
    373 	ifp->if_start = bridge_start;
    374 	ifp->if_stop = bridge_stop;
    375 	ifp->if_init = bridge_init;
    376 	ifp->if_type = IFT_BRIDGE;
    377 	ifp->if_addrlen = 0;
    378 	ifp->if_dlt = DLT_EN10MB;
    379 	ifp->if_hdrlen = ETHER_HDR_LEN;
    380 	IFQ_SET_READY(&ifp->if_snd);
    381 
    382 	sc->sc_fwd_pktq = pktq_create(IFQ_MAXLEN, bridge_forward, sc);
    383 	KASSERT(sc->sc_fwd_pktq != NULL);
    384 
    385 	bridge_sysctl_fwdq_setup(&ifp->if_sysctl_log, sc);
    386 
    387 	if_attach(ifp);
    388 
    389 	if_alloc_sadl(ifp);
    390 
    391 	s = splnet();
    392 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
    393 	splx(s);
    394 
    395 	return (0);
    396 }
    397 
    398 /*
    399  * bridge_clone_destroy:
    400  *
    401  *	Destroy a bridge instance.
    402  */
    403 static int
    404 bridge_clone_destroy(struct ifnet *ifp)
    405 {
    406 	struct bridge_softc *sc = ifp->if_softc;
    407 	struct bridge_iflist *bif;
    408 	int s;
    409 	uint64_t xc;
    410 
    411 	/* Must be called during IFF_RUNNING, i.e., before bridge_stop */
    412 	pktq_barrier(sc->sc_fwd_pktq);
    413 	xc = xc_broadcast(0, (xcfunc_t)nullop, NULL, NULL);
    414 	xc_wait(xc);
    415 
    416 	s = splnet();
    417 
    418 	bridge_stop(ifp, 1);
    419 
    420 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
    421 		bridge_delete_member(sc, bif);
    422 
    423 	LIST_REMOVE(sc, sc_list);
    424 
    425 	splx(s);
    426 
    427 	if_detach(ifp);
    428 
    429 	/* Should be called after if_detach for safe */
    430 	pktq_flush(sc->sc_fwd_pktq);
    431 	pktq_destroy(sc->sc_fwd_pktq);
    432 
    433 	/* Tear down the routing table. */
    434 	bridge_rtable_fini(sc);
    435 
    436 	free(sc, M_DEVBUF);
    437 
    438 	return (0);
    439 }
    440 
    441 static int
    442 bridge_sysctl_fwdq_maxlen(SYSCTLFN_ARGS)
    443 {
    444 	struct sysctlnode node = *rnode;
    445 	const struct bridge_softc *sc =	node.sysctl_data;
    446 	return sysctl_pktq_maxlen(SYSCTLFN_CALL(rnode), sc->sc_fwd_pktq);
    447 }
    448 
    449 #define	SYSCTL_BRIDGE_PKTQ(cn, c)					\
    450 	static int							\
    451 	bridge_sysctl_fwdq_##cn(SYSCTLFN_ARGS)				\
    452 	{								\
    453 		struct sysctlnode node = *rnode;			\
    454 		const struct bridge_softc *sc =	node.sysctl_data;	\
    455 		return sysctl_pktq_count(SYSCTLFN_CALL(rnode),		\
    456 					 sc->sc_fwd_pktq, c);		\
    457 	}
    458 
    459 SYSCTL_BRIDGE_PKTQ(items, PKTQ_NITEMS)
    460 SYSCTL_BRIDGE_PKTQ(drops, PKTQ_DROPS)
    461 
    462 static void
    463 bridge_sysctl_fwdq_setup(struct sysctllog **clog, struct bridge_softc *sc)
    464 {
    465 	const struct sysctlnode *cnode, *rnode;
    466 	sysctlfn len_func = NULL, maxlen_func = NULL, drops_func = NULL;
    467 	const char *ifname = sc->sc_if.if_xname;
    468 
    469 	len_func = bridge_sysctl_fwdq_items;
    470 	maxlen_func = bridge_sysctl_fwdq_maxlen;
    471 	drops_func = bridge_sysctl_fwdq_drops;
    472 
    473 	if (sysctl_createv(clog, 0, NULL, &rnode,
    474 			   CTLFLAG_PERMANENT,
    475 			   CTLTYPE_NODE, "interfaces",
    476 			   SYSCTL_DESCR("Per-interface controls"),
    477 			   NULL, 0, NULL, 0,
    478 			   CTL_NET, CTL_CREATE, CTL_EOL) != 0)
    479 		goto bad;
    480 
    481 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    482 			   CTLFLAG_PERMANENT,
    483 			   CTLTYPE_NODE, ifname,
    484 			   SYSCTL_DESCR("Interface controls"),
    485 			   NULL, 0, NULL, 0,
    486 			   CTL_CREATE, CTL_EOL) != 0)
    487 		goto bad;
    488 
    489 	if (sysctl_createv(clog, 0, &rnode, &rnode,
    490 			   CTLFLAG_PERMANENT,
    491 			   CTLTYPE_NODE, "fwdq",
    492 			   SYSCTL_DESCR("Protocol input queue controls"),
    493 			   NULL, 0, NULL, 0,
    494 			   CTL_CREATE, CTL_EOL) != 0)
    495 		goto bad;
    496 
    497 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    498 			   CTLFLAG_PERMANENT,
    499 			   CTLTYPE_INT, "len",
    500 			   SYSCTL_DESCR("Current forwarding queue length"),
    501 			   len_func, 0, (void *)sc, 0,
    502 			   CTL_CREATE, IFQCTL_LEN, CTL_EOL) != 0)
    503 		goto bad;
    504 
    505 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    506 			   CTLFLAG_PERMANENT|CTLFLAG_READWRITE,
    507 			   CTLTYPE_INT, "maxlen",
    508 			   SYSCTL_DESCR("Maximum allowed forwarding queue length"),
    509 			   maxlen_func, 0, (void *)sc, 0,
    510 			   CTL_CREATE, IFQCTL_MAXLEN, CTL_EOL) != 0)
    511 		goto bad;
    512 
    513 	if (sysctl_createv(clog, 0, &rnode, &cnode,
    514 			   CTLFLAG_PERMANENT,
    515 			   CTLTYPE_INT, "drops",
    516 			   SYSCTL_DESCR("Packets dropped due to full forwarding queue"),
    517 			   drops_func, 0, (void *)sc, 0,
    518 			   CTL_CREATE, IFQCTL_DROPS, CTL_EOL) != 0)
    519 		goto bad;
    520 
    521 	return;
    522 bad:
    523 	aprint_error("%s: could not attach sysctl nodes\n", ifname);
    524 	return;
    525 }
    526 
    527 /*
    528  * bridge_ioctl:
    529  *
    530  *	Handle a control request from the operator.
    531  */
    532 static int
    533 bridge_ioctl(struct ifnet *ifp, u_long cmd, void *data)
    534 {
    535 	struct bridge_softc *sc = ifp->if_softc;
    536 	struct lwp *l = curlwp;	/* XXX */
    537 	union {
    538 		struct ifbreq ifbreq;
    539 		struct ifbifconf ifbifconf;
    540 		struct ifbareq ifbareq;
    541 		struct ifbaconf ifbaconf;
    542 		struct ifbrparam ifbrparam;
    543 	} args;
    544 	struct ifdrv *ifd = (struct ifdrv *) data;
    545 	const struct bridge_control *bc = NULL; /* XXXGCC */
    546 	int s, error = 0;
    547 
    548 	/* Authorize command before calling splnet(). */
    549 	switch (cmd) {
    550 	case SIOCGDRVSPEC:
    551 	case SIOCSDRVSPEC:
    552 		if (ifd->ifd_cmd >= bridge_control_table_size) {
    553 			error = EINVAL;
    554 			return error;
    555 		}
    556 
    557 		bc = &bridge_control_table[ifd->ifd_cmd];
    558 
    559 		/* We only care about BC_F_SUSER at this point. */
    560 		if ((bc->bc_flags & BC_F_SUSER) == 0)
    561 			break;
    562 
    563 		error = kauth_authorize_network(l->l_cred,
    564 		    KAUTH_NETWORK_INTERFACE_BRIDGE,
    565 		    cmd == SIOCGDRVSPEC ?
    566 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_GETPRIV :
    567 		     KAUTH_REQ_NETWORK_INTERFACE_BRIDGE_SETPRIV,
    568 		     ifd, NULL, NULL);
    569 		if (error)
    570 			return (error);
    571 
    572 		break;
    573 	}
    574 
    575 	s = splnet();
    576 
    577 	switch (cmd) {
    578 	case SIOCGDRVSPEC:
    579 	case SIOCSDRVSPEC:
    580 		KASSERT(bc != NULL);
    581 		if (cmd == SIOCGDRVSPEC &&
    582 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
    583 			error = EINVAL;
    584 			break;
    585 		}
    586 		else if (cmd == SIOCSDRVSPEC &&
    587 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
    588 			error = EINVAL;
    589 			break;
    590 		}
    591 
    592 		/* BC_F_SUSER is checked above, before splnet(). */
    593 
    594 		if (ifd->ifd_len != bc->bc_argsize ||
    595 		    ifd->ifd_len > sizeof(args)) {
    596 			error = EINVAL;
    597 			break;
    598 		}
    599 
    600 		memset(&args, 0, sizeof(args));
    601 		if (bc->bc_flags & BC_F_COPYIN) {
    602 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
    603 			if (error)
    604 				break;
    605 		}
    606 
    607 		error = (*bc->bc_func)(sc, &args);
    608 		if (error)
    609 			break;
    610 
    611 		if (bc->bc_flags & BC_F_COPYOUT)
    612 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
    613 
    614 		break;
    615 
    616 	case SIOCSIFFLAGS:
    617 		if ((error = ifioctl_common(ifp, cmd, data)) != 0)
    618 			break;
    619 		switch (ifp->if_flags & (IFF_UP|IFF_RUNNING)) {
    620 		case IFF_RUNNING:
    621 			/*
    622 			 * If interface is marked down and it is running,
    623 			 * then stop and disable it.
    624 			 */
    625 			(*ifp->if_stop)(ifp, 1);
    626 			break;
    627 		case IFF_UP:
    628 			/*
    629 			 * If interface is marked up and it is stopped, then
    630 			 * start it.
    631 			 */
    632 			error = (*ifp->if_init)(ifp);
    633 			break;
    634 		default:
    635 			break;
    636 		}
    637 		break;
    638 
    639 	default:
    640 		error = ifioctl_common(ifp, cmd, data);
    641 		break;
    642 	}
    643 
    644 	splx(s);
    645 
    646 	return (error);
    647 }
    648 
    649 /*
    650  * bridge_lookup_member:
    651  *
    652  *	Lookup a bridge member interface.  Must be called at splnet().
    653  */
    654 static struct bridge_iflist *
    655 bridge_lookup_member(struct bridge_softc *sc, const char *name)
    656 {
    657 	struct bridge_iflist *bif;
    658 	struct ifnet *ifp;
    659 
    660 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    661 		ifp = bif->bif_ifp;
    662 		if (strcmp(ifp->if_xname, name) == 0)
    663 			return (bif);
    664 	}
    665 
    666 	return (NULL);
    667 }
    668 
    669 /*
    670  * bridge_lookup_member_if:
    671  *
    672  *	Lookup a bridge member interface by ifnet*.  Must be called at splnet().
    673  */
    674 static struct bridge_iflist *
    675 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
    676 {
    677 	struct bridge_iflist *bif;
    678 
    679 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    680 		if (bif->bif_ifp == member_ifp)
    681 			return (bif);
    682 	}
    683 
    684 	return (NULL);
    685 }
    686 
    687 /*
    688  * bridge_delete_member:
    689  *
    690  *	Delete the specified member interface.
    691  */
    692 static void
    693 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif)
    694 {
    695 	struct ifnet *ifs = bif->bif_ifp;
    696 
    697 	switch (ifs->if_type) {
    698 	case IFT_ETHER:
    699 		/*
    700 		 * Take the interface out of promiscuous mode.
    701 		 */
    702 		(void) ifpromisc(ifs, 0);
    703 		break;
    704 	default:
    705 #ifdef DIAGNOSTIC
    706 		panic("bridge_delete_member: impossible");
    707 #endif
    708 		break;
    709 	}
    710 
    711 	ifs->if_input = ether_input;
    712 	ifs->if_bridge = NULL;
    713 	LIST_REMOVE(bif, bif_next);
    714 
    715 	bridge_rtdelete(sc, ifs);
    716 
    717 	free(bif, M_DEVBUF);
    718 
    719 	if (sc->sc_if.if_flags & IFF_RUNNING)
    720 		bstp_initialization(sc);
    721 }
    722 
    723 static int
    724 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
    725 {
    726 	struct ifbreq *req = arg;
    727 	struct bridge_iflist *bif = NULL;
    728 	struct ifnet *ifs;
    729 	int error = 0;
    730 
    731 	ifs = ifunit(req->ifbr_ifsname);
    732 	if (ifs == NULL)
    733 		return (ENOENT);
    734 
    735 	if (sc->sc_if.if_mtu != ifs->if_mtu)
    736 		return (EINVAL);
    737 
    738 	if (ifs->if_bridge == sc)
    739 		return (EEXIST);
    740 
    741 	if (ifs->if_bridge != NULL)
    742 		return (EBUSY);
    743 
    744 	if (ifs->if_input != ether_input)
    745 		return EINVAL;
    746 
    747 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT);
    748 	if (bif == NULL)
    749 		return (ENOMEM);
    750 
    751 	switch (ifs->if_type) {
    752 	case IFT_ETHER:
    753 		/*
    754 		 * Place the interface into promiscuous mode.
    755 		 */
    756 		error = ifpromisc(ifs, 1);
    757 		if (error)
    758 			goto out;
    759 		break;
    760 	default:
    761 		error = EINVAL;
    762 		goto out;
    763 	}
    764 
    765 	bif->bif_ifp = ifs;
    766 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
    767 	bif->bif_priority = BSTP_DEFAULT_PORT_PRIORITY;
    768 	bif->bif_path_cost = BSTP_DEFAULT_PATH_COST;
    769 
    770 	ifs->if_bridge = sc;
    771 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
    772 	ifs->if_input = bridge_input;
    773 
    774 	if (sc->sc_if.if_flags & IFF_RUNNING)
    775 		bstp_initialization(sc);
    776 	else
    777 		bstp_stop(sc);
    778 
    779  out:
    780 	if (error) {
    781 		if (bif != NULL)
    782 			free(bif, M_DEVBUF);
    783 	}
    784 	return (error);
    785 }
    786 
    787 static int
    788 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
    789 {
    790 	struct ifbreq *req = arg;
    791 	struct bridge_iflist *bif;
    792 
    793 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    794 	if (bif == NULL)
    795 		return (ENOENT);
    796 
    797 	bridge_delete_member(sc, bif);
    798 
    799 	return (0);
    800 }
    801 
    802 static int
    803 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
    804 {
    805 	struct ifbreq *req = arg;
    806 	struct bridge_iflist *bif;
    807 
    808 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    809 	if (bif == NULL)
    810 		return (ENOENT);
    811 
    812 	req->ifbr_ifsflags = bif->bif_flags;
    813 	req->ifbr_state = bif->bif_state;
    814 	req->ifbr_priority = bif->bif_priority;
    815 	req->ifbr_path_cost = bif->bif_path_cost;
    816 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
    817 
    818 	return (0);
    819 }
    820 
    821 static int
    822 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
    823 {
    824 	struct ifbreq *req = arg;
    825 	struct bridge_iflist *bif;
    826 
    827 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
    828 	if (bif == NULL)
    829 		return (ENOENT);
    830 
    831 	if (req->ifbr_ifsflags & IFBIF_STP) {
    832 		switch (bif->bif_ifp->if_type) {
    833 		case IFT_ETHER:
    834 			/* These can do spanning tree. */
    835 			break;
    836 
    837 		default:
    838 			/* Nothing else can. */
    839 			return (EINVAL);
    840 		}
    841 	}
    842 
    843 	bif->bif_flags = req->ifbr_ifsflags;
    844 
    845 	if (sc->sc_if.if_flags & IFF_RUNNING)
    846 		bstp_initialization(sc);
    847 
    848 	return (0);
    849 }
    850 
    851 static int
    852 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
    853 {
    854 	struct ifbrparam *param = arg;
    855 
    856 	sc->sc_brtmax = param->ifbrp_csize;
    857 	bridge_rttrim(sc);
    858 
    859 	return (0);
    860 }
    861 
    862 static int
    863 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
    864 {
    865 	struct ifbrparam *param = arg;
    866 
    867 	param->ifbrp_csize = sc->sc_brtmax;
    868 
    869 	return (0);
    870 }
    871 
    872 static int
    873 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
    874 {
    875 	struct ifbifconf *bifc = arg;
    876 	struct bridge_iflist *bif;
    877 	struct ifbreq breq;
    878 	int count, len, error = 0;
    879 
    880 	count = 0;
    881 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
    882 		count++;
    883 
    884 	if (bifc->ifbic_len == 0) {
    885 		bifc->ifbic_len = sizeof(breq) * count;
    886 		return (0);
    887 	}
    888 
    889 	count = 0;
    890 	len = bifc->ifbic_len;
    891 	memset(&breq, 0, sizeof breq);
    892 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
    893 		if (len < sizeof(breq))
    894 			break;
    895 
    896 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
    897 		    sizeof(breq.ifbr_ifsname));
    898 		breq.ifbr_ifsflags = bif->bif_flags;
    899 		breq.ifbr_state = bif->bif_state;
    900 		breq.ifbr_priority = bif->bif_priority;
    901 		breq.ifbr_path_cost = bif->bif_path_cost;
    902 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
    903 		error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
    904 		if (error)
    905 			break;
    906 		count++;
    907 		len -= sizeof(breq);
    908 	}
    909 
    910 	bifc->ifbic_len = sizeof(breq) * count;
    911 	return (error);
    912 }
    913 
    914 static int
    915 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
    916 {
    917 	struct ifbaconf *bac = arg;
    918 	struct bridge_rtnode *brt;
    919 	struct ifbareq bareq;
    920 	int count = 0, error = 0, len;
    921 
    922 	if (bac->ifbac_len == 0)
    923 		return (0);
    924 
    925 	len = bac->ifbac_len;
    926 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
    927 		if (len < sizeof(bareq))
    928 			goto out;
    929 		memset(&bareq, 0, sizeof(bareq));
    930 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
    931 		    sizeof(bareq.ifba_ifsname));
    932 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
    933 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
    934 			bareq.ifba_expire = brt->brt_expire - time_uptime;
    935 		} else
    936 			bareq.ifba_expire = 0;
    937 		bareq.ifba_flags = brt->brt_flags;
    938 
    939 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
    940 		if (error)
    941 			goto out;
    942 		count++;
    943 		len -= sizeof(bareq);
    944 	}
    945  out:
    946 	bac->ifbac_len = sizeof(bareq) * count;
    947 	return (error);
    948 }
    949 
    950 static int
    951 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
    952 {
    953 	struct ifbareq *req = arg;
    954 	struct bridge_iflist *bif;
    955 	int error;
    956 
    957 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
    958 	if (bif == NULL)
    959 		return (ENOENT);
    960 
    961 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
    962 	    req->ifba_flags);
    963 
    964 	return (error);
    965 }
    966 
    967 static int
    968 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
    969 {
    970 	struct ifbrparam *param = arg;
    971 
    972 	sc->sc_brttimeout = param->ifbrp_ctime;
    973 
    974 	return (0);
    975 }
    976 
    977 static int
    978 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
    979 {
    980 	struct ifbrparam *param = arg;
    981 
    982 	param->ifbrp_ctime = sc->sc_brttimeout;
    983 
    984 	return (0);
    985 }
    986 
    987 static int
    988 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
    989 {
    990 	struct ifbareq *req = arg;
    991 
    992 	return (bridge_rtdaddr(sc, req->ifba_dst));
    993 }
    994 
    995 static int
    996 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
    997 {
    998 	struct ifbreq *req = arg;
    999 
   1000 	bridge_rtflush(sc, req->ifbr_ifsflags);
   1001 
   1002 	return (0);
   1003 }
   1004 
   1005 static int
   1006 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
   1007 {
   1008 	struct ifbrparam *param = arg;
   1009 
   1010 	param->ifbrp_prio = sc->sc_bridge_priority;
   1011 
   1012 	return (0);
   1013 }
   1014 
   1015 static int
   1016 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
   1017 {
   1018 	struct ifbrparam *param = arg;
   1019 
   1020 	sc->sc_bridge_priority = param->ifbrp_prio;
   1021 
   1022 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1023 		bstp_initialization(sc);
   1024 
   1025 	return (0);
   1026 }
   1027 
   1028 static int
   1029 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
   1030 {
   1031 	struct ifbrparam *param = arg;
   1032 
   1033 	param->ifbrp_hellotime = sc->sc_bridge_hello_time >> 8;
   1034 
   1035 	return (0);
   1036 }
   1037 
   1038 static int
   1039 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
   1040 {
   1041 	struct ifbrparam *param = arg;
   1042 
   1043 	if (param->ifbrp_hellotime == 0)
   1044 		return (EINVAL);
   1045 	sc->sc_bridge_hello_time = param->ifbrp_hellotime << 8;
   1046 
   1047 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1048 		bstp_initialization(sc);
   1049 
   1050 	return (0);
   1051 }
   1052 
   1053 static int
   1054 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
   1055 {
   1056 	struct ifbrparam *param = arg;
   1057 
   1058 	param->ifbrp_fwddelay = sc->sc_bridge_forward_delay >> 8;
   1059 
   1060 	return (0);
   1061 }
   1062 
   1063 static int
   1064 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
   1065 {
   1066 	struct ifbrparam *param = arg;
   1067 
   1068 	if (param->ifbrp_fwddelay == 0)
   1069 		return (EINVAL);
   1070 	sc->sc_bridge_forward_delay = param->ifbrp_fwddelay << 8;
   1071 
   1072 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1073 		bstp_initialization(sc);
   1074 
   1075 	return (0);
   1076 }
   1077 
   1078 static int
   1079 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
   1080 {
   1081 	struct ifbrparam *param = arg;
   1082 
   1083 	param->ifbrp_maxage = sc->sc_bridge_max_age >> 8;
   1084 
   1085 	return (0);
   1086 }
   1087 
   1088 static int
   1089 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
   1090 {
   1091 	struct ifbrparam *param = arg;
   1092 
   1093 	if (param->ifbrp_maxage == 0)
   1094 		return (EINVAL);
   1095 	sc->sc_bridge_max_age = param->ifbrp_maxage << 8;
   1096 
   1097 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1098 		bstp_initialization(sc);
   1099 
   1100 	return (0);
   1101 }
   1102 
   1103 static int
   1104 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
   1105 {
   1106 	struct ifbreq *req = arg;
   1107 	struct bridge_iflist *bif;
   1108 
   1109 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1110 	if (bif == NULL)
   1111 		return (ENOENT);
   1112 
   1113 	bif->bif_priority = req->ifbr_priority;
   1114 
   1115 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1116 		bstp_initialization(sc);
   1117 
   1118 	return (0);
   1119 }
   1120 
   1121 #if defined(BRIDGE_IPF)
   1122 static int
   1123 bridge_ioctl_gfilt(struct bridge_softc *sc, void *arg)
   1124 {
   1125 	struct ifbrparam *param = arg;
   1126 
   1127 	param->ifbrp_filter = sc->sc_filter_flags;
   1128 
   1129 	return (0);
   1130 }
   1131 
   1132 static int
   1133 bridge_ioctl_sfilt(struct bridge_softc *sc, void *arg)
   1134 {
   1135 	struct ifbrparam *param = arg;
   1136 	uint32_t nflags, oflags;
   1137 
   1138 	if (param->ifbrp_filter & ~IFBF_FILT_MASK)
   1139 		return (EINVAL);
   1140 
   1141 	nflags = param->ifbrp_filter;
   1142 	oflags = sc->sc_filter_flags;
   1143 
   1144 	if ((nflags & IFBF_FILT_USEIPF) && !(oflags & IFBF_FILT_USEIPF)) {
   1145 		pfil_add_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1146 			sc->sc_if.if_pfil);
   1147 	}
   1148 	if (!(nflags & IFBF_FILT_USEIPF) && (oflags & IFBF_FILT_USEIPF)) {
   1149 		pfil_remove_hook((void *)bridge_ipf, NULL, PFIL_IN|PFIL_OUT,
   1150 			sc->sc_if.if_pfil);
   1151 	}
   1152 
   1153 	sc->sc_filter_flags = nflags;
   1154 
   1155 	return (0);
   1156 }
   1157 #endif /* BRIDGE_IPF */
   1158 
   1159 static int
   1160 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
   1161 {
   1162 	struct ifbreq *req = arg;
   1163 	struct bridge_iflist *bif;
   1164 
   1165 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
   1166 	if (bif == NULL)
   1167 		return (ENOENT);
   1168 
   1169 	bif->bif_path_cost = req->ifbr_path_cost;
   1170 
   1171 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1172 		bstp_initialization(sc);
   1173 
   1174 	return (0);
   1175 }
   1176 
   1177 /*
   1178  * bridge_ifdetach:
   1179  *
   1180  *	Detach an interface from a bridge.  Called when a member
   1181  *	interface is detaching.
   1182  */
   1183 void
   1184 bridge_ifdetach(struct ifnet *ifp)
   1185 {
   1186 	struct bridge_softc *sc = ifp->if_bridge;
   1187 	struct ifbreq breq;
   1188 
   1189 	memset(&breq, 0, sizeof(breq));
   1190 	strlcpy(breq.ifbr_ifsname, ifp->if_xname, sizeof(breq.ifbr_ifsname));
   1191 
   1192 	(void) bridge_ioctl_del(sc, &breq);
   1193 }
   1194 
   1195 /*
   1196  * bridge_init:
   1197  *
   1198  *	Initialize a bridge interface.
   1199  */
   1200 static int
   1201 bridge_init(struct ifnet *ifp)
   1202 {
   1203 	struct bridge_softc *sc = ifp->if_softc;
   1204 
   1205 	if (ifp->if_flags & IFF_RUNNING)
   1206 		return (0);
   1207 
   1208 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
   1209 	    bridge_timer, sc);
   1210 
   1211 	ifp->if_flags |= IFF_RUNNING;
   1212 	bstp_initialization(sc);
   1213 	return (0);
   1214 }
   1215 
   1216 /*
   1217  * bridge_stop:
   1218  *
   1219  *	Stop the bridge interface.
   1220  */
   1221 static void
   1222 bridge_stop(struct ifnet *ifp, int disable)
   1223 {
   1224 	struct bridge_softc *sc = ifp->if_softc;
   1225 
   1226 	if ((ifp->if_flags & IFF_RUNNING) == 0)
   1227 		return;
   1228 
   1229 	callout_stop(&sc->sc_brcallout);
   1230 	bstp_stop(sc);
   1231 
   1232 	IF_PURGE(&ifp->if_snd);
   1233 
   1234 	bridge_rtflush(sc, IFBF_FLUSHDYN);
   1235 
   1236 	ifp->if_flags &= ~IFF_RUNNING;
   1237 }
   1238 
   1239 /*
   1240  * bridge_enqueue:
   1241  *
   1242  *	Enqueue a packet on a bridge member interface.
   1243  *
   1244  *	NOTE: must be called at splnet().
   1245  */
   1246 void
   1247 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m,
   1248     int runfilt)
   1249 {
   1250 	ALTQ_DECL(struct altq_pktattr pktattr;)
   1251 	int len, error;
   1252 	short mflags;
   1253 
   1254 	/*
   1255 	 * Clear any in-bound checksum flags for this packet.
   1256 	 */
   1257 	m->m_pkthdr.csum_flags = 0;
   1258 
   1259 	if (runfilt) {
   1260 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1261 		    dst_ifp, PFIL_OUT) != 0) {
   1262 			if (m != NULL)
   1263 				m_freem(m);
   1264 			return;
   1265 		}
   1266 		if (m == NULL)
   1267 			return;
   1268 	}
   1269 
   1270 #ifdef ALTQ
   1271 	/*
   1272 	 * If ALTQ is enabled on the member interface, do
   1273 	 * classification; the queueing discipline might
   1274 	 * not require classification, but might require
   1275 	 * the address family/header pointer in the pktattr.
   1276 	 */
   1277 	if (ALTQ_IS_ENABLED(&dst_ifp->if_snd)) {
   1278 		/* XXX IFT_ETHER */
   1279 		altq_etherclassify(&dst_ifp->if_snd, m, &pktattr);
   1280 	}
   1281 #endif /* ALTQ */
   1282 
   1283 	len = m->m_pkthdr.len;
   1284 	m->m_flags |= M_PROTO1;
   1285 	mflags = m->m_flags;
   1286 	IFQ_ENQUEUE(&dst_ifp->if_snd, m, &pktattr, error);
   1287 	if (error) {
   1288 		/* mbuf is already freed */
   1289 		sc->sc_if.if_oerrors++;
   1290 		return;
   1291 	}
   1292 
   1293 	sc->sc_if.if_opackets++;
   1294 	sc->sc_if.if_obytes += len;
   1295 
   1296 	dst_ifp->if_obytes += len;
   1297 
   1298 	if (mflags & M_MCAST) {
   1299 		sc->sc_if.if_omcasts++;
   1300 		dst_ifp->if_omcasts++;
   1301 	}
   1302 
   1303 	if ((dst_ifp->if_flags & IFF_OACTIVE) == 0)
   1304 		(*dst_ifp->if_start)(dst_ifp);
   1305 }
   1306 
   1307 /*
   1308  * bridge_output:
   1309  *
   1310  *	Send output from a bridge member interface.  This
   1311  *	performs the bridging function for locally originated
   1312  *	packets.
   1313  *
   1314  *	The mbuf has the Ethernet header already attached.  We must
   1315  *	enqueue or free the mbuf before returning.
   1316  */
   1317 int
   1318 bridge_output(struct ifnet *ifp, struct mbuf *m, const struct sockaddr *sa,
   1319     struct rtentry *rt)
   1320 {
   1321 	struct ether_header *eh;
   1322 	struct ifnet *dst_if;
   1323 	struct bridge_softc *sc;
   1324 	int s;
   1325 
   1326 	if (m->m_len < ETHER_HDR_LEN) {
   1327 		m = m_pullup(m, ETHER_HDR_LEN);
   1328 		if (m == NULL)
   1329 			return (0);
   1330 	}
   1331 
   1332 	eh = mtod(m, struct ether_header *);
   1333 	sc = ifp->if_bridge;
   1334 
   1335 	s = splnet();
   1336 
   1337 	/*
   1338 	 * If bridge is down, but the original output interface is up,
   1339 	 * go ahead and send out that interface.  Otherwise, the packet
   1340 	 * is dropped below.
   1341 	 */
   1342 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1343 		dst_if = ifp;
   1344 		goto sendunicast;
   1345 	}
   1346 
   1347 	/*
   1348 	 * If the packet is a multicast, or we don't know a better way to
   1349 	 * get there, send to all interfaces.
   1350 	 */
   1351 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
   1352 		dst_if = NULL;
   1353 	else
   1354 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1355 	if (dst_if == NULL) {
   1356 		struct bridge_iflist *bif;
   1357 		struct mbuf *mc;
   1358 		int used = 0;
   1359 
   1360 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1361 			dst_if = bif->bif_ifp;
   1362 			if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1363 				continue;
   1364 
   1365 			/*
   1366 			 * If this is not the original output interface,
   1367 			 * and the interface is participating in spanning
   1368 			 * tree, make sure the port is in a state that
   1369 			 * allows forwarding.
   1370 			 */
   1371 			if (dst_if != ifp &&
   1372 			    (bif->bif_flags & IFBIF_STP) != 0) {
   1373 				switch (bif->bif_state) {
   1374 				case BSTP_IFSTATE_BLOCKING:
   1375 				case BSTP_IFSTATE_LISTENING:
   1376 				case BSTP_IFSTATE_DISABLED:
   1377 					continue;
   1378 				}
   1379 			}
   1380 
   1381 			if (LIST_NEXT(bif, bif_next) == NULL) {
   1382 				used = 1;
   1383 				mc = m;
   1384 			} else {
   1385 				mc = m_copym(m, 0, M_COPYALL, M_NOWAIT);
   1386 				if (mc == NULL) {
   1387 					sc->sc_if.if_oerrors++;
   1388 					continue;
   1389 				}
   1390 			}
   1391 
   1392 			bridge_enqueue(sc, dst_if, mc, 0);
   1393 		}
   1394 		if (used == 0)
   1395 			m_freem(m);
   1396 		splx(s);
   1397 		return (0);
   1398 	}
   1399 
   1400  sendunicast:
   1401 	/*
   1402 	 * XXX Spanning tree consideration here?
   1403 	 */
   1404 
   1405 	if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1406 		m_freem(m);
   1407 		splx(s);
   1408 		return (0);
   1409 	}
   1410 
   1411 	bridge_enqueue(sc, dst_if, m, 0);
   1412 
   1413 	splx(s);
   1414 	return (0);
   1415 }
   1416 
   1417 /*
   1418  * bridge_start:
   1419  *
   1420  *	Start output on a bridge.
   1421  *
   1422  *	NOTE: This routine should never be called in this implementation.
   1423  */
   1424 static void
   1425 bridge_start(struct ifnet *ifp)
   1426 {
   1427 
   1428 	printf("%s: bridge_start() called\n", ifp->if_xname);
   1429 }
   1430 
   1431 /*
   1432  * bridge_forward:
   1433  *
   1434  *	The forwarding function of the bridge.
   1435  */
   1436 static void
   1437 bridge_forward(void *v)
   1438 {
   1439 	struct bridge_softc *sc = v;
   1440 	struct mbuf *m;
   1441 	struct bridge_iflist *bif;
   1442 	struct ifnet *src_if, *dst_if;
   1443 	struct ether_header *eh;
   1444 	int s;
   1445 
   1446 	KERNEL_LOCK(1, NULL);
   1447 	mutex_enter(softnet_lock);
   1448 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1449 		mutex_exit(softnet_lock);
   1450 		KERNEL_UNLOCK_ONE(NULL);
   1451 		return;
   1452 	}
   1453 
   1454 	s = splnet();
   1455 	while ((m = pktq_dequeue(sc->sc_fwd_pktq)) != NULL) {
   1456 		src_if = m->m_pkthdr.rcvif;
   1457 
   1458 		sc->sc_if.if_ipackets++;
   1459 		sc->sc_if.if_ibytes += m->m_pkthdr.len;
   1460 
   1461 		/*
   1462 		 * Look up the bridge_iflist.
   1463 		 */
   1464 		bif = bridge_lookup_member_if(sc, src_if);
   1465 		if (bif == NULL) {
   1466 			/* Interface is not a bridge member (anymore?) */
   1467 			m_freem(m);
   1468 			continue;
   1469 		}
   1470 
   1471 		if (bif->bif_flags & IFBIF_STP) {
   1472 			switch (bif->bif_state) {
   1473 			case BSTP_IFSTATE_BLOCKING:
   1474 			case BSTP_IFSTATE_LISTENING:
   1475 			case BSTP_IFSTATE_DISABLED:
   1476 				m_freem(m);
   1477 				continue;
   1478 			}
   1479 		}
   1480 
   1481 		eh = mtod(m, struct ether_header *);
   1482 
   1483 		/*
   1484 		 * If the interface is learning, and the source
   1485 		 * address is valid and not multicast, record
   1486 		 * the address.
   1487 		 */
   1488 		if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
   1489 		    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
   1490 		    (eh->ether_shost[0] == 0 &&
   1491 		     eh->ether_shost[1] == 0 &&
   1492 		     eh->ether_shost[2] == 0 &&
   1493 		     eh->ether_shost[3] == 0 &&
   1494 		     eh->ether_shost[4] == 0 &&
   1495 		     eh->ether_shost[5] == 0) == 0) {
   1496 			(void) bridge_rtupdate(sc, eh->ether_shost,
   1497 			    src_if, 0, IFBAF_DYNAMIC);
   1498 		}
   1499 
   1500 		if ((bif->bif_flags & IFBIF_STP) != 0 &&
   1501 		    bif->bif_state == BSTP_IFSTATE_LEARNING) {
   1502 			m_freem(m);
   1503 			continue;
   1504 		}
   1505 
   1506 		/*
   1507 		 * At this point, the port either doesn't participate
   1508 		 * in spanning tree or it is in the forwarding state.
   1509 		 */
   1510 
   1511 		/*
   1512 		 * If the packet is unicast, destined for someone on
   1513 		 * "this" side of the bridge, drop it.
   1514 		 */
   1515 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
   1516 			dst_if = bridge_rtlookup(sc, eh->ether_dhost);
   1517 			if (src_if == dst_if) {
   1518 				m_freem(m);
   1519 				continue;
   1520 			}
   1521 		} else {
   1522 			/* ...forward it to all interfaces. */
   1523 			sc->sc_if.if_imcasts++;
   1524 			dst_if = NULL;
   1525 		}
   1526 
   1527 		if (pfil_run_hooks(sc->sc_if.if_pfil, &m,
   1528 		    m->m_pkthdr.rcvif, PFIL_IN) != 0) {
   1529 			if (m != NULL)
   1530 				m_freem(m);
   1531 			continue;
   1532 		}
   1533 		if (m == NULL)
   1534 			continue;
   1535 
   1536 		if (dst_if == NULL) {
   1537 			bridge_broadcast(sc, src_if, m);
   1538 			continue;
   1539 		}
   1540 
   1541 		/*
   1542 		 * At this point, we're dealing with a unicast frame
   1543 		 * going to a different interface.
   1544 		 */
   1545 		if ((dst_if->if_flags & IFF_RUNNING) == 0) {
   1546 			m_freem(m);
   1547 			continue;
   1548 		}
   1549 		bif = bridge_lookup_member_if(sc, dst_if);
   1550 		if (bif == NULL) {
   1551 			/* Not a member of the bridge (anymore?) */
   1552 			m_freem(m);
   1553 			continue;
   1554 		}
   1555 
   1556 		if (bif->bif_flags & IFBIF_STP) {
   1557 			switch (bif->bif_state) {
   1558 			case BSTP_IFSTATE_DISABLED:
   1559 			case BSTP_IFSTATE_BLOCKING:
   1560 				m_freem(m);
   1561 				continue;
   1562 			}
   1563 		}
   1564 
   1565 		bridge_enqueue(sc, dst_if, m, 1);
   1566 	}
   1567 	splx(s);
   1568 	mutex_exit(softnet_lock);
   1569 	KERNEL_UNLOCK_ONE(NULL);
   1570 }
   1571 
   1572 static bool
   1573 bstp_state_before_learning(struct bridge_iflist *bif)
   1574 {
   1575 	if (bif->bif_flags & IFBIF_STP) {
   1576 		switch (bif->bif_state) {
   1577 		case BSTP_IFSTATE_BLOCKING:
   1578 		case BSTP_IFSTATE_LISTENING:
   1579 		case BSTP_IFSTATE_DISABLED:
   1580 			return true;
   1581 		}
   1582 	}
   1583 	return false;
   1584 }
   1585 
   1586 static bool
   1587 bridge_ourether(struct bridge_iflist *bif, struct ether_header *eh, int src)
   1588 {
   1589 	uint8_t *ether = src ? eh->ether_shost : eh->ether_dhost;
   1590 
   1591 	if (memcmp(CLLADDR(bif->bif_ifp->if_sadl), ether, ETHER_ADDR_LEN) == 0
   1592 #if NCARP > 0
   1593 	    || (bif->bif_ifp->if_carp &&
   1594 	        carp_ourether(bif->bif_ifp->if_carp, eh, IFT_ETHER, src) != NULL)
   1595 #endif /* NCARP > 0 */
   1596 	    )
   1597 		return true;
   1598 
   1599 	return false;
   1600 }
   1601 
   1602 /*
   1603  * bridge_input:
   1604  *
   1605  *	Receive input from a member interface.  Queue the packet for
   1606  *	bridging if it is not for us.
   1607  *	should be called at splnet()
   1608  */
   1609 static void
   1610 bridge_input(struct ifnet *ifp, struct mbuf *m)
   1611 {
   1612 	struct bridge_softc *sc = ifp->if_bridge;
   1613 	struct bridge_iflist *bif;
   1614 	struct ether_header *eh;
   1615 
   1616 	if ((sc->sc_if.if_flags & IFF_RUNNING) == 0) {
   1617 		ether_input(ifp, m);
   1618 		return;
   1619 	}
   1620 
   1621 	bif = bridge_lookup_member_if(sc, ifp);
   1622 	if (bif == NULL) {
   1623 		ether_input(ifp, m);
   1624 		return;
   1625 	}
   1626 
   1627 	eh = mtod(m, struct ether_header *);
   1628 
   1629 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
   1630 		if (memcmp(etherbroadcastaddr,
   1631 		    eh->ether_dhost, ETHER_ADDR_LEN) == 0)
   1632 			m->m_flags |= M_BCAST;
   1633 		else
   1634 			m->m_flags |= M_MCAST;
   1635 	}
   1636 
   1637 	/*
   1638 	 * A 'fast' path for packets addressed to interfaces that are
   1639 	 * part of this bridge.
   1640 	 */
   1641 	if (!(m->m_flags & (M_BCAST|M_MCAST)) &&
   1642 	    !bstp_state_before_learning(bif)) {
   1643 		struct bridge_iflist *_bif;
   1644 
   1645 		LIST_FOREACH(_bif, &sc->sc_iflist, bif_next) {
   1646 			/* It is destined for us. */
   1647 			if (bridge_ourether(_bif, eh, 0)) {
   1648 				if (_bif->bif_flags & IFBIF_LEARNING)
   1649 					(void) bridge_rtupdate(sc,
   1650 					    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
   1651 				m->m_pkthdr.rcvif = _bif->bif_ifp;
   1652 				ether_input(_bif->bif_ifp, m);
   1653 				return;
   1654 			}
   1655 
   1656 			/* We just received a packet that we sent out. */
   1657 			if (bridge_ourether(_bif, eh, 1)) {
   1658 				m_freem(m);
   1659 				return;
   1660 			}
   1661 		}
   1662 	}
   1663 
   1664 	/* Tap off 802.1D packets; they do not get forwarded. */
   1665 	if (bif->bif_flags & IFBIF_STP &&
   1666 	    memcmp(eh->ether_dhost, bstp_etheraddr, ETHER_ADDR_LEN) == 0) {
   1667 		bstp_input(sc, bif, m);
   1668 		return;
   1669 	}
   1670 
   1671 	/*
   1672 	 * A normal switch would discard the packet here, but that's not what
   1673 	 * we've done historically. This also prevents some obnoxious behaviour.
   1674 	 */
   1675 	if (bstp_state_before_learning(bif)) {
   1676 		ether_input(ifp, m);
   1677 		return;
   1678 	}
   1679 
   1680 	/* Queue the packet for bridge forwarding. */
   1681 	if (__predict_false(!pktq_enqueue(sc->sc_fwd_pktq, m, 0)))
   1682 		m_freem(m);
   1683 }
   1684 
   1685 /*
   1686  * bridge_broadcast:
   1687  *
   1688  *	Send a frame to all interfaces that are members of
   1689  *	the bridge, except for the one on which the packet
   1690  *	arrived.
   1691  */
   1692 static void
   1693 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
   1694     struct mbuf *m)
   1695 {
   1696 	struct bridge_iflist *bif;
   1697 	struct mbuf *mc;
   1698 	struct ifnet *dst_if;
   1699 	bool used, bmcast;
   1700 
   1701 	used = bmcast = m->m_flags & (M_BCAST|M_MCAST);
   1702 
   1703 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
   1704 		dst_if = bif->bif_ifp;
   1705 		if (dst_if == src_if)
   1706 			continue;
   1707 
   1708 		if (bif->bif_flags & IFBIF_STP) {
   1709 			switch (bif->bif_state) {
   1710 			case BSTP_IFSTATE_BLOCKING:
   1711 			case BSTP_IFSTATE_DISABLED:
   1712 				continue;
   1713 			}
   1714 		}
   1715 
   1716 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 && !bmcast)
   1717 			continue;
   1718 
   1719 		if ((dst_if->if_flags & IFF_RUNNING) == 0)
   1720 			continue;
   1721 
   1722 		if (!used && LIST_NEXT(bif, bif_next) == NULL) {
   1723 			mc = m;
   1724 			used = true;
   1725 		} else {
   1726 			mc = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
   1727 			if (mc == NULL) {
   1728 				sc->sc_if.if_oerrors++;
   1729 				continue;
   1730 			}
   1731 		}
   1732 
   1733 		bridge_enqueue(sc, dst_if, mc, 1);
   1734 	}
   1735 
   1736 	if (bmcast)
   1737 		ether_input(src_if, m);
   1738 	else if (!used)
   1739 		m_freem(m);
   1740 }
   1741 
   1742 /*
   1743  * bridge_rtupdate:
   1744  *
   1745  *	Add a bridge routing entry.
   1746  */
   1747 static int
   1748 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
   1749     struct ifnet *dst_if, int setflags, uint8_t flags)
   1750 {
   1751 	struct bridge_rtnode *brt;
   1752 	int error;
   1753 
   1754 	/*
   1755 	 * A route for this destination might already exist.  If so,
   1756 	 * update it, otherwise create a new one.
   1757 	 */
   1758 	if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
   1759 		if (sc->sc_brtcnt >= sc->sc_brtmax)
   1760 			return (ENOSPC);
   1761 
   1762 		/*
   1763 		 * Allocate a new bridge forwarding node, and
   1764 		 * initialize the expiration time and Ethernet
   1765 		 * address.
   1766 		 */
   1767 		brt = pool_get(&bridge_rtnode_pool, PR_NOWAIT);
   1768 		if (brt == NULL)
   1769 			return (ENOMEM);
   1770 
   1771 		memset(brt, 0, sizeof(*brt));
   1772 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1773 		brt->brt_flags = IFBAF_DYNAMIC;
   1774 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
   1775 
   1776 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
   1777 			pool_put(&bridge_rtnode_pool, brt);
   1778 			return (error);
   1779 		}
   1780 	}
   1781 
   1782 	brt->brt_ifp = dst_if;
   1783 	if (setflags) {
   1784 		brt->brt_flags = flags;
   1785 		if (flags & IFBAF_STATIC)
   1786 			brt->brt_expire = 0;
   1787 		else
   1788 			brt->brt_expire = time_uptime + sc->sc_brttimeout;
   1789 	}
   1790 
   1791 	return (0);
   1792 }
   1793 
   1794 /*
   1795  * bridge_rtlookup:
   1796  *
   1797  *	Lookup the destination interface for an address.
   1798  */
   1799 static struct ifnet *
   1800 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
   1801 {
   1802 	struct bridge_rtnode *brt;
   1803 
   1804 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
   1805 		return (NULL);
   1806 
   1807 	return (brt->brt_ifp);
   1808 }
   1809 
   1810 /*
   1811  * bridge_rttrim:
   1812  *
   1813  *	Trim the routine table so that we have a number
   1814  *	of routing entries less than or equal to the
   1815  *	maximum number.
   1816  */
   1817 static void
   1818 bridge_rttrim(struct bridge_softc *sc)
   1819 {
   1820 	struct bridge_rtnode *brt, *nbrt;
   1821 
   1822 	/* Make sure we actually need to do this. */
   1823 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   1824 		return;
   1825 
   1826 	/* Force an aging cycle; this might trim enough addresses. */
   1827 	bridge_rtage(sc);
   1828 	if (sc->sc_brtcnt <= sc->sc_brtmax)
   1829 		return;
   1830 
   1831 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1832 		nbrt = LIST_NEXT(brt, brt_list);
   1833 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1834 			bridge_rtnode_destroy(sc, brt);
   1835 			if (sc->sc_brtcnt <= sc->sc_brtmax)
   1836 				return;
   1837 		}
   1838 	}
   1839 }
   1840 
   1841 /*
   1842  * bridge_timer:
   1843  *
   1844  *	Aging timer for the bridge.
   1845  */
   1846 static void
   1847 bridge_timer(void *arg)
   1848 {
   1849 	struct bridge_softc *sc = arg;
   1850 	int s;
   1851 
   1852 	s = splnet();
   1853 	bridge_rtage(sc);
   1854 	splx(s);
   1855 
   1856 	if (sc->sc_if.if_flags & IFF_RUNNING)
   1857 		callout_reset(&sc->sc_brcallout,
   1858 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
   1859 }
   1860 
   1861 /*
   1862  * bridge_rtage:
   1863  *
   1864  *	Perform an aging cycle.
   1865  */
   1866 static void
   1867 bridge_rtage(struct bridge_softc *sc)
   1868 {
   1869 	struct bridge_rtnode *brt, *nbrt;
   1870 
   1871 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1872 		nbrt = LIST_NEXT(brt, brt_list);
   1873 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
   1874 			if (time_uptime >= brt->brt_expire)
   1875 				bridge_rtnode_destroy(sc, brt);
   1876 		}
   1877 	}
   1878 }
   1879 
   1880 /*
   1881  * bridge_rtflush:
   1882  *
   1883  *	Remove all dynamic addresses from the bridge.
   1884  */
   1885 static void
   1886 bridge_rtflush(struct bridge_softc *sc, int full)
   1887 {
   1888 	struct bridge_rtnode *brt, *nbrt;
   1889 
   1890 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1891 		nbrt = LIST_NEXT(brt, brt_list);
   1892 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
   1893 			bridge_rtnode_destroy(sc, brt);
   1894 	}
   1895 }
   1896 
   1897 /*
   1898  * bridge_rtdaddr:
   1899  *
   1900  *	Remove an address from the table.
   1901  */
   1902 static int
   1903 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
   1904 {
   1905 	struct bridge_rtnode *brt;
   1906 
   1907 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
   1908 		return (ENOENT);
   1909 
   1910 	bridge_rtnode_destroy(sc, brt);
   1911 	return (0);
   1912 }
   1913 
   1914 /*
   1915  * bridge_rtdelete:
   1916  *
   1917  *	Delete routes to a speicifc member interface.
   1918  */
   1919 static void
   1920 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp)
   1921 {
   1922 	struct bridge_rtnode *brt, *nbrt;
   1923 
   1924 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
   1925 		nbrt = LIST_NEXT(brt, brt_list);
   1926 		if (brt->brt_ifp == ifp)
   1927 			bridge_rtnode_destroy(sc, brt);
   1928 	}
   1929 }
   1930 
   1931 /*
   1932  * bridge_rtable_init:
   1933  *
   1934  *	Initialize the route table for this bridge.
   1935  */
   1936 static int
   1937 bridge_rtable_init(struct bridge_softc *sc)
   1938 {
   1939 	int i;
   1940 
   1941 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
   1942 	    M_DEVBUF, M_NOWAIT);
   1943 	if (sc->sc_rthash == NULL)
   1944 		return (ENOMEM);
   1945 
   1946 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
   1947 		LIST_INIT(&sc->sc_rthash[i]);
   1948 
   1949 	sc->sc_rthash_key = cprng_fast32();
   1950 
   1951 	LIST_INIT(&sc->sc_rtlist);
   1952 
   1953 	return (0);
   1954 }
   1955 
   1956 /*
   1957  * bridge_rtable_fini:
   1958  *
   1959  *	Deconstruct the route table for this bridge.
   1960  */
   1961 static void
   1962 bridge_rtable_fini(struct bridge_softc *sc)
   1963 {
   1964 
   1965 	free(sc->sc_rthash, M_DEVBUF);
   1966 }
   1967 
   1968 /*
   1969  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
   1970  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
   1971  */
   1972 #define	mix(a, b, c)							\
   1973 do {									\
   1974 	a -= b; a -= c; a ^= (c >> 13);					\
   1975 	b -= c; b -= a; b ^= (a << 8);					\
   1976 	c -= a; c -= b; c ^= (b >> 13);					\
   1977 	a -= b; a -= c; a ^= (c >> 12);					\
   1978 	b -= c; b -= a; b ^= (a << 16);					\
   1979 	c -= a; c -= b; c ^= (b >> 5);					\
   1980 	a -= b; a -= c; a ^= (c >> 3);					\
   1981 	b -= c; b -= a; b ^= (a << 10);					\
   1982 	c -= a; c -= b; c ^= (b >> 15);					\
   1983 } while (/*CONSTCOND*/0)
   1984 
   1985 static inline uint32_t
   1986 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
   1987 {
   1988 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
   1989 
   1990 	b += addr[5] << 8;
   1991 	b += addr[4];
   1992 	a += addr[3] << 24;
   1993 	a += addr[2] << 16;
   1994 	a += addr[1] << 8;
   1995 	a += addr[0];
   1996 
   1997 	mix(a, b, c);
   1998 
   1999 	return (c & BRIDGE_RTHASH_MASK);
   2000 }
   2001 
   2002 #undef mix
   2003 
   2004 /*
   2005  * bridge_rtnode_lookup:
   2006  *
   2007  *	Look up a bridge route node for the specified destination.
   2008  */
   2009 static struct bridge_rtnode *
   2010 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
   2011 {
   2012 	struct bridge_rtnode *brt;
   2013 	uint32_t hash;
   2014 	int dir;
   2015 
   2016 	hash = bridge_rthash(sc, addr);
   2017 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
   2018 		dir = memcmp(addr, brt->brt_addr, ETHER_ADDR_LEN);
   2019 		if (dir == 0)
   2020 			return (brt);
   2021 		if (dir > 0)
   2022 			return (NULL);
   2023 	}
   2024 
   2025 	return (NULL);
   2026 }
   2027 
   2028 /*
   2029  * bridge_rtnode_insert:
   2030  *
   2031  *	Insert the specified bridge node into the route table.  We
   2032  *	assume the entry is not already in the table.
   2033  */
   2034 static int
   2035 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2036 {
   2037 	struct bridge_rtnode *lbrt;
   2038 	uint32_t hash;
   2039 	int dir;
   2040 
   2041 	hash = bridge_rthash(sc, brt->brt_addr);
   2042 
   2043 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
   2044 	if (lbrt == NULL) {
   2045 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
   2046 		goto out;
   2047 	}
   2048 
   2049 	do {
   2050 		dir = memcmp(brt->brt_addr, lbrt->brt_addr, ETHER_ADDR_LEN);
   2051 		if (dir == 0)
   2052 			return (EEXIST);
   2053 		if (dir > 0) {
   2054 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
   2055 			goto out;
   2056 		}
   2057 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
   2058 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
   2059 			goto out;
   2060 		}
   2061 		lbrt = LIST_NEXT(lbrt, brt_hash);
   2062 	} while (lbrt != NULL);
   2063 
   2064 #ifdef DIAGNOSTIC
   2065 	panic("bridge_rtnode_insert: impossible");
   2066 #endif
   2067 
   2068  out:
   2069 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
   2070 	sc->sc_brtcnt++;
   2071 
   2072 	return (0);
   2073 }
   2074 
   2075 /*
   2076  * bridge_rtnode_destroy:
   2077  *
   2078  *	Destroy a bridge rtnode.
   2079  */
   2080 static void
   2081 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
   2082 {
   2083 	int s = splnet();
   2084 
   2085 	LIST_REMOVE(brt, brt_hash);
   2086 
   2087 	LIST_REMOVE(brt, brt_list);
   2088 	sc->sc_brtcnt--;
   2089 	pool_put(&bridge_rtnode_pool, brt);
   2090 
   2091 	splx(s);
   2092 }
   2093 
   2094 #if defined(BRIDGE_IPF)
   2095 extern pfil_head_t *inet_pfil_hook;                 /* XXX */
   2096 extern pfil_head_t *inet6_pfil_hook;                /* XXX */
   2097 
   2098 /*
   2099  * Send bridge packets through IPF if they are one of the types IPF can deal
   2100  * with, or if they are ARP or REVARP.  (IPF will pass ARP and REVARP without
   2101  * question.)
   2102  */
   2103 static int
   2104 bridge_ipf(void *arg, struct mbuf **mp, struct ifnet *ifp, int dir)
   2105 {
   2106 	int snap, error;
   2107 	struct ether_header *eh1, eh2;
   2108 	struct llc llc1;
   2109 	uint16_t ether_type;
   2110 
   2111 	snap = 0;
   2112 	error = -1;	/* Default error if not error == 0 */
   2113 	eh1 = mtod(*mp, struct ether_header *);
   2114 	ether_type = ntohs(eh1->ether_type);
   2115 
   2116 	/*
   2117 	 * Check for SNAP/LLC.
   2118 	 */
   2119         if (ether_type < ETHERMTU) {
   2120                 struct llc *llc2 = (struct llc *)(eh1 + 1);
   2121 
   2122                 if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
   2123                     llc2->llc_dsap == LLC_SNAP_LSAP &&
   2124                     llc2->llc_ssap == LLC_SNAP_LSAP &&
   2125                     llc2->llc_control == LLC_UI) {
   2126                 	ether_type = htons(llc2->llc_un.type_snap.ether_type);
   2127 			snap = 1;
   2128                 }
   2129         }
   2130 
   2131 	/*
   2132 	 * If we're trying to filter bridge traffic, don't look at anything
   2133 	 * other than IP and ARP traffic.  If the filter doesn't understand
   2134 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
   2135 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
   2136 	 * but of course we don't have an AppleTalk filter to begin with.
   2137 	 * (Note that since IPF doesn't understand ARP it will pass *ALL*
   2138 	 * ARP traffic.)
   2139 	 */
   2140 	switch (ether_type) {
   2141 		case ETHERTYPE_ARP:
   2142 		case ETHERTYPE_REVARP:
   2143 			return 0; /* Automatically pass */
   2144 		case ETHERTYPE_IP:
   2145 # ifdef INET6
   2146 		case ETHERTYPE_IPV6:
   2147 # endif /* INET6 */
   2148 			break;
   2149 		default:
   2150 			goto bad;
   2151 	}
   2152 
   2153 	/* Strip off the Ethernet header and keep a copy. */
   2154 	m_copydata(*mp, 0, ETHER_HDR_LEN, (void *) &eh2);
   2155 	m_adj(*mp, ETHER_HDR_LEN);
   2156 
   2157 	/* Strip off snap header, if present */
   2158 	if (snap) {
   2159 		m_copydata(*mp, 0, sizeof(struct llc), (void *) &llc1);
   2160 		m_adj(*mp, sizeof(struct llc));
   2161 	}
   2162 
   2163 	/*
   2164 	 * Check basic packet sanity and run IPF through pfil.
   2165 	 */
   2166 	KASSERT(!cpu_intr_p());
   2167 	switch (ether_type)
   2168 	{
   2169 	case ETHERTYPE_IP :
   2170 		error = (dir == PFIL_IN) ? bridge_ip_checkbasic(mp) : 0;
   2171 		if (error == 0)
   2172 			error = pfil_run_hooks(inet_pfil_hook, mp, ifp, dir);
   2173 		break;
   2174 # ifdef INET6
   2175 	case ETHERTYPE_IPV6 :
   2176 		error = (dir == PFIL_IN) ? bridge_ip6_checkbasic(mp) : 0;
   2177 		if (error == 0)
   2178 			error = pfil_run_hooks(inet6_pfil_hook, mp, ifp, dir);
   2179 		break;
   2180 # endif
   2181 	default :
   2182 		error = 0;
   2183 		break;
   2184 	}
   2185 
   2186 	if (*mp == NULL)
   2187 		return error;
   2188 	if (error != 0)
   2189 		goto bad;
   2190 
   2191 	error = -1;
   2192 
   2193 	/*
   2194 	 * Finally, put everything back the way it was and return
   2195 	 */
   2196 	if (snap) {
   2197 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
   2198 		if (*mp == NULL)
   2199 			return error;
   2200 		bcopy(&llc1, mtod(*mp, void *), sizeof(struct llc));
   2201 	}
   2202 
   2203 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
   2204 	if (*mp == NULL)
   2205 		return error;
   2206 	bcopy(&eh2, mtod(*mp, void *), ETHER_HDR_LEN);
   2207 
   2208 	return 0;
   2209 
   2210     bad:
   2211 	m_freem(*mp);
   2212 	*mp = NULL;
   2213 	return error;
   2214 }
   2215 
   2216 /*
   2217  * Perform basic checks on header size since
   2218  * IPF assumes ip_input has already processed
   2219  * it for it.  Cut-and-pasted from ip_input.c.
   2220  * Given how simple the IPv6 version is,
   2221  * does the IPv4 version really need to be
   2222  * this complicated?
   2223  *
   2224  * XXX Should we update ipstat here, or not?
   2225  * XXX Right now we update ipstat but not
   2226  * XXX csum_counter.
   2227  */
   2228 static int
   2229 bridge_ip_checkbasic(struct mbuf **mp)
   2230 {
   2231 	struct mbuf *m = *mp;
   2232 	struct ip *ip;
   2233 	int len, hlen;
   2234 
   2235 	if (*mp == NULL)
   2236 		return -1;
   2237 
   2238 	if (IP_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2239 		if ((m = m_copyup(m, sizeof(struct ip),
   2240 			(max_linkhdr + 3) & ~3)) == NULL) {
   2241 			/* XXXJRT new stat, please */
   2242 			ip_statinc(IP_STAT_TOOSMALL);
   2243 			goto bad;
   2244 		}
   2245 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
   2246 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
   2247 			ip_statinc(IP_STAT_TOOSMALL);
   2248 			goto bad;
   2249 		}
   2250 	}
   2251 	ip = mtod(m, struct ip *);
   2252 	if (ip == NULL) goto bad;
   2253 
   2254 	if (ip->ip_v != IPVERSION) {
   2255 		ip_statinc(IP_STAT_BADVERS);
   2256 		goto bad;
   2257 	}
   2258 	hlen = ip->ip_hl << 2;
   2259 	if (hlen < sizeof(struct ip)) { /* minimum header length */
   2260 		ip_statinc(IP_STAT_BADHLEN);
   2261 		goto bad;
   2262 	}
   2263 	if (hlen > m->m_len) {
   2264 		if ((m = m_pullup(m, hlen)) == 0) {
   2265 			ip_statinc(IP_STAT_BADHLEN);
   2266 			goto bad;
   2267 		}
   2268 		ip = mtod(m, struct ip *);
   2269 		if (ip == NULL) goto bad;
   2270 	}
   2271 
   2272         switch (m->m_pkthdr.csum_flags &
   2273                 ((m->m_pkthdr.rcvif->if_csum_flags_rx & M_CSUM_IPv4) |
   2274                  M_CSUM_IPv4_BAD)) {
   2275         case M_CSUM_IPv4|M_CSUM_IPv4_BAD:
   2276                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_bad); */
   2277                 goto bad;
   2278 
   2279         case M_CSUM_IPv4:
   2280                 /* Checksum was okay. */
   2281                 /* INET_CSUM_COUNTER_INCR(&ip_hwcsum_ok); */
   2282                 break;
   2283 
   2284         default:
   2285                 /* Must compute it ourselves. */
   2286                 /* INET_CSUM_COUNTER_INCR(&ip_swcsum); */
   2287                 if (in_cksum(m, hlen) != 0)
   2288                         goto bad;
   2289                 break;
   2290         }
   2291 
   2292         /* Retrieve the packet length. */
   2293         len = ntohs(ip->ip_len);
   2294 
   2295         /*
   2296          * Check for additional length bogosity
   2297          */
   2298         if (len < hlen) {
   2299 		ip_statinc(IP_STAT_BADLEN);
   2300                 goto bad;
   2301         }
   2302 
   2303         /*
   2304          * Check that the amount of data in the buffers
   2305          * is as at least much as the IP header would have us expect.
   2306          * Drop packet if shorter than we expect.
   2307          */
   2308         if (m->m_pkthdr.len < len) {
   2309 		ip_statinc(IP_STAT_TOOSHORT);
   2310                 goto bad;
   2311         }
   2312 
   2313 	/* Checks out, proceed */
   2314 	*mp = m;
   2315 	return 0;
   2316 
   2317     bad:
   2318 	*mp = m;
   2319 	return -1;
   2320 }
   2321 
   2322 # ifdef INET6
   2323 /*
   2324  * Same as above, but for IPv6.
   2325  * Cut-and-pasted from ip6_input.c.
   2326  * XXX Should we update ip6stat, or not?
   2327  */
   2328 static int
   2329 bridge_ip6_checkbasic(struct mbuf **mp)
   2330 {
   2331 	struct mbuf *m = *mp;
   2332 	struct ip6_hdr *ip6;
   2333 
   2334         /*
   2335          * If the IPv6 header is not aligned, slurp it up into a new
   2336          * mbuf with space for link headers, in the event we forward
   2337          * it.  Otherwise, if it is aligned, make sure the entire base
   2338          * IPv6 header is in the first mbuf of the chain.
   2339          */
   2340         if (IP6_HDR_ALIGNED_P(mtod(m, void *)) == 0) {
   2341                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2342                 if ((m = m_copyup(m, sizeof(struct ip6_hdr),
   2343                                   (max_linkhdr + 3) & ~3)) == NULL) {
   2344                         /* XXXJRT new stat, please */
   2345 			ip6_statinc(IP6_STAT_TOOSMALL);
   2346                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2347                         goto bad;
   2348                 }
   2349         } else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
   2350                 struct ifnet *inifp = m->m_pkthdr.rcvif;
   2351                 if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
   2352 			ip6_statinc(IP6_STAT_TOOSMALL);
   2353                         in6_ifstat_inc(inifp, ifs6_in_hdrerr);
   2354                         goto bad;
   2355                 }
   2356         }
   2357 
   2358         ip6 = mtod(m, struct ip6_hdr *);
   2359 
   2360         if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
   2361 		ip6_statinc(IP6_STAT_BADVERS);
   2362                 in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
   2363                 goto bad;
   2364         }
   2365 
   2366 	/* Checks out, proceed */
   2367 	*mp = m;
   2368 	return 0;
   2369 
   2370     bad:
   2371 	*mp = m;
   2372 	return -1;
   2373 }
   2374 # endif /* INET6 */
   2375 #endif /* BRIDGE_IPF */
   2376