xref: /freebsd/sys/net/if_bridge.c (revision 6356dba0b403daa023dec24559ab1f8e602e4f14)
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 #include "opt_carp.h"
83 
84 #include <sys/param.h>
85 #include <sys/mbuf.h>
86 #include <sys/malloc.h>
87 #include <sys/protosw.h>
88 #include <sys/systm.h>
89 #include <sys/time.h>
90 #include <sys/socket.h> /* for net/if.h */
91 #include <sys/sockio.h>
92 #include <sys/ctype.h>  /* string functions */
93 #include <sys/kernel.h>
94 #include <sys/random.h>
95 #include <sys/syslog.h>
96 #include <sys/sysctl.h>
97 #include <vm/uma.h>
98 #include <sys/module.h>
99 #include <sys/priv.h>
100 #include <sys/proc.h>
101 #include <sys/lock.h>
102 #include <sys/mutex.h>
103 #include <sys/vimage.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 
113 #include <netinet/in.h> /* for struct arpcom */
114 #include <netinet/in_systm.h>
115 #include <netinet/in_var.h>
116 #include <netinet/ip.h>
117 #include <netinet/ip_var.h>
118 #ifdef INET6
119 #include <netinet/ip6.h>
120 #include <netinet6/ip6_var.h>
121 #endif
122 #ifdef DEV_CARP
123 #include <netinet/ip_carp.h>
124 #endif
125 #include <machine/in_cksum.h>
126 #include <netinet/if_ether.h> /* for struct arpcom */
127 #include <net/bridgestp.h>
128 #include <net/if_bridgevar.h>
129 #include <net/if_llc.h>
130 #include <net/if_vlan_var.h>
131 
132 #include <net/route.h>
133 #include <netinet/ip_fw.h>
134 #include <netinet/ip_dummynet.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  * Maximum number of addresses to cache.
147  */
148 #ifndef BRIDGE_RTABLE_MAX
149 #define	BRIDGE_RTABLE_MAX		100
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  * Bridge interface list entry.
173  */
174 struct bridge_iflist {
175 	LIST_ENTRY(bridge_iflist) bif_next;
176 	struct ifnet		*bif_ifp;	/* member if */
177 	struct bstp_port	bif_stp;	/* STP state */
178 	uint32_t		bif_flags;	/* member if flags */
179 	int			bif_savedcaps;	/* saved capabilities */
180 	uint32_t		bif_addrmax;	/* max # of addresses */
181 	uint32_t		bif_addrcnt;	/* cur. # of addresses */
182 	uint32_t		bif_addrexceeded;/* # of address violations */
183 };
184 
185 /*
186  * Bridge route node.
187  */
188 struct bridge_rtnode {
189 	LIST_ENTRY(bridge_rtnode) brt_hash;	/* hash table linkage */
190 	LIST_ENTRY(bridge_rtnode) brt_list;	/* list linkage */
191 	struct bridge_iflist	*brt_dst;	/* destination if */
192 	unsigned long		brt_expire;	/* expiration time */
193 	uint8_t			brt_flags;	/* address flags */
194 	uint8_t			brt_addr[ETHER_ADDR_LEN];
195 	uint16_t		brt_vlan;	/* vlan id */
196 };
197 #define	brt_ifp			brt_dst->bif_ifp
198 
199 /*
200  * Software state for each bridge.
201  */
202 struct bridge_softc {
203 	struct ifnet		*sc_ifp;	/* make this an interface */
204 	LIST_ENTRY(bridge_softc) sc_list;
205 	struct mtx		sc_mtx;
206 	struct cv		sc_cv;
207 	uint32_t		sc_brtmax;	/* max # of addresses */
208 	uint32_t		sc_brtcnt;	/* cur. # of addresses */
209 	uint32_t		sc_brttimeout;	/* rt timeout in seconds */
210 	struct callout		sc_brcallout;	/* bridge callout */
211 	uint32_t		sc_iflist_ref;	/* refcount for sc_iflist */
212 	uint32_t		sc_iflist_xcnt;	/* refcount for sc_iflist */
213 	LIST_HEAD(, bridge_iflist) sc_iflist;	/* member interface list */
214 	LIST_HEAD(, bridge_rtnode) *sc_rthash;	/* our forwarding table */
215 	LIST_HEAD(, bridge_rtnode) sc_rtlist;	/* list version of above */
216 	uint32_t		sc_rthash_key;	/* key for hash */
217 	LIST_HEAD(, bridge_iflist) sc_spanlist;	/* span ports list */
218 	struct bstp_state	sc_stp;		/* STP state */
219 	uint32_t		sc_brtexceeded;	/* # of cache drops */
220 	u_char			sc_defaddr[6];	/* Default MAC address */
221 };
222 
223 static struct mtx 	bridge_list_mtx;
224 eventhandler_tag	bridge_detach_cookie = NULL;
225 
226 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
227 
228 uma_zone_t bridge_rtnode_zone;
229 
230 static int	bridge_clone_create(struct if_clone *, int, caddr_t);
231 static void	bridge_clone_destroy(struct ifnet *);
232 
233 static int	bridge_ioctl(struct ifnet *, u_long, caddr_t);
234 static void	bridge_mutecaps(struct bridge_softc *);
235 static void	bridge_set_ifcap(struct bridge_softc *, struct bridge_iflist *,
236 		    int);
237 static void	bridge_ifdetach(void *arg __unused, struct ifnet *);
238 static void	bridge_init(void *);
239 static void	bridge_dummynet(struct mbuf *, struct ifnet *);
240 static void	bridge_stop(struct ifnet *, int);
241 static void	bridge_start(struct ifnet *);
242 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
243 static int	bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
244 		    struct rtentry *);
245 static void	bridge_enqueue(struct bridge_softc *, struct ifnet *,
246 		    struct mbuf *);
247 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
248 
249 static void	bridge_forward(struct bridge_softc *, struct bridge_iflist *,
250 		    struct mbuf *m);
251 
252 static void	bridge_timer(void *);
253 
254 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
255 		    struct mbuf *, int);
256 static void	bridge_span(struct bridge_softc *, struct mbuf *);
257 
258 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
259 		    uint16_t, struct bridge_iflist *, int, uint8_t);
260 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *,
261 		    uint16_t);
262 static void	bridge_rttrim(struct bridge_softc *);
263 static void	bridge_rtage(struct bridge_softc *);
264 static void	bridge_rtflush(struct bridge_softc *, int);
265 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *,
266 		    uint16_t);
267 
268 static int	bridge_rtable_init(struct bridge_softc *);
269 static void	bridge_rtable_fini(struct bridge_softc *);
270 
271 static int	bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
272 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
273 		    const uint8_t *, uint16_t);
274 static int	bridge_rtnode_insert(struct bridge_softc *,
275 		    struct bridge_rtnode *);
276 static void	bridge_rtnode_destroy(struct bridge_softc *,
277 		    struct bridge_rtnode *);
278 static void	bridge_rtable_expire(struct ifnet *, int);
279 static void	bridge_state_change(struct ifnet *, int);
280 
281 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
282 		    const char *name);
283 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
284 		    struct ifnet *ifp);
285 static void	bridge_delete_member(struct bridge_softc *,
286 		    struct bridge_iflist *, int);
287 static void	bridge_delete_span(struct bridge_softc *,
288 		    struct bridge_iflist *);
289 
290 static int	bridge_ioctl_add(struct bridge_softc *, void *);
291 static int	bridge_ioctl_del(struct bridge_softc *, void *);
292 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
293 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
294 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
295 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
296 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
297 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
298 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
299 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
300 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
301 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
302 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
303 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
304 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
305 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
306 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
307 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
308 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
309 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
310 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
311 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
312 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
313 static int	bridge_ioctl_sifmaxaddr(struct bridge_softc *, void *);
314 static int	bridge_ioctl_addspan(struct bridge_softc *, void *);
315 static int	bridge_ioctl_delspan(struct bridge_softc *, void *);
316 static int	bridge_ioctl_gbparam(struct bridge_softc *, void *);
317 static int	bridge_ioctl_grte(struct bridge_softc *, void *);
318 static int	bridge_ioctl_gifsstp(struct bridge_softc *, void *);
319 static int	bridge_ioctl_sproto(struct bridge_softc *, void *);
320 static int	bridge_ioctl_stxhc(struct bridge_softc *, void *);
321 static int	bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
322 		    int);
323 static int	bridge_ip_checkbasic(struct mbuf **mp);
324 #ifdef INET6
325 static int	bridge_ip6_checkbasic(struct mbuf **mp);
326 #endif /* INET6 */
327 static int	bridge_fragment(struct ifnet *, struct mbuf *,
328 		    struct ether_header *, int, struct llc *);
329 
330 /* The default bridge vlan is 1 (IEEE 802.1Q-2003 Table 9-2) */
331 #define	VLANTAGOF(_m)	\
332     (_m->m_flags & M_VLANTAG) ? EVL_VLANOFTAG(_m->m_pkthdr.ether_vtag) : 1
333 
334 static struct bstp_cb_ops bridge_ops = {
335 	.bcb_state = bridge_state_change,
336 	.bcb_rtage = bridge_rtable_expire
337 };
338 
339 SYSCTL_DECL(_net_link);
340 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
341 
342 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
343 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
344 static int pfil_member = 1; /* run pfil hooks on the member interface */
345 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
346 static int pfil_ipfw_arp = 0;   /* layer2 filter with ipfw */
347 static int pfil_local_phys = 0; /* run pfil hooks on the physical interface for
348                                    locally destined packets */
349 static int log_stp   = 0;   /* log STP state changes */
350 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
351     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
352 SYSCTL_INT(_net_link_bridge, OID_AUTO, ipfw_arp, CTLFLAG_RW,
353     &pfil_ipfw_arp, 0, "Filter ARP packets through IPFW layer2");
354 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
355     &pfil_bridge, 0, "Packet filter on the bridge interface");
356 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
357     &pfil_member, 0, "Packet filter on the member interface");
358 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_local_phys, CTLFLAG_RW,
359     &pfil_local_phys, 0,
360     "Packet filter on the physical interface for locally destined packets");
361 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
362     &log_stp, 0, "Log STP state changes");
363 
364 struct bridge_control {
365 	int	(*bc_func)(struct bridge_softc *, void *);
366 	int	bc_argsize;
367 	int	bc_flags;
368 };
369 
370 #define	BC_F_COPYIN		0x01	/* copy arguments in */
371 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
372 #define	BC_F_SUSER		0x04	/* do super-user check */
373 
374 const struct bridge_control bridge_control_table[] = {
375 	{ bridge_ioctl_add,		sizeof(struct ifbreq),
376 	  BC_F_COPYIN|BC_F_SUSER },
377 	{ bridge_ioctl_del,		sizeof(struct ifbreq),
378 	  BC_F_COPYIN|BC_F_SUSER },
379 
380 	{ bridge_ioctl_gifflags,	sizeof(struct ifbreq),
381 	  BC_F_COPYIN|BC_F_COPYOUT },
382 	{ bridge_ioctl_sifflags,	sizeof(struct ifbreq),
383 	  BC_F_COPYIN|BC_F_SUSER },
384 
385 	{ bridge_ioctl_scache,		sizeof(struct ifbrparam),
386 	  BC_F_COPYIN|BC_F_SUSER },
387 	{ bridge_ioctl_gcache,		sizeof(struct ifbrparam),
388 	  BC_F_COPYOUT },
389 
390 	{ bridge_ioctl_gifs,		sizeof(struct ifbifconf),
391 	  BC_F_COPYIN|BC_F_COPYOUT },
392 	{ bridge_ioctl_rts,		sizeof(struct ifbaconf),
393 	  BC_F_COPYIN|BC_F_COPYOUT },
394 
395 	{ bridge_ioctl_saddr,		sizeof(struct ifbareq),
396 	  BC_F_COPYIN|BC_F_SUSER },
397 
398 	{ bridge_ioctl_sto,		sizeof(struct ifbrparam),
399 	  BC_F_COPYIN|BC_F_SUSER },
400 	{ bridge_ioctl_gto,		sizeof(struct ifbrparam),
401 	  BC_F_COPYOUT },
402 
403 	{ bridge_ioctl_daddr,		sizeof(struct ifbareq),
404 	  BC_F_COPYIN|BC_F_SUSER },
405 
406 	{ bridge_ioctl_flush,		sizeof(struct ifbreq),
407 	  BC_F_COPYIN|BC_F_SUSER },
408 
409 	{ bridge_ioctl_gpri,		sizeof(struct ifbrparam),
410 	  BC_F_COPYOUT },
411 	{ bridge_ioctl_spri,		sizeof(struct ifbrparam),
412 	  BC_F_COPYIN|BC_F_SUSER },
413 
414 	{ bridge_ioctl_ght,		sizeof(struct ifbrparam),
415 	  BC_F_COPYOUT },
416 	{ bridge_ioctl_sht,		sizeof(struct ifbrparam),
417 	  BC_F_COPYIN|BC_F_SUSER },
418 
419 	{ bridge_ioctl_gfd,		sizeof(struct ifbrparam),
420 	  BC_F_COPYOUT },
421 	{ bridge_ioctl_sfd,		sizeof(struct ifbrparam),
422 	  BC_F_COPYIN|BC_F_SUSER },
423 
424 	{ bridge_ioctl_gma,		sizeof(struct ifbrparam),
425 	  BC_F_COPYOUT },
426 	{ bridge_ioctl_sma,		sizeof(struct ifbrparam),
427 	  BC_F_COPYIN|BC_F_SUSER },
428 
429 	{ bridge_ioctl_sifprio,		sizeof(struct ifbreq),
430 	  BC_F_COPYIN|BC_F_SUSER },
431 
432 	{ bridge_ioctl_sifcost,		sizeof(struct ifbreq),
433 	  BC_F_COPYIN|BC_F_SUSER },
434 
435 	{ bridge_ioctl_addspan,		sizeof(struct ifbreq),
436 	  BC_F_COPYIN|BC_F_SUSER },
437 	{ bridge_ioctl_delspan,		sizeof(struct ifbreq),
438 	  BC_F_COPYIN|BC_F_SUSER },
439 
440 	{ bridge_ioctl_gbparam,		sizeof(struct ifbropreq),
441 	  BC_F_COPYOUT },
442 
443 	{ bridge_ioctl_grte,		sizeof(struct ifbrparam),
444 	  BC_F_COPYOUT },
445 
446 	{ bridge_ioctl_gifsstp,		sizeof(struct ifbpstpconf),
447 	  BC_F_COPYIN|BC_F_COPYOUT },
448 
449 	{ bridge_ioctl_sproto,		sizeof(struct ifbrparam),
450 	  BC_F_COPYIN|BC_F_SUSER },
451 
452 	{ bridge_ioctl_stxhc,		sizeof(struct ifbrparam),
453 	  BC_F_COPYIN|BC_F_SUSER },
454 
455 	{ bridge_ioctl_sifmaxaddr,	sizeof(struct ifbreq),
456 	  BC_F_COPYIN|BC_F_SUSER },
457 
458 };
459 const int bridge_control_table_size =
460     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
461 
462 LIST_HEAD(, bridge_softc) bridge_list;
463 
464 IFC_SIMPLE_DECLARE(bridge, 0);
465 
466 static int
467 bridge_modevent(module_t mod, int type, void *data)
468 {
469 
470 	switch (type) {
471 	case MOD_LOAD:
472 		mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
473 		if_clone_attach(&bridge_cloner);
474 		bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
475 		    sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
476 		    UMA_ALIGN_PTR, 0);
477 		LIST_INIT(&bridge_list);
478 		bridge_input_p = bridge_input;
479 		bridge_output_p = bridge_output;
480 		bridge_dn_p = bridge_dummynet;
481 		bridge_detach_cookie = EVENTHANDLER_REGISTER(
482 		    ifnet_departure_event, bridge_ifdetach, NULL,
483 		    EVENTHANDLER_PRI_ANY);
484 		break;
485 	case MOD_UNLOAD:
486 		EVENTHANDLER_DEREGISTER(ifnet_departure_event,
487 		    bridge_detach_cookie);
488 		if_clone_detach(&bridge_cloner);
489 		uma_zdestroy(bridge_rtnode_zone);
490 		bridge_input_p = NULL;
491 		bridge_output_p = NULL;
492 		bridge_dn_p = NULL;
493 		mtx_destroy(&bridge_list_mtx);
494 		break;
495 	default:
496 		return (EOPNOTSUPP);
497 	}
498 	return (0);
499 }
500 
501 static moduledata_t bridge_mod = {
502 	"if_bridge",
503 	bridge_modevent,
504 	0
505 };
506 
507 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
508 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
509 
510 /*
511  * handler for net.link.bridge.pfil_ipfw
512  */
513 static int
514 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
515 {
516 	int enable = pfil_ipfw;
517 	int error;
518 
519 	error = sysctl_handle_int(oidp, &enable, 0, req);
520 	enable = (enable) ? 1 : 0;
521 
522 	if (enable != pfil_ipfw) {
523 		pfil_ipfw = enable;
524 
525 		/*
526 		 * Disable pfil so that ipfw doesnt run twice, if the user
527 		 * really wants both then they can re-enable pfil_bridge and/or
528 		 * pfil_member. Also allow non-ip packets as ipfw can filter by
529 		 * layer2 type.
530 		 */
531 		if (pfil_ipfw) {
532 			pfil_onlyip = 0;
533 			pfil_bridge = 0;
534 			pfil_member = 0;
535 		}
536 	}
537 
538 	return (error);
539 }
540 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
541 	    &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
542 
543 /*
544  * bridge_clone_create:
545  *
546  *	Create a new bridge instance.
547  */
548 static int
549 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
550 {
551 	struct bridge_softc *sc, *sc2;
552 	struct ifnet *bifp, *ifp;
553 	int retry;
554 
555 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
556 	ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
557 	if (ifp == NULL) {
558 		free(sc, M_DEVBUF);
559 		return (ENOSPC);
560 	}
561 
562 	BRIDGE_LOCK_INIT(sc);
563 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
564 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
565 
566 	/* Initialize our routing table. */
567 	bridge_rtable_init(sc);
568 
569 	callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
570 
571 	LIST_INIT(&sc->sc_iflist);
572 	LIST_INIT(&sc->sc_spanlist);
573 
574 	ifp->if_softc = sc;
575 	if_initname(ifp, ifc->ifc_name, unit);
576 	ifp->if_flags = IFF_BROADCAST | IFF_SIMPLEX | IFF_MULTICAST;
577 	ifp->if_ioctl = bridge_ioctl;
578 	ifp->if_start = bridge_start;
579 	ifp->if_init = bridge_init;
580 	ifp->if_type = IFT_BRIDGE;
581 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
582 	ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
583 	IFQ_SET_READY(&ifp->if_snd);
584 
585 	/*
586 	 * Generate a random ethernet address with a locally administered
587 	 * address.
588 	 *
589 	 * Since we are using random ethernet addresses for the bridge, it is
590 	 * possible that we might have address collisions, so make sure that
591 	 * this hardware address isn't already in use on another bridge.
592 	 */
593 	for (retry = 1; retry != 0;) {
594 		arc4rand(sc->sc_defaddr, ETHER_ADDR_LEN, 1);
595 		sc->sc_defaddr[0] &= ~1;	/* clear multicast bit */
596 		sc->sc_defaddr[0] |= 2;		/* set the LAA bit */
597 		retry = 0;
598 		mtx_lock(&bridge_list_mtx);
599 		LIST_FOREACH(sc2, &bridge_list, sc_list) {
600 			bifp = sc2->sc_ifp;
601 			if (memcmp(sc->sc_defaddr,
602 			    IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
603 				retry = 1;
604 		}
605 		mtx_unlock(&bridge_list_mtx);
606 	}
607 
608 	bstp_attach(&sc->sc_stp, &bridge_ops);
609 	ether_ifattach(ifp, sc->sc_defaddr);
610 	/* Now undo some of the damage... */
611 	ifp->if_baudrate = 0;
612 	ifp->if_type = IFT_BRIDGE;
613 
614 	mtx_lock(&bridge_list_mtx);
615 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
616 	mtx_unlock(&bridge_list_mtx);
617 
618 	return (0);
619 }
620 
621 /*
622  * bridge_clone_destroy:
623  *
624  *	Destroy a bridge instance.
625  */
626 static void
627 bridge_clone_destroy(struct ifnet *ifp)
628 {
629 	struct bridge_softc *sc = ifp->if_softc;
630 	struct bridge_iflist *bif;
631 
632 	BRIDGE_LOCK(sc);
633 
634 	bridge_stop(ifp, 1);
635 	ifp->if_flags &= ~IFF_UP;
636 
637 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
638 		bridge_delete_member(sc, bif, 0);
639 
640 	while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
641 		bridge_delete_span(sc, bif);
642 	}
643 
644 	BRIDGE_UNLOCK(sc);
645 
646 	callout_drain(&sc->sc_brcallout);
647 
648 	mtx_lock(&bridge_list_mtx);
649 	LIST_REMOVE(sc, sc_list);
650 	mtx_unlock(&bridge_list_mtx);
651 
652 	bstp_detach(&sc->sc_stp);
653 	ether_ifdetach(ifp);
654 	if_free_type(ifp, IFT_ETHER);
655 
656 	/* Tear down the routing table. */
657 	bridge_rtable_fini(sc);
658 
659 	BRIDGE_LOCK_DESTROY(sc);
660 	free(sc, M_DEVBUF);
661 }
662 
663 /*
664  * bridge_ioctl:
665  *
666  *	Handle a control request from the operator.
667  */
668 static int
669 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
670 {
671 	struct bridge_softc *sc = ifp->if_softc;
672 	struct thread *td = curthread;
673 	union {
674 		struct ifbreq ifbreq;
675 		struct ifbifconf ifbifconf;
676 		struct ifbareq ifbareq;
677 		struct ifbaconf ifbaconf;
678 		struct ifbrparam ifbrparam;
679 		struct ifbropreq ifbropreq;
680 	} args;
681 	struct ifdrv *ifd = (struct ifdrv *) data;
682 	const struct bridge_control *bc;
683 	int error = 0;
684 
685 	switch (cmd) {
686 
687 	case SIOCADDMULTI:
688 	case SIOCDELMULTI:
689 		break;
690 
691 	case SIOCGDRVSPEC:
692 	case SIOCSDRVSPEC:
693 		if (ifd->ifd_cmd >= bridge_control_table_size) {
694 			error = EINVAL;
695 			break;
696 		}
697 		bc = &bridge_control_table[ifd->ifd_cmd];
698 
699 		if (cmd == SIOCGDRVSPEC &&
700 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
701 			error = EINVAL;
702 			break;
703 		}
704 		else if (cmd == SIOCSDRVSPEC &&
705 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
706 			error = EINVAL;
707 			break;
708 		}
709 
710 		if (bc->bc_flags & BC_F_SUSER) {
711 			error = priv_check(td, PRIV_NET_BRIDGE);
712 			if (error)
713 				break;
714 		}
715 
716 		if (ifd->ifd_len != bc->bc_argsize ||
717 		    ifd->ifd_len > sizeof(args)) {
718 			error = EINVAL;
719 			break;
720 		}
721 
722 		bzero(&args, sizeof(args));
723 		if (bc->bc_flags & BC_F_COPYIN) {
724 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
725 			if (error)
726 				break;
727 		}
728 
729 		BRIDGE_LOCK(sc);
730 		error = (*bc->bc_func)(sc, &args);
731 		BRIDGE_UNLOCK(sc);
732 		if (error)
733 			break;
734 
735 		if (bc->bc_flags & BC_F_COPYOUT)
736 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
737 
738 		break;
739 
740 	case SIOCSIFFLAGS:
741 		if (!(ifp->if_flags & IFF_UP) &&
742 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
743 			/*
744 			 * If interface is marked down and it is running,
745 			 * then stop and disable it.
746 			 */
747 			BRIDGE_LOCK(sc);
748 			bridge_stop(ifp, 1);
749 			BRIDGE_UNLOCK(sc);
750 		} else if ((ifp->if_flags & IFF_UP) &&
751 		    !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
752 			/*
753 			 * If interface is marked up and it is stopped, then
754 			 * start it.
755 			 */
756 			(*ifp->if_init)(sc);
757 		}
758 		break;
759 
760 	case SIOCSIFMTU:
761 		/* Do not allow the MTU to be changed on the bridge */
762 		error = EINVAL;
763 		break;
764 
765 	default:
766 		/*
767 		 * drop the lock as ether_ioctl() will call bridge_start() and
768 		 * cause the lock to be recursed.
769 		 */
770 		error = ether_ioctl(ifp, cmd, data);
771 		break;
772 	}
773 
774 	return (error);
775 }
776 
777 /*
778  * bridge_mutecaps:
779  *
780  *	Clear or restore unwanted capabilities on the member interface
781  */
782 static void
783 bridge_mutecaps(struct bridge_softc *sc)
784 {
785 	struct bridge_iflist *bif;
786 	int enabled, mask;
787 
788 	/* Initial bitmask of capabilities to test */
789 	mask = BRIDGE_IFCAPS_MASK;
790 
791 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
792 		/* Every member must support it or its disabled */
793 		mask &= bif->bif_savedcaps;
794 	}
795 
796 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
797 		enabled = bif->bif_ifp->if_capenable;
798 		/* strip off mask bits and enable them again if allowed */
799 		enabled &= ~BRIDGE_IFCAPS_MASK;
800 		enabled |= mask;
801 		/*
802 		 * Receive offload can only be enabled if all members also
803 		 * support send offload.
804 		 */
805 		if ((enabled & IFCAP_TSO) == 0)
806 			enabled &= ~IFCAP_LRO;
807 
808 		bridge_set_ifcap(sc, bif, enabled);
809 	}
810 
811 }
812 
813 static void
814 bridge_set_ifcap(struct bridge_softc *sc, struct bridge_iflist *bif, int set)
815 {
816 	struct ifnet *ifp = bif->bif_ifp;
817 	struct ifreq ifr;
818 	int error;
819 
820 	bzero(&ifr, sizeof(ifr));
821 	ifr.ifr_reqcap = set;
822 
823 	if (ifp->if_capenable != set) {
824 		IFF_LOCKGIANT(ifp);
825 		error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
826 		IFF_UNLOCKGIANT(ifp);
827 		if (error)
828 			if_printf(sc->sc_ifp,
829 			    "error setting interface capabilities on %s\n",
830 			    ifp->if_xname);
831 	}
832 }
833 
834 /*
835  * bridge_lookup_member:
836  *
837  *	Lookup a bridge member interface.
838  */
839 static struct bridge_iflist *
840 bridge_lookup_member(struct bridge_softc *sc, const char *name)
841 {
842 	struct bridge_iflist *bif;
843 	struct ifnet *ifp;
844 
845 	BRIDGE_LOCK_ASSERT(sc);
846 
847 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
848 		ifp = bif->bif_ifp;
849 		if (strcmp(ifp->if_xname, name) == 0)
850 			return (bif);
851 	}
852 
853 	return (NULL);
854 }
855 
856 /*
857  * bridge_lookup_member_if:
858  *
859  *	Lookup a bridge member interface by ifnet*.
860  */
861 static struct bridge_iflist *
862 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
863 {
864 	struct bridge_iflist *bif;
865 
866 	BRIDGE_LOCK_ASSERT(sc);
867 
868 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
869 		if (bif->bif_ifp == member_ifp)
870 			return (bif);
871 	}
872 
873 	return (NULL);
874 }
875 
876 /*
877  * bridge_delete_member:
878  *
879  *	Delete the specified member interface.
880  */
881 static void
882 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
883     int gone)
884 {
885 	struct ifnet *ifs = bif->bif_ifp;
886 	struct ifnet *fif = NULL;
887 
888 	BRIDGE_LOCK_ASSERT(sc);
889 
890 	if (!gone) {
891 		switch (ifs->if_type) {
892 		case IFT_ETHER:
893 		case IFT_L2VLAN:
894 			/*
895 			 * Take the interface out of promiscuous mode.
896 			 */
897 			(void) ifpromisc(ifs, 0);
898 			break;
899 
900 		case IFT_GIF:
901 			break;
902 
903 		default:
904 #ifdef DIAGNOSTIC
905 			panic("bridge_delete_member: impossible");
906 #endif
907 			break;
908 		}
909 		/* reneable any interface capabilities */
910 		bridge_set_ifcap(sc, bif, bif->bif_savedcaps);
911 	}
912 
913 	if (bif->bif_flags & IFBIF_STP)
914 		bstp_disable(&bif->bif_stp);
915 
916 	ifs->if_bridge = NULL;
917 	BRIDGE_XLOCK(sc);
918 	LIST_REMOVE(bif, bif_next);
919 	BRIDGE_XDROP(sc);
920 
921 	/*
922 	 * If removing the interface that gave the bridge its mac address, set
923 	 * the mac address of the bridge to the address of the next member, or
924 	 * to its default address if no members are left.
925 	 */
926 	if (!memcmp(IF_LLADDR(sc->sc_ifp), IF_LLADDR(ifs), ETHER_ADDR_LEN)) {
927 		if (LIST_EMPTY(&sc->sc_iflist))
928 			bcopy(sc->sc_defaddr,
929 			    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
930 		else {
931 			fif = LIST_FIRST(&sc->sc_iflist)->bif_ifp;
932 			bcopy(IF_LLADDR(fif),
933 			    IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
934 		}
935 	}
936 
937 	bridge_mutecaps(sc);	/* recalcuate now this interface is removed */
938 	bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
939 	KASSERT(bif->bif_addrcnt == 0,
940 	    ("%s: %d bridge routes referenced", __func__, bif->bif_addrcnt));
941 
942 	BRIDGE_UNLOCK(sc);
943 	bstp_destroy(&bif->bif_stp);	/* prepare to free */
944 	BRIDGE_LOCK(sc);
945 	free(bif, M_DEVBUF);
946 }
947 
948 /*
949  * bridge_delete_span:
950  *
951  *	Delete the specified span interface.
952  */
953 static void
954 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
955 {
956 	BRIDGE_LOCK_ASSERT(sc);
957 
958 	KASSERT(bif->bif_ifp->if_bridge == NULL,
959 	    ("%s: not a span interface", __func__));
960 
961 	LIST_REMOVE(bif, bif_next);
962 	free(bif, M_DEVBUF);
963 }
964 
965 static int
966 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
967 {
968 	struct ifbreq *req = arg;
969 	struct bridge_iflist *bif = NULL;
970 	struct ifnet *ifs;
971 	int error = 0;
972 
973 	ifs = ifunit(req->ifbr_ifsname);
974 	if (ifs == NULL)
975 		return (ENOENT);
976 	if (ifs->if_ioctl == NULL)	/* must be supported */
977 		return (EINVAL);
978 
979 	/* If it's in the span list, it can't be a member. */
980 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
981 		if (ifs == bif->bif_ifp)
982 			return (EBUSY);
983 
984 	/* Allow the first Ethernet member to define the MTU */
985 	if (ifs->if_type != IFT_GIF) {
986 		if (LIST_EMPTY(&sc->sc_iflist))
987 			sc->sc_ifp->if_mtu = ifs->if_mtu;
988 		else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
989 			if_printf(sc->sc_ifp, "invalid MTU for %s\n",
990 			    ifs->if_xname);
991 			return (EINVAL);
992 		}
993 	}
994 
995 	if (ifs->if_bridge == sc)
996 		return (EEXIST);
997 
998 	if (ifs->if_bridge != NULL)
999 		return (EBUSY);
1000 
1001 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1002 	if (bif == NULL)
1003 		return (ENOMEM);
1004 
1005 	bif->bif_ifp = ifs;
1006 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
1007 	bif->bif_savedcaps = ifs->if_capenable;
1008 
1009 	switch (ifs->if_type) {
1010 	case IFT_ETHER:
1011 	case IFT_L2VLAN:
1012 		/*
1013 		 * Place the interface into promiscuous mode.
1014 		 */
1015 		error = ifpromisc(ifs, 1);
1016 		if (error)
1017 			goto out;
1018 		break;
1019 
1020 	case IFT_GIF:
1021 		break;
1022 
1023 	default:
1024 		error = EINVAL;
1025 		goto out;
1026 	}
1027 
1028 	/*
1029 	 * Assign the interface's MAC address to the bridge if it's the first
1030 	 * member and the MAC address of the bridge has not been changed from
1031 	 * the default randomly generated one.
1032 	 */
1033 	if (LIST_EMPTY(&sc->sc_iflist) &&
1034 	    !memcmp(IF_LLADDR(sc->sc_ifp), sc->sc_defaddr, ETHER_ADDR_LEN))
1035 		bcopy(IF_LLADDR(ifs), IF_LLADDR(sc->sc_ifp), ETHER_ADDR_LEN);
1036 
1037 	ifs->if_bridge = sc;
1038 	bstp_create(&sc->sc_stp, &bif->bif_stp, bif->bif_ifp);
1039 	/*
1040 	 * XXX: XLOCK HERE!?!
1041 	 *
1042 	 * NOTE: insert_***HEAD*** should be safe for the traversals.
1043 	 */
1044 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
1045 
1046 	/* Set interface capabilities to the intersection set of all members */
1047 	bridge_mutecaps(sc);
1048 out:
1049 	if (error) {
1050 		if (bif != NULL)
1051 			free(bif, M_DEVBUF);
1052 	}
1053 	return (error);
1054 }
1055 
1056 static int
1057 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
1058 {
1059 	struct ifbreq *req = arg;
1060 	struct bridge_iflist *bif;
1061 
1062 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1063 	if (bif == NULL)
1064 		return (ENOENT);
1065 
1066 	bridge_delete_member(sc, bif, 0);
1067 
1068 	return (0);
1069 }
1070 
1071 static int
1072 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
1073 {
1074 	struct ifbreq *req = arg;
1075 	struct bridge_iflist *bif;
1076 	struct bstp_port *bp;
1077 
1078 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1079 	if (bif == NULL)
1080 		return (ENOENT);
1081 
1082 	bp = &bif->bif_stp;
1083 	req->ifbr_ifsflags = bif->bif_flags;
1084 	req->ifbr_state = bp->bp_state;
1085 	req->ifbr_priority = bp->bp_priority;
1086 	req->ifbr_path_cost = bp->bp_path_cost;
1087 	req->ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1088 	req->ifbr_proto = bp->bp_protover;
1089 	req->ifbr_role = bp->bp_role;
1090 	req->ifbr_stpflags = bp->bp_flags;
1091 	req->ifbr_addrcnt = bif->bif_addrcnt;
1092 	req->ifbr_addrmax = bif->bif_addrmax;
1093 	req->ifbr_addrexceeded = bif->bif_addrexceeded;
1094 
1095 	/* Copy STP state options as flags */
1096 	if (bp->bp_operedge)
1097 		req->ifbr_ifsflags |= IFBIF_BSTP_EDGE;
1098 	if (bp->bp_flags & BSTP_PORT_AUTOEDGE)
1099 		req->ifbr_ifsflags |= IFBIF_BSTP_AUTOEDGE;
1100 	if (bp->bp_ptp_link)
1101 		req->ifbr_ifsflags |= IFBIF_BSTP_PTP;
1102 	if (bp->bp_flags & BSTP_PORT_AUTOPTP)
1103 		req->ifbr_ifsflags |= IFBIF_BSTP_AUTOPTP;
1104 	if (bp->bp_flags & BSTP_PORT_ADMEDGE)
1105 		req->ifbr_ifsflags |= IFBIF_BSTP_ADMEDGE;
1106 	if (bp->bp_flags & BSTP_PORT_ADMCOST)
1107 		req->ifbr_ifsflags |= IFBIF_BSTP_ADMCOST;
1108 	return (0);
1109 }
1110 
1111 static int
1112 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1113 {
1114 	struct ifbreq *req = arg;
1115 	struct bridge_iflist *bif;
1116 	struct bstp_port *bp;
1117 	int error;
1118 
1119 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1120 	if (bif == NULL)
1121 		return (ENOENT);
1122 	bp = &bif->bif_stp;
1123 
1124 	if (req->ifbr_ifsflags & IFBIF_SPAN)
1125 		/* SPAN is readonly */
1126 		return (EINVAL);
1127 
1128 	if (req->ifbr_ifsflags & IFBIF_STP) {
1129 		if ((bif->bif_flags & IFBIF_STP) == 0) {
1130 			error = bstp_enable(&bif->bif_stp);
1131 			if (error)
1132 				return (error);
1133 		}
1134 	} else {
1135 		if ((bif->bif_flags & IFBIF_STP) != 0)
1136 			bstp_disable(&bif->bif_stp);
1137 	}
1138 
1139 	/* Pass on STP flags */
1140 	bstp_set_edge(bp, req->ifbr_ifsflags & IFBIF_BSTP_EDGE ? 1 : 0);
1141 	bstp_set_autoedge(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOEDGE ? 1 : 0);
1142 	bstp_set_ptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_PTP ? 1 : 0);
1143 	bstp_set_autoptp(bp, req->ifbr_ifsflags & IFBIF_BSTP_AUTOPTP ? 1 : 0);
1144 
1145 	/* Save the bits relating to the bridge */
1146 	bif->bif_flags = req->ifbr_ifsflags & IFBIFMASK;
1147 
1148 	return (0);
1149 }
1150 
1151 static int
1152 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1153 {
1154 	struct ifbrparam *param = arg;
1155 
1156 	sc->sc_brtmax = param->ifbrp_csize;
1157 	bridge_rttrim(sc);
1158 
1159 	return (0);
1160 }
1161 
1162 static int
1163 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1164 {
1165 	struct ifbrparam *param = arg;
1166 
1167 	param->ifbrp_csize = sc->sc_brtmax;
1168 
1169 	return (0);
1170 }
1171 
1172 static int
1173 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1174 {
1175 	struct ifbifconf *bifc = arg;
1176 	struct bridge_iflist *bif;
1177 	struct ifbreq breq;
1178 	char *buf, *outbuf;
1179 	int count, buflen, len, error = 0;
1180 
1181 	count = 0;
1182 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1183 		count++;
1184 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1185 		count++;
1186 
1187 	buflen = sizeof(breq) * count;
1188 	if (bifc->ifbic_len == 0) {
1189 		bifc->ifbic_len = buflen;
1190 		return (0);
1191 	}
1192 	BRIDGE_UNLOCK(sc);
1193 	outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1194 	BRIDGE_LOCK(sc);
1195 
1196 	count = 0;
1197 	buf = outbuf;
1198 	len = min(bifc->ifbic_len, buflen);
1199 	bzero(&breq, sizeof(breq));
1200 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1201 		if (len < sizeof(breq))
1202 			break;
1203 
1204 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1205 		    sizeof(breq.ifbr_ifsname));
1206 		/* Fill in the ifbreq structure */
1207 		error = bridge_ioctl_gifflags(sc, &breq);
1208 		if (error)
1209 			break;
1210 		memcpy(buf, &breq, sizeof(breq));
1211 		count++;
1212 		buf += sizeof(breq);
1213 		len -= sizeof(breq);
1214 	}
1215 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1216 		if (len < sizeof(breq))
1217 			break;
1218 
1219 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1220 		    sizeof(breq.ifbr_ifsname));
1221 		breq.ifbr_ifsflags = bif->bif_flags;
1222 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xfff;
1223 		memcpy(buf, &breq, sizeof(breq));
1224 		count++;
1225 		buf += sizeof(breq);
1226 		len -= sizeof(breq);
1227 	}
1228 
1229 	BRIDGE_UNLOCK(sc);
1230 	bifc->ifbic_len = sizeof(breq) * count;
1231 	error = copyout(outbuf, bifc->ifbic_req, bifc->ifbic_len);
1232 	BRIDGE_LOCK(sc);
1233 	free(outbuf, M_TEMP);
1234 	return (error);
1235 }
1236 
1237 static int
1238 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1239 {
1240 	struct ifbaconf *bac = arg;
1241 	struct bridge_rtnode *brt;
1242 	struct ifbareq bareq;
1243 	char *buf, *outbuf;
1244 	int count, buflen, len, error = 0;
1245 
1246 	if (bac->ifbac_len == 0)
1247 		return (0);
1248 
1249 	count = 0;
1250 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list)
1251 		count++;
1252 	buflen = sizeof(bareq) * count;
1253 
1254 	BRIDGE_UNLOCK(sc);
1255 	outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1256 	BRIDGE_LOCK(sc);
1257 
1258 	count = 0;
1259 	buf = outbuf;
1260 	len = min(bac->ifbac_len, buflen);
1261 	bzero(&bareq, sizeof(bareq));
1262 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1263 		if (len < sizeof(bareq))
1264 			goto out;
1265 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1266 		    sizeof(bareq.ifba_ifsname));
1267 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1268 		bareq.ifba_vlan = brt->brt_vlan;
1269 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1270 				time_uptime < brt->brt_expire)
1271 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1272 		else
1273 			bareq.ifba_expire = 0;
1274 		bareq.ifba_flags = brt->brt_flags;
1275 
1276 		memcpy(buf, &bareq, sizeof(bareq));
1277 		count++;
1278 		buf += sizeof(bareq);
1279 		len -= sizeof(bareq);
1280 	}
1281 out:
1282 	BRIDGE_UNLOCK(sc);
1283 	bac->ifbac_len = sizeof(bareq) * count;
1284 	error = copyout(outbuf, bac->ifbac_req, bac->ifbac_len);
1285 	BRIDGE_LOCK(sc);
1286 	free(outbuf, M_TEMP);
1287 	return (error);
1288 }
1289 
1290 static int
1291 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1292 {
1293 	struct ifbareq *req = arg;
1294 	struct bridge_iflist *bif;
1295 	int error;
1296 
1297 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
1298 	if (bif == NULL)
1299 		return (ENOENT);
1300 
1301 	error = bridge_rtupdate(sc, req->ifba_dst, req->ifba_vlan, bif, 1,
1302 	    req->ifba_flags);
1303 
1304 	return (error);
1305 }
1306 
1307 static int
1308 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1309 {
1310 	struct ifbrparam *param = arg;
1311 
1312 	sc->sc_brttimeout = param->ifbrp_ctime;
1313 	return (0);
1314 }
1315 
1316 static int
1317 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1318 {
1319 	struct ifbrparam *param = arg;
1320 
1321 	param->ifbrp_ctime = sc->sc_brttimeout;
1322 	return (0);
1323 }
1324 
1325 static int
1326 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1327 {
1328 	struct ifbareq *req = arg;
1329 
1330 	return (bridge_rtdaddr(sc, req->ifba_dst, req->ifba_vlan));
1331 }
1332 
1333 static int
1334 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1335 {
1336 	struct ifbreq *req = arg;
1337 
1338 	bridge_rtflush(sc, req->ifbr_ifsflags);
1339 	return (0);
1340 }
1341 
1342 static int
1343 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1344 {
1345 	struct ifbrparam *param = arg;
1346 	struct bstp_state *bs = &sc->sc_stp;
1347 
1348 	param->ifbrp_prio = bs->bs_bridge_priority;
1349 	return (0);
1350 }
1351 
1352 static int
1353 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1354 {
1355 	struct ifbrparam *param = arg;
1356 
1357 	return (bstp_set_priority(&sc->sc_stp, param->ifbrp_prio));
1358 }
1359 
1360 static int
1361 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1362 {
1363 	struct ifbrparam *param = arg;
1364 	struct bstp_state *bs = &sc->sc_stp;
1365 
1366 	param->ifbrp_hellotime = bs->bs_bridge_htime >> 8;
1367 	return (0);
1368 }
1369 
1370 static int
1371 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1372 {
1373 	struct ifbrparam *param = arg;
1374 
1375 	return (bstp_set_htime(&sc->sc_stp, param->ifbrp_hellotime));
1376 }
1377 
1378 static int
1379 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1380 {
1381 	struct ifbrparam *param = arg;
1382 	struct bstp_state *bs = &sc->sc_stp;
1383 
1384 	param->ifbrp_fwddelay = bs->bs_bridge_fdelay >> 8;
1385 	return (0);
1386 }
1387 
1388 static int
1389 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1390 {
1391 	struct ifbrparam *param = arg;
1392 
1393 	return (bstp_set_fdelay(&sc->sc_stp, param->ifbrp_fwddelay));
1394 }
1395 
1396 static int
1397 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1398 {
1399 	struct ifbrparam *param = arg;
1400 	struct bstp_state *bs = &sc->sc_stp;
1401 
1402 	param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1403 	return (0);
1404 }
1405 
1406 static int
1407 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1408 {
1409 	struct ifbrparam *param = arg;
1410 
1411 	return (bstp_set_maxage(&sc->sc_stp, param->ifbrp_maxage));
1412 }
1413 
1414 static int
1415 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1416 {
1417 	struct ifbreq *req = arg;
1418 	struct bridge_iflist *bif;
1419 
1420 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1421 	if (bif == NULL)
1422 		return (ENOENT);
1423 
1424 	return (bstp_set_port_priority(&bif->bif_stp, req->ifbr_priority));
1425 }
1426 
1427 static int
1428 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1429 {
1430 	struct ifbreq *req = arg;
1431 	struct bridge_iflist *bif;
1432 
1433 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1434 	if (bif == NULL)
1435 		return (ENOENT);
1436 
1437 	return (bstp_set_path_cost(&bif->bif_stp, req->ifbr_path_cost));
1438 }
1439 
1440 static int
1441 bridge_ioctl_sifmaxaddr(struct bridge_softc *sc, void *arg)
1442 {
1443 	struct ifbreq *req = arg;
1444 	struct bridge_iflist *bif;
1445 
1446 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1447 	if (bif == NULL)
1448 		return (ENOENT);
1449 
1450 	bif->bif_addrmax = req->ifbr_addrmax;
1451 	return (0);
1452 }
1453 
1454 static int
1455 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1456 {
1457 	struct ifbreq *req = arg;
1458 	struct bridge_iflist *bif = NULL;
1459 	struct ifnet *ifs;
1460 
1461 	ifs = ifunit(req->ifbr_ifsname);
1462 	if (ifs == NULL)
1463 		return (ENOENT);
1464 
1465 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1466 		if (ifs == bif->bif_ifp)
1467 			return (EBUSY);
1468 
1469 	if (ifs->if_bridge != NULL)
1470 		return (EBUSY);
1471 
1472 	switch (ifs->if_type) {
1473 		case IFT_ETHER:
1474 		case IFT_GIF:
1475 		case IFT_L2VLAN:
1476 			break;
1477 		default:
1478 			return (EINVAL);
1479 	}
1480 
1481 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1482 	if (bif == NULL)
1483 		return (ENOMEM);
1484 
1485 	bif->bif_ifp = ifs;
1486 	bif->bif_flags = IFBIF_SPAN;
1487 
1488 	LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1489 
1490 	return (0);
1491 }
1492 
1493 static int
1494 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1495 {
1496 	struct ifbreq *req = arg;
1497 	struct bridge_iflist *bif;
1498 	struct ifnet *ifs;
1499 
1500 	ifs = ifunit(req->ifbr_ifsname);
1501 	if (ifs == NULL)
1502 		return (ENOENT);
1503 
1504 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1505 		if (ifs == bif->bif_ifp)
1506 			break;
1507 
1508 	if (bif == NULL)
1509 		return (ENOENT);
1510 
1511 	bridge_delete_span(sc, bif);
1512 
1513 	return (0);
1514 }
1515 
1516 static int
1517 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1518 {
1519 	struct ifbropreq *req = arg;
1520 	struct bstp_state *bs = &sc->sc_stp;
1521 	struct bstp_port *root_port;
1522 
1523 	req->ifbop_maxage = bs->bs_bridge_max_age >> 8;
1524 	req->ifbop_hellotime = bs->bs_bridge_htime >> 8;
1525 	req->ifbop_fwddelay = bs->bs_bridge_fdelay >> 8;
1526 
1527 	root_port = bs->bs_root_port;
1528 	if (root_port == NULL)
1529 		req->ifbop_root_port = 0;
1530 	else
1531 		req->ifbop_root_port = root_port->bp_ifp->if_index;
1532 
1533 	req->ifbop_holdcount = bs->bs_txholdcount;
1534 	req->ifbop_priority = bs->bs_bridge_priority;
1535 	req->ifbop_protocol = bs->bs_protover;
1536 	req->ifbop_root_path_cost = bs->bs_root_pv.pv_cost;
1537 	req->ifbop_bridgeid = bs->bs_bridge_pv.pv_dbridge_id;
1538 	req->ifbop_designated_root = bs->bs_root_pv.pv_root_id;
1539 	req->ifbop_designated_bridge = bs->bs_root_pv.pv_dbridge_id;
1540 	req->ifbop_last_tc_time.tv_sec = bs->bs_last_tc_time.tv_sec;
1541 	req->ifbop_last_tc_time.tv_usec = bs->bs_last_tc_time.tv_usec;
1542 
1543 	return (0);
1544 }
1545 
1546 static int
1547 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1548 {
1549 	struct ifbrparam *param = arg;
1550 
1551 	param->ifbrp_cexceeded = sc->sc_brtexceeded;
1552 	return (0);
1553 }
1554 
1555 static int
1556 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1557 {
1558 	struct ifbpstpconf *bifstp = arg;
1559 	struct bridge_iflist *bif;
1560 	struct bstp_port *bp;
1561 	struct ifbpstpreq bpreq;
1562 	char *buf, *outbuf;
1563 	int count, buflen, len, error = 0;
1564 
1565 	count = 0;
1566 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1567 		if ((bif->bif_flags & IFBIF_STP) != 0)
1568 			count++;
1569 	}
1570 
1571 	buflen = sizeof(bpreq) * count;
1572 	if (bifstp->ifbpstp_len == 0) {
1573 		bifstp->ifbpstp_len = buflen;
1574 		return (0);
1575 	}
1576 
1577 	BRIDGE_UNLOCK(sc);
1578 	outbuf = malloc(buflen, M_TEMP, M_WAITOK | M_ZERO);
1579 	BRIDGE_LOCK(sc);
1580 
1581 	count = 0;
1582 	buf = outbuf;
1583 	len = min(bifstp->ifbpstp_len, buflen);
1584 	bzero(&bpreq, sizeof(bpreq));
1585 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1586 		if (len < sizeof(bpreq))
1587 			break;
1588 
1589 		if ((bif->bif_flags & IFBIF_STP) == 0)
1590 			continue;
1591 
1592 		bp = &bif->bif_stp;
1593 		bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xfff;
1594 		bpreq.ifbp_fwd_trans = bp->bp_forward_transitions;
1595 		bpreq.ifbp_design_cost = bp->bp_desg_pv.pv_cost;
1596 		bpreq.ifbp_design_port = bp->bp_desg_pv.pv_port_id;
1597 		bpreq.ifbp_design_bridge = bp->bp_desg_pv.pv_dbridge_id;
1598 		bpreq.ifbp_design_root = bp->bp_desg_pv.pv_root_id;
1599 
1600 		memcpy(buf, &bpreq, sizeof(bpreq));
1601 		count++;
1602 		buf += sizeof(bpreq);
1603 		len -= sizeof(bpreq);
1604 	}
1605 
1606 	BRIDGE_UNLOCK(sc);
1607 	bifstp->ifbpstp_len = sizeof(bpreq) * count;
1608 	error = copyout(outbuf, bifstp->ifbpstp_req, bifstp->ifbpstp_len);
1609 	BRIDGE_LOCK(sc);
1610 	free(outbuf, M_TEMP);
1611 	return (error);
1612 }
1613 
1614 static int
1615 bridge_ioctl_sproto(struct bridge_softc *sc, void *arg)
1616 {
1617 	struct ifbrparam *param = arg;
1618 
1619 	return (bstp_set_protocol(&sc->sc_stp, param->ifbrp_proto));
1620 }
1621 
1622 static int
1623 bridge_ioctl_stxhc(struct bridge_softc *sc, void *arg)
1624 {
1625 	struct ifbrparam *param = arg;
1626 
1627 	return (bstp_set_holdcount(&sc->sc_stp, param->ifbrp_txhc));
1628 }
1629 
1630 /*
1631  * bridge_ifdetach:
1632  *
1633  *	Detach an interface from a bridge.  Called when a member
1634  *	interface is detaching.
1635  */
1636 static void
1637 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1638 {
1639 	struct bridge_softc *sc = ifp->if_bridge;
1640 	struct bridge_iflist *bif;
1641 
1642 	/* Check if the interface is a bridge member */
1643 	if (sc != NULL) {
1644 		BRIDGE_LOCK(sc);
1645 
1646 		bif = bridge_lookup_member_if(sc, ifp);
1647 		if (bif != NULL)
1648 			bridge_delete_member(sc, bif, 1);
1649 
1650 		BRIDGE_UNLOCK(sc);
1651 		return;
1652 	}
1653 
1654 	/* Check if the interface is a span port */
1655 	mtx_lock(&bridge_list_mtx);
1656 	LIST_FOREACH(sc, &bridge_list, sc_list) {
1657 		BRIDGE_LOCK(sc);
1658 		LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1659 			if (ifp == bif->bif_ifp) {
1660 				bridge_delete_span(sc, bif);
1661 				break;
1662 			}
1663 
1664 		BRIDGE_UNLOCK(sc);
1665 	}
1666 	mtx_unlock(&bridge_list_mtx);
1667 }
1668 
1669 /*
1670  * bridge_init:
1671  *
1672  *	Initialize a bridge interface.
1673  */
1674 static void
1675 bridge_init(void *xsc)
1676 {
1677 	struct bridge_softc *sc = (struct bridge_softc *)xsc;
1678 	struct ifnet *ifp = sc->sc_ifp;
1679 
1680 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1681 		return;
1682 
1683 	BRIDGE_LOCK(sc);
1684 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1685 	    bridge_timer, sc);
1686 
1687 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
1688 	bstp_init(&sc->sc_stp);		/* Initialize Spanning Tree */
1689 
1690 	BRIDGE_UNLOCK(sc);
1691 }
1692 
1693 /*
1694  * bridge_stop:
1695  *
1696  *	Stop the bridge interface.
1697  */
1698 static void
1699 bridge_stop(struct ifnet *ifp, int disable)
1700 {
1701 	struct bridge_softc *sc = ifp->if_softc;
1702 
1703 	BRIDGE_LOCK_ASSERT(sc);
1704 
1705 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1706 		return;
1707 
1708 	callout_stop(&sc->sc_brcallout);
1709 	bstp_stop(&sc->sc_stp);
1710 
1711 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1712 
1713 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1714 }
1715 
1716 /*
1717  * bridge_enqueue:
1718  *
1719  *	Enqueue a packet on a bridge member interface.
1720  *
1721  */
1722 static void
1723 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1724 {
1725 	int len, err = 0;
1726 	short mflags;
1727 	struct mbuf *m0;
1728 
1729 	len = m->m_pkthdr.len;
1730 	mflags = m->m_flags;
1731 
1732 	/* We may be sending a fragment so traverse the mbuf */
1733 	for (; m; m = m0) {
1734 		m0 = m->m_nextpkt;
1735 		m->m_nextpkt = NULL;
1736 
1737 		/*
1738 		 * If underlying interface can not do VLAN tag insertion itself
1739 		 * then attach a packet tag that holds it.
1740 		 */
1741 		if ((m->m_flags & M_VLANTAG) &&
1742 		    (dst_ifp->if_capenable & IFCAP_VLAN_HWTAGGING) == 0) {
1743 			m = ether_vlanencap(m, m->m_pkthdr.ether_vtag);
1744 			if (m == NULL) {
1745 				if_printf(dst_ifp,
1746 				    "unable to prepend VLAN header\n");
1747 				dst_ifp->if_oerrors++;
1748 				continue;
1749 			}
1750 			m->m_flags &= ~M_VLANTAG;
1751 		}
1752 
1753 		if (err == 0)
1754 			IFQ_ENQUEUE(&dst_ifp->if_snd, m, err);
1755 	}
1756 
1757 	if (err == 0) {
1758 
1759 		sc->sc_ifp->if_opackets++;
1760 		sc->sc_ifp->if_obytes += len;
1761 
1762 		dst_ifp->if_obytes += len;
1763 
1764 		if (mflags & M_MCAST) {
1765 			sc->sc_ifp->if_omcasts++;
1766 			dst_ifp->if_omcasts++;
1767 		}
1768 	}
1769 
1770 	if ((dst_ifp->if_drv_flags & IFF_DRV_OACTIVE) == 0)
1771 		(*dst_ifp->if_start)(dst_ifp);
1772 }
1773 
1774 /*
1775  * bridge_dummynet:
1776  *
1777  * 	Receive a queued packet from dummynet and pass it on to the output
1778  * 	interface.
1779  *
1780  *	The mbuf has the Ethernet header already attached.
1781  */
1782 static void
1783 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1784 {
1785 	struct bridge_softc *sc;
1786 
1787 	sc = ifp->if_bridge;
1788 
1789 	/*
1790 	 * The packet didnt originate from a member interface. This should only
1791 	 * ever happen if a member interface is removed while packets are
1792 	 * queued for it.
1793 	 */
1794 	if (sc == NULL) {
1795 		m_freem(m);
1796 		return;
1797 	}
1798 
1799 	if (PFIL_HOOKED(&inet_pfil_hook)
1800 #ifdef INET6
1801 	    || PFIL_HOOKED(&inet6_pfil_hook)
1802 #endif
1803 	    ) {
1804 		if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1805 			return;
1806 		if (m == NULL)
1807 			return;
1808 	}
1809 
1810 	bridge_enqueue(sc, ifp, m);
1811 }
1812 
1813 /*
1814  * bridge_output:
1815  *
1816  *	Send output from a bridge member interface.  This
1817  *	performs the bridging function for locally originated
1818  *	packets.
1819  *
1820  *	The mbuf has the Ethernet header already attached.  We must
1821  *	enqueue or free the mbuf before returning.
1822  */
1823 static int
1824 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1825     struct rtentry *rt)
1826 {
1827 	struct ether_header *eh;
1828 	struct ifnet *dst_if;
1829 	struct bridge_softc *sc;
1830 	uint16_t vlan;
1831 
1832 	if (m->m_len < ETHER_HDR_LEN) {
1833 		m = m_pullup(m, ETHER_HDR_LEN);
1834 		if (m == NULL)
1835 			return (0);
1836 	}
1837 
1838 	eh = mtod(m, struct ether_header *);
1839 	sc = ifp->if_bridge;
1840 	vlan = VLANTAGOF(m);
1841 
1842 	BRIDGE_LOCK(sc);
1843 
1844 	/*
1845 	 * If bridge is down, but the original output interface is up,
1846 	 * go ahead and send out that interface.  Otherwise, the packet
1847 	 * is dropped below.
1848 	 */
1849 	if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1850 		dst_if = ifp;
1851 		goto sendunicast;
1852 	}
1853 
1854 	/*
1855 	 * If the packet is a multicast, or we don't know a better way to
1856 	 * get there, send to all interfaces.
1857 	 */
1858 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
1859 		dst_if = NULL;
1860 	else
1861 		dst_if = bridge_rtlookup(sc, eh->ether_dhost, vlan);
1862 	if (dst_if == NULL) {
1863 		struct bridge_iflist *bif;
1864 		struct mbuf *mc;
1865 		int error = 0, used = 0;
1866 
1867 		bridge_span(sc, m);
1868 
1869 		BRIDGE_LOCK2REF(sc, error);
1870 		if (error) {
1871 			m_freem(m);
1872 			return (0);
1873 		}
1874 
1875 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1876 			dst_if = bif->bif_ifp;
1877 
1878 			if (dst_if->if_type == IFT_GIF)
1879 				continue;
1880 			if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1881 				continue;
1882 
1883 			/*
1884 			 * If this is not the original output interface,
1885 			 * and the interface is participating in spanning
1886 			 * tree, make sure the port is in a state that
1887 			 * allows forwarding.
1888 			 */
1889 			if (dst_if != ifp && (bif->bif_flags & IFBIF_STP) &&
1890 			    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1891 				continue;
1892 
1893 			if (LIST_NEXT(bif, bif_next) == NULL) {
1894 				used = 1;
1895 				mc = m;
1896 			} else {
1897 				mc = m_copypacket(m, M_DONTWAIT);
1898 				if (mc == NULL) {
1899 					sc->sc_ifp->if_oerrors++;
1900 					continue;
1901 				}
1902 			}
1903 
1904 			bridge_enqueue(sc, dst_if, mc);
1905 		}
1906 		if (used == 0)
1907 			m_freem(m);
1908 		BRIDGE_UNREF(sc);
1909 		return (0);
1910 	}
1911 
1912 sendunicast:
1913 	/*
1914 	 * XXX Spanning tree consideration here?
1915 	 */
1916 
1917 	bridge_span(sc, m);
1918 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1919 		m_freem(m);
1920 		BRIDGE_UNLOCK(sc);
1921 		return (0);
1922 	}
1923 
1924 	BRIDGE_UNLOCK(sc);
1925 	bridge_enqueue(sc, dst_if, m);
1926 	return (0);
1927 }
1928 
1929 /*
1930  * bridge_start:
1931  *
1932  *	Start output on a bridge.
1933  *
1934  */
1935 static void
1936 bridge_start(struct ifnet *ifp)
1937 {
1938 	struct bridge_softc *sc;
1939 	struct mbuf *m;
1940 	struct ether_header *eh;
1941 	struct ifnet *dst_if;
1942 
1943 	sc = ifp->if_softc;
1944 
1945 	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1946 	for (;;) {
1947 		IFQ_DEQUEUE(&ifp->if_snd, m);
1948 		if (m == 0)
1949 			break;
1950 		ETHER_BPF_MTAP(ifp, m);
1951 
1952 		eh = mtod(m, struct ether_header *);
1953 		dst_if = NULL;
1954 
1955 		BRIDGE_LOCK(sc);
1956 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1957 			dst_if = bridge_rtlookup(sc, eh->ether_dhost, 1);
1958 		}
1959 
1960 		if (dst_if == NULL)
1961 			bridge_broadcast(sc, ifp, m, 0);
1962 		else {
1963 			BRIDGE_UNLOCK(sc);
1964 			bridge_enqueue(sc, dst_if, m);
1965 		}
1966 	}
1967 	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
1968 }
1969 
1970 /*
1971  * bridge_forward:
1972  *
1973  *	The forwarding function of the bridge.
1974  *
1975  *	NOTE: Releases the lock on return.
1976  */
1977 static void
1978 bridge_forward(struct bridge_softc *sc, struct bridge_iflist *sbif,
1979     struct mbuf *m)
1980 {
1981 	struct bridge_iflist *dbif;
1982 	struct ifnet *src_if, *dst_if, *ifp;
1983 	struct ether_header *eh;
1984 	uint16_t vlan;
1985 	uint8_t *dst;
1986 	int error;
1987 
1988 	src_if = m->m_pkthdr.rcvif;
1989 	ifp = sc->sc_ifp;
1990 
1991 	ifp->if_ipackets++;
1992 	ifp->if_ibytes += m->m_pkthdr.len;
1993 	vlan = VLANTAGOF(m);
1994 
1995 	if ((sbif->bif_flags & IFBIF_STP) &&
1996 	    sbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
1997 		goto drop;
1998 
1999 	eh = mtod(m, struct ether_header *);
2000 	dst = eh->ether_dhost;
2001 
2002 	/* If the interface is learning, record the address. */
2003 	if (sbif->bif_flags & IFBIF_LEARNING) {
2004 		error = bridge_rtupdate(sc, eh->ether_shost, vlan,
2005 		    sbif, 0, IFBAF_DYNAMIC);
2006 		/*
2007 		 * If the interface has addresses limits then deny any source
2008 		 * that is not in the cache.
2009 		 */
2010 		if (error && sbif->bif_addrmax)
2011 			goto drop;
2012 	}
2013 
2014 	if ((sbif->bif_flags & IFBIF_STP) != 0 &&
2015 	    sbif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING)
2016 		goto drop;
2017 
2018 	/*
2019 	 * At this point, the port either doesn't participate
2020 	 * in spanning tree or it is in the forwarding state.
2021 	 */
2022 
2023 	/*
2024 	 * If the packet is unicast, destined for someone on
2025 	 * "this" side of the bridge, drop it.
2026 	 */
2027 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
2028 		dst_if = bridge_rtlookup(sc, dst, vlan);
2029 		if (src_if == dst_if)
2030 			goto drop;
2031 	} else {
2032 		/*
2033 		 * Check if its a reserved multicast address, any address
2034 		 * listed in 802.1D section 7.12.6 may not be forwarded by the
2035 		 * bridge.
2036 		 * This is currently 01-80-C2-00-00-00 to 01-80-C2-00-00-0F
2037 		 */
2038 		if (dst[0] == 0x01 && dst[1] == 0x80 &&
2039 		    dst[2] == 0xc2 && dst[3] == 0x00 &&
2040 		    dst[4] == 0x00 && dst[5] <= 0x0f)
2041 			goto drop;
2042 
2043 		/* ...forward it to all interfaces. */
2044 		ifp->if_imcasts++;
2045 		dst_if = NULL;
2046 	}
2047 
2048 	/*
2049 	 * If we have a destination interface which is a member of our bridge,
2050 	 * OR this is a unicast packet, push it through the bpf(4) machinery.
2051 	 * For broadcast or multicast packets, don't bother because it will
2052 	 * be reinjected into ether_input. We do this before we pass the packets
2053 	 * through the pfil(9) framework, as it is possible that pfil(9) will
2054 	 * drop the packet, or possibly modify it, making it difficult to debug
2055 	 * firewall issues on the bridge.
2056 	 */
2057 	if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
2058 		ETHER_BPF_MTAP(ifp, m);
2059 
2060 	/* run the packet filter */
2061 	if (PFIL_HOOKED(&inet_pfil_hook)
2062 #ifdef INET6
2063 	    || PFIL_HOOKED(&inet6_pfil_hook)
2064 #endif
2065 	    ) {
2066 		BRIDGE_UNLOCK(sc);
2067 		if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
2068 			return;
2069 		if (m == NULL)
2070 			return;
2071 		BRIDGE_LOCK(sc);
2072 	}
2073 
2074 	if (dst_if == NULL) {
2075 		bridge_broadcast(sc, src_if, m, 1);
2076 		return;
2077 	}
2078 
2079 	/*
2080 	 * At this point, we're dealing with a unicast frame
2081 	 * going to a different interface.
2082 	 */
2083 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2084 		goto drop;
2085 
2086 	dbif = bridge_lookup_member_if(sc, dst_if);
2087 	if (dbif == NULL)
2088 		/* Not a member of the bridge (anymore?) */
2089 		goto drop;
2090 
2091 	/* Private segments can not talk to each other */
2092 	if (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE)
2093 		goto drop;
2094 
2095 	if ((dbif->bif_flags & IFBIF_STP) &&
2096 	    dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2097 		goto drop;
2098 
2099 	BRIDGE_UNLOCK(sc);
2100 
2101 	if (PFIL_HOOKED(&inet_pfil_hook)
2102 #ifdef INET6
2103 	    || PFIL_HOOKED(&inet6_pfil_hook)
2104 #endif
2105 	    ) {
2106 		if (bridge_pfil(&m, ifp, dst_if, PFIL_OUT) != 0)
2107 			return;
2108 		if (m == NULL)
2109 			return;
2110 	}
2111 
2112 	bridge_enqueue(sc, dst_if, m);
2113 	return;
2114 
2115 drop:
2116 	BRIDGE_UNLOCK(sc);
2117 	m_freem(m);
2118 }
2119 
2120 /*
2121  * bridge_input:
2122  *
2123  *	Receive input from a member interface.  Queue the packet for
2124  *	bridging if it is not for us.
2125  */
2126 static struct mbuf *
2127 bridge_input(struct ifnet *ifp, struct mbuf *m)
2128 {
2129 	struct bridge_softc *sc = ifp->if_bridge;
2130 	struct bridge_iflist *bif, *bif2;
2131 	struct ifnet *bifp;
2132 	struct ether_header *eh;
2133 	struct mbuf *mc, *mc2;
2134 	uint16_t vlan;
2135 	int error;
2136 
2137 	if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2138 		return (m);
2139 
2140 	bifp = sc->sc_ifp;
2141 	vlan = VLANTAGOF(m);
2142 
2143 	/*
2144 	 * Implement support for bridge monitoring. If this flag has been
2145 	 * set on this interface, discard the packet once we push it through
2146 	 * the bpf(4) machinery, but before we do, increment the byte and
2147 	 * packet counters associated with this interface.
2148 	 */
2149 	if ((bifp->if_flags & IFF_MONITOR) != 0) {
2150 		m->m_pkthdr.rcvif  = bifp;
2151 		ETHER_BPF_MTAP(bifp, m);
2152 		bifp->if_ipackets++;
2153 		bifp->if_ibytes += m->m_pkthdr.len;
2154 		m_freem(m);
2155 		return (NULL);
2156 	}
2157 	BRIDGE_LOCK(sc);
2158 	bif = bridge_lookup_member_if(sc, ifp);
2159 	if (bif == NULL) {
2160 		BRIDGE_UNLOCK(sc);
2161 		return (m);
2162 	}
2163 
2164 	eh = mtod(m, struct ether_header *);
2165 
2166 	bridge_span(sc, m);
2167 
2168 	if (m->m_flags & (M_BCAST|M_MCAST)) {
2169 		/* Tap off 802.1D packets; they do not get forwarded. */
2170 		if (memcmp(eh->ether_dhost, bstp_etheraddr,
2171 		    ETHER_ADDR_LEN) == 0) {
2172 			m = bstp_input(&bif->bif_stp, ifp, m);
2173 			if (m == NULL) {
2174 				BRIDGE_UNLOCK(sc);
2175 				return (NULL);
2176 			}
2177 		}
2178 
2179 		if ((bif->bif_flags & IFBIF_STP) &&
2180 		    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2181 			BRIDGE_UNLOCK(sc);
2182 			return (m);
2183 		}
2184 
2185 		/*
2186 		 * Make a deep copy of the packet and enqueue the copy
2187 		 * for bridge processing; return the original packet for
2188 		 * local processing.
2189 		 */
2190 		mc = m_dup(m, M_DONTWAIT);
2191 		if (mc == NULL) {
2192 			BRIDGE_UNLOCK(sc);
2193 			return (m);
2194 		}
2195 
2196 		/* Perform the bridge forwarding function with the copy. */
2197 		bridge_forward(sc, bif, mc);
2198 
2199 		/*
2200 		 * Reinject the mbuf as arriving on the bridge so we have a
2201 		 * chance at claiming multicast packets. We can not loop back
2202 		 * here from ether_input as a bridge is never a member of a
2203 		 * bridge.
2204 		 */
2205 		KASSERT(bifp->if_bridge == NULL,
2206 		    ("loop created in bridge_input"));
2207 		mc2 = m_dup(m, M_DONTWAIT);
2208 		if (mc2 != NULL) {
2209 			/* Keep the layer3 header aligned */
2210 			int i = min(mc2->m_pkthdr.len, max_protohdr);
2211 			mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2212 		}
2213 		if (mc2 != NULL) {
2214 			mc2->m_pkthdr.rcvif = bifp;
2215 			(*bifp->if_input)(bifp, mc2);
2216 		}
2217 
2218 		/* Return the original packet for local processing. */
2219 		return (m);
2220 	}
2221 
2222 	if ((bif->bif_flags & IFBIF_STP) &&
2223 	    bif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING) {
2224 		BRIDGE_UNLOCK(sc);
2225 		return (m);
2226 	}
2227 
2228 #ifdef DEV_CARP
2229 #   define OR_CARP_CHECK_WE_ARE_DST(iface) \
2230 	|| ((iface)->if_carp \
2231 	    && carp_forus((iface)->if_carp, eh->ether_dhost))
2232 #   define OR_CARP_CHECK_WE_ARE_SRC(iface) \
2233 	|| ((iface)->if_carp \
2234 	    && carp_forus((iface)->if_carp, eh->ether_shost))
2235 #else
2236 #   define OR_CARP_CHECK_WE_ARE_DST(iface)
2237 #   define OR_CARP_CHECK_WE_ARE_SRC(iface)
2238 #endif
2239 
2240 #ifdef INET6
2241 #   define OR_PFIL_HOOKED_INET6 \
2242 	|| PFIL_HOOKED(&inet6_pfil_hook)
2243 #else
2244 #   define OR_PFIL_HOOKED_INET6
2245 #endif
2246 
2247 #define GRAB_OUR_PACKETS(iface) \
2248 	if ((iface)->if_type == IFT_GIF) \
2249 		continue; \
2250 	/* It is destined for us. */ \
2251 	if (memcmp(IF_LLADDR((iface)), eh->ether_dhost,  ETHER_ADDR_LEN) == 0 \
2252 	    OR_CARP_CHECK_WE_ARE_DST((iface))				\
2253 	    ) {								\
2254 		if ((iface)->if_type == IFT_BRIDGE) {			\
2255 			ETHER_BPF_MTAP(iface, m);			\
2256 			iface->if_ipackets++;				\
2257 			/* Filter on the physical interface. */		\
2258 			if (pfil_local_phys &&				\
2259 			    (PFIL_HOOKED(&inet_pfil_hook)		\
2260 			     OR_PFIL_HOOKED_INET6)) {			\
2261 				if (bridge_pfil(&m, NULL, ifp,		\
2262 				    PFIL_IN) != 0 || m == NULL) {	\
2263 					BRIDGE_UNLOCK(sc);		\
2264 					return (NULL);			\
2265 				}					\
2266 			}						\
2267 		}							\
2268 		if (bif->bif_flags & IFBIF_LEARNING) {			\
2269 			error = bridge_rtupdate(sc, eh->ether_shost,	\
2270 			    vlan, bif, 0, IFBAF_DYNAMIC);		\
2271 			if (error && bif->bif_addrmax) {		\
2272 				BRIDGE_UNLOCK(sc);			\
2273 				m_freem(m);				\
2274 				return (NULL);				\
2275 			}						\
2276 		}							\
2277 		m->m_pkthdr.rcvif = iface;				\
2278 		BRIDGE_UNLOCK(sc);					\
2279 		return (m);						\
2280 	}								\
2281 									\
2282 	/* We just received a packet that we sent out. */		\
2283 	if (memcmp(IF_LLADDR((iface)), eh->ether_shost, ETHER_ADDR_LEN) == 0 \
2284 	    OR_CARP_CHECK_WE_ARE_SRC((iface))			\
2285 	    ) {								\
2286 		BRIDGE_UNLOCK(sc);					\
2287 		m_freem(m);						\
2288 		return (NULL);						\
2289 	}
2290 
2291 	/*
2292 	 * Unicast.  Make sure it's not for the bridge.
2293 	 */
2294 	do { GRAB_OUR_PACKETS(bifp) } while (0);
2295 
2296 	/*
2297 	 * Give a chance for ifp at first priority. This will help when	the
2298 	 * packet comes through the interface like VLAN's with the same MACs
2299 	 * on several interfaces from the same bridge. This also will save
2300 	 * some CPU cycles in case the destination interface and the input
2301 	 * interface (eq ifp) are the same.
2302 	 */
2303 	do { GRAB_OUR_PACKETS(ifp) } while (0);
2304 
2305 	/* Now check the all bridge members. */
2306 	LIST_FOREACH(bif2, &sc->sc_iflist, bif_next) {
2307 		GRAB_OUR_PACKETS(bif2->bif_ifp)
2308 	}
2309 
2310 #undef OR_CARP_CHECK_WE_ARE_DST
2311 #undef OR_CARP_CHECK_WE_ARE_SRC
2312 #undef OR_PFIL_HOOKED_INET6
2313 #undef GRAB_OUR_PACKETS
2314 
2315 	/* Perform the bridge forwarding function. */
2316 	bridge_forward(sc, bif, m);
2317 
2318 	return (NULL);
2319 }
2320 
2321 /*
2322  * bridge_broadcast:
2323  *
2324  *	Send a frame to all interfaces that are members of
2325  *	the bridge, except for the one on which the packet
2326  *	arrived.
2327  *
2328  *	NOTE: Releases the lock on return.
2329  */
2330 static void
2331 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2332     struct mbuf *m, int runfilt)
2333 {
2334 	struct bridge_iflist *dbif, *sbif;
2335 	struct mbuf *mc;
2336 	struct ifnet *dst_if;
2337 	int error = 0, used = 0, i;
2338 
2339 	sbif = bridge_lookup_member_if(sc, src_if);
2340 
2341 	BRIDGE_LOCK2REF(sc, error);
2342 	if (error) {
2343 		m_freem(m);
2344 		return;
2345 	}
2346 
2347 	/* Filter on the bridge interface before broadcasting */
2348 	if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2349 #ifdef INET6
2350 	    || PFIL_HOOKED(&inet6_pfil_hook)
2351 #endif
2352 	    )) {
2353 		if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2354 			goto out;
2355 		if (m == NULL)
2356 			goto out;
2357 	}
2358 
2359 	LIST_FOREACH(dbif, &sc->sc_iflist, bif_next) {
2360 		dst_if = dbif->bif_ifp;
2361 		if (dst_if == src_if)
2362 			continue;
2363 
2364 		/* Private segments can not talk to each other */
2365 		if (sbif && (sbif->bif_flags & dbif->bif_flags & IFBIF_PRIVATE))
2366 			continue;
2367 
2368 		if ((dbif->bif_flags & IFBIF_STP) &&
2369 		    dbif->bif_stp.bp_state == BSTP_IFSTATE_DISCARDING)
2370 			continue;
2371 
2372 		if ((dbif->bif_flags & IFBIF_DISCOVER) == 0 &&
2373 		    (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2374 			continue;
2375 
2376 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2377 			continue;
2378 
2379 		if (LIST_NEXT(dbif, bif_next) == NULL) {
2380 			mc = m;
2381 			used = 1;
2382 		} else {
2383 			mc = m_dup(m, M_DONTWAIT);
2384 			if (mc == NULL) {
2385 				sc->sc_ifp->if_oerrors++;
2386 				continue;
2387 			}
2388 		}
2389 
2390 		/*
2391 		 * Filter on the output interface. Pass a NULL bridge interface
2392 		 * pointer so we do not redundantly filter on the bridge for
2393 		 * each interface we broadcast on.
2394 		 */
2395 		if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2396 #ifdef INET6
2397 		    || PFIL_HOOKED(&inet6_pfil_hook)
2398 #endif
2399 		    )) {
2400 			if (used == 0) {
2401 				/* Keep the layer3 header aligned */
2402 				i = min(mc->m_pkthdr.len, max_protohdr);
2403 				mc = m_copyup(mc, i, ETHER_ALIGN);
2404 				if (mc == NULL) {
2405 					sc->sc_ifp->if_oerrors++;
2406 					continue;
2407 				}
2408 			}
2409 			if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2410 				continue;
2411 			if (mc == NULL)
2412 				continue;
2413 		}
2414 
2415 		bridge_enqueue(sc, dst_if, mc);
2416 	}
2417 	if (used == 0)
2418 		m_freem(m);
2419 
2420 out:
2421 	BRIDGE_UNREF(sc);
2422 }
2423 
2424 /*
2425  * bridge_span:
2426  *
2427  *	Duplicate a packet out one or more interfaces that are in span mode,
2428  *	the original mbuf is unmodified.
2429  */
2430 static void
2431 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2432 {
2433 	struct bridge_iflist *bif;
2434 	struct ifnet *dst_if;
2435 	struct mbuf *mc;
2436 
2437 	if (LIST_EMPTY(&sc->sc_spanlist))
2438 		return;
2439 
2440 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2441 		dst_if = bif->bif_ifp;
2442 
2443 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2444 			continue;
2445 
2446 		mc = m_copypacket(m, M_DONTWAIT);
2447 		if (mc == NULL) {
2448 			sc->sc_ifp->if_oerrors++;
2449 			continue;
2450 		}
2451 
2452 		bridge_enqueue(sc, dst_if, mc);
2453 	}
2454 }
2455 
2456 /*
2457  * bridge_rtupdate:
2458  *
2459  *	Add a bridge routing entry.
2460  */
2461 static int
2462 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst, uint16_t vlan,
2463     struct bridge_iflist *bif, int setflags, uint8_t flags)
2464 {
2465 	struct bridge_rtnode *brt;
2466 	int error;
2467 
2468 	BRIDGE_LOCK_ASSERT(sc);
2469 
2470 	/* Check the source address is valid and not multicast. */
2471 	if (ETHER_IS_MULTICAST(dst) ||
2472 	    (dst[0] == 0 && dst[1] == 0 && dst[2] == 0 &&
2473 	     dst[3] == 0 && dst[4] == 0 && dst[5] == 0) != 0)
2474 		return (EINVAL);
2475 
2476 	/* 802.1p frames map to vlan 1 */
2477 	if (vlan == 0)
2478 		vlan = 1;
2479 
2480 	/*
2481 	 * A route for this destination might already exist.  If so,
2482 	 * update it, otherwise create a new one.
2483 	 */
2484 	if ((brt = bridge_rtnode_lookup(sc, dst, vlan)) == NULL) {
2485 		if (sc->sc_brtcnt >= sc->sc_brtmax) {
2486 			sc->sc_brtexceeded++;
2487 			return (ENOSPC);
2488 		}
2489 		/* Check per interface address limits (if enabled) */
2490 		if (bif->bif_addrmax && bif->bif_addrcnt >= bif->bif_addrmax) {
2491 			bif->bif_addrexceeded++;
2492 			return (ENOSPC);
2493 		}
2494 
2495 		/*
2496 		 * Allocate a new bridge forwarding node, and
2497 		 * initialize the expiration time and Ethernet
2498 		 * address.
2499 		 */
2500 		brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2501 		if (brt == NULL)
2502 			return (ENOMEM);
2503 
2504 		if (bif->bif_flags & IFBIF_STICKY)
2505 			brt->brt_flags = IFBAF_STICKY;
2506 		else
2507 			brt->brt_flags = IFBAF_DYNAMIC;
2508 
2509 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2510 		brt->brt_vlan = vlan;
2511 
2512 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2513 			uma_zfree(bridge_rtnode_zone, brt);
2514 			return (error);
2515 		}
2516 		brt->brt_dst = bif;
2517 		bif->bif_addrcnt++;
2518 	}
2519 
2520 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
2521 	    brt->brt_dst != bif) {
2522 		brt->brt_dst->bif_addrcnt--;
2523 		brt->brt_dst = bif;
2524 		brt->brt_dst->bif_addrcnt++;
2525 	}
2526 
2527 	if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2528 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
2529 	if (setflags)
2530 		brt->brt_flags = flags;
2531 
2532 	return (0);
2533 }
2534 
2535 /*
2536  * bridge_rtlookup:
2537  *
2538  *	Lookup the destination interface for an address.
2539  */
2540 static struct ifnet *
2541 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2542 {
2543 	struct bridge_rtnode *brt;
2544 
2545 	BRIDGE_LOCK_ASSERT(sc);
2546 
2547 	if ((brt = bridge_rtnode_lookup(sc, addr, vlan)) == NULL)
2548 		return (NULL);
2549 
2550 	return (brt->brt_ifp);
2551 }
2552 
2553 /*
2554  * bridge_rttrim:
2555  *
2556  *	Trim the routine table so that we have a number
2557  *	of routing entries less than or equal to the
2558  *	maximum number.
2559  */
2560 static void
2561 bridge_rttrim(struct bridge_softc *sc)
2562 {
2563 	struct bridge_rtnode *brt, *nbrt;
2564 
2565 	BRIDGE_LOCK_ASSERT(sc);
2566 
2567 	/* Make sure we actually need to do this. */
2568 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2569 		return;
2570 
2571 	/* Force an aging cycle; this might trim enough addresses. */
2572 	bridge_rtage(sc);
2573 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2574 		return;
2575 
2576 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2577 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2578 			bridge_rtnode_destroy(sc, brt);
2579 			if (sc->sc_brtcnt <= sc->sc_brtmax)
2580 				return;
2581 		}
2582 	}
2583 }
2584 
2585 /*
2586  * bridge_timer:
2587  *
2588  *	Aging timer for the bridge.
2589  */
2590 static void
2591 bridge_timer(void *arg)
2592 {
2593 	struct bridge_softc *sc = arg;
2594 
2595 	BRIDGE_LOCK_ASSERT(sc);
2596 
2597 	bridge_rtage(sc);
2598 
2599 	if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2600 		callout_reset(&sc->sc_brcallout,
2601 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
2602 }
2603 
2604 /*
2605  * bridge_rtage:
2606  *
2607  *	Perform an aging cycle.
2608  */
2609 static void
2610 bridge_rtage(struct bridge_softc *sc)
2611 {
2612 	struct bridge_rtnode *brt, *nbrt;
2613 
2614 	BRIDGE_LOCK_ASSERT(sc);
2615 
2616 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2617 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2618 			if (time_uptime >= brt->brt_expire)
2619 				bridge_rtnode_destroy(sc, brt);
2620 		}
2621 	}
2622 }
2623 
2624 /*
2625  * bridge_rtflush:
2626  *
2627  *	Remove all dynamic addresses from the bridge.
2628  */
2629 static void
2630 bridge_rtflush(struct bridge_softc *sc, int full)
2631 {
2632 	struct bridge_rtnode *brt, *nbrt;
2633 
2634 	BRIDGE_LOCK_ASSERT(sc);
2635 
2636 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2637 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2638 			bridge_rtnode_destroy(sc, brt);
2639 	}
2640 }
2641 
2642 /*
2643  * bridge_rtdaddr:
2644  *
2645  *	Remove an address from the table.
2646  */
2647 static int
2648 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2649 {
2650 	struct bridge_rtnode *brt;
2651 	int found = 0;
2652 
2653 	BRIDGE_LOCK_ASSERT(sc);
2654 
2655 	/*
2656 	 * If vlan is zero then we want to delete for all vlans so the lookup
2657 	 * may return more than one.
2658 	 */
2659 	while ((brt = bridge_rtnode_lookup(sc, addr, vlan)) != NULL) {
2660 		bridge_rtnode_destroy(sc, brt);
2661 		found = 1;
2662 	}
2663 
2664 	return (found ? 0 : ENOENT);
2665 }
2666 
2667 /*
2668  * bridge_rtdelete:
2669  *
2670  *	Delete routes to a speicifc member interface.
2671  */
2672 static void
2673 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2674 {
2675 	struct bridge_rtnode *brt, *nbrt;
2676 
2677 	BRIDGE_LOCK_ASSERT(sc);
2678 
2679 	LIST_FOREACH_SAFE(brt, &sc->sc_rtlist, brt_list, nbrt) {
2680 		if (brt->brt_ifp == ifp && (full ||
2681 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2682 			bridge_rtnode_destroy(sc, brt);
2683 	}
2684 }
2685 
2686 /*
2687  * bridge_rtable_init:
2688  *
2689  *	Initialize the route table for this bridge.
2690  */
2691 static int
2692 bridge_rtable_init(struct bridge_softc *sc)
2693 {
2694 	int i;
2695 
2696 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2697 	    M_DEVBUF, M_NOWAIT);
2698 	if (sc->sc_rthash == NULL)
2699 		return (ENOMEM);
2700 
2701 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2702 		LIST_INIT(&sc->sc_rthash[i]);
2703 
2704 	sc->sc_rthash_key = arc4random();
2705 
2706 	LIST_INIT(&sc->sc_rtlist);
2707 
2708 	return (0);
2709 }
2710 
2711 /*
2712  * bridge_rtable_fini:
2713  *
2714  *	Deconstruct the route table for this bridge.
2715  */
2716 static void
2717 bridge_rtable_fini(struct bridge_softc *sc)
2718 {
2719 
2720 	KASSERT(sc->sc_brtcnt == 0,
2721 	    ("%s: %d bridge routes referenced", __func__, sc->sc_brtcnt));
2722 	free(sc->sc_rthash, M_DEVBUF);
2723 }
2724 
2725 /*
2726  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2727  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2728  */
2729 #define	mix(a, b, c)							\
2730 do {									\
2731 	a -= b; a -= c; a ^= (c >> 13);					\
2732 	b -= c; b -= a; b ^= (a << 8);					\
2733 	c -= a; c -= b; c ^= (b >> 13);					\
2734 	a -= b; a -= c; a ^= (c >> 12);					\
2735 	b -= c; b -= a; b ^= (a << 16);					\
2736 	c -= a; c -= b; c ^= (b >> 5);					\
2737 	a -= b; a -= c; a ^= (c >> 3);					\
2738 	b -= c; b -= a; b ^= (a << 10);					\
2739 	c -= a; c -= b; c ^= (b >> 15);					\
2740 } while (/*CONSTCOND*/0)
2741 
2742 static __inline uint32_t
2743 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2744 {
2745 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2746 
2747 	b += addr[5] << 8;
2748 	b += addr[4];
2749 	a += addr[3] << 24;
2750 	a += addr[2] << 16;
2751 	a += addr[1] << 8;
2752 	a += addr[0];
2753 
2754 	mix(a, b, c);
2755 
2756 	return (c & BRIDGE_RTHASH_MASK);
2757 }
2758 
2759 #undef mix
2760 
2761 static int
2762 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2763 {
2764 	int i, d;
2765 
2766 	for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2767 		d = ((int)a[i]) - ((int)b[i]);
2768 	}
2769 
2770 	return (d);
2771 }
2772 
2773 /*
2774  * bridge_rtnode_lookup:
2775  *
2776  *	Look up a bridge route node for the specified destination. Compare the
2777  *	vlan id or if zero then just return the first match.
2778  */
2779 static struct bridge_rtnode *
2780 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr, uint16_t vlan)
2781 {
2782 	struct bridge_rtnode *brt;
2783 	uint32_t hash;
2784 	int dir;
2785 
2786 	BRIDGE_LOCK_ASSERT(sc);
2787 
2788 	hash = bridge_rthash(sc, addr);
2789 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2790 		dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2791 		if (dir == 0 && (brt->brt_vlan == vlan || vlan == 0))
2792 			return (brt);
2793 		if (dir > 0)
2794 			return (NULL);
2795 	}
2796 
2797 	return (NULL);
2798 }
2799 
2800 /*
2801  * bridge_rtnode_insert:
2802  *
2803  *	Insert the specified bridge node into the route table.  We
2804  *	assume the entry is not already in the table.
2805  */
2806 static int
2807 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2808 {
2809 	struct bridge_rtnode *lbrt;
2810 	uint32_t hash;
2811 	int dir;
2812 
2813 	BRIDGE_LOCK_ASSERT(sc);
2814 
2815 	hash = bridge_rthash(sc, brt->brt_addr);
2816 
2817 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2818 	if (lbrt == NULL) {
2819 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2820 		goto out;
2821 	}
2822 
2823 	do {
2824 		dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2825 		if (dir == 0 && brt->brt_vlan == lbrt->brt_vlan)
2826 			return (EEXIST);
2827 		if (dir > 0) {
2828 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2829 			goto out;
2830 		}
2831 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2832 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2833 			goto out;
2834 		}
2835 		lbrt = LIST_NEXT(lbrt, brt_hash);
2836 	} while (lbrt != NULL);
2837 
2838 #ifdef DIAGNOSTIC
2839 	panic("bridge_rtnode_insert: impossible");
2840 #endif
2841 
2842 out:
2843 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2844 	sc->sc_brtcnt++;
2845 
2846 	return (0);
2847 }
2848 
2849 /*
2850  * bridge_rtnode_destroy:
2851  *
2852  *	Destroy a bridge rtnode.
2853  */
2854 static void
2855 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2856 {
2857 	BRIDGE_LOCK_ASSERT(sc);
2858 
2859 	LIST_REMOVE(brt, brt_hash);
2860 
2861 	LIST_REMOVE(brt, brt_list);
2862 	sc->sc_brtcnt--;
2863 	brt->brt_dst->bif_addrcnt--;
2864 	uma_zfree(bridge_rtnode_zone, brt);
2865 }
2866 
2867 /*
2868  * bridge_rtable_expire:
2869  *
2870  *	Set the expiry time for all routes on an interface.
2871  */
2872 static void
2873 bridge_rtable_expire(struct ifnet *ifp, int age)
2874 {
2875 	struct bridge_softc *sc = ifp->if_bridge;
2876 	struct bridge_rtnode *brt;
2877 
2878 	BRIDGE_LOCK(sc);
2879 
2880 	/*
2881 	 * If the age is zero then flush, otherwise set all the expiry times to
2882 	 * age for the interface
2883 	 */
2884 	if (age == 0)
2885 		bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2886 	else {
2887 		LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
2888 			/* Cap the expiry time to 'age' */
2889 			if (brt->brt_ifp == ifp &&
2890 			    brt->brt_expire > time_uptime + age &&
2891 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2892 				brt->brt_expire = time_uptime + age;
2893 		}
2894 	}
2895 	BRIDGE_UNLOCK(sc);
2896 }
2897 
2898 /*
2899  * bridge_state_change:
2900  *
2901  *	Callback from the bridgestp code when a port changes states.
2902  */
2903 static void
2904 bridge_state_change(struct ifnet *ifp, int state)
2905 {
2906 	struct bridge_softc *sc = ifp->if_bridge;
2907 	static const char *stpstates[] = {
2908 		"disabled",
2909 		"listening",
2910 		"learning",
2911 		"forwarding",
2912 		"blocking",
2913 		"discarding"
2914 	};
2915 
2916 	if (log_stp)
2917 		log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2918 		    sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2919 }
2920 
2921 /*
2922  * Send bridge packets through pfil if they are one of the types pfil can deal
2923  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2924  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2925  * that interface.
2926  */
2927 static int
2928 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2929 {
2930 	int snap, error, i, hlen;
2931 	struct ether_header *eh1, eh2;
2932 	struct ip_fw_args args;
2933 	struct ip *ip;
2934 	struct llc llc1;
2935 	u_int16_t ether_type;
2936 
2937 	snap = 0;
2938 	error = -1;	/* Default error if not error == 0 */
2939 
2940 #if 0
2941 	/* we may return with the IP fields swapped, ensure its not shared */
2942 	KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2943 #endif
2944 
2945 	if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2946 		return (0); /* filtering is disabled */
2947 
2948 	i = min((*mp)->m_pkthdr.len, max_protohdr);
2949 	if ((*mp)->m_len < i) {
2950 	    *mp = m_pullup(*mp, i);
2951 	    if (*mp == NULL) {
2952 		printf("%s: m_pullup failed\n", __func__);
2953 		return (-1);
2954 	    }
2955 	}
2956 
2957 	eh1 = mtod(*mp, struct ether_header *);
2958 	ether_type = ntohs(eh1->ether_type);
2959 
2960 	/*
2961 	 * Check for SNAP/LLC.
2962 	 */
2963 	if (ether_type < ETHERMTU) {
2964 		struct llc *llc2 = (struct llc *)(eh1 + 1);
2965 
2966 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2967 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
2968 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
2969 		    llc2->llc_control == LLC_UI) {
2970 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
2971 			snap = 1;
2972 		}
2973 	}
2974 
2975 	/*
2976 	 * If we're trying to filter bridge traffic, don't look at anything
2977 	 * other than IP and ARP traffic.  If the filter doesn't understand
2978 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2979 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2980 	 * but of course we don't have an AppleTalk filter to begin with.
2981 	 * (Note that since pfil doesn't understand ARP it will pass *ALL*
2982 	 * ARP traffic.)
2983 	 */
2984 	switch (ether_type) {
2985 		case ETHERTYPE_ARP:
2986 		case ETHERTYPE_REVARP:
2987 			if (pfil_ipfw_arp == 0)
2988 				return (0); /* Automatically pass */
2989 			break;
2990 
2991 		case ETHERTYPE_IP:
2992 #ifdef INET6
2993 		case ETHERTYPE_IPV6:
2994 #endif /* INET6 */
2995 			break;
2996 		default:
2997 			/*
2998 			 * Check to see if the user wants to pass non-ip
2999 			 * packets, these will not be checked by pfil(9) and
3000 			 * passed unconditionally so the default is to drop.
3001 			 */
3002 			if (pfil_onlyip)
3003 				goto bad;
3004 	}
3005 
3006 	/* Strip off the Ethernet header and keep a copy. */
3007 	m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
3008 	m_adj(*mp, ETHER_HDR_LEN);
3009 
3010 	/* Strip off snap header, if present */
3011 	if (snap) {
3012 		m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
3013 		m_adj(*mp, sizeof(struct llc));
3014 	}
3015 
3016 	/*
3017 	 * Check the IP header for alignment and errors
3018 	 */
3019 	if (dir == PFIL_IN) {
3020 		switch (ether_type) {
3021 			case ETHERTYPE_IP:
3022 				error = bridge_ip_checkbasic(mp);
3023 				break;
3024 #ifdef INET6
3025 			case ETHERTYPE_IPV6:
3026 				error = bridge_ip6_checkbasic(mp);
3027 				break;
3028 #endif /* INET6 */
3029 			default:
3030 				error = 0;
3031 		}
3032 		if (error)
3033 			goto bad;
3034 	}
3035 
3036 	if (IPFW_LOADED && pfil_ipfw != 0 && dir == PFIL_OUT && ifp != NULL) {
3037 		error = -1;
3038 		args.rule = ip_dn_claim_rule(*mp);
3039 		if (args.rule != NULL && V_fw_one_pass)
3040 			goto ipfwpass; /* packet already partially processed */
3041 
3042 		args.m = *mp;
3043 		args.oif = ifp;
3044 		args.next_hop = NULL;
3045 		args.eh = &eh2;
3046 		args.inp = NULL;	/* used by ipfw uid/gid/jail rules */
3047 		i = ip_fw_chk_ptr(&args);
3048 		*mp = args.m;
3049 
3050 		if (*mp == NULL)
3051 			return (error);
3052 
3053 		if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
3054 
3055 			/* put the Ethernet header back on */
3056 			M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3057 			if (*mp == NULL)
3058 				return (error);
3059 			bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3060 
3061 			/*
3062 			 * Pass the pkt to dummynet, which consumes it. The
3063 			 * packet will return to us via bridge_dummynet().
3064 			 */
3065 			args.oif = ifp;
3066 			ip_dn_io_ptr(mp, DN_TO_IFB_FWD, &args);
3067 			return (error);
3068 		}
3069 
3070 		if (i != IP_FW_PASS) /* drop */
3071 			goto bad;
3072 	}
3073 
3074 ipfwpass:
3075 	error = 0;
3076 
3077 	/*
3078 	 * Run the packet through pfil
3079 	 */
3080 	switch (ether_type) {
3081 	case ETHERTYPE_IP:
3082 		/*
3083 		 * before calling the firewall, swap fields the same as
3084 		 * IP does. here we assume the header is contiguous
3085 		 */
3086 		ip = mtod(*mp, struct ip *);
3087 
3088 		ip->ip_len = ntohs(ip->ip_len);
3089 		ip->ip_off = ntohs(ip->ip_off);
3090 
3091 		/*
3092 		 * Run pfil on the member interface and the bridge, both can
3093 		 * be skipped by clearing pfil_member or pfil_bridge.
3094 		 *
3095 		 * Keep the order:
3096 		 *   in_if -> bridge_if -> out_if
3097 		 */
3098 		if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3099 			error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3100 					dir, NULL);
3101 
3102 		if (*mp == NULL || error != 0) /* filter may consume */
3103 			break;
3104 
3105 		if (pfil_member && ifp != NULL)
3106 			error = pfil_run_hooks(&inet_pfil_hook, mp, ifp,
3107 					dir, NULL);
3108 
3109 		if (*mp == NULL || error != 0) /* filter may consume */
3110 			break;
3111 
3112 		if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3113 			error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
3114 					dir, NULL);
3115 
3116 		if (*mp == NULL || error != 0) /* filter may consume */
3117 			break;
3118 
3119 		/* check if we need to fragment the packet */
3120 		if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
3121 			i = (*mp)->m_pkthdr.len;
3122 			if (i > ifp->if_mtu) {
3123 				error = bridge_fragment(ifp, *mp, &eh2, snap,
3124 					    &llc1);
3125 				return (error);
3126 			}
3127 		}
3128 
3129 		/* Recalculate the ip checksum and restore byte ordering */
3130 		ip = mtod(*mp, struct ip *);
3131 		hlen = ip->ip_hl << 2;
3132 		if (hlen < sizeof(struct ip))
3133 			goto bad;
3134 		if (hlen > (*mp)->m_len) {
3135 			if ((*mp = m_pullup(*mp, hlen)) == 0)
3136 				goto bad;
3137 			ip = mtod(*mp, struct ip *);
3138 			if (ip == NULL)
3139 				goto bad;
3140 		}
3141 		ip->ip_len = htons(ip->ip_len);
3142 		ip->ip_off = htons(ip->ip_off);
3143 		ip->ip_sum = 0;
3144 		if (hlen == sizeof(struct ip))
3145 			ip->ip_sum = in_cksum_hdr(ip);
3146 		else
3147 			ip->ip_sum = in_cksum(*mp, hlen);
3148 
3149 		break;
3150 #ifdef INET6
3151 	case ETHERTYPE_IPV6:
3152 		if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
3153 			error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3154 					dir, NULL);
3155 
3156 		if (*mp == NULL || error != 0) /* filter may consume */
3157 			break;
3158 
3159 		if (pfil_member && ifp != NULL)
3160 			error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
3161 					dir, NULL);
3162 
3163 		if (*mp == NULL || error != 0) /* filter may consume */
3164 			break;
3165 
3166 		if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
3167 			error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
3168 					dir, NULL);
3169 		break;
3170 #endif
3171 	default:
3172 		error = 0;
3173 		break;
3174 	}
3175 
3176 	if (*mp == NULL)
3177 		return (error);
3178 	if (error != 0)
3179 		goto bad;
3180 
3181 	error = -1;
3182 
3183 	/*
3184 	 * Finally, put everything back the way it was and return
3185 	 */
3186 	if (snap) {
3187 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3188 		if (*mp == NULL)
3189 			return (error);
3190 		bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3191 	}
3192 
3193 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3194 	if (*mp == NULL)
3195 		return (error);
3196 	bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3197 
3198 	return (0);
3199 
3200 bad:
3201 	m_freem(*mp);
3202 	*mp = NULL;
3203 	return (error);
3204 }
3205 
3206 /*
3207  * Perform basic checks on header size since
3208  * pfil assumes ip_input has already processed
3209  * it for it.  Cut-and-pasted from ip_input.c.
3210  * Given how simple the IPv6 version is,
3211  * does the IPv4 version really need to be
3212  * this complicated?
3213  *
3214  * XXX Should we update ipstat here, or not?
3215  * XXX Right now we update ipstat but not
3216  * XXX csum_counter.
3217  */
3218 static int
3219 bridge_ip_checkbasic(struct mbuf **mp)
3220 {
3221 	struct mbuf *m = *mp;
3222 	struct ip *ip;
3223 	int len, hlen;
3224 	u_short sum;
3225 
3226 	if (*mp == NULL)
3227 		return (-1);
3228 
3229 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3230 		if ((m = m_copyup(m, sizeof(struct ip),
3231 			(max_linkhdr + 3) & ~3)) == NULL) {
3232 			/* XXXJRT new stat, please */
3233 			V_ipstat.ips_toosmall++;
3234 			goto bad;
3235 		}
3236 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
3237 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3238 			V_ipstat.ips_toosmall++;
3239 			goto bad;
3240 		}
3241 	}
3242 	ip = mtod(m, struct ip *);
3243 	if (ip == NULL) goto bad;
3244 
3245 	if (ip->ip_v != IPVERSION) {
3246 		V_ipstat.ips_badvers++;
3247 		goto bad;
3248 	}
3249 	hlen = ip->ip_hl << 2;
3250 	if (hlen < sizeof(struct ip)) { /* minimum header length */
3251 		V_ipstat.ips_badhlen++;
3252 		goto bad;
3253 	}
3254 	if (hlen > m->m_len) {
3255 		if ((m = m_pullup(m, hlen)) == 0) {
3256 			V_ipstat.ips_badhlen++;
3257 			goto bad;
3258 		}
3259 		ip = mtod(m, struct ip *);
3260 		if (ip == NULL) goto bad;
3261 	}
3262 
3263 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3264 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3265 	} else {
3266 		if (hlen == sizeof(struct ip)) {
3267 			sum = in_cksum_hdr(ip);
3268 		} else {
3269 			sum = in_cksum(m, hlen);
3270 		}
3271 	}
3272 	if (sum) {
3273 		V_ipstat.ips_badsum++;
3274 		goto bad;
3275 	}
3276 
3277 	/* Retrieve the packet length. */
3278 	len = ntohs(ip->ip_len);
3279 
3280 	/*
3281 	 * Check for additional length bogosity
3282 	 */
3283 	if (len < hlen) {
3284 		V_ipstat.ips_badlen++;
3285 		goto bad;
3286 	}
3287 
3288 	/*
3289 	 * Check that the amount of data in the buffers
3290 	 * is as at least much as the IP header would have us expect.
3291 	 * Drop packet if shorter than we expect.
3292 	 */
3293 	if (m->m_pkthdr.len < len) {
3294 		V_ipstat.ips_tooshort++;
3295 		goto bad;
3296 	}
3297 
3298 	/* Checks out, proceed */
3299 	*mp = m;
3300 	return (0);
3301 
3302 bad:
3303 	*mp = m;
3304 	return (-1);
3305 }
3306 
3307 #ifdef INET6
3308 /*
3309  * Same as above, but for IPv6.
3310  * Cut-and-pasted from ip6_input.c.
3311  * XXX Should we update ip6stat, or not?
3312  */
3313 static int
3314 bridge_ip6_checkbasic(struct mbuf **mp)
3315 {
3316 	struct mbuf *m = *mp;
3317 	struct ip6_hdr *ip6;
3318 
3319 	/*
3320 	 * If the IPv6 header is not aligned, slurp it up into a new
3321 	 * mbuf with space for link headers, in the event we forward
3322 	 * it.  Otherwise, if it is aligned, make sure the entire base
3323 	 * IPv6 header is in the first mbuf of the chain.
3324 	 */
3325 	if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3326 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3327 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3328 			    (max_linkhdr + 3) & ~3)) == NULL) {
3329 			/* XXXJRT new stat, please */
3330 			V_ip6stat.ip6s_toosmall++;
3331 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3332 			goto bad;
3333 		}
3334 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3335 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3336 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3337 			V_ip6stat.ip6s_toosmall++;
3338 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3339 			goto bad;
3340 		}
3341 	}
3342 
3343 	ip6 = mtod(m, struct ip6_hdr *);
3344 
3345 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3346 		V_ip6stat.ip6s_badvers++;
3347 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3348 		goto bad;
3349 	}
3350 
3351 	/* Checks out, proceed */
3352 	*mp = m;
3353 	return (0);
3354 
3355 bad:
3356 	*mp = m;
3357 	return (-1);
3358 }
3359 #endif /* INET6 */
3360 
3361 /*
3362  * bridge_fragment:
3363  *
3364  *	Return a fragmented mbuf chain.
3365  */
3366 static int
3367 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3368     int snap, struct llc *llc)
3369 {
3370 	struct mbuf *m0;
3371 	struct ip *ip;
3372 	int error = -1;
3373 
3374 	if (m->m_len < sizeof(struct ip) &&
3375 	    (m = m_pullup(m, sizeof(struct ip))) == NULL)
3376 		goto out;
3377 	ip = mtod(m, struct ip *);
3378 
3379 	error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3380 		    CSUM_DELAY_IP);
3381 	if (error)
3382 		goto out;
3383 
3384 	/* walk the chain and re-add the Ethernet header */
3385 	for (m0 = m; m0; m0 = m0->m_nextpkt) {
3386 		if (error == 0) {
3387 			if (snap) {
3388 				M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3389 				if (m0 == NULL) {
3390 					error = ENOBUFS;
3391 					continue;
3392 				}
3393 				bcopy(llc, mtod(m0, caddr_t),
3394 				    sizeof(struct llc));
3395 			}
3396 			M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3397 			if (m0 == NULL) {
3398 				error = ENOBUFS;
3399 				continue;
3400 			}
3401 			bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3402 		} else
3403 			m_freem(m);
3404 	}
3405 
3406 	if (error == 0)
3407 		V_ipstat.ips_fragmented++;
3408 
3409 	return (error);
3410 
3411 out:
3412 	if (m != NULL)
3413 		m_freem(m);
3414 	return (error);
3415 }
3416