xref: /freebsd/sys/net/if_bridge.c (revision e4e9813eb92cd7c4d4b819a8fbed5cbd3d92f5d8)
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  * 3. All advertising materials mentioning features or use of this software
51  *    must display the following acknowledgement:
52  *	This product includes software developed by Jason L. Wright
53  * 4. The name of the author may not be used to endorse or promote products
54  *    derived from this software without specific prior written permission.
55  *
56  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
57  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
58  * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE
59  * DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT,
60  * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
61  * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
62  * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
63  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
64  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN
65  * ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
66  * POSSIBILITY OF SUCH DAMAGE.
67  *
68  * OpenBSD: if_bridge.c,v 1.60 2001/06/15 03:38:33 itojun Exp
69  */
70 
71 /*
72  * Network interface bridge support.
73  *
74  * TODO:
75  *
76  *	- Currently only supports Ethernet-like interfaces (Ethernet,
77  *	  802.11, VLANs on Ethernet, etc.)  Figure out a nice way
78  *	  to bridge other types of interfaces (FDDI-FDDI, and maybe
79  *	  consider heterogenous bridges).
80  */
81 
82 #include <sys/cdefs.h>
83 __FBSDID("$FreeBSD$");
84 
85 #include "opt_inet.h"
86 #include "opt_inet6.h"
87 #include "opt_carp.h"
88 
89 #include <sys/param.h>
90 #include <sys/mbuf.h>
91 #include <sys/malloc.h>
92 #include <sys/protosw.h>
93 #include <sys/systm.h>
94 #include <sys/time.h>
95 #include <sys/socket.h> /* for net/if.h */
96 #include <sys/sockio.h>
97 #include <sys/ctype.h>  /* string functions */
98 #include <sys/kernel.h>
99 #include <sys/random.h>
100 #include <sys/syslog.h>
101 #include <sys/sysctl.h>
102 #include <vm/uma.h>
103 #include <sys/module.h>
104 #include <sys/proc.h>
105 #include <sys/lock.h>
106 #include <sys/mutex.h>
107 
108 #include <net/bpf.h>
109 #include <net/if.h>
110 #include <net/if_clone.h>
111 #include <net/if_dl.h>
112 #include <net/if_types.h>
113 #include <net/if_var.h>
114 #include <net/pfil.h>
115 
116 #include <netinet/in.h> /* for struct arpcom */
117 #include <netinet/in_systm.h>
118 #include <netinet/in_var.h>
119 #include <netinet/ip.h>
120 #include <netinet/ip_var.h>
121 #ifdef INET6
122 #include <netinet/ip6.h>
123 #include <netinet6/ip6_var.h>
124 #endif
125 #ifdef DEV_CARP
126 #include <netinet/ip_carp.h>
127 #endif
128 #include <machine/in_cksum.h>
129 #include <netinet/if_ether.h> /* for struct arpcom */
130 #include <net/bridgestp.h>
131 #include <net/if_bridgevar.h>
132 #include <net/if_llc.h>
133 
134 #include <net/route.h>
135 #include <netinet/ip_fw.h>
136 #include <netinet/ip_dummynet.h>
137 
138 /*
139  * Size of the route hash table.  Must be a power of two.
140  */
141 #ifndef BRIDGE_RTHASH_SIZE
142 #define	BRIDGE_RTHASH_SIZE		1024
143 #endif
144 
145 #define	BRIDGE_RTHASH_MASK		(BRIDGE_RTHASH_SIZE - 1)
146 
147 /*
148  * Maximum number of addresses to cache.
149  */
150 #ifndef BRIDGE_RTABLE_MAX
151 #define	BRIDGE_RTABLE_MAX		100
152 #endif
153 
154 /*
155  * Timeout (in seconds) for entries learned dynamically.
156  */
157 #ifndef BRIDGE_RTABLE_TIMEOUT
158 #define	BRIDGE_RTABLE_TIMEOUT		(20 * 60)	/* same as ARP */
159 #endif
160 
161 /*
162  * Number of seconds between walks of the route list.
163  */
164 #ifndef BRIDGE_RTABLE_PRUNE_PERIOD
165 #define	BRIDGE_RTABLE_PRUNE_PERIOD	(5 * 60)
166 #endif
167 
168 /*
169  * List of capabilities to mask on the member interface.
170  */
171 #define	BRIDGE_IFCAPS_MASK		IFCAP_TXCSUM
172 
173 /*
174  * Bridge interface list entry.
175  */
176 struct bridge_iflist {
177 	LIST_ENTRY(bridge_iflist) bif_next;
178 	struct ifnet		*bif_ifp;	/* member if */
179 	struct bstp_port	bif_stp;	/* STP state */
180 	uint32_t		bif_flags;	/* member if flags */
181 	int			bif_mutecap;	/* member muted caps */
182 };
183 
184 /*
185  * Bridge route node.
186  */
187 struct bridge_rtnode {
188 	LIST_ENTRY(bridge_rtnode) brt_hash;	/* hash table linkage */
189 	LIST_ENTRY(bridge_rtnode) brt_list;	/* list linkage */
190 	struct ifnet		*brt_ifp;	/* destination if */
191 	unsigned long		brt_expire;	/* expiration time */
192 	uint8_t			brt_flags;	/* address flags */
193 	uint8_t			brt_addr[ETHER_ADDR_LEN];
194 };
195 
196 /*
197  * Software state for each bridge.
198  */
199 struct bridge_softc {
200 	struct ifnet		*sc_ifp;	/* make this an interface */
201 	LIST_ENTRY(bridge_softc) sc_list;
202 	struct mtx		sc_mtx;
203 	struct cv		sc_cv;
204 	uint32_t		sc_brtmax;	/* max # of addresses */
205 	uint32_t		sc_brtcnt;	/* cur. # of addresses */
206 	uint32_t		sc_brttimeout;	/* rt timeout in seconds */
207 	struct callout		sc_brcallout;	/* bridge callout */
208 	uint32_t		sc_iflist_ref;	/* refcount for sc_iflist */
209 	uint32_t		sc_iflist_xcnt;	/* refcount for sc_iflist */
210 	LIST_HEAD(, bridge_iflist) sc_iflist;	/* member interface list */
211 	LIST_HEAD(, bridge_rtnode) *sc_rthash;	/* our forwarding table */
212 	LIST_HEAD(, bridge_rtnode) sc_rtlist;	/* list version of above */
213 	uint32_t		sc_rthash_key;	/* key for hash */
214 	LIST_HEAD(, bridge_iflist) sc_spanlist;	/* span ports list */
215 	struct bstp_state	sc_stp;		/* STP state */
216 	uint32_t		sc_brtexceeded;	/* # of cache drops */
217 };
218 
219 static struct mtx 	bridge_list_mtx;
220 eventhandler_tag	bridge_detach_cookie = NULL;
221 
222 int	bridge_rtable_prune_period = BRIDGE_RTABLE_PRUNE_PERIOD;
223 
224 uma_zone_t bridge_rtnode_zone;
225 
226 static int	bridge_clone_create(struct if_clone *, int, caddr_t);
227 static void	bridge_clone_destroy(struct ifnet *);
228 
229 static int	bridge_ioctl(struct ifnet *, u_long, caddr_t);
230 static void	bridge_mutecaps(struct bridge_iflist *, int);
231 static void	bridge_ifdetach(void *arg __unused, struct ifnet *);
232 static void	bridge_init(void *);
233 static void	bridge_dummynet(struct mbuf *, struct ifnet *);
234 static void	bridge_stop(struct ifnet *, int);
235 static void	bridge_start(struct ifnet *);
236 static struct mbuf *bridge_input(struct ifnet *, struct mbuf *);
237 static int	bridge_output(struct ifnet *, struct mbuf *, struct sockaddr *,
238 		    struct rtentry *);
239 static void	bridge_enqueue(struct bridge_softc *, struct ifnet *,
240 		    struct mbuf *);
241 static void	bridge_rtdelete(struct bridge_softc *, struct ifnet *ifp, int);
242 
243 static void	bridge_forward(struct bridge_softc *, struct mbuf *m);
244 
245 static void	bridge_timer(void *);
246 
247 static void	bridge_broadcast(struct bridge_softc *, struct ifnet *,
248 		    struct mbuf *, int);
249 static void	bridge_span(struct bridge_softc *, struct mbuf *);
250 
251 static int	bridge_rtupdate(struct bridge_softc *, const uint8_t *,
252 		    struct ifnet *, int, uint8_t);
253 static struct ifnet *bridge_rtlookup(struct bridge_softc *, const uint8_t *);
254 static void	bridge_rttrim(struct bridge_softc *);
255 static void	bridge_rtage(struct bridge_softc *);
256 static void	bridge_rtflush(struct bridge_softc *, int);
257 static int	bridge_rtdaddr(struct bridge_softc *, const uint8_t *);
258 
259 static int	bridge_rtable_init(struct bridge_softc *);
260 static void	bridge_rtable_fini(struct bridge_softc *);
261 
262 static int	bridge_rtnode_addr_cmp(const uint8_t *, const uint8_t *);
263 static struct bridge_rtnode *bridge_rtnode_lookup(struct bridge_softc *,
264 		    const uint8_t *);
265 static int	bridge_rtnode_insert(struct bridge_softc *,
266 		    struct bridge_rtnode *);
267 static void	bridge_rtnode_destroy(struct bridge_softc *,
268 		    struct bridge_rtnode *);
269 static void	bridge_state_change(struct ifnet *, int);
270 
271 static struct bridge_iflist *bridge_lookup_member(struct bridge_softc *,
272 		    const char *name);
273 static struct bridge_iflist *bridge_lookup_member_if(struct bridge_softc *,
274 		    struct ifnet *ifp);
275 static void	bridge_delete_member(struct bridge_softc *,
276 		    struct bridge_iflist *, int);
277 static void	bridge_delete_span(struct bridge_softc *,
278 		    struct bridge_iflist *);
279 
280 static int	bridge_ioctl_add(struct bridge_softc *, void *);
281 static int	bridge_ioctl_del(struct bridge_softc *, void *);
282 static int	bridge_ioctl_gifflags(struct bridge_softc *, void *);
283 static int	bridge_ioctl_sifflags(struct bridge_softc *, void *);
284 static int	bridge_ioctl_scache(struct bridge_softc *, void *);
285 static int	bridge_ioctl_gcache(struct bridge_softc *, void *);
286 static int	bridge_ioctl_gifs(struct bridge_softc *, void *);
287 static int	bridge_ioctl_rts(struct bridge_softc *, void *);
288 static int	bridge_ioctl_saddr(struct bridge_softc *, void *);
289 static int	bridge_ioctl_sto(struct bridge_softc *, void *);
290 static int	bridge_ioctl_gto(struct bridge_softc *, void *);
291 static int	bridge_ioctl_daddr(struct bridge_softc *, void *);
292 static int	bridge_ioctl_flush(struct bridge_softc *, void *);
293 static int	bridge_ioctl_gpri(struct bridge_softc *, void *);
294 static int	bridge_ioctl_spri(struct bridge_softc *, void *);
295 static int	bridge_ioctl_ght(struct bridge_softc *, void *);
296 static int	bridge_ioctl_sht(struct bridge_softc *, void *);
297 static int	bridge_ioctl_gfd(struct bridge_softc *, void *);
298 static int	bridge_ioctl_sfd(struct bridge_softc *, void *);
299 static int	bridge_ioctl_gma(struct bridge_softc *, void *);
300 static int	bridge_ioctl_sma(struct bridge_softc *, void *);
301 static int	bridge_ioctl_sifprio(struct bridge_softc *, void *);
302 static int	bridge_ioctl_sifcost(struct bridge_softc *, void *);
303 static int	bridge_ioctl_addspan(struct bridge_softc *, void *);
304 static int	bridge_ioctl_delspan(struct bridge_softc *, void *);
305 static int	bridge_ioctl_gbparam(struct bridge_softc *, void *);
306 static int	bridge_ioctl_grte(struct bridge_softc *, void *);
307 static int	bridge_ioctl_gifsstp(struct bridge_softc *, void *);
308 static int	bridge_pfil(struct mbuf **, struct ifnet *, struct ifnet *,
309 		    int);
310 static int	bridge_ip_checkbasic(struct mbuf **mp);
311 #ifdef INET6
312 static int	bridge_ip6_checkbasic(struct mbuf **mp);
313 #endif /* INET6 */
314 static int	bridge_fragment(struct ifnet *, struct mbuf *,
315 		    struct ether_header *, int, struct llc *);
316 
317 SYSCTL_DECL(_net_link);
318 SYSCTL_NODE(_net_link, IFT_BRIDGE, bridge, CTLFLAG_RW, 0, "Bridge");
319 
320 static int pfil_onlyip = 1; /* only pass IP[46] packets when pfil is enabled */
321 static int pfil_bridge = 1; /* run pfil hooks on the bridge interface */
322 static int pfil_member = 1; /* run pfil hooks on the member interface */
323 static int pfil_ipfw = 0;   /* layer2 filter with ipfw */
324 static int log_stp   = 0;   /* log STP state changes */
325 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_onlyip, CTLFLAG_RW,
326     &pfil_onlyip, 0, "Only pass IP packets when pfil is enabled");
327 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_bridge, CTLFLAG_RW,
328     &pfil_bridge, 0, "Packet filter on the bridge interface");
329 SYSCTL_INT(_net_link_bridge, OID_AUTO, pfil_member, CTLFLAG_RW,
330     &pfil_member, 0, "Packet filter on the member interface");
331 SYSCTL_INT(_net_link_bridge, OID_AUTO, log_stp, CTLFLAG_RW,
332     &log_stp, 0, "Log STP state changes");
333 
334 struct bridge_control {
335 	int	(*bc_func)(struct bridge_softc *, void *);
336 	int	bc_argsize;
337 	int	bc_flags;
338 };
339 
340 #define	BC_F_COPYIN		0x01	/* copy arguments in */
341 #define	BC_F_COPYOUT		0x02	/* copy arguments out */
342 #define	BC_F_SUSER		0x04	/* do super-user check */
343 
344 const struct bridge_control bridge_control_table[] = {
345 	{ bridge_ioctl_add,		sizeof(struct ifbreq),
346 	  BC_F_COPYIN|BC_F_SUSER },
347 	{ bridge_ioctl_del,		sizeof(struct ifbreq),
348 	  BC_F_COPYIN|BC_F_SUSER },
349 
350 	{ bridge_ioctl_gifflags,	sizeof(struct ifbreq),
351 	  BC_F_COPYIN|BC_F_COPYOUT },
352 	{ bridge_ioctl_sifflags,	sizeof(struct ifbreq),
353 	  BC_F_COPYIN|BC_F_SUSER },
354 
355 	{ bridge_ioctl_scache,		sizeof(struct ifbrparam),
356 	  BC_F_COPYIN|BC_F_SUSER },
357 	{ bridge_ioctl_gcache,		sizeof(struct ifbrparam),
358 	  BC_F_COPYOUT },
359 
360 	{ bridge_ioctl_gifs,		sizeof(struct ifbifconf),
361 	  BC_F_COPYIN|BC_F_COPYOUT },
362 	{ bridge_ioctl_rts,		sizeof(struct ifbaconf),
363 	  BC_F_COPYIN|BC_F_COPYOUT },
364 
365 	{ bridge_ioctl_saddr,		sizeof(struct ifbareq),
366 	  BC_F_COPYIN|BC_F_SUSER },
367 
368 	{ bridge_ioctl_sto,		sizeof(struct ifbrparam),
369 	  BC_F_COPYIN|BC_F_SUSER },
370 	{ bridge_ioctl_gto,		sizeof(struct ifbrparam),
371 	  BC_F_COPYOUT },
372 
373 	{ bridge_ioctl_daddr,		sizeof(struct ifbareq),
374 	  BC_F_COPYIN|BC_F_SUSER },
375 
376 	{ bridge_ioctl_flush,		sizeof(struct ifbreq),
377 	  BC_F_COPYIN|BC_F_SUSER },
378 
379 	{ bridge_ioctl_gpri,		sizeof(struct ifbrparam),
380 	  BC_F_COPYOUT },
381 	{ bridge_ioctl_spri,		sizeof(struct ifbrparam),
382 	  BC_F_COPYIN|BC_F_SUSER },
383 
384 	{ bridge_ioctl_ght,		sizeof(struct ifbrparam),
385 	  BC_F_COPYOUT },
386 	{ bridge_ioctl_sht,		sizeof(struct ifbrparam),
387 	  BC_F_COPYIN|BC_F_SUSER },
388 
389 	{ bridge_ioctl_gfd,		sizeof(struct ifbrparam),
390 	  BC_F_COPYOUT },
391 	{ bridge_ioctl_sfd,		sizeof(struct ifbrparam),
392 	  BC_F_COPYIN|BC_F_SUSER },
393 
394 	{ bridge_ioctl_gma,		sizeof(struct ifbrparam),
395 	  BC_F_COPYOUT },
396 	{ bridge_ioctl_sma,		sizeof(struct ifbrparam),
397 	  BC_F_COPYIN|BC_F_SUSER },
398 
399 	{ bridge_ioctl_sifprio,		sizeof(struct ifbreq),
400 	  BC_F_COPYIN|BC_F_SUSER },
401 
402 	{ bridge_ioctl_sifcost,		sizeof(struct ifbreq),
403 	  BC_F_COPYIN|BC_F_SUSER },
404 
405 	{ bridge_ioctl_addspan,		sizeof(struct ifbreq),
406 	  BC_F_COPYIN|BC_F_SUSER },
407 	{ bridge_ioctl_delspan,		sizeof(struct ifbreq),
408 	  BC_F_COPYIN|BC_F_SUSER },
409 
410 	{ bridge_ioctl_gbparam,		sizeof(struct ifbropreq),
411 	  BC_F_COPYOUT },
412 
413 	{ bridge_ioctl_grte,		sizeof(struct ifbrparam),
414 	  BC_F_COPYOUT },
415 
416 	{ bridge_ioctl_gifsstp,		sizeof(struct ifbpstpconf),
417 	  BC_F_COPYOUT },
418 };
419 const int bridge_control_table_size =
420     sizeof(bridge_control_table) / sizeof(bridge_control_table[0]);
421 
422 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
423 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
424 
425 LIST_HEAD(, bridge_softc) bridge_list;
426 
427 IFC_SIMPLE_DECLARE(bridge, 0);
428 
429 static int
430 bridge_modevent(module_t mod, int type, void *data)
431 {
432 
433 	switch (type) {
434 	case MOD_LOAD:
435 		mtx_init(&bridge_list_mtx, "if_bridge list", NULL, MTX_DEF);
436 		if_clone_attach(&bridge_cloner);
437 		bridge_rtnode_zone = uma_zcreate("bridge_rtnode",
438 		    sizeof(struct bridge_rtnode), NULL, NULL, NULL, NULL,
439 		    UMA_ALIGN_PTR, 0);
440 		LIST_INIT(&bridge_list);
441 		bridge_input_p = bridge_input;
442 		bridge_output_p = bridge_output;
443 		bridge_dn_p = bridge_dummynet;
444 		bstp_linkstate_p = bstp_linkstate;
445 		bridge_detach_cookie = EVENTHANDLER_REGISTER(
446 		    ifnet_departure_event, bridge_ifdetach, NULL,
447 		    EVENTHANDLER_PRI_ANY);
448 		break;
449 	case MOD_UNLOAD:
450 		EVENTHANDLER_DEREGISTER(ifnet_departure_event,
451 		    bridge_detach_cookie);
452 		if_clone_detach(&bridge_cloner);
453 		uma_zdestroy(bridge_rtnode_zone);
454 		bridge_input_p = NULL;
455 		bridge_output_p = NULL;
456 		bridge_dn_p = NULL;
457 		bstp_linkstate_p = NULL;
458 		mtx_destroy(&bridge_list_mtx);
459 		break;
460 	default:
461 		return (EOPNOTSUPP);
462 	}
463 	return (0);
464 }
465 
466 static moduledata_t bridge_mod = {
467 	"if_bridge",
468 	bridge_modevent,
469 	0
470 };
471 
472 DECLARE_MODULE(if_bridge, bridge_mod, SI_SUB_PSEUDO, SI_ORDER_ANY);
473 MODULE_DEPEND(if_bridge, bridgestp, 1, 1, 1);
474 
475 /*
476  * handler for net.link.bridge.pfil_ipfw
477  */
478 static int
479 sysctl_pfil_ipfw(SYSCTL_HANDLER_ARGS)
480 {
481 	int enable = pfil_ipfw;
482 	int error;
483 
484 	error = sysctl_handle_int(oidp, &enable, 0, req);
485 	enable = (enable) ? 1 : 0;
486 
487 	if (enable != pfil_ipfw) {
488 		pfil_ipfw = enable;
489 
490 		/*
491 		 * Disable pfil so that ipfw doesnt run twice, if the user
492 		 * really wants both then they can re-enable pfil_bridge and/or
493 		 * pfil_member. Also allow non-ip packets as ipfw can filter by
494 		 * layer2 type.
495 		 */
496 		if (pfil_ipfw) {
497 			pfil_onlyip = 0;
498 			pfil_bridge = 0;
499 			pfil_member = 0;
500 		}
501 	}
502 
503 	return (error);
504 }
505 SYSCTL_PROC(_net_link_bridge, OID_AUTO, ipfw, CTLTYPE_INT|CTLFLAG_RW,
506 	    &pfil_ipfw, 0, &sysctl_pfil_ipfw, "I", "Layer2 filter with IPFW");
507 
508 /*
509  * bridge_clone_create:
510  *
511  *	Create a new bridge instance.
512  */
513 static int
514 bridge_clone_create(struct if_clone *ifc, int unit, caddr_t params)
515 {
516 	struct bridge_softc *sc, *sc2;
517 	struct ifnet *bifp, *ifp;
518 	u_char eaddr[6];
519 	int retry;
520 
521 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
522 	BRIDGE_LOCK_INIT(sc);
523 	ifp = sc->sc_ifp = if_alloc(IFT_ETHER);
524 	if (ifp == NULL) {
525 		free(sc, M_DEVBUF);
526 		return (ENOSPC);
527 	}
528 
529 	sc->sc_brtmax = BRIDGE_RTABLE_MAX;
530 	sc->sc_brttimeout = BRIDGE_RTABLE_TIMEOUT;
531 	getmicrotime(&(sc->sc_stp.bs_last_tc_time));
532 
533 	/* Initialize our routing table. */
534 	bridge_rtable_init(sc);
535 
536 	callout_init_mtx(&sc->sc_brcallout, &sc->sc_mtx, 0);
537 
538 	LIST_INIT(&sc->sc_iflist);
539 	LIST_INIT(&sc->sc_spanlist);
540 
541 	ifp->if_softc = sc;
542 	if_initname(ifp, ifc->ifc_name, unit);
543 	ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
544 	ifp->if_ioctl = bridge_ioctl;
545 	ifp->if_start = bridge_start;
546 	ifp->if_init = bridge_init;
547 	ifp->if_type = IFT_BRIDGE;
548 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
549 	ifp->if_snd.ifq_drv_maxlen = ifqmaxlen;
550 	IFQ_SET_READY(&ifp->if_snd);
551 
552 	/*
553 	 * Generate a random ethernet address with a locally administered
554 	 * address.
555 	 *
556 	 * Since we are using random ethernet addresses for the bridge, it is
557 	 * possible that we might have address collisions, so make sure that
558 	 * this hardware address isn't already in use on another bridge.
559 	 */
560 	for (retry = 1; retry != 0;) {
561 		arc4rand(eaddr, ETHER_ADDR_LEN, 1);
562 		eaddr[0] &= ~1;		/* clear multicast bit */
563 		eaddr[0] |= 2;		/* set the LAA bit */
564 		retry = 0;
565 		mtx_lock(&bridge_list_mtx);
566 		LIST_FOREACH(sc2, &bridge_list, sc_list) {
567 			bifp = sc2->sc_ifp;
568 			if (memcmp(eaddr, IF_LLADDR(bifp), ETHER_ADDR_LEN) == 0)
569 				retry = 1;
570 		}
571 		mtx_unlock(&bridge_list_mtx);
572 	}
573 
574 	bstp_attach(&sc->sc_stp, bridge_state_change);
575 	ether_ifattach(ifp, eaddr);
576 	/* Now undo some of the damage... */
577 	ifp->if_baudrate = 0;
578 	ifp->if_type = IFT_BRIDGE;
579 
580 	mtx_lock(&bridge_list_mtx);
581 	LIST_INSERT_HEAD(&bridge_list, sc, sc_list);
582 	mtx_unlock(&bridge_list_mtx);
583 
584 	return (0);
585 }
586 
587 /*
588  * bridge_clone_destroy:
589  *
590  *	Destroy a bridge instance.
591  */
592 static void
593 bridge_clone_destroy(struct ifnet *ifp)
594 {
595 	struct bridge_softc *sc = ifp->if_softc;
596 	struct bridge_iflist *bif;
597 
598 	BRIDGE_LOCK(sc);
599 
600 	bridge_stop(ifp, 1);
601 	ifp->if_flags &= ~IFF_UP;
602 
603 	while ((bif = LIST_FIRST(&sc->sc_iflist)) != NULL)
604 		bridge_delete_member(sc, bif, 0);
605 
606 	while ((bif = LIST_FIRST(&sc->sc_spanlist)) != NULL) {
607 		bridge_delete_span(sc, bif);
608 	}
609 
610 	BRIDGE_UNLOCK(sc);
611 
612 	callout_drain(&sc->sc_brcallout);
613 
614 	mtx_lock(&bridge_list_mtx);
615 	LIST_REMOVE(sc, sc_list);
616 	mtx_unlock(&bridge_list_mtx);
617 
618 	bstp_detach(&sc->sc_stp);
619 	ether_ifdetach(ifp);
620 	if_free_type(ifp, IFT_ETHER);
621 
622 	/* Tear down the routing table. */
623 	bridge_rtable_fini(sc);
624 
625 	BRIDGE_LOCK_DESTROY(sc);
626 	free(sc, M_DEVBUF);
627 }
628 
629 /*
630  * bridge_ioctl:
631  *
632  *	Handle a control request from the operator.
633  */
634 static int
635 bridge_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
636 {
637 	struct bridge_softc *sc = ifp->if_softc;
638 	struct thread *td = curthread;
639 	union {
640 		struct ifbreq ifbreq;
641 		struct ifbifconf ifbifconf;
642 		struct ifbareq ifbareq;
643 		struct ifbaconf ifbaconf;
644 		struct ifbrparam ifbrparam;
645 	} args;
646 	struct ifdrv *ifd = (struct ifdrv *) data;
647 	const struct bridge_control *bc;
648 	int error = 0;
649 
650 	BRIDGE_LOCK(sc);
651 
652 	switch (cmd) {
653 
654 	case SIOCADDMULTI:
655 	case SIOCDELMULTI:
656 		break;
657 
658 	case SIOCGDRVSPEC:
659 	case SIOCSDRVSPEC:
660 		if (ifd->ifd_cmd >= bridge_control_table_size) {
661 			error = EINVAL;
662 			break;
663 		}
664 		bc = &bridge_control_table[ifd->ifd_cmd];
665 
666 		if (cmd == SIOCGDRVSPEC &&
667 		    (bc->bc_flags & BC_F_COPYOUT) == 0) {
668 			error = EINVAL;
669 			break;
670 		}
671 		else if (cmd == SIOCSDRVSPEC &&
672 		    (bc->bc_flags & BC_F_COPYOUT) != 0) {
673 			error = EINVAL;
674 			break;
675 		}
676 
677 		if (bc->bc_flags & BC_F_SUSER) {
678 			error = suser(td);
679 			if (error)
680 				break;
681 		}
682 
683 		if (ifd->ifd_len != bc->bc_argsize ||
684 		    ifd->ifd_len > sizeof(args)) {
685 			error = EINVAL;
686 			break;
687 		}
688 
689 		bzero(&args, sizeof(args));
690 		if (bc->bc_flags & BC_F_COPYIN) {
691 			error = copyin(ifd->ifd_data, &args, ifd->ifd_len);
692 			if (error)
693 				break;
694 		}
695 
696 		error = (*bc->bc_func)(sc, &args);
697 		if (error)
698 			break;
699 
700 		if (bc->bc_flags & BC_F_COPYOUT)
701 			error = copyout(&args, ifd->ifd_data, ifd->ifd_len);
702 
703 		break;
704 
705 	case SIOCSIFFLAGS:
706 		if (!(ifp->if_flags & IFF_UP) &&
707 		    (ifp->if_drv_flags & IFF_DRV_RUNNING)) {
708 			/*
709 			 * If interface is marked down and it is running,
710 			 * then stop and disable it.
711 			 */
712 			bridge_stop(ifp, 1);
713 		} else if ((ifp->if_flags & IFF_UP) &&
714 		    !(ifp->if_drv_flags & IFF_DRV_RUNNING)) {
715 			/*
716 			 * If interface is marked up and it is stopped, then
717 			 * start it.
718 			 */
719 			BRIDGE_UNLOCK(sc);
720 			(*ifp->if_init)(sc);
721 		}
722 		break;
723 
724 	case SIOCSIFMTU:
725 		/* Do not allow the MTU to be changed on the bridge */
726 		error = EINVAL;
727 		break;
728 
729 	default:
730 		/*
731 		 * drop the lock as ether_ioctl() will call bridge_start() and
732 		 * cause the lock to be recursed.
733 		 */
734 		BRIDGE_UNLOCK(sc);
735 		error = ether_ioctl(ifp, cmd, data);
736 		break;
737 	}
738 
739 	if (BRIDGE_LOCKED(sc))
740 		BRIDGE_UNLOCK(sc);
741 
742 	return (error);
743 }
744 
745 /*
746  * bridge_mutecaps:
747  *
748  *	Clear or restore unwanted capabilities on the member interface
749  */
750 static void
751 bridge_mutecaps(struct bridge_iflist *bif, int mute)
752 {
753 	struct ifnet *ifp = bif->bif_ifp;
754 	struct ifreq ifr;
755 	int error;
756 
757 	if (ifp->if_ioctl == NULL)
758 		return;
759 
760 	bzero(&ifr, sizeof(ifr));
761 	ifr.ifr_reqcap = ifp->if_capenable;
762 
763 	if (mute) {
764 		/* mask off and save capabilities */
765 		bif->bif_mutecap = ifr.ifr_reqcap & BRIDGE_IFCAPS_MASK;
766 		if (bif->bif_mutecap != 0)
767 			ifr.ifr_reqcap &= ~BRIDGE_IFCAPS_MASK;
768 	} else
769 		/* restore muted capabilities */
770 		ifr.ifr_reqcap |= bif->bif_mutecap;
771 
772 
773 	if (bif->bif_mutecap != 0) {
774 		IFF_LOCKGIANT(ifp);
775 		error = (*ifp->if_ioctl)(ifp, SIOCSIFCAP, (caddr_t)&ifr);
776 		IFF_UNLOCKGIANT(ifp);
777 	}
778 }
779 
780 /*
781  * bridge_lookup_member:
782  *
783  *	Lookup a bridge member interface.
784  */
785 static struct bridge_iflist *
786 bridge_lookup_member(struct bridge_softc *sc, const char *name)
787 {
788 	struct bridge_iflist *bif;
789 	struct ifnet *ifp;
790 
791 	BRIDGE_LOCK_ASSERT(sc);
792 
793 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
794 		ifp = bif->bif_ifp;
795 		if (strcmp(ifp->if_xname, name) == 0)
796 			return (bif);
797 	}
798 
799 	return (NULL);
800 }
801 
802 /*
803  * bridge_lookup_member_if:
804  *
805  *	Lookup a bridge member interface by ifnet*.
806  */
807 static struct bridge_iflist *
808 bridge_lookup_member_if(struct bridge_softc *sc, struct ifnet *member_ifp)
809 {
810 	struct bridge_iflist *bif;
811 
812 	BRIDGE_LOCK_ASSERT(sc);
813 
814 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
815 		if (bif->bif_ifp == member_ifp)
816 			return (bif);
817 	}
818 
819 	return (NULL);
820 }
821 
822 /*
823  * bridge_delete_member:
824  *
825  *	Delete the specified member interface.
826  */
827 static void
828 bridge_delete_member(struct bridge_softc *sc, struct bridge_iflist *bif,
829     int gone)
830 {
831 	struct ifnet *ifs = bif->bif_ifp;
832 
833 	BRIDGE_LOCK_ASSERT(sc);
834 
835 	if (!gone) {
836 		switch (ifs->if_type) {
837 		case IFT_ETHER:
838 		case IFT_L2VLAN:
839 			/*
840 			 * Take the interface out of promiscuous mode.
841 			 */
842 			(void) ifpromisc(ifs, 0);
843 			bridge_mutecaps(bif, 0);
844 			break;
845 
846 		case IFT_GIF:
847 			break;
848 
849 		default:
850 #ifdef DIAGNOSTIC
851 			panic("bridge_delete_member: impossible");
852 #endif
853 			break;
854 		}
855 	}
856 
857 	if (bif->bif_flags & IFBIF_STP)
858 		bstp_delete(&bif->bif_stp);
859 
860 	ifs->if_bridge = NULL;
861 	BRIDGE_XLOCK(sc);
862 	LIST_REMOVE(bif, bif_next);
863 	BRIDGE_XDROP(sc);
864 
865 	bridge_rtdelete(sc, ifs, IFBF_FLUSHALL);
866 
867 	BRIDGE_UNLOCK(sc);
868 	bstp_drain(&bif->bif_stp);	/* prepare to free */
869 	BRIDGE_LOCK(sc);
870 	free(bif, M_DEVBUF);
871 }
872 
873 /*
874  * bridge_delete_span:
875  *
876  *	Delete the specified span interface.
877  */
878 static void
879 bridge_delete_span(struct bridge_softc *sc, struct bridge_iflist *bif)
880 {
881 	BRIDGE_LOCK_ASSERT(sc);
882 
883 	KASSERT(bif->bif_ifp->if_bridge == NULL,
884 	    ("%s: not a span interface", __func__));
885 
886 	LIST_REMOVE(bif, bif_next);
887 	free(bif, M_DEVBUF);
888 }
889 
890 static int
891 bridge_ioctl_add(struct bridge_softc *sc, void *arg)
892 {
893 	struct ifbreq *req = arg;
894 	struct bridge_iflist *bif = NULL;
895 	struct ifnet *ifs;
896 	int error = 0;
897 
898 	BRIDGE_LOCK_ASSERT(sc);
899 
900 	ifs = ifunit(req->ifbr_ifsname);
901 	if (ifs == NULL)
902 		return (ENOENT);
903 
904 	/* If it's in the span list, it can't be a member. */
905 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
906 		if (ifs == bif->bif_ifp)
907 			return (EBUSY);
908 
909 	/* Allow the first Ethernet member to define the MTU */
910 	if (ifs->if_type != IFT_GIF) {
911 		if (LIST_EMPTY(&sc->sc_iflist))
912 			sc->sc_ifp->if_mtu = ifs->if_mtu;
913 		else if (sc->sc_ifp->if_mtu != ifs->if_mtu) {
914 			if_printf(sc->sc_ifp, "invalid MTU for %s\n",
915 			    ifs->if_xname);
916 			return (EINVAL);
917 		}
918 	}
919 
920 	if (ifs->if_bridge == sc)
921 		return (EEXIST);
922 
923 	if (ifs->if_bridge != NULL)
924 		return (EBUSY);
925 
926 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
927 	if (bif == NULL)
928 		return (ENOMEM);
929 
930 	bif->bif_ifp = ifs;
931 	bif->bif_flags = IFBIF_LEARNING | IFBIF_DISCOVER;
932 
933 	switch (ifs->if_type) {
934 	case IFT_ETHER:
935 	case IFT_L2VLAN:
936 		/*
937 		 * Place the interface into promiscuous mode.
938 		 */
939 		error = ifpromisc(ifs, 1);
940 		if (error)
941 			goto out;
942 
943 		bridge_mutecaps(bif, 1);
944 		break;
945 
946 	case IFT_GIF:
947 		break;
948 
949 	default:
950 		error = EINVAL;
951 		goto out;
952 	}
953 
954 	ifs->if_bridge = sc;
955 	/*
956 	 * XXX: XLOCK HERE!?!
957 	 *
958 	 * NOTE: insert_***HEAD*** should be safe for the traversals.
959 	 */
960 	LIST_INSERT_HEAD(&sc->sc_iflist, bif, bif_next);
961 
962 out:
963 	if (error) {
964 		if (bif != NULL)
965 			free(bif, M_DEVBUF);
966 	}
967 	return (error);
968 }
969 
970 static int
971 bridge_ioctl_del(struct bridge_softc *sc, void *arg)
972 {
973 	struct ifbreq *req = arg;
974 	struct bridge_iflist *bif;
975 
976 	BRIDGE_LOCK_ASSERT(sc);
977 
978 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
979 	if (bif == NULL)
980 		return (ENOENT);
981 
982 	bridge_delete_member(sc, bif, 0);
983 
984 	return (0);
985 }
986 
987 static int
988 bridge_ioctl_gifflags(struct bridge_softc *sc, void *arg)
989 {
990 	struct ifbreq *req = arg;
991 	struct bridge_iflist *bif;
992 
993 	BRIDGE_LOCK_ASSERT(sc);
994 
995 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
996 	if (bif == NULL)
997 		return (ENOENT);
998 
999 	req->ifbr_ifsflags = bif->bif_flags;
1000 	req->ifbr_state = bif->bif_stp.bp_state;
1001 	req->ifbr_priority = bif->bif_stp.bp_priority;
1002 	req->ifbr_path_cost = bif->bif_stp.bp_path_cost;
1003 	req->ifbr_portno = bif->bif_ifp->if_index & 0xff;
1004 
1005 	return (0);
1006 }
1007 
1008 static int
1009 bridge_ioctl_sifflags(struct bridge_softc *sc, void *arg)
1010 {
1011 	struct ifbreq *req = arg;
1012 	struct bridge_iflist *bif;
1013 	int error;
1014 
1015 	BRIDGE_LOCK_ASSERT(sc);
1016 
1017 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1018 	if (bif == NULL)
1019 		return (ENOENT);
1020 
1021 	if (req->ifbr_ifsflags & IFBIF_SPAN)
1022 		/* SPAN is readonly */
1023 		return (EINVAL);
1024 
1025 	if (req->ifbr_ifsflags & IFBIF_STP) {
1026 		if ((bif->bif_flags & IFBIF_STP) == 0) {
1027 			error = bstp_add(&sc->sc_stp, &bif->bif_stp,
1028 				    bif->bif_ifp);
1029 			if (error)
1030 				return (error);
1031 		}
1032 	} else {
1033 		if ((bif->bif_flags & IFBIF_STP) != 0)
1034 			bstp_delete(&bif->bif_stp);
1035 	}
1036 
1037 	bif->bif_flags = req->ifbr_ifsflags;
1038 
1039 	return (0);
1040 }
1041 
1042 static int
1043 bridge_ioctl_scache(struct bridge_softc *sc, void *arg)
1044 {
1045 	struct ifbrparam *param = arg;
1046 
1047 	BRIDGE_LOCK_ASSERT(sc);
1048 
1049 	sc->sc_brtmax = param->ifbrp_csize;
1050 	bridge_rttrim(sc);
1051 
1052 	return (0);
1053 }
1054 
1055 static int
1056 bridge_ioctl_gcache(struct bridge_softc *sc, void *arg)
1057 {
1058 	struct ifbrparam *param = arg;
1059 
1060 	BRIDGE_LOCK_ASSERT(sc);
1061 
1062 	param->ifbrp_csize = sc->sc_brtmax;
1063 
1064 	return (0);
1065 }
1066 
1067 static int
1068 bridge_ioctl_gifs(struct bridge_softc *sc, void *arg)
1069 {
1070 	struct ifbifconf *bifc = arg;
1071 	struct bridge_iflist *bif;
1072 	struct ifbreq breq;
1073 	int count, len, error = 0;
1074 
1075 	BRIDGE_LOCK_ASSERT(sc);
1076 
1077 	count = 0;
1078 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next)
1079 		count++;
1080 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1081 		count++;
1082 
1083 	if (bifc->ifbic_len == 0) {
1084 		bifc->ifbic_len = sizeof(breq) * count;
1085 		return (0);
1086 	}
1087 
1088 	count = 0;
1089 	len = bifc->ifbic_len;
1090 	bzero(&breq, sizeof(breq));
1091 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1092 		if (len < sizeof(breq))
1093 			break;
1094 
1095 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1096 		    sizeof(breq.ifbr_ifsname));
1097 		breq.ifbr_ifsflags = bif->bif_flags;
1098 		breq.ifbr_state = bif->bif_stp.bp_state;
1099 		breq.ifbr_priority = bif->bif_stp.bp_priority;
1100 		breq.ifbr_path_cost = bif->bif_stp.bp_path_cost;
1101 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
1102 		error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
1103 		if (error)
1104 			break;
1105 		count++;
1106 		len -= sizeof(breq);
1107 	}
1108 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
1109 		if (len < sizeof(breq))
1110 			break;
1111 
1112 		strlcpy(breq.ifbr_ifsname, bif->bif_ifp->if_xname,
1113 		    sizeof(breq.ifbr_ifsname));
1114 		breq.ifbr_ifsflags = bif->bif_flags;
1115 		breq.ifbr_state = bif->bif_stp.bp_state;
1116 		breq.ifbr_priority = bif->bif_stp.bp_priority;
1117 		breq.ifbr_path_cost = bif->bif_stp.bp_path_cost;
1118 		breq.ifbr_portno = bif->bif_ifp->if_index & 0xff;
1119 		error = copyout(&breq, bifc->ifbic_req + count, sizeof(breq));
1120 		if (error)
1121 			break;
1122 		count++;
1123 		len -= sizeof(breq);
1124 	}
1125 
1126 	bifc->ifbic_len = sizeof(breq) * count;
1127 	return (error);
1128 }
1129 
1130 static int
1131 bridge_ioctl_rts(struct bridge_softc *sc, void *arg)
1132 {
1133 	struct ifbaconf *bac = arg;
1134 	struct bridge_rtnode *brt;
1135 	struct ifbareq bareq;
1136 	int count = 0, error = 0, len;
1137 
1138 	BRIDGE_LOCK_ASSERT(sc);
1139 
1140 	if (bac->ifbac_len == 0)
1141 		return (0);
1142 
1143 	len = bac->ifbac_len;
1144 	bzero(&bareq, sizeof(bareq));
1145 	LIST_FOREACH(brt, &sc->sc_rtlist, brt_list) {
1146 		if (len < sizeof(bareq))
1147 			goto out;
1148 		strlcpy(bareq.ifba_ifsname, brt->brt_ifp->if_xname,
1149 		    sizeof(bareq.ifba_ifsname));
1150 		memcpy(bareq.ifba_dst, brt->brt_addr, sizeof(brt->brt_addr));
1151 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC &&
1152 				time_uptime < brt->brt_expire)
1153 			bareq.ifba_expire = brt->brt_expire - time_uptime;
1154 		else
1155 			bareq.ifba_expire = 0;
1156 		bareq.ifba_flags = brt->brt_flags;
1157 
1158 		error = copyout(&bareq, bac->ifbac_req + count, sizeof(bareq));
1159 		if (error)
1160 			goto out;
1161 		count++;
1162 		len -= sizeof(bareq);
1163 	}
1164 out:
1165 	bac->ifbac_len = sizeof(bareq) * count;
1166 	return (error);
1167 }
1168 
1169 static int
1170 bridge_ioctl_saddr(struct bridge_softc *sc, void *arg)
1171 {
1172 	struct ifbareq *req = arg;
1173 	struct bridge_iflist *bif;
1174 	int error;
1175 
1176 	BRIDGE_LOCK_ASSERT(sc);
1177 
1178 	bif = bridge_lookup_member(sc, req->ifba_ifsname);
1179 	if (bif == NULL)
1180 		return (ENOENT);
1181 
1182 	error = bridge_rtupdate(sc, req->ifba_dst, bif->bif_ifp, 1,
1183 	    req->ifba_flags);
1184 
1185 	return (error);
1186 }
1187 
1188 static int
1189 bridge_ioctl_sto(struct bridge_softc *sc, void *arg)
1190 {
1191 	struct ifbrparam *param = arg;
1192 
1193 	BRIDGE_LOCK_ASSERT(sc);
1194 
1195 	sc->sc_brttimeout = param->ifbrp_ctime;
1196 
1197 	return (0);
1198 }
1199 
1200 static int
1201 bridge_ioctl_gto(struct bridge_softc *sc, void *arg)
1202 {
1203 	struct ifbrparam *param = arg;
1204 
1205 	BRIDGE_LOCK_ASSERT(sc);
1206 
1207 	param->ifbrp_ctime = sc->sc_brttimeout;
1208 
1209 	return (0);
1210 }
1211 
1212 static int
1213 bridge_ioctl_daddr(struct bridge_softc *sc, void *arg)
1214 {
1215 	struct ifbareq *req = arg;
1216 
1217 	BRIDGE_LOCK_ASSERT(sc);
1218 
1219 	return (bridge_rtdaddr(sc, req->ifba_dst));
1220 }
1221 
1222 static int
1223 bridge_ioctl_flush(struct bridge_softc *sc, void *arg)
1224 {
1225 	struct ifbreq *req = arg;
1226 
1227 	BRIDGE_LOCK_ASSERT(sc);
1228 
1229 	bridge_rtflush(sc, req->ifbr_ifsflags);
1230 
1231 	return (0);
1232 }
1233 
1234 static int
1235 bridge_ioctl_gpri(struct bridge_softc *sc, void *arg)
1236 {
1237 	struct ifbrparam *param = arg;
1238 	struct bstp_state *bs = &sc->sc_stp;
1239 
1240 	BRIDGE_LOCK_ASSERT(sc);
1241 
1242 	param->ifbrp_prio = bs->bs_bridge_priority;
1243 
1244 	return (0);
1245 }
1246 
1247 static int
1248 bridge_ioctl_spri(struct bridge_softc *sc, void *arg)
1249 {
1250 	struct ifbrparam *param = arg;
1251 	struct bstp_state *bs = &sc->sc_stp;
1252 
1253 	BRIDGE_LOCK_ASSERT(sc);
1254 
1255 	bs->bs_bridge_priority = param->ifbrp_prio;
1256 	bstp_reinit(bs);
1257 
1258 	return (0);
1259 }
1260 
1261 static int
1262 bridge_ioctl_ght(struct bridge_softc *sc, void *arg)
1263 {
1264 	struct ifbrparam *param = arg;
1265 	struct bstp_state *bs = &sc->sc_stp;
1266 
1267 	BRIDGE_LOCK_ASSERT(sc);
1268 
1269 	param->ifbrp_hellotime = bs->bs_bridge_hello_time >> 8;
1270 
1271 	return (0);
1272 }
1273 
1274 static int
1275 bridge_ioctl_sht(struct bridge_softc *sc, void *arg)
1276 {
1277 	struct ifbrparam *param = arg;
1278 	struct bstp_state *bs = &sc->sc_stp;
1279 
1280 	BRIDGE_LOCK_ASSERT(sc);
1281 
1282 	if (param->ifbrp_hellotime == 0)
1283 		return (EINVAL);
1284 	bs->bs_bridge_hello_time = param->ifbrp_hellotime << 8;
1285 	bstp_reinit(bs);
1286 
1287 	return (0);
1288 }
1289 
1290 static int
1291 bridge_ioctl_gfd(struct bridge_softc *sc, void *arg)
1292 {
1293 	struct ifbrparam *param = arg;
1294 	struct bstp_state *bs = &sc->sc_stp;
1295 
1296 	BRIDGE_LOCK_ASSERT(sc);
1297 
1298 	param->ifbrp_fwddelay = bs->bs_bridge_forward_delay >> 8;
1299 
1300 	return (0);
1301 }
1302 
1303 static int
1304 bridge_ioctl_sfd(struct bridge_softc *sc, void *arg)
1305 {
1306 	struct ifbrparam *param = arg;
1307 	struct bstp_state *bs = &sc->sc_stp;
1308 
1309 	BRIDGE_LOCK_ASSERT(sc);
1310 
1311 	if (param->ifbrp_fwddelay == 0)
1312 		return (EINVAL);
1313 	bs->bs_bridge_forward_delay = param->ifbrp_fwddelay << 8;
1314 	bstp_reinit(bs);
1315 
1316 	return (0);
1317 }
1318 
1319 static int
1320 bridge_ioctl_gma(struct bridge_softc *sc, void *arg)
1321 {
1322 	struct ifbrparam *param = arg;
1323 	struct bstp_state *bs = &sc->sc_stp;
1324 
1325 	BRIDGE_LOCK_ASSERT(sc);
1326 
1327 	param->ifbrp_maxage = bs->bs_bridge_max_age >> 8;
1328 
1329 	return (0);
1330 }
1331 
1332 static int
1333 bridge_ioctl_sma(struct bridge_softc *sc, void *arg)
1334 {
1335 	struct ifbrparam *param = arg;
1336 	struct bstp_state *bs = &sc->sc_stp;
1337 
1338 	BRIDGE_LOCK_ASSERT(sc);
1339 
1340 	if (param->ifbrp_maxage == 0)
1341 		return (EINVAL);
1342 	bs->bs_bridge_max_age = param->ifbrp_maxage << 8;
1343 	bstp_reinit(bs);
1344 
1345 	return (0);
1346 }
1347 
1348 static int
1349 bridge_ioctl_sifprio(struct bridge_softc *sc, void *arg)
1350 {
1351 	struct ifbreq *req = arg;
1352 	struct bridge_iflist *bif;
1353 
1354 	BRIDGE_LOCK_ASSERT(sc);
1355 
1356 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1357 	if (bif == NULL)
1358 		return (ENOENT);
1359 
1360 	bif->bif_stp.bp_priority = req->ifbr_priority;
1361 	bstp_reinit(&sc->sc_stp);
1362 
1363 	return (0);
1364 }
1365 
1366 static int
1367 bridge_ioctl_sifcost(struct bridge_softc *sc, void *arg)
1368 {
1369 	struct ifbreq *req = arg;
1370 	struct bridge_iflist *bif;
1371 
1372 	BRIDGE_LOCK_ASSERT(sc);
1373 
1374 	bif = bridge_lookup_member(sc, req->ifbr_ifsname);
1375 	if (bif == NULL)
1376 		return (ENOENT);
1377 
1378 	bif->bif_stp.bp_path_cost = req->ifbr_path_cost;
1379 	bstp_reinit(&sc->sc_stp);
1380 
1381 	return (0);
1382 }
1383 
1384 static int
1385 bridge_ioctl_addspan(struct bridge_softc *sc, void *arg)
1386 {
1387 	struct ifbreq *req = arg;
1388 	struct bridge_iflist *bif = NULL;
1389 	struct ifnet *ifs;
1390 
1391 	BRIDGE_LOCK_ASSERT(sc);
1392 
1393 	ifs = ifunit(req->ifbr_ifsname);
1394 	if (ifs == NULL)
1395 		return (ENOENT);
1396 
1397 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1398 		if (ifs == bif->bif_ifp)
1399 			return (EBUSY);
1400 
1401 	if (ifs->if_bridge != NULL)
1402 		return (EBUSY);
1403 
1404 	switch (ifs->if_type) {
1405 		case IFT_ETHER:
1406 		case IFT_GIF:
1407 		case IFT_L2VLAN:
1408 			break;
1409 		default:
1410 			return (EINVAL);
1411 	}
1412 
1413 	bif = malloc(sizeof(*bif), M_DEVBUF, M_NOWAIT|M_ZERO);
1414 	if (bif == NULL)
1415 		return (ENOMEM);
1416 
1417 	bif->bif_ifp = ifs;
1418 	bif->bif_flags = IFBIF_SPAN;
1419 
1420 	LIST_INSERT_HEAD(&sc->sc_spanlist, bif, bif_next);
1421 
1422 	return (0);
1423 }
1424 
1425 static int
1426 bridge_ioctl_delspan(struct bridge_softc *sc, void *arg)
1427 {
1428 	struct ifbreq *req = arg;
1429 	struct bridge_iflist *bif;
1430 	struct ifnet *ifs;
1431 
1432 	BRIDGE_LOCK_ASSERT(sc);
1433 
1434 	ifs = ifunit(req->ifbr_ifsname);
1435 	if (ifs == NULL)
1436 		return (ENOENT);
1437 
1438 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1439 		if (ifs == bif->bif_ifp)
1440 			break;
1441 
1442 	if (bif == NULL)
1443 		return (ENOENT);
1444 
1445 	bridge_delete_span(sc, bif);
1446 
1447 	return (0);
1448 }
1449 
1450 static int
1451 bridge_ioctl_gbparam(struct bridge_softc *sc, void *arg)
1452 {
1453 	struct ifbropreq *req = arg;
1454 	struct bstp_port *root_port;
1455 
1456 	BRIDGE_LOCK_ASSERT(sc);
1457 
1458 	req->ifbop_maxage = sc->sc_stp.bs_max_age;
1459 	req->ifbop_hellotime = sc->sc_stp.bs_hello_time;
1460 	req->ifbop_fwddelay = sc->sc_stp.bs_forward_delay;
1461 
1462 	root_port = sc->sc_stp.bs_root_port;
1463 	if (root_port == NULL)
1464 		req->ifbop_root_port = 0;
1465 	else
1466 		req->ifbop_root_port = root_port->bp_ifp->if_index;
1467 
1468 	req->ifbop_root_path_cost = sc->sc_stp.bs_root_path_cost;
1469 	req->ifbop_designated_root = sc->sc_stp.bs_designated_root;
1470 	req->ifbop_last_tc_time.tv_sec = sc->sc_stp.bs_last_tc_time.tv_sec;
1471 	req->ifbop_last_tc_time.tv_usec = sc->sc_stp.bs_last_tc_time.tv_usec;
1472 
1473 	return (0);
1474 }
1475 
1476 static int
1477 bridge_ioctl_grte(struct bridge_softc *sc, void *arg)
1478 {
1479 	struct ifbrparam *param = arg;
1480 
1481 	BRIDGE_LOCK_ASSERT(sc);
1482 
1483 	param->ifbrp_cexceeded = sc->sc_brtexceeded;
1484 
1485 	return (0);
1486 }
1487 
1488 static int
1489 bridge_ioctl_gifsstp(struct bridge_softc *sc, void *arg)
1490 {
1491 	struct ifbpstpconf *bifstp = arg;
1492 	struct bridge_iflist *bif;
1493 	struct ifbpstpreq bpreq;
1494 	int count, len, error = 0;
1495 
1496 	BRIDGE_LOCK_ASSERT(sc);
1497 
1498 	count = 0;
1499 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1500 		if ((bif->bif_flags & IFBIF_STP) != 0)
1501 			count++;
1502 	}
1503 
1504 	if (bifstp->ifbpstp_len == 0) {
1505 		bifstp->ifbpstp_len = sizeof(bpreq) * count;
1506 		return (0);
1507 	}
1508 
1509 	count = 0;
1510 	len = bifstp->ifbpstp_len;
1511 	bzero(&bpreq, sizeof(bpreq));
1512 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1513 		if (len < sizeof(bpreq))
1514 			break;
1515 
1516 		if ((bif->bif_flags & IFBIF_STP) == 0)
1517 			continue;
1518 
1519 		bpreq.ifbp_portno = bif->bif_ifp->if_index & 0xff;
1520 		bpreq.ifbp_fwd_trans = bif->bif_stp.bp_forward_transitions;
1521 		bpreq.ifbp_design_cost = bif->bif_stp.bp_designated_cost;
1522 		bpreq.ifbp_design_port = bif->bif_stp.bp_designated_port;
1523 		bpreq.ifbp_design_bridge = bif->bif_stp.bp_designated_bridge;
1524 		bpreq.ifbp_design_root = bif->bif_stp.bp_designated_root;
1525 
1526 		error = copyout(&bpreq, bifstp->ifbpstp_req + count,
1527 				sizeof(bpreq));
1528 		if (error != 0)
1529 			break;
1530 
1531 		count++;
1532 		len -= sizeof(bpreq);
1533 	}
1534 
1535 	bifstp->ifbpstp_len = sizeof(bpreq) * count;
1536 	return (error);
1537 }
1538 
1539 /*
1540  * bridge_ifdetach:
1541  *
1542  *	Detach an interface from a bridge.  Called when a member
1543  *	interface is detaching.
1544  */
1545 static void
1546 bridge_ifdetach(void *arg __unused, struct ifnet *ifp)
1547 {
1548 	struct bridge_softc *sc = ifp->if_bridge;
1549 	struct bridge_iflist *bif;
1550 
1551 	/* Check if the interface is a bridge member */
1552 	if (sc != NULL) {
1553 		BRIDGE_LOCK(sc);
1554 
1555 		bif = bridge_lookup_member_if(sc, ifp);
1556 		if (bif != NULL)
1557 			bridge_delete_member(sc, bif, 1);
1558 
1559 		BRIDGE_UNLOCK(sc);
1560 		return;
1561 	}
1562 
1563 	/* Check if the interface is a span port */
1564 	mtx_lock(&bridge_list_mtx);
1565 	LIST_FOREACH(sc, &bridge_list, sc_list) {
1566 		BRIDGE_LOCK(sc);
1567 		LIST_FOREACH(bif, &sc->sc_spanlist, bif_next)
1568 			if (ifp == bif->bif_ifp) {
1569 				bridge_delete_span(sc, bif);
1570 				break;
1571 			}
1572 
1573 		BRIDGE_UNLOCK(sc);
1574 	}
1575 	mtx_unlock(&bridge_list_mtx);
1576 }
1577 
1578 /*
1579  * bridge_init:
1580  *
1581  *	Initialize a bridge interface.
1582  */
1583 static void
1584 bridge_init(void *xsc)
1585 {
1586 	struct bridge_softc *sc = (struct bridge_softc *)xsc;
1587 	struct ifnet *ifp = sc->sc_ifp;
1588 
1589 	if (ifp->if_drv_flags & IFF_DRV_RUNNING)
1590 		return;
1591 
1592 	BRIDGE_LOCK(sc);
1593 	callout_reset(&sc->sc_brcallout, bridge_rtable_prune_period * hz,
1594 	    bridge_timer, sc);
1595 
1596 	ifp->if_drv_flags |= IFF_DRV_RUNNING;
1597 	bstp_init(&sc->sc_stp);		/* Initialize Spanning Tree */
1598 
1599 	BRIDGE_UNLOCK(sc);
1600 }
1601 
1602 /*
1603  * bridge_stop:
1604  *
1605  *	Stop the bridge interface.
1606  */
1607 static void
1608 bridge_stop(struct ifnet *ifp, int disable)
1609 {
1610 	struct bridge_softc *sc = ifp->if_softc;
1611 
1612 	BRIDGE_LOCK_ASSERT(sc);
1613 
1614 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
1615 		return;
1616 
1617 	callout_stop(&sc->sc_brcallout);
1618 	bstp_stop(&sc->sc_stp);
1619 
1620 	bridge_rtflush(sc, IFBF_FLUSHDYN);
1621 
1622 	ifp->if_drv_flags &= ~IFF_DRV_RUNNING;
1623 }
1624 
1625 /*
1626  * bridge_enqueue:
1627  *
1628  *	Enqueue a packet on a bridge member interface.
1629  *
1630  */
1631 static void
1632 bridge_enqueue(struct bridge_softc *sc, struct ifnet *dst_ifp, struct mbuf *m)
1633 {
1634 	int len, err = 0;
1635 	short mflags;
1636 	struct mbuf *m0;
1637 
1638 	len = m->m_pkthdr.len;
1639 	mflags = m->m_flags;
1640 
1641 	/* We may be sending a fragment so traverse the mbuf */
1642 	for (; m; m = m0) {
1643 		m0 = m->m_nextpkt;
1644 		m->m_nextpkt = NULL;
1645 
1646 		if (err == 0)
1647 			IFQ_ENQUEUE(&dst_ifp->if_snd, m, err);
1648 	}
1649 
1650 	if (err == 0) {
1651 
1652 		sc->sc_ifp->if_opackets++;
1653 		sc->sc_ifp->if_obytes += len;
1654 
1655 		dst_ifp->if_obytes += len;
1656 
1657 		if (mflags & M_MCAST) {
1658 			sc->sc_ifp->if_omcasts++;
1659 			dst_ifp->if_omcasts++;
1660 		}
1661 	}
1662 
1663 	if ((dst_ifp->if_drv_flags & IFF_DRV_OACTIVE) == 0)
1664 		(*dst_ifp->if_start)(dst_ifp);
1665 }
1666 
1667 /*
1668  * bridge_dummynet:
1669  *
1670  * 	Receive a queued packet from dummynet and pass it on to the output
1671  * 	interface.
1672  *
1673  *	The mbuf has the Ethernet header already attached.
1674  */
1675 static void
1676 bridge_dummynet(struct mbuf *m, struct ifnet *ifp)
1677 {
1678 	struct bridge_softc *sc;
1679 
1680 	sc = ifp->if_bridge;
1681 
1682 	/*
1683 	 * The packet didnt originate from a member interface. This should only
1684 	 * ever happen if a member interface is removed while packets are
1685 	 * queued for it.
1686 	 */
1687 	if (sc == NULL) {
1688 		m_freem(m);
1689 		return;
1690 	}
1691 
1692 	if (PFIL_HOOKED(&inet_pfil_hook)
1693 #ifdef INET6
1694 	    || PFIL_HOOKED(&inet6_pfil_hook)
1695 #endif
1696 	    ) {
1697 		if (bridge_pfil(&m, sc->sc_ifp, ifp, PFIL_OUT) != 0)
1698 			return;
1699 		if (m == NULL)
1700 			return;
1701 	}
1702 
1703 	bridge_enqueue(sc, ifp, m);
1704 }
1705 
1706 /*
1707  * bridge_output:
1708  *
1709  *	Send output from a bridge member interface.  This
1710  *	performs the bridging function for locally originated
1711  *	packets.
1712  *
1713  *	The mbuf has the Ethernet header already attached.  We must
1714  *	enqueue or free the mbuf before returning.
1715  */
1716 static int
1717 bridge_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *sa,
1718     struct rtentry *rt)
1719 {
1720 	struct ether_header *eh;
1721 	struct ifnet *dst_if;
1722 	struct bridge_softc *sc;
1723 
1724 	if (m->m_len < ETHER_HDR_LEN) {
1725 		m = m_pullup(m, ETHER_HDR_LEN);
1726 		if (m == NULL)
1727 			return (0);
1728 	}
1729 
1730 	eh = mtod(m, struct ether_header *);
1731 	sc = ifp->if_bridge;
1732 
1733 	BRIDGE_LOCK(sc);
1734 
1735 	/*
1736 	 * If bridge is down, but the original output interface is up,
1737 	 * go ahead and send out that interface.  Otherwise, the packet
1738 	 * is dropped below.
1739 	 */
1740 	if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1741 		dst_if = ifp;
1742 		goto sendunicast;
1743 	}
1744 
1745 	/*
1746 	 * If the packet is a multicast, or we don't know a better way to
1747 	 * get there, send to all interfaces.
1748 	 */
1749 	if (ETHER_IS_MULTICAST(eh->ether_dhost))
1750 		dst_if = NULL;
1751 	else
1752 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1753 	if (dst_if == NULL) {
1754 		struct bridge_iflist *bif;
1755 		struct mbuf *mc;
1756 		int error = 0, used = 0;
1757 
1758 		BRIDGE_LOCK2REF(sc, error);
1759 		if (error) {
1760 			m_freem(m);
1761 			return (0);
1762 		}
1763 
1764 		bridge_span(sc, m);
1765 
1766 		LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
1767 			dst_if = bif->bif_ifp;
1768 
1769 			if (dst_if->if_type == IFT_GIF)
1770 				continue;
1771 			if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
1772 				continue;
1773 
1774 			/*
1775 			 * If this is not the original output interface,
1776 			 * and the interface is participating in spanning
1777 			 * tree, make sure the port is in a state that
1778 			 * allows forwarding.
1779 			 */
1780 			if (dst_if != ifp &&
1781 			    (bif->bif_flags & IFBIF_STP) != 0) {
1782 				switch (bif->bif_stp.bp_state) {
1783 				case BSTP_IFSTATE_BLOCKING:
1784 				case BSTP_IFSTATE_LISTENING:
1785 				case BSTP_IFSTATE_DISABLED:
1786 					continue;
1787 				}
1788 			}
1789 
1790 			if (LIST_NEXT(bif, bif_next) == NULL) {
1791 				used = 1;
1792 				mc = m;
1793 			} else {
1794 				mc = m_copypacket(m, M_DONTWAIT);
1795 				if (mc == NULL) {
1796 					sc->sc_ifp->if_oerrors++;
1797 					continue;
1798 				}
1799 			}
1800 
1801 			bridge_enqueue(sc, dst_if, mc);
1802 		}
1803 		if (used == 0)
1804 			m_freem(m);
1805 		BRIDGE_UNREF(sc);
1806 		return (0);
1807 	}
1808 
1809 sendunicast:
1810 	/*
1811 	 * XXX Spanning tree consideration here?
1812 	 */
1813 
1814 	bridge_span(sc, m);
1815 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1816 		m_freem(m);
1817 		BRIDGE_UNLOCK(sc);
1818 		return (0);
1819 	}
1820 
1821 	BRIDGE_UNLOCK(sc);
1822 	bridge_enqueue(sc, dst_if, m);
1823 	return (0);
1824 }
1825 
1826 /*
1827  * bridge_start:
1828  *
1829  *	Start output on a bridge.
1830  *
1831  */
1832 static void
1833 bridge_start(struct ifnet *ifp)
1834 {
1835 	struct bridge_softc *sc;
1836 	struct mbuf *m;
1837 	struct ether_header *eh;
1838 	struct ifnet *dst_if;
1839 
1840 	sc = ifp->if_softc;
1841 
1842 	ifp->if_drv_flags |= IFF_DRV_OACTIVE;
1843 	for (;;) {
1844 		IFQ_DEQUEUE(&ifp->if_snd, m);
1845 		if (m == 0)
1846 			break;
1847 		BPF_MTAP(ifp, m);
1848 
1849 		eh = mtod(m, struct ether_header *);
1850 		dst_if = NULL;
1851 
1852 		BRIDGE_LOCK(sc);
1853 		if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1854 			dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1855 		}
1856 
1857 		if (dst_if == NULL)
1858 			bridge_broadcast(sc, ifp, m, 0);
1859 		else {
1860 			BRIDGE_UNLOCK(sc);
1861 			bridge_enqueue(sc, dst_if, m);
1862 		}
1863 	}
1864 	ifp->if_drv_flags &= ~IFF_DRV_OACTIVE;
1865 }
1866 
1867 /*
1868  * bridge_forward:
1869  *
1870  *	The forwarding function of the bridge.
1871  *
1872  *	NOTE: Releases the lock on return.
1873  */
1874 static void
1875 bridge_forward(struct bridge_softc *sc, struct mbuf *m)
1876 {
1877 	struct bridge_iflist *bif;
1878 	struct ifnet *src_if, *dst_if, *ifp;
1879 	struct ether_header *eh;
1880 
1881 	src_if = m->m_pkthdr.rcvif;
1882 	BRIDGE_LOCK_ASSERT(sc);
1883 	ifp = sc->sc_ifp;
1884 
1885 	sc->sc_ifp->if_ipackets++;
1886 	sc->sc_ifp->if_ibytes += m->m_pkthdr.len;
1887 
1888 	/*
1889 	 * Look up the bridge_iflist.
1890 	 */
1891 	bif = bridge_lookup_member_if(sc, src_if);
1892 	if (bif == NULL) {
1893 		/* Interface is not a bridge member (anymore?) */
1894 		BRIDGE_UNLOCK(sc);
1895 		m_freem(m);
1896 		return;
1897 	}
1898 
1899 	if (bif->bif_flags & IFBIF_STP) {
1900 		switch (bif->bif_stp.bp_state) {
1901 		case BSTP_IFSTATE_BLOCKING:
1902 		case BSTP_IFSTATE_LISTENING:
1903 		case BSTP_IFSTATE_DISABLED:
1904 			BRIDGE_UNLOCK(sc);
1905 			m_freem(m);
1906 			return;
1907 		}
1908 	}
1909 
1910 	eh = mtod(m, struct ether_header *);
1911 
1912 	/*
1913 	 * If the interface is learning, and the source
1914 	 * address is valid and not multicast, record
1915 	 * the address.
1916 	 */
1917 	if ((bif->bif_flags & IFBIF_LEARNING) != 0 &&
1918 	    ETHER_IS_MULTICAST(eh->ether_shost) == 0 &&
1919 	    (eh->ether_shost[0] == 0 &&
1920 	     eh->ether_shost[1] == 0 &&
1921 	     eh->ether_shost[2] == 0 &&
1922 	     eh->ether_shost[3] == 0 &&
1923 	     eh->ether_shost[4] == 0 &&
1924 	     eh->ether_shost[5] == 0) == 0) {
1925 		(void) bridge_rtupdate(sc, eh->ether_shost,
1926 		    src_if, 0, IFBAF_DYNAMIC);
1927 	}
1928 
1929 	if ((bif->bif_flags & IFBIF_STP) != 0 &&
1930 	    bif->bif_stp.bp_state == BSTP_IFSTATE_LEARNING) {
1931 		m_freem(m);
1932 		BRIDGE_UNLOCK(sc);
1933 		return;
1934 	}
1935 
1936 	/*
1937 	 * At this point, the port either doesn't participate
1938 	 * in spanning tree or it is in the forwarding state.
1939 	 */
1940 
1941 	/*
1942 	 * If the packet is unicast, destined for someone on
1943 	 * "this" side of the bridge, drop it.
1944 	 */
1945 	if ((m->m_flags & (M_BCAST|M_MCAST)) == 0) {
1946 		dst_if = bridge_rtlookup(sc, eh->ether_dhost);
1947 		if (src_if == dst_if) {
1948 			BRIDGE_UNLOCK(sc);
1949 			m_freem(m);
1950 			return;
1951 		}
1952 	} else {
1953 		/* ...forward it to all interfaces. */
1954 		sc->sc_ifp->if_imcasts++;
1955 		dst_if = NULL;
1956 	}
1957 
1958 	/*
1959 	 * If we have a destination interface which is a member of our bridge,
1960 	 * OR this is a unicast packet, push it through the bpf(4) machinery.
1961 	 * For broadcast or multicast packets, don't bother because it will
1962 	 * be reinjected into ether_input. We do this before we pass the packets
1963 	 * through the pfil(9) framework, as it is possible that pfil(9) will
1964 	 * drop the packet, or possibly modify it, making it difficult to debug
1965 	 * firewall issues on the bridge.
1966 	 */
1967 	if (dst_if != NULL || (m->m_flags & (M_BCAST | M_MCAST)) == 0)
1968 		BPF_MTAP(ifp, m);
1969 
1970 	/* run the packet filter */
1971 	if (PFIL_HOOKED(&inet_pfil_hook)
1972 #ifdef INET6
1973 	    || PFIL_HOOKED(&inet6_pfil_hook)
1974 #endif
1975 	    ) {
1976 		BRIDGE_UNLOCK(sc);
1977 		if (bridge_pfil(&m, ifp, src_if, PFIL_IN) != 0)
1978 			return;
1979 		if (m == NULL)
1980 			return;
1981 		BRIDGE_LOCK(sc);
1982 	}
1983 
1984 	if (dst_if == NULL) {
1985 		bridge_broadcast(sc, src_if, m, 1);
1986 		return;
1987 	}
1988 
1989 	/*
1990 	 * At this point, we're dealing with a unicast frame
1991 	 * going to a different interface.
1992 	 */
1993 	if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0) {
1994 		BRIDGE_UNLOCK(sc);
1995 		m_freem(m);
1996 		return;
1997 	}
1998 	bif = bridge_lookup_member_if(sc, dst_if);
1999 	if (bif == NULL) {
2000 		/* Not a member of the bridge (anymore?) */
2001 		BRIDGE_UNLOCK(sc);
2002 		m_freem(m);
2003 		return;
2004 	}
2005 
2006 	if (bif->bif_flags & IFBIF_STP) {
2007 		switch (bif->bif_stp.bp_state) {
2008 		case BSTP_IFSTATE_DISABLED:
2009 		case BSTP_IFSTATE_BLOCKING:
2010 			BRIDGE_UNLOCK(sc);
2011 			m_freem(m);
2012 			return;
2013 		}
2014 	}
2015 
2016 	BRIDGE_UNLOCK(sc);
2017 
2018 	if (PFIL_HOOKED(&inet_pfil_hook)
2019 #ifdef INET6
2020 	    || PFIL_HOOKED(&inet6_pfil_hook)
2021 #endif
2022 	    ) {
2023 		if (bridge_pfil(&m, sc->sc_ifp, dst_if, PFIL_OUT) != 0)
2024 			return;
2025 		if (m == NULL)
2026 			return;
2027 	}
2028 
2029 	bridge_enqueue(sc, dst_if, m);
2030 }
2031 
2032 /*
2033  * bridge_input:
2034  *
2035  *	Receive input from a member interface.  Queue the packet for
2036  *	bridging if it is not for us.
2037  */
2038 static struct mbuf *
2039 bridge_input(struct ifnet *ifp, struct mbuf *m)
2040 {
2041 	struct bridge_softc *sc = ifp->if_bridge;
2042 	struct bridge_iflist *bif;
2043 	struct ifnet *bifp;
2044 	struct ether_header *eh;
2045 	struct mbuf *mc, *mc2;
2046 
2047 	if ((sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING) == 0)
2048 		return (m);
2049 
2050 	bifp = sc->sc_ifp;
2051 
2052 	/*
2053 	 * Implement support for bridge monitoring. If this flag has been
2054 	 * set on this interface, discard the packet once we push it through
2055 	 * the bpf(4) machinery, but before we do, increment the byte and
2056 	 * packet counters associated with this interface.
2057 	 */
2058 	if ((bifp->if_flags & IFF_MONITOR) != 0) {
2059 		m->m_pkthdr.rcvif  = bifp;
2060 		BPF_MTAP(bifp, m);
2061 		bifp->if_ipackets++;
2062 		bifp->if_ibytes += m->m_pkthdr.len;
2063 		m_free(m);
2064 		return (NULL);
2065 	}
2066 	BRIDGE_LOCK(sc);
2067 	bif = bridge_lookup_member_if(sc, ifp);
2068 	if (bif == NULL) {
2069 		BRIDGE_UNLOCK(sc);
2070 		return (m);
2071 	}
2072 
2073 	eh = mtod(m, struct ether_header *);
2074 
2075 	if (memcmp(eh->ether_dhost, IF_LLADDR(bifp),
2076 	    ETHER_ADDR_LEN) == 0) {
2077 		/*
2078 		 * If the packet is for us, set the packets source as the
2079 		 * bridge, and return the packet back to ether_input for
2080 		 * local processing.
2081 		 */
2082 
2083 		/* Mark the packet as arriving on the bridge interface */
2084 		m->m_pkthdr.rcvif = bifp;
2085 		BPF_MTAP(bifp, m);
2086 		bifp->if_ipackets++;
2087 
2088 		BRIDGE_UNLOCK(sc);
2089 		return (m);
2090 	}
2091 
2092 	bridge_span(sc, m);
2093 
2094 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
2095 		/* Tap off 802.1D packets; they do not get forwarded. */
2096 		if (memcmp(eh->ether_dhost, bstp_etheraddr,
2097 		    ETHER_ADDR_LEN) == 0) {
2098 			m = bstp_input(&bif->bif_stp, ifp, m);
2099 			if (m == NULL) {
2100 				BRIDGE_UNLOCK(sc);
2101 				return (NULL);
2102 			}
2103 		}
2104 
2105 		if (bif->bif_flags & IFBIF_STP) {
2106 			switch (bif->bif_stp.bp_state) {
2107 			case BSTP_IFSTATE_BLOCKING:
2108 			case BSTP_IFSTATE_LISTENING:
2109 			case BSTP_IFSTATE_DISABLED:
2110 				BRIDGE_UNLOCK(sc);
2111 				return (m);
2112 			}
2113 		}
2114 
2115 		if (bcmp(etherbroadcastaddr, eh->ether_dhost,
2116 		    sizeof(etherbroadcastaddr)) == 0)
2117 			m->m_flags |= M_BCAST;
2118 		else
2119 			m->m_flags |= M_MCAST;
2120 
2121 		/*
2122 		 * Make a deep copy of the packet and enqueue the copy
2123 		 * for bridge processing; return the original packet for
2124 		 * local processing.
2125 		 */
2126 		mc = m_dup(m, M_DONTWAIT);
2127 		if (mc == NULL) {
2128 			BRIDGE_UNLOCK(sc);
2129 			return (m);
2130 		}
2131 
2132 		/* Perform the bridge forwarding function with the copy. */
2133 		bridge_forward(sc, mc);
2134 
2135 		/*
2136 		 * Reinject the mbuf as arriving on the bridge so we have a
2137 		 * chance at claiming multicast packets. We can not loop back
2138 		 * here from ether_input as a bridge is never a member of a
2139 		 * bridge.
2140 		 */
2141 		KASSERT(bifp->if_bridge == NULL,
2142 		    ("loop created in bridge_input"));
2143 		mc2 = m_dup(m, M_DONTWAIT);
2144 		if (mc2 != NULL) {
2145 			/* Keep the layer3 header aligned */
2146 			int i = min(mc2->m_pkthdr.len, max_protohdr);
2147 			mc2 = m_copyup(mc2, i, ETHER_ALIGN);
2148 		}
2149 		if (mc2 != NULL) {
2150 			mc2->m_pkthdr.rcvif = bifp;
2151 			(*bifp->if_input)(bifp, mc2);
2152 		}
2153 
2154 		/* Return the original packet for local processing. */
2155 		return (m);
2156 	}
2157 
2158 	if (bif->bif_flags & IFBIF_STP) {
2159 		switch (bif->bif_stp.bp_state) {
2160 		case BSTP_IFSTATE_BLOCKING:
2161 		case BSTP_IFSTATE_LISTENING:
2162 		case BSTP_IFSTATE_DISABLED:
2163 			BRIDGE_UNLOCK(sc);
2164 			return (m);
2165 		}
2166 	}
2167 
2168 	/*
2169 	 * Unicast.  Make sure it's not for us.
2170 	 */
2171 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2172 		if (bif->bif_ifp->if_type == IFT_GIF)
2173 			continue;
2174 		/* It is destined for us. */
2175 		if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_dhost,
2176 		    ETHER_ADDR_LEN) == 0
2177 #ifdef DEV_CARP
2178 		    || (bif->bif_ifp->if_carp
2179 			&& carp_forus(bif->bif_ifp->if_carp, eh->ether_dhost))
2180 #endif
2181 		    ) {
2182 			if (bif->bif_flags & IFBIF_LEARNING)
2183 				(void) bridge_rtupdate(sc,
2184 				    eh->ether_shost, ifp, 0, IFBAF_DYNAMIC);
2185 			m->m_pkthdr.rcvif = bif->bif_ifp;
2186 			BRIDGE_UNLOCK(sc);
2187 			return (m);
2188 		}
2189 
2190 		/* We just received a packet that we sent out. */
2191 		if (memcmp(IF_LLADDR(bif->bif_ifp), eh->ether_shost,
2192 		    ETHER_ADDR_LEN) == 0
2193 #ifdef DEV_CARP
2194 		    || (bif->bif_ifp->if_carp
2195 			&& carp_forus(bif->bif_ifp->if_carp, eh->ether_shost))
2196 #endif
2197 		    ) {
2198 			BRIDGE_UNLOCK(sc);
2199 			m_freem(m);
2200 			return (NULL);
2201 		}
2202 	}
2203 
2204 	/* Perform the bridge forwarding function. */
2205 	bridge_forward(sc, m);
2206 
2207 	return (NULL);
2208 }
2209 
2210 /*
2211  * bridge_broadcast:
2212  *
2213  *	Send a frame to all interfaces that are members of
2214  *	the bridge, except for the one on which the packet
2215  *	arrived.
2216  *
2217  *	NOTE: Releases the lock on return.
2218  */
2219 static void
2220 bridge_broadcast(struct bridge_softc *sc, struct ifnet *src_if,
2221     struct mbuf *m, int runfilt)
2222 {
2223 	struct bridge_iflist *bif;
2224 	struct mbuf *mc;
2225 	struct ifnet *dst_if;
2226 	int error = 0, used = 0, i;
2227 
2228 	BRIDGE_LOCK_ASSERT(sc);
2229 	BRIDGE_LOCK2REF(sc, error);
2230 	if (error) {
2231 		m_freem(m);
2232 		return;
2233 	}
2234 
2235 	/* Filter on the bridge interface before broadcasting */
2236 	if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2237 #ifdef INET6
2238 	    || PFIL_HOOKED(&inet6_pfil_hook)
2239 #endif
2240 	    )) {
2241 		if (bridge_pfil(&m, sc->sc_ifp, NULL, PFIL_OUT) != 0)
2242 			goto out;
2243 		if (m == NULL)
2244 			goto out;
2245 	}
2246 
2247 	LIST_FOREACH(bif, &sc->sc_iflist, bif_next) {
2248 		dst_if = bif->bif_ifp;
2249 		if (dst_if == src_if)
2250 			continue;
2251 
2252 		if (bif->bif_flags & IFBIF_STP) {
2253 			switch (bif->bif_stp.bp_state) {
2254 			case BSTP_IFSTATE_BLOCKING:
2255 			case BSTP_IFSTATE_DISABLED:
2256 				continue;
2257 			}
2258 		}
2259 
2260 		if ((bif->bif_flags & IFBIF_DISCOVER) == 0 &&
2261 		    (m->m_flags & (M_BCAST|M_MCAST)) == 0)
2262 			continue;
2263 
2264 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2265 			continue;
2266 
2267 		if (LIST_NEXT(bif, bif_next) == NULL) {
2268 			mc = m;
2269 			used = 1;
2270 		} else {
2271 			mc = m_dup(m, M_DONTWAIT);
2272 			if (mc == NULL) {
2273 				sc->sc_ifp->if_oerrors++;
2274 				continue;
2275 			}
2276 		}
2277 
2278 		/*
2279 		 * Filter on the output interface. Pass a NULL bridge interface
2280 		 * pointer so we do not redundantly filter on the bridge for
2281 		 * each interface we broadcast on.
2282 		 */
2283 		if (runfilt && (PFIL_HOOKED(&inet_pfil_hook)
2284 #ifdef INET6
2285 		    || PFIL_HOOKED(&inet6_pfil_hook)
2286 #endif
2287 		    )) {
2288 			if (used == 0) {
2289 				/* Keep the layer3 header aligned */
2290 				i = min(mc->m_pkthdr.len, max_protohdr);
2291 				mc = m_copyup(mc, i, ETHER_ALIGN);
2292 				if (mc == NULL) {
2293 					sc->sc_ifp->if_oerrors++;
2294 					continue;
2295 				}
2296 			}
2297 			if (bridge_pfil(&mc, NULL, dst_if, PFIL_OUT) != 0)
2298 				continue;
2299 			if (mc == NULL)
2300 				continue;
2301 		}
2302 
2303 		bridge_enqueue(sc, dst_if, mc);
2304 	}
2305 	if (used == 0)
2306 		m_freem(m);
2307 
2308 out:
2309 	BRIDGE_UNREF(sc);
2310 }
2311 
2312 /*
2313  * bridge_span:
2314  *
2315  *	Duplicate a packet out one or more interfaces that are in span mode,
2316  *	the original mbuf is unmodified.
2317  */
2318 static void
2319 bridge_span(struct bridge_softc *sc, struct mbuf *m)
2320 {
2321 	struct bridge_iflist *bif;
2322 	struct ifnet *dst_if;
2323 	struct mbuf *mc;
2324 
2325 	if (LIST_EMPTY(&sc->sc_spanlist))
2326 		return;
2327 
2328 	LIST_FOREACH(bif, &sc->sc_spanlist, bif_next) {
2329 		dst_if = bif->bif_ifp;
2330 
2331 		if ((dst_if->if_drv_flags & IFF_DRV_RUNNING) == 0)
2332 			continue;
2333 
2334 		mc = m_copypacket(m, M_DONTWAIT);
2335 		if (mc == NULL) {
2336 			sc->sc_ifp->if_oerrors++;
2337 			continue;
2338 		}
2339 
2340 		bridge_enqueue(sc, dst_if, mc);
2341 	}
2342 }
2343 
2344 /*
2345  * bridge_rtupdate:
2346  *
2347  *	Add a bridge routing entry.
2348  */
2349 static int
2350 bridge_rtupdate(struct bridge_softc *sc, const uint8_t *dst,
2351     struct ifnet *dst_if, int setflags, uint8_t flags)
2352 {
2353 	struct bridge_rtnode *brt;
2354 	int error;
2355 
2356 	BRIDGE_LOCK_ASSERT(sc);
2357 
2358 	/*
2359 	 * A route for this destination might already exist.  If so,
2360 	 * update it, otherwise create a new one.
2361 	 */
2362 	if ((brt = bridge_rtnode_lookup(sc, dst)) == NULL) {
2363 		if (sc->sc_brtcnt >= sc->sc_brtmax) {
2364 			sc->sc_brtexceeded++;
2365 			return (ENOSPC);
2366 		}
2367 
2368 		/*
2369 		 * Allocate a new bridge forwarding node, and
2370 		 * initialize the expiration time and Ethernet
2371 		 * address.
2372 		 */
2373 		brt = uma_zalloc(bridge_rtnode_zone, M_NOWAIT | M_ZERO);
2374 		if (brt == NULL)
2375 			return (ENOMEM);
2376 
2377 		brt->brt_flags = IFBAF_DYNAMIC;
2378 		memcpy(brt->brt_addr, dst, ETHER_ADDR_LEN);
2379 
2380 		if ((error = bridge_rtnode_insert(sc, brt)) != 0) {
2381 			uma_zfree(bridge_rtnode_zone, brt);
2382 			return (error);
2383 		}
2384 	}
2385 
2386 	if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2387 		brt->brt_ifp = dst_if;
2388 	if ((flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2389 		brt->brt_expire = time_uptime + sc->sc_brttimeout;
2390 	if (setflags)
2391 		brt->brt_flags = flags;
2392 
2393 	return (0);
2394 }
2395 
2396 /*
2397  * bridge_rtlookup:
2398  *
2399  *	Lookup the destination interface for an address.
2400  */
2401 static struct ifnet *
2402 bridge_rtlookup(struct bridge_softc *sc, const uint8_t *addr)
2403 {
2404 	struct bridge_rtnode *brt;
2405 
2406 	BRIDGE_LOCK_ASSERT(sc);
2407 
2408 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2409 		return (NULL);
2410 
2411 	return (brt->brt_ifp);
2412 }
2413 
2414 /*
2415  * bridge_rttrim:
2416  *
2417  *	Trim the routine table so that we have a number
2418  *	of routing entries less than or equal to the
2419  *	maximum number.
2420  */
2421 static void
2422 bridge_rttrim(struct bridge_softc *sc)
2423 {
2424 	struct bridge_rtnode *brt, *nbrt;
2425 
2426 	BRIDGE_LOCK_ASSERT(sc);
2427 
2428 	/* Make sure we actually need to do this. */
2429 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2430 		return;
2431 
2432 	/* Force an aging cycle; this might trim enough addresses. */
2433 	bridge_rtage(sc);
2434 	if (sc->sc_brtcnt <= sc->sc_brtmax)
2435 		return;
2436 
2437 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2438 		nbrt = LIST_NEXT(brt, brt_list);
2439 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2440 			bridge_rtnode_destroy(sc, brt);
2441 			if (sc->sc_brtcnt <= sc->sc_brtmax)
2442 				return;
2443 		}
2444 	}
2445 }
2446 
2447 /*
2448  * bridge_timer:
2449  *
2450  *	Aging timer for the bridge.
2451  */
2452 static void
2453 bridge_timer(void *arg)
2454 {
2455 	struct bridge_softc *sc = arg;
2456 
2457 	BRIDGE_LOCK_ASSERT(sc);
2458 
2459 	bridge_rtage(sc);
2460 
2461 	if (sc->sc_ifp->if_drv_flags & IFF_DRV_RUNNING)
2462 		callout_reset(&sc->sc_brcallout,
2463 		    bridge_rtable_prune_period * hz, bridge_timer, sc);
2464 }
2465 
2466 /*
2467  * bridge_rtage:
2468  *
2469  *	Perform an aging cycle.
2470  */
2471 static void
2472 bridge_rtage(struct bridge_softc *sc)
2473 {
2474 	struct bridge_rtnode *brt, *nbrt;
2475 
2476 	BRIDGE_LOCK_ASSERT(sc);
2477 
2478 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2479 		nbrt = LIST_NEXT(brt, brt_list);
2480 		if ((brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC) {
2481 			if (time_uptime >= brt->brt_expire)
2482 				bridge_rtnode_destroy(sc, brt);
2483 		}
2484 	}
2485 }
2486 
2487 /*
2488  * bridge_rtflush:
2489  *
2490  *	Remove all dynamic addresses from the bridge.
2491  */
2492 static void
2493 bridge_rtflush(struct bridge_softc *sc, int full)
2494 {
2495 	struct bridge_rtnode *brt, *nbrt;
2496 
2497 	BRIDGE_LOCK_ASSERT(sc);
2498 
2499 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2500 		nbrt = LIST_NEXT(brt, brt_list);
2501 		if (full || (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC)
2502 			bridge_rtnode_destroy(sc, brt);
2503 	}
2504 }
2505 
2506 /*
2507  * bridge_rtdaddr:
2508  *
2509  *	Remove an address from the table.
2510  */
2511 static int
2512 bridge_rtdaddr(struct bridge_softc *sc, const uint8_t *addr)
2513 {
2514 	struct bridge_rtnode *brt;
2515 
2516 	BRIDGE_LOCK_ASSERT(sc);
2517 
2518 	if ((brt = bridge_rtnode_lookup(sc, addr)) == NULL)
2519 		return (ENOENT);
2520 
2521 	bridge_rtnode_destroy(sc, brt);
2522 	return (0);
2523 }
2524 
2525 /*
2526  * bridge_rtdelete:
2527  *
2528  *	Delete routes to a speicifc member interface.
2529  */
2530 static void
2531 bridge_rtdelete(struct bridge_softc *sc, struct ifnet *ifp, int full)
2532 {
2533 	struct bridge_rtnode *brt, *nbrt;
2534 
2535 	BRIDGE_LOCK_ASSERT(sc);
2536 
2537 	for (brt = LIST_FIRST(&sc->sc_rtlist); brt != NULL; brt = nbrt) {
2538 		nbrt = LIST_NEXT(brt, brt_list);
2539 		if (brt->brt_ifp == ifp && (full ||
2540 			    (brt->brt_flags & IFBAF_TYPEMASK) == IFBAF_DYNAMIC))
2541 			bridge_rtnode_destroy(sc, brt);
2542 	}
2543 }
2544 
2545 /*
2546  * bridge_rtable_init:
2547  *
2548  *	Initialize the route table for this bridge.
2549  */
2550 static int
2551 bridge_rtable_init(struct bridge_softc *sc)
2552 {
2553 	int i;
2554 
2555 	sc->sc_rthash = malloc(sizeof(*sc->sc_rthash) * BRIDGE_RTHASH_SIZE,
2556 	    M_DEVBUF, M_NOWAIT);
2557 	if (sc->sc_rthash == NULL)
2558 		return (ENOMEM);
2559 
2560 	for (i = 0; i < BRIDGE_RTHASH_SIZE; i++)
2561 		LIST_INIT(&sc->sc_rthash[i]);
2562 
2563 	sc->sc_rthash_key = arc4random();
2564 
2565 	LIST_INIT(&sc->sc_rtlist);
2566 
2567 	return (0);
2568 }
2569 
2570 /*
2571  * bridge_rtable_fini:
2572  *
2573  *	Deconstruct the route table for this bridge.
2574  */
2575 static void
2576 bridge_rtable_fini(struct bridge_softc *sc)
2577 {
2578 
2579 	free(sc->sc_rthash, M_DEVBUF);
2580 }
2581 
2582 /*
2583  * The following hash function is adapted from "Hash Functions" by Bob Jenkins
2584  * ("Algorithm Alley", Dr. Dobbs Journal, September 1997).
2585  */
2586 #define	mix(a, b, c)							\
2587 do {									\
2588 	a -= b; a -= c; a ^= (c >> 13);					\
2589 	b -= c; b -= a; b ^= (a << 8);					\
2590 	c -= a; c -= b; c ^= (b >> 13);					\
2591 	a -= b; a -= c; a ^= (c >> 12);					\
2592 	b -= c; b -= a; b ^= (a << 16);					\
2593 	c -= a; c -= b; c ^= (b >> 5);					\
2594 	a -= b; a -= c; a ^= (c >> 3);					\
2595 	b -= c; b -= a; b ^= (a << 10);					\
2596 	c -= a; c -= b; c ^= (b >> 15);					\
2597 } while (/*CONSTCOND*/0)
2598 
2599 static __inline uint32_t
2600 bridge_rthash(struct bridge_softc *sc, const uint8_t *addr)
2601 {
2602 	uint32_t a = 0x9e3779b9, b = 0x9e3779b9, c = sc->sc_rthash_key;
2603 
2604 	b += addr[5] << 8;
2605 	b += addr[4];
2606 	a += addr[3] << 24;
2607 	a += addr[2] << 16;
2608 	a += addr[1] << 8;
2609 	a += addr[0];
2610 
2611 	mix(a, b, c);
2612 
2613 	return (c & BRIDGE_RTHASH_MASK);
2614 }
2615 
2616 #undef mix
2617 
2618 static int
2619 bridge_rtnode_addr_cmp(const uint8_t *a, const uint8_t *b)
2620 {
2621 	int i, d;
2622 
2623 	for (i = 0, d = 0; i < ETHER_ADDR_LEN && d == 0; i++) {
2624 		d = ((int)a[i]) - ((int)b[i]);
2625 	}
2626 
2627 	return (d);
2628 }
2629 
2630 /*
2631  * bridge_rtnode_lookup:
2632  *
2633  *	Look up a bridge route node for the specified destination.
2634  */
2635 static struct bridge_rtnode *
2636 bridge_rtnode_lookup(struct bridge_softc *sc, const uint8_t *addr)
2637 {
2638 	struct bridge_rtnode *brt;
2639 	uint32_t hash;
2640 	int dir;
2641 
2642 	BRIDGE_LOCK_ASSERT(sc);
2643 
2644 	hash = bridge_rthash(sc, addr);
2645 	LIST_FOREACH(brt, &sc->sc_rthash[hash], brt_hash) {
2646 		dir = bridge_rtnode_addr_cmp(addr, brt->brt_addr);
2647 		if (dir == 0)
2648 			return (brt);
2649 		if (dir > 0)
2650 			return (NULL);
2651 	}
2652 
2653 	return (NULL);
2654 }
2655 
2656 /*
2657  * bridge_rtnode_insert:
2658  *
2659  *	Insert the specified bridge node into the route table.  We
2660  *	assume the entry is not already in the table.
2661  */
2662 static int
2663 bridge_rtnode_insert(struct bridge_softc *sc, struct bridge_rtnode *brt)
2664 {
2665 	struct bridge_rtnode *lbrt;
2666 	uint32_t hash;
2667 	int dir;
2668 
2669 	BRIDGE_LOCK_ASSERT(sc);
2670 
2671 	hash = bridge_rthash(sc, brt->brt_addr);
2672 
2673 	lbrt = LIST_FIRST(&sc->sc_rthash[hash]);
2674 	if (lbrt == NULL) {
2675 		LIST_INSERT_HEAD(&sc->sc_rthash[hash], brt, brt_hash);
2676 		goto out;
2677 	}
2678 
2679 	do {
2680 		dir = bridge_rtnode_addr_cmp(brt->brt_addr, lbrt->brt_addr);
2681 		if (dir == 0)
2682 			return (EEXIST);
2683 		if (dir > 0) {
2684 			LIST_INSERT_BEFORE(lbrt, brt, brt_hash);
2685 			goto out;
2686 		}
2687 		if (LIST_NEXT(lbrt, brt_hash) == NULL) {
2688 			LIST_INSERT_AFTER(lbrt, brt, brt_hash);
2689 			goto out;
2690 		}
2691 		lbrt = LIST_NEXT(lbrt, brt_hash);
2692 	} while (lbrt != NULL);
2693 
2694 #ifdef DIAGNOSTIC
2695 	panic("bridge_rtnode_insert: impossible");
2696 #endif
2697 
2698 out:
2699 	LIST_INSERT_HEAD(&sc->sc_rtlist, brt, brt_list);
2700 	sc->sc_brtcnt++;
2701 
2702 	return (0);
2703 }
2704 
2705 /*
2706  * bridge_rtnode_destroy:
2707  *
2708  *	Destroy a bridge rtnode.
2709  */
2710 static void
2711 bridge_rtnode_destroy(struct bridge_softc *sc, struct bridge_rtnode *brt)
2712 {
2713 	BRIDGE_LOCK_ASSERT(sc);
2714 
2715 	LIST_REMOVE(brt, brt_hash);
2716 
2717 	LIST_REMOVE(brt, brt_list);
2718 	sc->sc_brtcnt--;
2719 	uma_zfree(bridge_rtnode_zone, brt);
2720 }
2721 
2722 /*
2723  * bridge_state_change:
2724  *
2725  *	Callback from the bridgestp code when a port changes states.
2726  */
2727 static void
2728 bridge_state_change(struct ifnet *ifp, int state)
2729 {
2730 	struct bridge_softc *sc = ifp->if_bridge;
2731 	static const char *stpstates[] = {
2732 		"disabled",
2733 		"listening",
2734 		"learning",
2735 		"forwarding",
2736 		"blocking",
2737 	};
2738 
2739 	if (log_stp)
2740 		log(LOG_NOTICE, "%s: state changed to %s on %s\n",
2741 		    sc->sc_ifp->if_xname, stpstates[state], ifp->if_xname);
2742 
2743 	/* if the port is blocking then remove any routes to it */
2744 	switch (state) {
2745 		case BSTP_IFSTATE_DISABLED:
2746 		case BSTP_IFSTATE_BLOCKING:
2747 			BRIDGE_LOCK(sc);
2748 			bridge_rtdelete(sc, ifp, IFBF_FLUSHDYN);
2749 			BRIDGE_UNLOCK(sc);
2750 	}
2751 }
2752 
2753 /*
2754  * Send bridge packets through pfil if they are one of the types pfil can deal
2755  * with, or if they are ARP or REVARP.  (pfil will pass ARP and REVARP without
2756  * question.) If *bifp or *ifp are NULL then packet filtering is skipped for
2757  * that interface.
2758  */
2759 static int
2760 bridge_pfil(struct mbuf **mp, struct ifnet *bifp, struct ifnet *ifp, int dir)
2761 {
2762 	int snap, error, i, hlen;
2763 	struct ether_header *eh1, eh2;
2764 	struct ip_fw_args args;
2765 	struct ip *ip;
2766 	struct llc llc1;
2767 	u_int16_t ether_type;
2768 
2769 	snap = 0;
2770 	error = -1;	/* Default error if not error == 0 */
2771 
2772 	/* we may return with the IP fields swapped, ensure its not shared */
2773 	KASSERT(M_WRITABLE(*mp), ("%s: modifying a shared mbuf", __func__));
2774 
2775 	if (pfil_bridge == 0 && pfil_member == 0 && pfil_ipfw == 0)
2776 		return (0); /* filtering is disabled */
2777 
2778 	i = min((*mp)->m_pkthdr.len, max_protohdr);
2779 	if ((*mp)->m_len < i) {
2780 	    *mp = m_pullup(*mp, i);
2781 	    if (*mp == NULL) {
2782 		printf("%s: m_pullup failed\n", __func__);
2783 		return (-1);
2784 	    }
2785 	}
2786 
2787 	eh1 = mtod(*mp, struct ether_header *);
2788 	ether_type = ntohs(eh1->ether_type);
2789 
2790 	/*
2791 	 * Check for SNAP/LLC.
2792 	 */
2793 	if (ether_type < ETHERMTU) {
2794 		struct llc *llc2 = (struct llc *)(eh1 + 1);
2795 
2796 		if ((*mp)->m_len >= ETHER_HDR_LEN + 8 &&
2797 		    llc2->llc_dsap == LLC_SNAP_LSAP &&
2798 		    llc2->llc_ssap == LLC_SNAP_LSAP &&
2799 		    llc2->llc_control == LLC_UI) {
2800 			ether_type = htons(llc2->llc_un.type_snap.ether_type);
2801 			snap = 1;
2802 		}
2803 	}
2804 
2805 	/*
2806 	 * If we're trying to filter bridge traffic, don't look at anything
2807 	 * other than IP and ARP traffic.  If the filter doesn't understand
2808 	 * IPv6, don't allow IPv6 through the bridge either.  This is lame
2809 	 * since if we really wanted, say, an AppleTalk filter, we are hosed,
2810 	 * but of course we don't have an AppleTalk filter to begin with.
2811 	 * (Note that since pfil doesn't understand ARP it will pass *ALL*
2812 	 * ARP traffic.)
2813 	 */
2814 	switch (ether_type) {
2815 		case ETHERTYPE_ARP:
2816 		case ETHERTYPE_REVARP:
2817 			return (0); /* Automatically pass */
2818 		case ETHERTYPE_IP:
2819 #ifdef INET6
2820 		case ETHERTYPE_IPV6:
2821 #endif /* INET6 */
2822 			break;
2823 		default:
2824 			/*
2825 			 * Check to see if the user wants to pass non-ip
2826 			 * packets, these will not be checked by pfil(9) and
2827 			 * passed unconditionally so the default is to drop.
2828 			 */
2829 			if (pfil_onlyip)
2830 				goto bad;
2831 	}
2832 
2833 	/* Strip off the Ethernet header and keep a copy. */
2834 	m_copydata(*mp, 0, ETHER_HDR_LEN, (caddr_t) &eh2);
2835 	m_adj(*mp, ETHER_HDR_LEN);
2836 
2837 	/* Strip off snap header, if present */
2838 	if (snap) {
2839 		m_copydata(*mp, 0, sizeof(struct llc), (caddr_t) &llc1);
2840 		m_adj(*mp, sizeof(struct llc));
2841 	}
2842 
2843 	/*
2844 	 * Check the IP header for alignment and errors
2845 	 */
2846 	if (dir == PFIL_IN) {
2847 		switch (ether_type) {
2848 			case ETHERTYPE_IP:
2849 				error = bridge_ip_checkbasic(mp);
2850 				break;
2851 #ifdef INET6
2852 			case ETHERTYPE_IPV6:
2853 				error = bridge_ip6_checkbasic(mp);
2854 				break;
2855 #endif /* INET6 */
2856 			default:
2857 				error = 0;
2858 		}
2859 		if (error)
2860 			goto bad;
2861 	}
2862 
2863 	if (IPFW_LOADED && pfil_ipfw != 0 && dir == PFIL_OUT && ifp != NULL) {
2864 		error = -1;
2865 		args.rule = ip_dn_claim_rule(*mp);
2866 		if (args.rule != NULL && fw_one_pass)
2867 			goto ipfwpass; /* packet already partially processed */
2868 
2869 		args.m = *mp;
2870 		args.oif = ifp;
2871 		args.next_hop = NULL;
2872 		args.eh = &eh2;
2873 		args.inp = NULL;	/* used by ipfw uid/gid/jail rules */
2874 		i = ip_fw_chk_ptr(&args);
2875 		*mp = args.m;
2876 
2877 		if (*mp == NULL)
2878 			return (error);
2879 
2880 		if (DUMMYNET_LOADED && (i == IP_FW_DUMMYNET)) {
2881 
2882 			/* put the Ethernet header back on */
2883 			M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
2884 			if (*mp == NULL)
2885 				return (error);
2886 			bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
2887 
2888 			/*
2889 			 * Pass the pkt to dummynet, which consumes it. The
2890 			 * packet will return to us via bridge_dummynet().
2891 			 */
2892 			args.oif = ifp;
2893 			ip_dn_io_ptr(*mp, DN_TO_IFB_FWD, &args);
2894 			return (error);
2895 		}
2896 
2897 		if (i != IP_FW_PASS) /* drop */
2898 			goto bad;
2899 	}
2900 
2901 ipfwpass:
2902 	error = 0;
2903 
2904 	/*
2905 	 * Run the packet through pfil
2906 	 */
2907 	switch (ether_type) {
2908 	case ETHERTYPE_IP:
2909 		/*
2910 		 * before calling the firewall, swap fields the same as
2911 		 * IP does. here we assume the header is contiguous
2912 		 */
2913 		ip = mtod(*mp, struct ip *);
2914 
2915 		ip->ip_len = ntohs(ip->ip_len);
2916 		ip->ip_off = ntohs(ip->ip_off);
2917 
2918 		/*
2919 		 * Run pfil on the member interface and the bridge, both can
2920 		 * be skipped by clearing pfil_member or pfil_bridge.
2921 		 *
2922 		 * Keep the order:
2923 		 *   in_if -> bridge_if -> out_if
2924 		 */
2925 		if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
2926 			error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
2927 					dir, NULL);
2928 
2929 		if (*mp == NULL || error != 0) /* filter may consume */
2930 			break;
2931 
2932 		if (pfil_member && ifp != NULL)
2933 			error = pfil_run_hooks(&inet_pfil_hook, mp, ifp,
2934 					dir, NULL);
2935 
2936 		if (*mp == NULL || error != 0) /* filter may consume */
2937 			break;
2938 
2939 		if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
2940 			error = pfil_run_hooks(&inet_pfil_hook, mp, bifp,
2941 					dir, NULL);
2942 
2943 		if (*mp == NULL || error != 0) /* filter may consume */
2944 			break;
2945 
2946 		/* check if we need to fragment the packet */
2947 		if (pfil_member && ifp != NULL && dir == PFIL_OUT) {
2948 			i = (*mp)->m_pkthdr.len;
2949 			if (i > ifp->if_mtu) {
2950 				error = bridge_fragment(ifp, *mp, &eh2, snap,
2951 					    &llc1);
2952 				return (error);
2953 			}
2954 		}
2955 
2956 		/* Recalculate the ip checksum and restore byte ordering */
2957 		ip = mtod(*mp, struct ip *);
2958 		hlen = ip->ip_hl << 2;
2959 		if (hlen < sizeof(struct ip))
2960 			goto bad;
2961 		if (hlen > (*mp)->m_len) {
2962 			if ((*mp = m_pullup(*mp, hlen)) == 0)
2963 				goto bad;
2964 			ip = mtod(*mp, struct ip *);
2965 			if (ip == NULL)
2966 				goto bad;
2967 		}
2968 		ip->ip_len = htons(ip->ip_len);
2969 		ip->ip_off = htons(ip->ip_off);
2970 		ip->ip_sum = 0;
2971 		if (hlen == sizeof(struct ip))
2972 			ip->ip_sum = in_cksum_hdr(ip);
2973 		else
2974 			ip->ip_sum = in_cksum(*mp, hlen);
2975 
2976 		break;
2977 #ifdef INET6
2978 	case ETHERTYPE_IPV6:
2979 		if (pfil_bridge && dir == PFIL_OUT && bifp != NULL)
2980 			error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
2981 					dir, NULL);
2982 
2983 		if (*mp == NULL || error != 0) /* filter may consume */
2984 			break;
2985 
2986 		if (pfil_member && ifp != NULL)
2987 			error = pfil_run_hooks(&inet6_pfil_hook, mp, ifp,
2988 					dir, NULL);
2989 
2990 		if (*mp == NULL || error != 0) /* filter may consume */
2991 			break;
2992 
2993 		if (pfil_bridge && dir == PFIL_IN && bifp != NULL)
2994 			error = pfil_run_hooks(&inet6_pfil_hook, mp, bifp,
2995 					dir, NULL);
2996 		break;
2997 #endif
2998 	default:
2999 		error = 0;
3000 		break;
3001 	}
3002 
3003 	if (*mp == NULL)
3004 		return (error);
3005 	if (error != 0)
3006 		goto bad;
3007 
3008 	error = -1;
3009 
3010 	/*
3011 	 * Finally, put everything back the way it was and return
3012 	 */
3013 	if (snap) {
3014 		M_PREPEND(*mp, sizeof(struct llc), M_DONTWAIT);
3015 		if (*mp == NULL)
3016 			return (error);
3017 		bcopy(&llc1, mtod(*mp, caddr_t), sizeof(struct llc));
3018 	}
3019 
3020 	M_PREPEND(*mp, ETHER_HDR_LEN, M_DONTWAIT);
3021 	if (*mp == NULL)
3022 		return (error);
3023 	bcopy(&eh2, mtod(*mp, caddr_t), ETHER_HDR_LEN);
3024 
3025 	return (0);
3026 
3027 bad:
3028 	m_freem(*mp);
3029 	*mp = NULL;
3030 	return (error);
3031 }
3032 
3033 /*
3034  * Perform basic checks on header size since
3035  * pfil assumes ip_input has already processed
3036  * it for it.  Cut-and-pasted from ip_input.c.
3037  * Given how simple the IPv6 version is,
3038  * does the IPv4 version really need to be
3039  * this complicated?
3040  *
3041  * XXX Should we update ipstat here, or not?
3042  * XXX Right now we update ipstat but not
3043  * XXX csum_counter.
3044  */
3045 static int
3046 bridge_ip_checkbasic(struct mbuf **mp)
3047 {
3048 	struct mbuf *m = *mp;
3049 	struct ip *ip;
3050 	int len, hlen;
3051 	u_short sum;
3052 
3053 	if (*mp == NULL)
3054 		return (-1);
3055 
3056 	if (IP_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3057 		if ((m = m_copyup(m, sizeof(struct ip),
3058 			(max_linkhdr + 3) & ~3)) == NULL) {
3059 			/* XXXJRT new stat, please */
3060 			ipstat.ips_toosmall++;
3061 			goto bad;
3062 		}
3063 	} else if (__predict_false(m->m_len < sizeof (struct ip))) {
3064 		if ((m = m_pullup(m, sizeof (struct ip))) == NULL) {
3065 			ipstat.ips_toosmall++;
3066 			goto bad;
3067 		}
3068 	}
3069 	ip = mtod(m, struct ip *);
3070 	if (ip == NULL) goto bad;
3071 
3072 	if (ip->ip_v != IPVERSION) {
3073 		ipstat.ips_badvers++;
3074 		goto bad;
3075 	}
3076 	hlen = ip->ip_hl << 2;
3077 	if (hlen < sizeof(struct ip)) { /* minimum header length */
3078 		ipstat.ips_badhlen++;
3079 		goto bad;
3080 	}
3081 	if (hlen > m->m_len) {
3082 		if ((m = m_pullup(m, hlen)) == 0) {
3083 			ipstat.ips_badhlen++;
3084 			goto bad;
3085 		}
3086 		ip = mtod(m, struct ip *);
3087 		if (ip == NULL) goto bad;
3088 	}
3089 
3090 	if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) {
3091 		sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID);
3092 	} else {
3093 		if (hlen == sizeof(struct ip)) {
3094 			sum = in_cksum_hdr(ip);
3095 		} else {
3096 			sum = in_cksum(m, hlen);
3097 		}
3098 	}
3099 	if (sum) {
3100 		ipstat.ips_badsum++;
3101 		goto bad;
3102 	}
3103 
3104 	/* Retrieve the packet length. */
3105 	len = ntohs(ip->ip_len);
3106 
3107 	/*
3108 	 * Check for additional length bogosity
3109 	 */
3110 	if (len < hlen) {
3111 		ipstat.ips_badlen++;
3112 		goto bad;
3113 	}
3114 
3115 	/*
3116 	 * Check that the amount of data in the buffers
3117 	 * is as at least much as the IP header would have us expect.
3118 	 * Drop packet if shorter than we expect.
3119 	 */
3120 	if (m->m_pkthdr.len < len) {
3121 		ipstat.ips_tooshort++;
3122 		goto bad;
3123 	}
3124 
3125 	/* Checks out, proceed */
3126 	*mp = m;
3127 	return (0);
3128 
3129 bad:
3130 	*mp = m;
3131 	return (-1);
3132 }
3133 
3134 #ifdef INET6
3135 /*
3136  * Same as above, but for IPv6.
3137  * Cut-and-pasted from ip6_input.c.
3138  * XXX Should we update ip6stat, or not?
3139  */
3140 static int
3141 bridge_ip6_checkbasic(struct mbuf **mp)
3142 {
3143 	struct mbuf *m = *mp;
3144 	struct ip6_hdr *ip6;
3145 
3146 	/*
3147 	 * If the IPv6 header is not aligned, slurp it up into a new
3148 	 * mbuf with space for link headers, in the event we forward
3149 	 * it.  Otherwise, if it is aligned, make sure the entire base
3150 	 * IPv6 header is in the first mbuf of the chain.
3151 	 */
3152 	if (IP6_HDR_ALIGNED_P(mtod(m, caddr_t)) == 0) {
3153 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3154 		if ((m = m_copyup(m, sizeof(struct ip6_hdr),
3155 			    (max_linkhdr + 3) & ~3)) == NULL) {
3156 			/* XXXJRT new stat, please */
3157 			ip6stat.ip6s_toosmall++;
3158 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3159 			goto bad;
3160 		}
3161 	} else if (__predict_false(m->m_len < sizeof(struct ip6_hdr))) {
3162 		struct ifnet *inifp = m->m_pkthdr.rcvif;
3163 		if ((m = m_pullup(m, sizeof(struct ip6_hdr))) == NULL) {
3164 			ip6stat.ip6s_toosmall++;
3165 			in6_ifstat_inc(inifp, ifs6_in_hdrerr);
3166 			goto bad;
3167 		}
3168 	}
3169 
3170 	ip6 = mtod(m, struct ip6_hdr *);
3171 
3172 	if ((ip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
3173 		ip6stat.ip6s_badvers++;
3174 		in6_ifstat_inc(m->m_pkthdr.rcvif, ifs6_in_hdrerr);
3175 		goto bad;
3176 	}
3177 
3178 	/* Checks out, proceed */
3179 	*mp = m;
3180 	return (0);
3181 
3182 bad:
3183 	*mp = m;
3184 	return (-1);
3185 }
3186 #endif /* INET6 */
3187 
3188 /*
3189  * bridge_fragment:
3190  *
3191  *	Return a fragmented mbuf chain.
3192  */
3193 static int
3194 bridge_fragment(struct ifnet *ifp, struct mbuf *m, struct ether_header *eh,
3195     int snap, struct llc *llc)
3196 {
3197 	struct mbuf *m0;
3198 	struct ip *ip;
3199 	int error = -1;
3200 
3201 	if (m->m_len < sizeof(struct ip) &&
3202 	    (m = m_pullup(m, sizeof(struct ip))) == NULL)
3203 		goto out;
3204 	ip = mtod(m, struct ip *);
3205 
3206 	error = ip_fragment(ip, &m, ifp->if_mtu, ifp->if_hwassist,
3207 		    CSUM_DELAY_IP);
3208 	if (error)
3209 		goto out;
3210 
3211 	/* walk the chain and re-add the Ethernet header */
3212 	for (m0 = m; m0; m0 = m0->m_nextpkt) {
3213 		if (error == 0) {
3214 			if (snap) {
3215 				M_PREPEND(m0, sizeof(struct llc), M_DONTWAIT);
3216 				if (m0 == NULL) {
3217 					error = ENOBUFS;
3218 					continue;
3219 				}
3220 				bcopy(llc, mtod(m0, caddr_t),
3221 				    sizeof(struct llc));
3222 			}
3223 			M_PREPEND(m0, ETHER_HDR_LEN, M_DONTWAIT);
3224 			if (m0 == NULL) {
3225 				error = ENOBUFS;
3226 				continue;
3227 			}
3228 			bcopy(eh, mtod(m0, caddr_t), ETHER_HDR_LEN);
3229 		} else
3230 			m_freem(m);
3231 	}
3232 
3233 	if (error == 0)
3234 		ipstat.ips_fragmented++;
3235 
3236 	return (error);
3237 
3238 out:
3239 	if (m != NULL)
3240 		m_freem(m);
3241 	return (error);
3242 }
3243