xref: /openbsd-src/sys/net/if_bpe.c (revision 4b70baf6e17fc8b27fc1f7fa7929335753fa94c3)
1 /*	$OpenBSD: if_bpe.c,v 1.5 2019/04/23 10:53:45 dlg Exp $ */
2 /*
3  * Copyright (c) 2018 David Gwynne <dlg@openbsd.org>
4  *
5  * Permission to use, copy, modify, and distribute this software for any
6  * purpose with or without fee is hereby granted, provided that the above
7  * copyright notice and this permission notice appear in all copies.
8  *
9  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
10  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
11  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
12  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
13  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
14  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
15  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
16  */
17 
18 #include "bpfilter.h"
19 #include "pf.h"
20 
21 #include <sys/param.h>
22 #include <sys/systm.h>
23 #include <sys/kernel.h>
24 #include <sys/mbuf.h>
25 #include <sys/socket.h>
26 #include <sys/ioctl.h>
27 #include <sys/timeout.h>
28 #include <sys/pool.h>
29 #include <sys/tree.h>
30 
31 #include <net/if.h>
32 #include <net/if_var.h>
33 #include <net/if_dl.h>
34 #include <net/if_media.h>
35 #include <net/if_types.h>
36 #include <net/rtable.h>
37 
38 #include <netinet/in.h>
39 #include <netinet/if_ether.h>
40 
41 /* for bridge stuff */
42 #include <net/if_bridge.h>
43 
44 
45 #if NBPFILTER > 0
46 #include <net/bpf.h>
47 #endif
48 
49 #include <net/if_bpe.h>
50 
51 #define PBB_ITAG_ISID		0x00ffffff
52 #define PBB_ITAG_ISID_MIN	0x00000000
53 #define PBB_ITAG_ISID_MAX	0x00ffffff
54 #define PBB_ITAG_RES2		0x03000000	/* must be zero on input */
55 #define PBB_ITAG_RES1		0x04000000	/* ignore on input */
56 #define PBB_ITAG_UCA		0x08000000
57 #define PBB_ITAG_DEI		0x10000000
58 #define PBB_ITAG_PCP_SHIFT	29
59 #define PBB_ITAG_PCP_MASK	(0x7U << PBB_ITAG_PCP_SHIFT)
60 
61 #define BPE_BRIDGE_AGE_TMO	100 /* seconds */
62 
63 struct bpe_key {
64 	int			k_if;
65 	uint32_t		k_isid;
66 
67 	RBT_ENTRY(bpe_tunnel)	k_entry;
68 };
69 
70 RBT_HEAD(bpe_tree, bpe_key);
71 
72 static inline int bpe_cmp(const struct bpe_key *, const struct bpe_key *);
73 
74 RBT_PROTOTYPE(bpe_tree, bpe_key, k_entry, bpe_cmp);
75 RBT_GENERATE(bpe_tree, bpe_key, k_entry, bpe_cmp);
76 
77 struct bpe_entry {
78 	struct ether_addr	be_c_da; /* customer address - must be first */
79 	struct ether_addr	be_b_da; /* bridge address */
80 	unsigned int		be_type;
81 #define BPE_ENTRY_DYNAMIC		0
82 #define BPE_ENTRY_STATIC		1
83 	struct refcnt		be_refs;
84 	time_t			be_age;
85 
86 	RBT_ENTRY(bpe_entry)	be_entry;
87 };
88 
89 RBT_HEAD(bpe_map, bpe_entry);
90 
91 static inline int bpe_entry_cmp(const struct bpe_entry *,
92     const struct bpe_entry *);
93 
94 RBT_PROTOTYPE(bpe_map, bpe_entry, be_entry, bpe_entry_cmp);
95 RBT_GENERATE(bpe_map, bpe_entry, be_entry, bpe_entry_cmp);
96 
97 struct bpe_softc {
98 	struct bpe_key		sc_key; /* must be first */
99 	struct arpcom		sc_ac;
100 	struct ifmedia		sc_media;
101 	int			sc_txhprio;
102 	int			sc_rxhprio;
103 	uint8_t			sc_group[ETHER_ADDR_LEN];
104 
105 	void *			sc_lh_cookie;
106 	void *			sc_dh_cookie;
107 
108 	struct bpe_map		sc_bridge_map;
109 	struct rwlock		sc_bridge_lock;
110 	unsigned int		sc_bridge_num;
111 	unsigned int		sc_bridge_max;
112 	int			sc_bridge_tmo; /* seconds */
113 	struct timeout		sc_bridge_age;
114 };
115 
116 void		bpeattach(int);
117 
118 static int	bpe_clone_create(struct if_clone *, int);
119 static int	bpe_clone_destroy(struct ifnet *);
120 
121 static void	bpe_start(struct ifnet *);
122 static int	bpe_ioctl(struct ifnet *, u_long, caddr_t);
123 static int	bpe_media_get(struct bpe_softc *, struct ifreq *);
124 static int	bpe_up(struct bpe_softc *);
125 static int	bpe_down(struct bpe_softc *);
126 static int	bpe_multi(struct bpe_softc *, struct ifnet *, u_long);
127 static int	bpe_set_vnetid(struct bpe_softc *, const struct ifreq *);
128 static void	bpe_set_group(struct bpe_softc *, uint32_t);
129 static int	bpe_set_parent(struct bpe_softc *, const struct if_parent *);
130 static int	bpe_get_parent(struct bpe_softc *, struct if_parent *);
131 static int	bpe_del_parent(struct bpe_softc *);
132 static void	bpe_link_hook(void *);
133 static void	bpe_link_state(struct bpe_softc *, u_char, uint64_t);
134 static void	bpe_detach_hook(void *);
135 
136 static void	bpe_input_map(struct bpe_softc *,
137 		    const uint8_t *, const uint8_t *);
138 static void	bpe_bridge_age(void *);
139 
140 static struct if_clone bpe_cloner =
141     IF_CLONE_INITIALIZER("bpe", bpe_clone_create, bpe_clone_destroy);
142 
143 static struct bpe_tree bpe_interfaces = RBT_INITIALIZER();
144 static struct rwlock bpe_lock = RWLOCK_INITIALIZER("bpeifs");
145 static struct pool bpe_entry_pool;
146 
147 #define ether_cmp(_a, _b)	memcmp((_a), (_b), ETHER_ADDR_LEN)
148 #define ether_is_eq(_a, _b)	(ether_cmp((_a), (_b)) == 0)
149 #define ether_is_bcast(_a)	ether_is_eq((_a), etherbroadcastaddr)
150 
151 void
152 bpeattach(int count)
153 {
154 	if_clone_attach(&bpe_cloner);
155 }
156 
157 static int
158 bpe_clone_create(struct if_clone *ifc, int unit)
159 {
160 	struct bpe_softc *sc;
161 	struct ifnet *ifp;
162 
163 	if (bpe_entry_pool.pr_size == 0) {
164 		pool_init(&bpe_entry_pool, sizeof(struct bpe_entry), 0,
165 		    IPL_NONE, 0, "bpepl", NULL);
166 	}
167 
168 	sc = malloc(sizeof(*sc), M_DEVBUF, M_WAITOK|M_ZERO);
169 	ifp = &sc->sc_ac.ac_if;
170 
171 	snprintf(ifp->if_xname, sizeof(ifp->if_xname), "%s%d",
172 	    ifc->ifc_name, unit);
173 
174 	sc->sc_key.k_if = 0;
175 	sc->sc_key.k_isid = 0;
176 	bpe_set_group(sc, 0);
177 
178 	sc->sc_txhprio = IF_HDRPRIO_PACKET;
179 	sc->sc_txhprio = IF_HDRPRIO_OUTER;
180 
181 	rw_init(&sc->sc_bridge_lock, "bpebr");
182 	RBT_INIT(bpe_map, &sc->sc_bridge_map);
183 	sc->sc_bridge_num = 0;
184 	sc->sc_bridge_max = 100; /* XXX */
185 	sc->sc_bridge_tmo = 240;
186 	timeout_set_proc(&sc->sc_bridge_age, bpe_bridge_age, sc);
187 
188 	ifp->if_softc = sc;
189 	ifp->if_hardmtu = ETHER_MAX_HARDMTU_LEN;
190 	ifp->if_ioctl = bpe_ioctl;
191 	ifp->if_start = bpe_start;
192 	ifp->if_flags = IFF_BROADCAST | IFF_MULTICAST;
193 	ifp->if_xflags = IFXF_CLONED;
194 	IFQ_SET_MAXLEN(&ifp->if_snd, IFQ_MAXLEN);
195 	ether_fakeaddr(ifp);
196 
197 	if_counters_alloc(ifp);
198 	if_attach(ifp);
199 	ether_ifattach(ifp);
200 
201 	return (0);
202 }
203 
204 static int
205 bpe_clone_destroy(struct ifnet *ifp)
206 {
207 	struct bpe_softc *sc = ifp->if_softc;
208 
209 	NET_LOCK();
210 	if (ISSET(ifp->if_flags, IFF_RUNNING))
211 		bpe_down(sc);
212 	NET_UNLOCK();
213 
214 	ifmedia_delete_instance(&sc->sc_media, IFM_INST_ANY);
215 	ether_ifdetach(ifp);
216 	if_detach(ifp);
217 
218 	free(sc, M_DEVBUF, sizeof(*sc));
219 
220 	return (0);
221 }
222 
223 static inline int
224 bpe_entry_valid(struct bpe_softc *sc, const struct bpe_entry *be)
225 {
226 	time_t diff;
227 
228 	if (be == NULL)
229 		return (0);
230 
231 	if (be->be_type == BPE_ENTRY_STATIC)
232 		return (1);
233 
234 	diff = time_uptime - be->be_age;
235 	if (diff < sc->sc_bridge_tmo)
236 		return (1);
237 
238 	return (0);
239 }
240 
241 static void
242 bpe_start(struct ifnet *ifp)
243 {
244 	struct bpe_softc *sc = ifp->if_softc;
245 	struct ifnet *ifp0;
246 	struct mbuf *m0, *m;
247 	struct ether_header *ceh;
248 	struct ether_header *beh;
249 	uint32_t itag, *itagp;
250 	int hlen = sizeof(*beh) + sizeof(*itagp);
251 #if NBPFILTER > 0
252 	caddr_t if_bpf;
253 #endif
254 	int txprio;
255 	uint8_t prio;
256 
257 	ifp0 = if_get(sc->sc_key.k_if);
258 	if (ifp0 == NULL || !ISSET(ifp0->if_flags, IFF_RUNNING)) {
259 		ifq_purge(&ifp->if_snd);
260 		goto done;
261 	}
262 
263 	txprio = sc->sc_txhprio;
264 
265 	while ((m0 = ifq_dequeue(&ifp->if_snd)) != NULL) {
266 #if NBPFILTER > 0
267 		if_bpf = ifp->if_bpf;
268 		if (if_bpf)
269 			bpf_mtap_ether(if_bpf, m0, BPF_DIRECTION_OUT);
270 #endif
271 
272 		ceh = mtod(m0, struct ether_header *);
273 
274 		/* force prepend of a whole mbuf because of alignment */
275 		m = m_get(M_DONTWAIT, m0->m_type);
276 		if (m == NULL) {
277 			m_freem(m0);
278 			continue;
279 		}
280 
281 		M_MOVE_PKTHDR(m, m0);
282 		m->m_next = m0;
283 
284 		m_align(m, 0);
285 		m->m_len = 0;
286 
287 		m = m_prepend(m, hlen, M_DONTWAIT);
288 		if (m == NULL)
289 			continue;
290 
291 		beh = mtod(m, struct ether_header *);
292 
293 		if (ether_is_bcast(ceh->ether_dhost)) {
294 			memcpy(beh->ether_dhost, sc->sc_group,
295 			    sizeof(beh->ether_dhost));
296 		} else {
297 			struct bpe_entry *be;
298 
299 			rw_enter_read(&sc->sc_bridge_lock);
300 			be = RBT_FIND(bpe_map, &sc->sc_bridge_map,
301 			    (struct bpe_entry *)ceh->ether_dhost);
302 			if (bpe_entry_valid(sc, be)) {
303 				memcpy(beh->ether_dhost, &be->be_b_da,
304 				    sizeof(beh->ether_dhost));
305 			} else {
306 				/* "flood" to unknown hosts */
307 				memcpy(beh->ether_dhost, sc->sc_group,
308 				    sizeof(beh->ether_dhost));
309 			}
310 			rw_exit_read(&sc->sc_bridge_lock);
311 		}
312 
313 		memcpy(beh->ether_shost, ((struct arpcom *)ifp0)->ac_enaddr,
314 		    sizeof(beh->ether_shost));
315 		beh->ether_type = htons(ETHERTYPE_PBB);
316 
317 		prio = (txprio == IF_HDRPRIO_PACKET) ?
318 		    m->m_pkthdr.pf.prio : txprio;
319 
320 		itag = sc->sc_key.k_isid;
321 		itag |= prio << PBB_ITAG_PCP_SHIFT;
322 		itagp = (uint32_t *)(beh + 1);
323 
324 		htobem32(itagp, itag);
325 
326 		if_enqueue(ifp0, m);
327 	}
328 
329 done:
330 	if_put(ifp0);
331 }
332 
333 static void
334 bpe_bridge_age(void *arg)
335 {
336 	struct bpe_softc *sc = arg;
337 	struct bpe_entry *be, *nbe;
338 	time_t diff;
339 
340 	timeout_add_sec(&sc->sc_bridge_age, BPE_BRIDGE_AGE_TMO);
341 
342 	rw_enter_write(&sc->sc_bridge_lock);
343 	RBT_FOREACH_SAFE(be, bpe_map, &sc->sc_bridge_map, nbe) {
344 		if (be->be_type != BPE_ENTRY_DYNAMIC)
345 			continue;
346 
347 		diff = time_uptime - be->be_age;
348 		if (diff < sc->sc_bridge_tmo)
349 			continue;
350 
351 		sc->sc_bridge_num--;
352 		RBT_REMOVE(bpe_map, &sc->sc_bridge_map, be);
353 		if (refcnt_rele(&be->be_refs))
354 			pool_put(&bpe_entry_pool, be);
355 	}
356 	rw_exit_write(&sc->sc_bridge_lock);
357 }
358 
359 static int
360 bpe_rtfind(struct bpe_softc *sc, struct ifbaconf *baconf)
361 {
362 	struct ifnet *ifp = &sc->sc_ac.ac_if;
363 	struct bpe_entry *be;
364 	struct ifbareq bareq;
365 	caddr_t uaddr, end;
366 	int error;
367 	time_t age;
368 	struct sockaddr_dl *sdl;
369 
370 	if (baconf->ifbac_len == 0) {
371 		/* single read is atomic */
372 		baconf->ifbac_len = sc->sc_bridge_num * sizeof(bareq);
373 		return (0);
374 	}
375 
376 	uaddr = baconf->ifbac_buf;
377 	end = uaddr + baconf->ifbac_len;
378 
379 	rw_enter_read(&sc->sc_bridge_lock);
380 	RBT_FOREACH(be, bpe_map, &sc->sc_bridge_map) {
381 		if (uaddr >= end)
382 			break;
383 
384 		memcpy(bareq.ifba_name, ifp->if_xname,
385 		    sizeof(bareq.ifba_name));
386 		memcpy(bareq.ifba_ifsname, ifp->if_xname,
387 		    sizeof(bareq.ifba_ifsname));
388 		memcpy(&bareq.ifba_dst, &be->be_c_da,
389 		    sizeof(bareq.ifba_dst));
390 
391 		memset(&bareq.ifba_dstsa, 0, sizeof(bareq.ifba_dstsa));
392 
393 		bzero(&bareq.ifba_dstsa, sizeof(bareq.ifba_dstsa));
394 		sdl = (struct sockaddr_dl *)&bareq.ifba_dstsa;
395 		sdl->sdl_len = sizeof(sdl);
396 		sdl->sdl_family = AF_LINK;
397 		sdl->sdl_index = 0;
398 		sdl->sdl_type = IFT_ETHER;
399 		sdl->sdl_nlen = 0;
400 		sdl->sdl_alen = sizeof(be->be_b_da);
401 		CTASSERT(sizeof(sdl->sdl_data) >= sizeof(be->be_b_da));
402 		memcpy(sdl->sdl_data, &be->be_b_da, sizeof(be->be_b_da));
403 
404 		switch (be->be_type) {
405 		case BPE_ENTRY_DYNAMIC:
406 			age = time_uptime - be->be_age;
407 			bareq.ifba_age = MIN(age, 0xff);
408 			bareq.ifba_flags = IFBAF_DYNAMIC;
409 			break;
410 		case BPE_ENTRY_STATIC:
411 			bareq.ifba_age = 0;
412 			bareq.ifba_flags = IFBAF_STATIC;
413 			break;
414 		}
415 
416 		error = copyout(&bareq, uaddr, sizeof(bareq));
417 		if (error != 0) {
418 			rw_exit_read(&sc->sc_bridge_lock);
419 			return (error);
420 		}
421 
422 		uaddr += sizeof(bareq);
423 	}
424 	baconf->ifbac_len = sc->sc_bridge_num * sizeof(bareq);
425 	rw_exit_read(&sc->sc_bridge_lock);
426 
427 	return (0);
428 }
429 
430 static void
431 bpe_flush_map(struct bpe_softc *sc, uint32_t flags)
432 {
433 	struct bpe_entry *be, *nbe;
434 
435 	rw_enter_write(&sc->sc_bridge_lock);
436 	RBT_FOREACH_SAFE(be, bpe_map, &sc->sc_bridge_map, nbe) {
437 		if (flags == IFBF_FLUSHDYN &&
438 		    be->be_type != BPE_ENTRY_DYNAMIC)
439 			continue;
440 
441 		RBT_REMOVE(bpe_map, &sc->sc_bridge_map, be);
442 		if (refcnt_rele(&be->be_refs))
443 			pool_put(&bpe_entry_pool, be);
444 	}
445 	rw_exit_write(&sc->sc_bridge_lock);
446 }
447 
448 static int
449 bpe_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
450 {
451 	struct bpe_softc *sc = ifp->if_softc;
452 	struct ifreq *ifr = (struct ifreq *)data;
453 	struct ifbrparam *bparam = (struct ifbrparam *)data;
454 	int error = 0;
455 
456 	switch (cmd) {
457 	case SIOCSIFFLAGS:
458 		if (ISSET(ifp->if_flags, IFF_UP)) {
459 			if (!ISSET(ifp->if_flags, IFF_RUNNING))
460 				error = bpe_up(sc);
461 			else
462 				error = 0;
463 		} else {
464 			if (ISSET(ifp->if_flags, IFF_RUNNING))
465 				error = bpe_down(sc);
466 		}
467 		break;
468 
469 	case SIOCSVNETID:
470 		error = bpe_set_vnetid(sc, ifr);
471 	case SIOCGVNETID:
472 		ifr->ifr_vnetid = sc->sc_key.k_isid;
473 		break;
474 
475 	case SIOCSIFPARENT:
476 		error = bpe_set_parent(sc, (struct if_parent *)data);
477 		break;
478 	case SIOCGIFPARENT:
479 		error = bpe_get_parent(sc, (struct if_parent *)data);
480 		break;
481 	case SIOCDIFPARENT:
482 		error = bpe_del_parent(sc);
483 		break;
484 
485 	case SIOCSTXHPRIO:
486 		error = if_txhprio_l2_check(ifr->ifr_hdrprio);
487 		if (error != 0)
488 			break;
489 
490 		sc->sc_txhprio = ifr->ifr_hdrprio;
491 		break;
492 	case SIOCGTXHPRIO:
493 		ifr->ifr_hdrprio = sc->sc_txhprio;
494 		break;
495 
496 	case SIOCSRXHPRIO:
497 		error = if_rxhprio_l2_check(ifr->ifr_hdrprio);
498 		if (error != 0)
499 			break;
500 
501 		sc->sc_rxhprio = ifr->ifr_hdrprio;
502 		break;
503 	case SIOCGRXHPRIO:
504 		ifr->ifr_hdrprio = sc->sc_rxhprio;
505 		break;
506 
507 	case SIOCGIFMEDIA:
508 		error = bpe_media_get(sc, ifr);
509 		break;
510 
511 	case SIOCBRDGSCACHE:
512 		error = suser(curproc);
513 		if (error != 0)
514 			break;
515 
516 		if (bparam->ifbrp_csize < 1) {
517 			error = EINVAL;
518 			break;
519 		}
520 
521 		/* commit */
522 		sc->sc_bridge_max = bparam->ifbrp_csize;
523 		break;
524 	case SIOCBRDGGCACHE:
525 		bparam->ifbrp_csize = sc->sc_bridge_max;
526 		break;
527 
528 	case SIOCBRDGSTO:
529 		error = suser(curproc);
530 		if (error != 0)
531 			break;
532 
533 		if (bparam->ifbrp_ctime < 8 ||
534 		    bparam->ifbrp_ctime > 3600) {
535 			error = EINVAL;
536 			break;
537 		}
538 		sc->sc_bridge_tmo = bparam->ifbrp_ctime;
539 		break;
540 	case SIOCBRDGGTO:
541 		bparam->ifbrp_ctime = sc->sc_bridge_tmo;
542 		break;
543 
544 	case SIOCBRDGRTS:
545 		error = bpe_rtfind(sc, (struct ifbaconf *)data);
546 		break;
547 	case SIOCBRDGFLUSH:
548 		error = suser(curproc);
549 		if (error != 0)
550 			break;
551 
552 		bpe_flush_map(sc,
553 		    ((struct ifbreq *)data)->ifbr_ifsflags);
554 		break;
555 
556 	default:
557 		error = ether_ioctl(ifp, &sc->sc_ac, cmd, data);
558 		break;
559 	}
560 
561 	return (error);
562 }
563 
564 static int
565 bpe_media_get(struct bpe_softc *sc, struct ifreq *ifr)
566 {
567 	struct ifnet *ifp0;
568 	int error;
569 
570 	ifp0 = if_get(sc->sc_key.k_if);
571 	if (ifp0 != NULL)
572 		error = (*ifp0->if_ioctl)(ifp0, SIOCGIFMEDIA, (caddr_t)ifr);
573 	else
574 		error = ENOTTY;
575 	if_put(ifp0);
576 
577 	return (error);
578 }
579 
580 static int
581 bpe_up(struct bpe_softc *sc)
582 {
583 	struct ifnet *ifp = &sc->sc_ac.ac_if;
584 	struct ifnet *ifp0;
585 	struct bpe_softc *osc;
586 	int error = 0;
587 	u_int hardmtu;
588 	u_int hlen = sizeof(struct ether_header) + sizeof(uint32_t);
589 
590 	KASSERT(!ISSET(ifp->if_flags, IFF_RUNNING));
591 	NET_ASSERT_LOCKED();
592 
593 	ifp0 = if_get(sc->sc_key.k_if);
594 	if (ifp0 == NULL)
595 		return (ENXIO);
596 
597 	/* check again if bpe will work on top of the parent */
598 	if (ifp0->if_type != IFT_ETHER) {
599 		error = EPROTONOSUPPORT;
600 		goto put;
601 	}
602 
603 	hardmtu = ifp0->if_hardmtu;
604 	if (hardmtu < hlen) {
605 		error = ENOBUFS;
606 		goto put;
607 	}
608 	hardmtu -= hlen;
609 	if (ifp->if_mtu > hardmtu) {
610 		error = ENOBUFS;
611 		goto put;
612 	}
613 
614 	/* parent is fine, let's prepare the bpe to handle packets */
615 	ifp->if_hardmtu = hardmtu;
616 	SET(ifp->if_flags, ifp0->if_flags & IFF_SIMPLEX);
617 
618 	/* commit the interface */
619 	error = rw_enter(&bpe_lock, RW_WRITE | RW_INTR);
620 	if (error != 0)
621 		goto scrub;
622 
623 	osc = (struct bpe_softc *)RBT_INSERT(bpe_tree, &bpe_interfaces,
624 	    (struct bpe_key *)sc);
625 	rw_exit(&bpe_lock);
626 
627 	if (osc != NULL) {
628 		error = EADDRINUSE;
629 		goto scrub;
630 	}
631 
632 	if (bpe_multi(sc, ifp0, SIOCADDMULTI) != 0) {
633 		error = ENOTCONN;
634 		goto remove;
635 	}
636 
637 	/* Register callback for physical link state changes */
638 	sc->sc_lh_cookie = hook_establish(ifp0->if_linkstatehooks, 1,
639 	    bpe_link_hook, sc);
640 
641 	/* Register callback if parent wants to unregister */
642 	sc->sc_dh_cookie = hook_establish(ifp0->if_detachhooks, 0,
643 	    bpe_detach_hook, sc);
644 
645 	/* we're running now */
646 	SET(ifp->if_flags, IFF_RUNNING);
647 	bpe_link_state(sc, ifp0->if_link_state, ifp0->if_baudrate);
648 
649 	if_put(ifp0);
650 
651 	timeout_add_sec(&sc->sc_bridge_age, BPE_BRIDGE_AGE_TMO);
652 
653 	return (0);
654 
655 remove:
656 	rw_enter(&bpe_lock, RW_WRITE);
657 	RBT_REMOVE(bpe_tree, &bpe_interfaces, (struct bpe_key *)sc);
658 	rw_exit(&bpe_lock);
659 scrub:
660 	CLR(ifp->if_flags, IFF_SIMPLEX);
661 	ifp->if_hardmtu = 0xffff;
662 put:
663 	if_put(ifp0);
664 
665 	return (error);
666 }
667 
668 static int
669 bpe_down(struct bpe_softc *sc)
670 {
671 	struct ifnet *ifp = &sc->sc_ac.ac_if;
672 	struct ifnet *ifp0;
673 
674 	NET_ASSERT_LOCKED();
675 
676 	CLR(ifp->if_flags, IFF_RUNNING);
677 
678 	ifp0 = if_get(sc->sc_key.k_if);
679 	if (ifp0 != NULL) {
680 		hook_disestablish(ifp0->if_detachhooks, sc->sc_dh_cookie);
681 		hook_disestablish(ifp0->if_linkstatehooks, sc->sc_lh_cookie);
682 		bpe_multi(sc, ifp0, SIOCDELMULTI);
683 	}
684 	if_put(ifp0);
685 
686 	rw_enter(&bpe_lock, RW_WRITE);
687 	RBT_REMOVE(bpe_tree, &bpe_interfaces, (struct bpe_key *)sc);
688 	rw_exit(&bpe_lock);
689 
690 	CLR(ifp->if_flags, IFF_SIMPLEX);
691 	ifp->if_hardmtu = 0xffff;
692 
693 	return (0);
694 }
695 
696 static int
697 bpe_multi(struct bpe_softc *sc, struct ifnet *ifp0, u_long cmd)
698 {
699 	struct ifreq ifr;
700 	struct sockaddr *sa;
701 
702 	/* make it convincing */
703 	CTASSERT(sizeof(ifr.ifr_name) == sizeof(ifp0->if_xname));
704 	memcpy(ifr.ifr_name, ifp0->if_xname, sizeof(ifr.ifr_name));
705 
706 	sa = &ifr.ifr_addr;
707 	CTASSERT(sizeof(sa->sa_data) >= sizeof(sc->sc_group));
708 
709 	sa->sa_family = AF_UNSPEC;
710 	memcpy(sa->sa_data, sc->sc_group, sizeof(sc->sc_group));
711 
712 	return ((*ifp0->if_ioctl)(ifp0, cmd, (caddr_t)&ifr));
713 }
714 
715 static void
716 bpe_set_group(struct bpe_softc *sc, uint32_t isid)
717 {
718 	uint8_t *group = sc->sc_group;
719 
720 	group[0] = 0x01;
721 	group[1] = 0x1e;
722 	group[2] = 0x83;
723 	group[3] = isid >> 16;
724 	group[4] = isid >> 8;
725 	group[5] = isid >> 0;
726 }
727 
728 static int
729 bpe_set_vnetid(struct bpe_softc *sc, const struct ifreq *ifr)
730 {
731 	struct ifnet *ifp = &sc->sc_ac.ac_if;
732 	uint32_t isid;
733 
734 	if (ifr->ifr_vnetid < PBB_ITAG_ISID_MIN ||
735 	    ifr->ifr_vnetid > PBB_ITAG_ISID_MAX)
736 		return (EINVAL);
737 
738 	isid = ifr->ifr_vnetid;
739 	if (isid == sc->sc_key.k_isid)
740 		return (0);
741 
742 	if (ISSET(ifp->if_flags, IFF_RUNNING))
743 		return (EBUSY);
744 
745 	/* commit */
746 	sc->sc_key.k_isid = isid;
747 	bpe_set_group(sc, isid);
748 	bpe_flush_map(sc, IFBF_FLUSHALL);
749 
750 	return (0);
751 }
752 
753 static int
754 bpe_set_parent(struct bpe_softc *sc, const struct if_parent *p)
755 {
756 	struct ifnet *ifp = &sc->sc_ac.ac_if;
757 	struct ifnet *ifp0;
758 
759 	ifp0 = ifunit(p->ifp_parent); /* doesn't need an if_put */
760 	if (ifp0 == NULL)
761 		return (ENXIO);
762 
763 	if (ifp0->if_type != IFT_ETHER)
764 		return (ENXIO);
765 
766 	if (ifp0->if_index == sc->sc_key.k_if)
767 		return (0);
768 
769 	if (ISSET(ifp->if_flags, IFF_RUNNING))
770 		return (EBUSY);
771 
772 	/* commit */
773 	sc->sc_key.k_if = ifp0->if_index;
774 	bpe_flush_map(sc, IFBF_FLUSHALL);
775 
776 	return (0);
777 }
778 
779 static int
780 bpe_get_parent(struct bpe_softc *sc, struct if_parent *p)
781 {
782 	struct ifnet *ifp0;
783 	int error = 0;
784 
785 	ifp0 = if_get(sc->sc_key.k_if);
786 	if (ifp0 == NULL)
787 		error = EADDRNOTAVAIL;
788 	else
789 		memcpy(p->ifp_parent, ifp0->if_xname, sizeof(p->ifp_parent));
790 	if_put(ifp0);
791 
792 	return (error);
793 }
794 
795 static int
796 bpe_del_parent(struct bpe_softc *sc)
797 {
798 	struct ifnet *ifp = &sc->sc_ac.ac_if;
799 
800 	if (ISSET(ifp->if_flags, IFF_RUNNING))
801 		return (EBUSY);
802 
803 	/* commit */
804 	sc->sc_key.k_if = 0;
805 	bpe_flush_map(sc, IFBF_FLUSHALL);
806 
807 	return (0);
808 }
809 
810 static inline struct bpe_softc *
811 bpe_find(struct ifnet *ifp0, uint32_t isid)
812 {
813 	struct bpe_key k = { .k_if = ifp0->if_index, .k_isid = isid };
814 	struct bpe_softc *sc;
815 
816 	rw_enter_read(&bpe_lock);
817 	sc = (struct bpe_softc *)RBT_FIND(bpe_tree, &bpe_interfaces, &k);
818 	rw_exit_read(&bpe_lock);
819 
820 	return (sc);
821 }
822 
823 static void
824 bpe_input_map(struct bpe_softc *sc, const uint8_t *ba, const uint8_t *ca)
825 {
826 	struct bpe_entry *be;
827 	int new = 0;
828 
829 	if (ETHER_IS_MULTICAST(ca))
830 		return;
831 
832 	/* remember where it came from */
833 	rw_enter_read(&sc->sc_bridge_lock);
834 	be = RBT_FIND(bpe_map, &sc->sc_bridge_map, (struct bpe_entry *)ca);
835 	if (be == NULL)
836 		new = 1;
837 	else {
838 		be->be_age = time_uptime; /* only a little bit racy */
839 
840 		if (be->be_type != BPE_ENTRY_DYNAMIC ||
841 		    ether_is_eq(ba, &be->be_b_da))
842 			be = NULL;
843 		else
844 			refcnt_take(&be->be_refs);
845 	}
846 	rw_exit_read(&sc->sc_bridge_lock);
847 
848 	if (new) {
849 		struct bpe_entry *obe;
850 		unsigned int num;
851 
852 		be = pool_get(&bpe_entry_pool, PR_NOWAIT);
853 		if (be == NULL) {
854 			/* oh well */
855 			return;
856 		}
857 
858 		memcpy(&be->be_c_da, ca, sizeof(be->be_c_da));
859 		memcpy(&be->be_b_da, ba, sizeof(be->be_b_da));
860 		be->be_type = BPE_ENTRY_DYNAMIC;
861 		refcnt_init(&be->be_refs);
862 		be->be_age = time_uptime;
863 
864 		rw_enter_write(&sc->sc_bridge_lock);
865 		num = sc->sc_bridge_num;
866 		if (++num > sc->sc_bridge_max)
867 			obe = be;
868 		else {
869 			/* try and give the ref to the map */
870 			obe = RBT_INSERT(bpe_map, &sc->sc_bridge_map, be);
871 			if (obe == NULL) {
872 				/* count the insert */
873 				sc->sc_bridge_num = num;
874 			}
875 		}
876 		rw_exit_write(&sc->sc_bridge_lock);
877 
878 		if (obe != NULL)
879 			pool_put(&bpe_entry_pool, obe);
880 	} else if (be != NULL) {
881 		rw_enter_write(&sc->sc_bridge_lock);
882 		memcpy(&be->be_b_da, ba, sizeof(be->be_b_da));
883 		rw_exit_write(&sc->sc_bridge_lock);
884 
885 		if (refcnt_rele(&be->be_refs)) {
886 			/* ioctl may have deleted the entry */
887 			pool_put(&bpe_entry_pool, be);
888 		}
889 	}
890 }
891 
892 void
893 bpe_input(struct ifnet *ifp0, struct mbuf *m)
894 {
895 	struct bpe_softc *sc;
896 	struct ifnet *ifp;
897 	struct ether_header *beh, *ceh;
898 	uint32_t *itagp, itag;
899 	unsigned int hlen = sizeof(*beh) + sizeof(*itagp) + sizeof(*ceh);
900 	struct mbuf *n;
901 	int off;
902 	int prio;
903 
904 	if (m->m_len < hlen) {
905 		m = m_pullup(m, hlen);
906 		if (m == NULL) {
907 			/* pbb short ++ */
908 			return;
909 		}
910 	}
911 
912 	beh = mtod(m, struct ether_header *);
913 	itagp = (uint32_t *)(beh + 1);
914 	itag = bemtoh32(itagp);
915 
916 	if (itag & PBB_ITAG_RES2) {
917 		/* dropped by res2 ++ */
918 		goto drop;
919 	}
920 
921 	sc = bpe_find(ifp0, itag & PBB_ITAG_ISID);
922 	if (sc == NULL) {
923 		/* no interface found */
924 		goto drop;
925 	}
926 
927 	ceh = (struct ether_header *)(itagp + 1);
928 
929 	bpe_input_map(sc, beh->ether_shost, ceh->ether_shost);
930 
931 	m_adj(m, sizeof(*beh) + sizeof(*itagp));
932 
933 	n = m_getptr(m, sizeof(*ceh), &off);
934 	if (n == NULL) {
935 		/* no data ++ */
936 		goto drop;
937 	}
938 
939 	if (!ALIGNED_POINTER(mtod(n, caddr_t) + off, uint32_t)) {
940 		/* unaligned ++ */
941 		n = m_dup_pkt(m, ETHER_ALIGN, M_NOWAIT);
942 		m_freem(m);
943 		if (n == NULL)
944 			return;
945 
946 		m = n;
947 	}
948 
949 	ifp = &sc->sc_ac.ac_if;
950 
951 	prio = sc->sc_rxhprio;
952 	switch (prio) {
953 	case IF_HDRPRIO_PACKET:
954 		break;
955 	case IF_HDRPRIO_OUTER:
956 		m->m_pkthdr.pf.prio = (itag & PBB_ITAG_PCP_MASK) >>
957 		    PBB_ITAG_PCP_SHIFT;
958 		break;
959 	default:
960 		m->m_pkthdr.pf.prio = prio;
961 		break;
962 	}
963 
964 	m->m_flags &= ~(M_BCAST|M_MCAST);
965 	m->m_pkthdr.ph_ifidx = ifp->if_index;
966 	m->m_pkthdr.ph_rtableid = ifp->if_rdomain;
967 
968 #if NPF > 0
969 	pf_pkt_addr_changed(m);
970 #endif
971 
972 	if_vinput(ifp, m);
973 	return;
974 
975 drop:
976 	m_freem(m);
977 }
978 
979 void
980 bpe_detach_hook(void *arg)
981 {
982 	struct bpe_softc *sc = arg;
983 	struct ifnet *ifp = &sc->sc_ac.ac_if;
984 
985 	if (ISSET(ifp->if_flags, IFF_RUNNING)) {
986 		bpe_down(sc);
987 		CLR(ifp->if_flags, IFF_UP);
988 	}
989 
990 	sc->sc_key.k_if = 0;
991 }
992 
993 static void
994 bpe_link_hook(void *arg)
995 {
996 	struct bpe_softc *sc = arg;
997 	struct ifnet *ifp0;
998 	u_char link = LINK_STATE_DOWN;
999 	uint64_t baud = 0;
1000 
1001 	ifp0 = if_get(sc->sc_key.k_if);
1002 	if (ifp0 != NULL) {
1003 		link = ifp0->if_link_state;
1004 		baud = ifp0->if_baudrate;
1005 	}
1006 	if_put(ifp0);
1007 
1008 	bpe_link_state(sc, link, baud);
1009 }
1010 
1011 void
1012 bpe_link_state(struct bpe_softc *sc, u_char link, uint64_t baud)
1013 {
1014 	struct ifnet *ifp = &sc->sc_ac.ac_if;
1015 
1016 	if (ifp->if_link_state == link)
1017 		return;
1018 
1019 	ifp->if_link_state = link;
1020 	ifp->if_baudrate = baud;
1021 
1022 	if_link_state_change(ifp);
1023 }
1024 
1025 static inline int
1026 bpe_cmp(const struct bpe_key *a, const struct bpe_key *b)
1027 {
1028 	if (a->k_if > b->k_if)
1029 		return (1);
1030 	if (a->k_if < b->k_if)
1031 		return (-1);
1032 	if (a->k_isid > b->k_isid)
1033 		return (1);
1034 	if (a->k_isid < b->k_isid)
1035 		return (-1);
1036 
1037 	return (0);
1038 }
1039 
1040 static inline int
1041 bpe_entry_cmp(const struct bpe_entry *a, const struct bpe_entry *b)
1042 {
1043 	return memcmp(&a->be_c_da, &b->be_c_da, sizeof(a->be_c_da));
1044 }
1045