xref: /openbsd-src/sys/net/if_trunk.c (revision 43003dfe3ad45d1698bed8a37f2b0f5b14f20d4f)
1 /*	$OpenBSD: if_trunk.c,v 1.69 2009/09/17 13:13:56 claudio Exp $	*/
2 
3 /*
4  * Copyright (c) 2005, 2006, 2007 Reyk Floeter <reyk@openbsd.org>
5  *
6  * Permission to use, copy, modify, and distribute this software for any
7  * purpose with or without fee is hereby granted, provided that the above
8  * copyright notice and this permission notice appear in all copies.
9  *
10  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
11  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
12  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
13  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
14  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
15  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
16  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
17  */
18 
19 #include "bpfilter.h"
20 #include "trunk.h"
21 
22 #include <sys/param.h>
23 #include <sys/kernel.h>
24 #include <sys/malloc.h>
25 #include <sys/mbuf.h>
26 #include <sys/queue.h>
27 #include <sys/socket.h>
28 #include <sys/sockio.h>
29 #include <sys/sysctl.h>
30 #include <sys/systm.h>
31 #include <sys/proc.h>
32 #include <sys/hash.h>
33 
34 #include <dev/rndvar.h>
35 
36 #include <net/if.h>
37 #include <net/if_arp.h>
38 #include <net/if_dl.h>
39 #include <net/if_llc.h>
40 #include <net/if_media.h>
41 #include <net/if_types.h>
42 #if NBPFILTER > 0
43 #include <net/bpf.h>
44 #endif
45 
46 #ifdef INET
47 #include <netinet/in.h>
48 #include <netinet/in_systm.h>
49 #include <netinet/if_ether.h>
50 #include <netinet/ip.h>
51 #endif
52 
53 #ifdef INET6
54 #include <netinet/ip6.h>
55 #endif
56 
57 #include <net/if_vlan_var.h>
58 #include <net/if_trunk.h>
59 #include <net/trunklacp.h>
60 
61 
62 SLIST_HEAD(__trhead, trunk_softc) trunk_list;	/* list of trunks */
63 
64 extern struct ifaddr **ifnet_addrs;
65 
66 void	 trunkattach(int);
67 int	 trunk_clone_create(struct if_clone *, int);
68 int	 trunk_clone_destroy(struct ifnet *);
69 void	 trunk_lladdr(struct arpcom *, u_int8_t *);
70 int	 trunk_capabilities(struct trunk_softc *);
71 void	 trunk_port_lladdr(struct trunk_port *, u_int8_t *);
72 int	 trunk_port_create(struct trunk_softc *, struct ifnet *);
73 int	 trunk_port_destroy(struct trunk_port *);
74 void	 trunk_port_watchdog(struct ifnet *);
75 void	 trunk_port_state(void *);
76 int	 trunk_port_ioctl(struct ifnet *, u_long, caddr_t);
77 struct trunk_port *trunk_port_get(struct trunk_softc *, struct ifnet *);
78 int	 trunk_port_checkstacking(struct trunk_softc *);
79 void	 trunk_port2req(struct trunk_port *, struct trunk_reqport *);
80 int	 trunk_ioctl(struct ifnet *, u_long, caddr_t);
81 int	 trunk_ether_addmulti(struct trunk_softc *, struct ifreq *);
82 int	 trunk_ether_delmulti(struct trunk_softc *, struct ifreq *);
83 void	 trunk_ether_purgemulti(struct trunk_softc *);
84 int	 trunk_ether_cmdmulti(struct trunk_port *, u_long);
85 int	 trunk_ioctl_allports(struct trunk_softc *, u_long, caddr_t);
86 void	 trunk_start(struct ifnet *);
87 void	 trunk_init(struct ifnet *);
88 void	 trunk_stop(struct ifnet *);
89 void	 trunk_watchdog(struct ifnet *);
90 int	 trunk_media_change(struct ifnet *);
91 void	 trunk_media_status(struct ifnet *, struct ifmediareq *);
92 struct trunk_port *trunk_link_active(struct trunk_softc *,
93 	    struct trunk_port *);
94 const void *trunk_gethdr(struct mbuf *, u_int, u_int, void *);
95 
96 struct if_clone trunk_cloner =
97     IF_CLONE_INITIALIZER("trunk", trunk_clone_create, trunk_clone_destroy);
98 
99 /* Simple round robin */
100 int	 trunk_rr_attach(struct trunk_softc *);
101 int	 trunk_rr_detach(struct trunk_softc *);
102 void	 trunk_rr_port_destroy(struct trunk_port *);
103 int	 trunk_rr_start(struct trunk_softc *, struct mbuf *);
104 int	 trunk_rr_input(struct trunk_softc *, struct trunk_port *,
105 	    struct ether_header *, struct mbuf *);
106 
107 /* Active failover */
108 int	 trunk_fail_attach(struct trunk_softc *);
109 int	 trunk_fail_detach(struct trunk_softc *);
110 int	 trunk_fail_start(struct trunk_softc *, struct mbuf *);
111 int	 trunk_fail_input(struct trunk_softc *, struct trunk_port *,
112 	    struct ether_header *, struct mbuf *);
113 
114 /* Loadbalancing */
115 int	 trunk_lb_attach(struct trunk_softc *);
116 int	 trunk_lb_detach(struct trunk_softc *);
117 int	 trunk_lb_port_create(struct trunk_port *);
118 void	 trunk_lb_port_destroy(struct trunk_port *);
119 int	 trunk_lb_start(struct trunk_softc *, struct mbuf *);
120 int	 trunk_lb_input(struct trunk_softc *, struct trunk_port *,
121 	    struct ether_header *, struct mbuf *);
122 int	 trunk_lb_porttable(struct trunk_softc *, struct trunk_port *);
123 
124 /* Broadcast mode */
125 int	 trunk_bcast_attach(struct trunk_softc *);
126 int	 trunk_bcast_detach(struct trunk_softc *);
127 int	 trunk_bcast_start(struct trunk_softc *, struct mbuf *);
128 int	 trunk_bcast_input(struct trunk_softc *, struct trunk_port *,
129 	    struct ether_header *, struct mbuf *);
130 
131 /* 802.3ad LACP */
132 int	 trunk_lacp_attach(struct trunk_softc *);
133 int	 trunk_lacp_detach(struct trunk_softc *);
134 int	 trunk_lacp_start(struct trunk_softc *, struct mbuf *);
135 int	 trunk_lacp_input(struct trunk_softc *, struct trunk_port *,
136 	    struct ether_header *, struct mbuf *);
137 
138 /* Trunk protocol table */
139 static const struct {
140 	enum trunk_proto	ti_proto;
141 	int			(*ti_attach)(struct trunk_softc *);
142 } trunk_protos[] = {
143 	{ TRUNK_PROTO_ROUNDROBIN,	trunk_rr_attach },
144 	{ TRUNK_PROTO_FAILOVER,		trunk_fail_attach },
145 	{ TRUNK_PROTO_LOADBALANCE,	trunk_lb_attach },
146 	{ TRUNK_PROTO_BROADCAST,	trunk_bcast_attach },
147 	{ TRUNK_PROTO_LACP,		trunk_lacp_attach },
148 	{ TRUNK_PROTO_NONE,		NULL }
149 };
150 
151 void
152 trunkattach(int count)
153 {
154 	SLIST_INIT(&trunk_list);
155 	if_clone_attach(&trunk_cloner);
156 }
157 
158 int
159 trunk_clone_create(struct if_clone *ifc, int unit)
160 {
161 	struct trunk_softc *tr;
162 	struct ifnet *ifp;
163 	int i, error = 0;
164 
165 	if ((tr = malloc(sizeof(struct trunk_softc),
166 	    M_DEVBUF, M_NOWAIT|M_ZERO)) == NULL)
167 		return (ENOMEM);
168 
169 	tr->tr_unit = unit;
170 	tr->tr_proto = TRUNK_PROTO_NONE;
171 	for (i = 0; trunk_protos[i].ti_proto != TRUNK_PROTO_NONE; i++) {
172 		if (trunk_protos[i].ti_proto == TRUNK_PROTO_DEFAULT) {
173 			tr->tr_proto = trunk_protos[i].ti_proto;
174 			if ((error = trunk_protos[i].ti_attach(tr)) != 0) {
175 				free(tr, M_DEVBUF);
176 				return (error);
177 			}
178 			break;
179 		}
180 	}
181 	SLIST_INIT(&tr->tr_ports);
182 
183 	/* Initialise pseudo media types */
184 	ifmedia_init(&tr->tr_media, 0, trunk_media_change,
185 	    trunk_media_status);
186 	ifmedia_add(&tr->tr_media, IFM_ETHER | IFM_AUTO, 0, NULL);
187 	ifmedia_set(&tr->tr_media, IFM_ETHER | IFM_AUTO);
188 
189 	ifp = &tr->tr_ac.ac_if;
190 	ifp->if_carp = NULL;
191 	ifp->if_type = IFT_ETHER;
192 	ifp->if_softc = tr;
193 	ifp->if_start = trunk_start;
194 	ifp->if_watchdog = trunk_watchdog;
195 	ifp->if_ioctl = trunk_ioctl;
196 	ifp->if_output = ether_output;
197 	ifp->if_flags = IFF_SIMPLEX | IFF_BROADCAST | IFF_MULTICAST;
198 	ifp->if_capabilities = trunk_capabilities(tr);
199 
200 	IFQ_SET_MAXLEN(&ifp->if_snd, ifqmaxlen);
201 	IFQ_SET_READY(&ifp->if_snd);
202 
203 	snprintf(ifp->if_xname, sizeof(ifp->if_xname), "%s%d",
204 	    ifc->ifc_name, unit);
205 
206 	/*
207 	 * Attach as an ordinary ethernet device, childs will be attached
208 	 * as special device IFT_IEEE8023ADLAG.
209 	 */
210 	if_attach(ifp);
211 	ether_ifattach(ifp);
212 
213 	/* Insert into the global list of trunks */
214 	SLIST_INSERT_HEAD(&trunk_list, tr, tr_entries);
215 
216 	return (0);
217 }
218 
219 int
220 trunk_clone_destroy(struct ifnet *ifp)
221 {
222 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
223 	struct trunk_port *tp;
224 	int error, s;
225 
226 	/* Remove any multicast groups that we may have joined. */
227 	trunk_ether_purgemulti(tr);
228 
229 	s = splnet();
230 
231 	/* Shutdown and remove trunk ports, return on error */
232 	while ((tp = SLIST_FIRST(&tr->tr_ports)) != NULL) {
233 		if ((error = trunk_port_destroy(tp)) != 0) {
234 			splx(s);
235 			return (error);
236 		}
237 	}
238 
239 	ifmedia_delete_instance(&tr->tr_media, IFM_INST_ANY);
240 	ether_ifdetach(ifp);
241 	if_detach(ifp);
242 
243 	SLIST_REMOVE(&trunk_list, tr, trunk_softc, tr_entries);
244 	free(tr, M_DEVBUF);
245 
246 	splx(s);
247 
248 	return (0);
249 }
250 
251 void
252 trunk_lladdr(struct arpcom *ac, u_int8_t *lladdr)
253 {
254 	struct ifnet *ifp = &ac->ac_if;
255 	struct ifaddr *ifa;
256 	struct sockaddr_dl *sdl;
257 
258 	ifa = ifnet_addrs[ifp->if_index];
259 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
260 	sdl->sdl_type = IFT_ETHER;
261 	sdl->sdl_alen = ETHER_ADDR_LEN;
262 	bcopy(lladdr, LLADDR(sdl), ETHER_ADDR_LEN);
263 	bcopy(lladdr, ac->ac_enaddr, ETHER_ADDR_LEN);
264 }
265 
266 int
267 trunk_capabilities(struct trunk_softc *tr)
268 {
269 	struct trunk_port *tp;
270 	int cap = ~0, priv;
271 
272 	/* Preserve private capabilities */
273 	priv = tr->tr_capabilities & IFCAP_TRUNK_MASK;
274 
275 	/* Get capabilities from the trunk ports */
276 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries)
277 		cap &= tp->tp_capabilities;
278 
279 	if (tr->tr_ifflags & IFF_DEBUG) {
280 		printf("%s: capabilities 0x%08x\n",
281 		    tr->tr_ifname, cap == ~0 ? priv : (cap | priv));
282 	}
283 
284 	return (cap == ~0 ? priv : (cap | priv));
285 }
286 
287 void
288 trunk_port_lladdr(struct trunk_port *tp, u_int8_t *lladdr)
289 {
290 	struct ifnet *ifp = tp->tp_if;
291 	struct ifaddr *ifa;
292 	struct ifreq ifr;
293 
294 	/* Set the link layer address */
295 	trunk_lladdr((struct arpcom *)ifp, lladdr);
296 
297 	/* Reset the port to update the lladdr */
298 	if (ifp->if_flags & IFF_UP) {
299 		int s = splnet();
300 		ifp->if_flags &= ~IFF_UP;
301 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
302 		ifp->if_flags |= IFF_UP;
303 		(*ifp->if_ioctl)(ifp, SIOCSIFFLAGS, (caddr_t)&ifr);
304 		splx(s);
305 		TAILQ_FOREACH(ifa, &ifp->if_addrlist, ifa_list) {
306 			if (ifa->ifa_addr != NULL &&
307 			    ifa->ifa_addr->sa_family == AF_INET)
308 				arp_ifinit((struct arpcom *)ifp, ifa);
309 		}
310 	}
311 }
312 
313 int
314 trunk_port_create(struct trunk_softc *tr, struct ifnet *ifp)
315 {
316 	struct trunk_softc *tr_ptr;
317 	struct trunk_port *tp;
318 	int error = 0;
319 
320 	/* Limit the maximal number of trunk ports */
321 	if (tr->tr_count >= TRUNK_MAX_PORTS)
322 		return (ENOSPC);
323 
324 	/* New trunk port has to be in an idle state */
325 	if (ifp->if_flags & IFF_OACTIVE)
326 		return (EBUSY);
327 
328 	/* Check if port has already been associated to a trunk */
329 	if (trunk_port_get(NULL, ifp) != NULL)
330 		return (EBUSY);
331 
332 	/* XXX Disallow non-ethernet interfaces (this should be any of 802) */
333 	if (ifp->if_type != IFT_ETHER)
334 		return (EPROTONOSUPPORT);
335 
336 	if ((error = ifpromisc(ifp, 1)) != 0)
337 		return (error);
338 
339 	if ((tp = malloc(sizeof(struct trunk_port),
340 	    M_DEVBUF, M_NOWAIT|M_ZERO)) == NULL)
341 		return (ENOMEM);
342 
343 	/* Check if port is a stacked trunk */
344 	SLIST_FOREACH(tr_ptr, &trunk_list, tr_entries) {
345 		if (ifp == &tr_ptr->tr_ac.ac_if) {
346 			tp->tp_flags |= TRUNK_PORT_STACK;
347 			if (trunk_port_checkstacking(tr_ptr) >=
348 			    TRUNK_MAX_STACKING) {
349 				free(tp, M_DEVBUF);
350 				return (E2BIG);
351 			}
352 		}
353 	}
354 
355 	/* Change the interface type */
356 	tp->tp_iftype = ifp->if_type;
357 	ifp->if_type = IFT_IEEE8023ADLAG;
358 	ifp->if_tp = (caddr_t)tp;
359 	tp->tp_watchdog = ifp->if_watchdog;
360 	ifp->if_watchdog = trunk_port_watchdog;
361 	tp->tp_ioctl = ifp->if_ioctl;
362 	ifp->if_ioctl = trunk_port_ioctl;
363 
364 	tp->tp_if = ifp;
365 	tp->tp_trunk = tr;
366 
367 	/* Save port link layer address */
368 	bcopy(((struct arpcom *)ifp)->ac_enaddr, tp->tp_lladdr, ETHER_ADDR_LEN);
369 
370 	if (SLIST_EMPTY(&tr->tr_ports)) {
371 		tr->tr_primary = tp;
372 		tp->tp_flags |= TRUNK_PORT_MASTER;
373 		trunk_lladdr(&tr->tr_ac, tp->tp_lladdr);
374 	}
375 
376 	/* Update link layer address for this port */
377 	trunk_port_lladdr(tp, tr->tr_primary->tp_lladdr);
378 
379 	/* Insert into the list of ports */
380 	SLIST_INSERT_HEAD(&tr->tr_ports, tp, tp_entries);
381 	tr->tr_count++;
382 
383 	/* Update trunk capabilities */
384 	tr->tr_capabilities = trunk_capabilities(tr);
385 
386 	/* Add multicast addresses to this port */
387 	trunk_ether_cmdmulti(tp, SIOCADDMULTI);
388 
389 	/* Register callback for physical link state changes */
390 	if (ifp->if_linkstatehooks != NULL)
391 		tp->lh_cookie = hook_establish(ifp->if_linkstatehooks, 1,
392 		    trunk_port_state, tp);
393 
394 	if (tr->tr_port_create != NULL)
395 		error = (*tr->tr_port_create)(tp);
396 
397 	return (error);
398 }
399 
400 int
401 trunk_port_checkstacking(struct trunk_softc *tr)
402 {
403 	struct trunk_softc *tr_ptr;
404 	struct trunk_port *tp;
405 	int m = 0;
406 
407 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries) {
408 		if (tp->tp_flags & TRUNK_PORT_STACK) {
409 			tr_ptr = (struct trunk_softc *)tp->tp_if->if_softc;
410 			m = MAX(m, trunk_port_checkstacking(tr_ptr));
411 		}
412 	}
413 
414 	return (m + 1);
415 }
416 
417 int
418 trunk_port_destroy(struct trunk_port *tp)
419 {
420 	struct trunk_softc *tr = (struct trunk_softc *)tp->tp_trunk;
421 	struct trunk_port *tp_ptr;
422 	struct ifnet *ifp = tp->tp_if;
423 
424 	if (tr->tr_port_destroy != NULL)
425 		(*tr->tr_port_destroy)(tp);
426 
427 	/* Remove multicast addresses from this port */
428 	trunk_ether_cmdmulti(tp, SIOCDELMULTI);
429 
430 	/* Port has to be down */
431 	if (ifp->if_flags & IFF_UP)
432 		if_down(ifp);
433 
434 	ifpromisc(ifp, 0);
435 
436 	/* Restore interface */
437 	ifp->if_type = tp->tp_iftype;
438 	ifp->if_watchdog = tp->tp_watchdog;
439 	ifp->if_ioctl = tp->tp_ioctl;
440 	ifp->if_tp = NULL;
441 
442 	if (ifp->if_linkstatehooks != NULL)
443 		hook_disestablish(ifp->if_linkstatehooks, tp->lh_cookie);
444 
445 	/* Finally, remove the port from the trunk */
446 	SLIST_REMOVE(&tr->tr_ports, tp, trunk_port, tp_entries);
447 	tr->tr_count--;
448 
449 	/* Update the primary interface */
450 	if (tp == tr->tr_primary) {
451 		u_int8_t lladdr[ETHER_ADDR_LEN];
452 
453 		if ((tp_ptr = SLIST_FIRST(&tr->tr_ports)) == NULL) {
454 			bzero(&lladdr, ETHER_ADDR_LEN);
455 		} else {
456 			bcopy(((struct arpcom *)tp_ptr->tp_if)->ac_enaddr,
457 			    lladdr, ETHER_ADDR_LEN);
458 			tp_ptr->tp_flags = TRUNK_PORT_MASTER;
459 		}
460 		trunk_lladdr(&tr->tr_ac, lladdr);
461 		tr->tr_primary = tp_ptr;
462 
463 		/* Update link layer address for each port */
464 		SLIST_FOREACH(tp_ptr, &tr->tr_ports, tp_entries)
465 			trunk_port_lladdr(tp_ptr, lladdr);
466 	}
467 
468 	/* Reset the port lladdr */
469 	trunk_port_lladdr(tp, tp->tp_lladdr);
470 
471 	free(tp, M_DEVBUF);
472 
473 	/* Update trunk capabilities */
474 	tr->tr_capabilities = trunk_capabilities(tr);
475 
476 	return (0);
477 }
478 
479 void
480 trunk_port_watchdog(struct ifnet *ifp)
481 {
482 	struct trunk_port *tp;
483 
484 	/* Should be checked by the caller */
485 	if (ifp->if_type != IFT_IEEE8023ADLAG)
486 		return;
487 	if ((tp = (struct trunk_port *)ifp->if_tp) == NULL ||
488 	    tp->tp_trunk == NULL)
489 		return;
490 
491 	if (tp->tp_watchdog != NULL)
492 		(*tp->tp_watchdog)(ifp);
493 }
494 
495 
496 int
497 trunk_port_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
498 {
499 	struct trunk_reqport *rp = (struct trunk_reqport *)data;
500 	struct trunk_softc *tr;
501 	struct trunk_port *tp = NULL;
502 	int s, error = 0;
503 
504 	s = splnet();
505 
506 	/* Should be checked by the caller */
507 	if (ifp->if_type != IFT_IEEE8023ADLAG ||
508 	    (tp = (struct trunk_port *)ifp->if_tp) == NULL ||
509 	    (tr = (struct trunk_softc *)tp->tp_trunk) == NULL) {
510 		error = EINVAL;
511 		goto fallback;
512 	}
513 
514 	switch (cmd) {
515 	case SIOCGTRUNKPORT:
516 		if (rp->rp_portname[0] == '\0' ||
517 		    ifunit(rp->rp_portname) != ifp) {
518 			error = EINVAL;
519 			break;
520 		}
521 
522 		/* Search in all trunks if the global flag is set */
523 		if ((tp = trunk_port_get(rp->rp_flags & TRUNK_PORT_GLOBAL ?
524 		    NULL : tr, ifp)) == NULL) {
525 			error = ENOENT;
526 			break;
527 		}
528 
529 		trunk_port2req(tp, rp);
530 		break;
531 	default:
532 		error = ENOTTY;
533 		goto fallback;
534 	}
535 
536 	splx(s);
537 	return (error);
538 
539  fallback:
540 	splx(s);
541 
542 	if (tp != NULL)
543 		error = (*tp->tp_ioctl)(ifp, cmd, data);
544 
545 	return (error);
546 }
547 
548 void
549 trunk_port_ifdetach(struct ifnet *ifp)
550 {
551 	struct trunk_port *tp;
552 
553 	if ((tp = (struct trunk_port *)ifp->if_tp) == NULL)
554 		return;
555 
556 	trunk_port_destroy(tp);
557 }
558 
559 struct trunk_port *
560 trunk_port_get(struct trunk_softc *tr, struct ifnet *ifp)
561 {
562 	struct trunk_port *tp;
563 	struct trunk_softc *tr_ptr;
564 
565 	if (tr != NULL) {
566 		/* Search port in specified trunk */
567 		SLIST_FOREACH(tp, &tr->tr_ports, tp_entries) {
568 			if (tp->tp_if == ifp)
569 				return (tp);
570 		}
571 	} else {
572 		/* Search all trunks for the selected port */
573 		SLIST_FOREACH(tr_ptr, &trunk_list, tr_entries) {
574 			SLIST_FOREACH(tp, &tr_ptr->tr_ports, tp_entries) {
575 				if (tp->tp_if == ifp)
576 					return (tp);
577 			}
578 		}
579 	}
580 
581 	return (NULL);
582 }
583 
584 void
585 trunk_port2req(struct trunk_port *tp, struct trunk_reqport *rp)
586 {
587 	struct trunk_softc *tr = (struct trunk_softc *)tp->tp_trunk;
588 
589 	strlcpy(rp->rp_ifname, tr->tr_ifname, sizeof(rp->rp_ifname));
590 	strlcpy(rp->rp_portname, tp->tp_if->if_xname, sizeof(rp->rp_portname));
591 	rp->rp_prio = tp->tp_prio;
592 	if (tr->tr_portreq != NULL)
593 		(*tr->tr_portreq)(tp, (caddr_t)&rp->rp_psc);
594 
595 	/* Add protocol specific flags */
596 	switch (tr->tr_proto) {
597 	case TRUNK_PROTO_FAILOVER:
598 		rp->rp_flags = tp->tp_flags;
599 		if (tp == trunk_link_active(tr, tr->tr_primary))
600 			rp->rp_flags |= TRUNK_PORT_ACTIVE;
601 		break;
602 
603 	case TRUNK_PROTO_ROUNDROBIN:
604 	case TRUNK_PROTO_LOADBALANCE:
605 	case TRUNK_PROTO_BROADCAST:
606 		rp->rp_flags = tp->tp_flags;
607 		if (TRUNK_PORTACTIVE(tp))
608 			rp->rp_flags |= TRUNK_PORT_ACTIVE;
609 		break;
610 
611 	case TRUNK_PROTO_LACP:
612 		/* LACP has a different definition of active */
613 		rp->rp_flags = lacp_port_status(tp);
614 		break;
615 	default:
616 		break;
617 	}
618 }
619 
620 int
621 trunk_ioctl(struct ifnet *ifp, u_long cmd, caddr_t data)
622 {
623 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
624 	struct trunk_reqall *ra = (struct trunk_reqall *)data;
625 	struct trunk_reqport *rp = (struct trunk_reqport *)data, rpbuf;
626 	struct ifreq *ifr = (struct ifreq *)data;
627 	struct ifaddr *ifa = (struct ifaddr *)data;
628 	struct trunk_port *tp;
629 	struct ifnet *tpif;
630 	int s, i, error = 0;
631 
632 	s = splnet();
633 
634 	bzero(&rpbuf, sizeof(rpbuf));
635 
636 	switch (cmd) {
637 	case SIOCGTRUNK:
638 		ra->ra_proto = tr->tr_proto;
639 		if (tr->tr_req != NULL)
640 			(*tr->tr_req)(tr, (caddr_t)&ra->ra_psc);
641 		ra->ra_ports = i = 0;
642 		tp = SLIST_FIRST(&tr->tr_ports);
643 		while (tp && ra->ra_size >=
644 		    i + sizeof(struct trunk_reqport)) {
645 			trunk_port2req(tp, &rpbuf);
646 			error = copyout(&rpbuf, (caddr_t)ra->ra_port + i,
647 			    sizeof(struct trunk_reqport));
648 			if (error)
649 				break;
650 			i += sizeof(struct trunk_reqport);
651 			ra->ra_ports++;
652 			tp = SLIST_NEXT(tp, tp_entries);
653 		}
654 		break;
655 	case SIOCSTRUNK:
656 		if ((error = suser(curproc, 0)) != 0) {
657 			error = EPERM;
658 			break;
659 		}
660 		if (ra->ra_proto >= TRUNK_PROTO_MAX) {
661 			error = EPROTONOSUPPORT;
662 			break;
663 		}
664 		if (tr->tr_proto != TRUNK_PROTO_NONE)
665 			error = tr->tr_detach(tr);
666 		if (error != 0)
667 			break;
668 		for (i = 0; i < (sizeof(trunk_protos) /
669 		    sizeof(trunk_protos[0])); i++) {
670 			if (trunk_protos[i].ti_proto == ra->ra_proto) {
671 				if (tr->tr_ifflags & IFF_DEBUG)
672 					printf("%s: using proto %u\n",
673 					    tr->tr_ifname,
674 					    trunk_protos[i].ti_proto);
675 				tr->tr_proto = trunk_protos[i].ti_proto;
676 				if (tr->tr_proto != TRUNK_PROTO_NONE)
677 					error = trunk_protos[i].ti_attach(tr);
678 				goto out;
679 			}
680 		}
681 		error = EPROTONOSUPPORT;
682 		break;
683 	case SIOCGTRUNKPORT:
684 		if (rp->rp_portname[0] == '\0' ||
685 		    (tpif = ifunit(rp->rp_portname)) == NULL) {
686 			error = EINVAL;
687 			break;
688 		}
689 
690 		/* Search in all trunks if the global flag is set */
691 		if ((tp = trunk_port_get(rp->rp_flags & TRUNK_PORT_GLOBAL ?
692 		    NULL : tr, tpif)) == NULL) {
693 			error = ENOENT;
694 			break;
695 		}
696 
697 		trunk_port2req(tp, rp);
698 		break;
699 	case SIOCSTRUNKPORT:
700 		if ((error = suser(curproc, 0)) != 0) {
701 			error = EPERM;
702 			break;
703 		}
704 		if (rp->rp_portname[0] == '\0' ||
705 		    (tpif = ifunit(rp->rp_portname)) == NULL) {
706 			error = EINVAL;
707 			break;
708 		}
709 		error = trunk_port_create(tr, tpif);
710 		break;
711 	case SIOCSTRUNKDELPORT:
712 		if ((error = suser(curproc, 0)) != 0) {
713 			error = EPERM;
714 			break;
715 		}
716 		if (rp->rp_portname[0] == '\0' ||
717 		    (tpif = ifunit(rp->rp_portname)) == NULL) {
718 			error = EINVAL;
719 			break;
720 		}
721 
722 		/* Search in all trunks if the global flag is set */
723 		if ((tp = trunk_port_get(rp->rp_flags & TRUNK_PORT_GLOBAL ?
724 		    NULL : tr, tpif)) == NULL) {
725 			error = ENOENT;
726 			break;
727 		}
728 
729 		error = trunk_port_destroy(tp);
730 		break;
731 	case SIOCSIFADDR:
732 		ifp->if_flags |= IFF_UP;
733 #ifdef INET
734 		if (ifa->ifa_addr->sa_family == AF_INET)
735 			arp_ifinit(&tr->tr_ac, ifa);
736 #endif /* INET */
737 		error = ENETRESET;
738 		break;
739 	case SIOCSIFFLAGS:
740 		error = ENETRESET;
741 		break;
742 	case SIOCADDMULTI:
743 		error = trunk_ether_addmulti(tr, ifr);
744 		break;
745 	case SIOCDELMULTI:
746 		error = trunk_ether_delmulti(tr, ifr);
747 		break;
748 	case SIOCSIFMEDIA:
749 	case SIOCGIFMEDIA:
750 		error = ifmedia_ioctl(ifp, ifr, &tr->tr_media, cmd);
751 		break;
752 	case SIOCSIFLLADDR:
753 		/* Update the port lladdrs as well */
754 		SLIST_FOREACH(tp, &tr->tr_ports, tp_entries)
755 			trunk_port_lladdr(tp, ifr->ifr_addr.sa_data);
756 		error = ENETRESET;
757 		break;
758 	default:
759 		error = ether_ioctl(ifp, &tr->tr_ac, cmd, data);
760 	}
761 
762 	if (error == ENETRESET) {
763 		if (ifp->if_flags & IFF_UP) {
764 			if ((ifp->if_flags & IFF_RUNNING) == 0)
765 				trunk_init(ifp);
766 		} else {
767 			if (ifp->if_flags & IFF_RUNNING)
768 				trunk_stop(ifp);
769 		}
770 		error = 0;
771 	}
772 
773  out:
774 	splx(s);
775 	return (error);
776 }
777 
778 int
779 trunk_ether_addmulti(struct trunk_softc *tr, struct ifreq *ifr)
780 {
781 	struct trunk_mc *mc;
782 	u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN];
783 	int error;
784 
785 	/* Ignore ENETRESET error code */
786 	if ((error = ether_addmulti(ifr, &tr->tr_ac)) != ENETRESET)
787 		return (error);
788 
789 	if ((mc = malloc(sizeof(*mc), M_DEVBUF, M_NOWAIT)) == NULL) {
790 		error = ENOMEM;
791 		goto failed;
792 	}
793 
794 	ether_multiaddr(&ifr->ifr_addr, addrlo, addrhi);
795 	ETHER_LOOKUP_MULTI(addrlo, addrhi, &tr->tr_ac, mc->mc_enm);
796 	bcopy(&ifr->ifr_addr, &mc->mc_addr, ifr->ifr_addr.sa_len);
797 	SLIST_INSERT_HEAD(&tr->tr_mc_head, mc, mc_entries);
798 
799 	if ((error = trunk_ioctl_allports(tr, SIOCADDMULTI,
800 	    (caddr_t)ifr)) != 0) {
801 		trunk_ether_delmulti(tr, ifr);
802 		return (error);
803 	}
804 
805 	return (error);
806 
807  failed:
808 	ether_delmulti(ifr, &tr->tr_ac);
809 
810 	return (error);
811 }
812 
813 int
814 trunk_ether_delmulti(struct trunk_softc *tr, struct ifreq *ifr)
815 {
816 	struct ether_multi *enm;
817 	struct trunk_mc *mc;
818 	u_int8_t addrlo[ETHER_ADDR_LEN], addrhi[ETHER_ADDR_LEN];
819 	int error;
820 
821 	if ((error = ether_multiaddr(&ifr->ifr_addr, addrlo, addrhi)) != 0)
822 		return (error);
823 	ETHER_LOOKUP_MULTI(addrlo, addrhi, &tr->tr_ac, enm);
824 	if (enm == NULL)
825 		return (EINVAL);
826 
827 	SLIST_FOREACH(mc, &tr->tr_mc_head, mc_entries)
828 		if (mc->mc_enm == enm)
829 			break;
830 
831 	/* We won't delete entries we didn't add */
832 	if (mc == NULL)
833 		return (EINVAL);
834 
835 	if ((error = ether_delmulti(ifr, &tr->tr_ac)) != ENETRESET)
836 		return (error);
837 
838 	if ((error = trunk_ioctl_allports(tr, SIOCDELMULTI,
839 	    (caddr_t)ifr)) != 0) {
840 		/* XXX At least one port failed to remove the address */
841 		if (tr->tr_ifflags & IFF_DEBUG) {
842 			printf("%s: failed to remove multicast address "
843 			    "on all ports\n", tr->tr_ifname);
844 		}
845 	}
846 
847 	SLIST_REMOVE(&tr->tr_mc_head, mc, trunk_mc, mc_entries);
848 	free(mc, M_DEVBUF);
849 
850 	return (0);
851 }
852 
853 void
854 trunk_ether_purgemulti(struct trunk_softc *tr)
855 {
856 	struct trunk_mc *mc;
857 	struct trunk_ifreq ifs;
858 	struct ifreq *ifr = &ifs.ifreq.ifreq;
859 
860 	while ((mc = SLIST_FIRST(&tr->tr_mc_head)) != NULL) {
861 		bcopy(&mc->mc_addr, &ifr->ifr_addr, mc->mc_addr.ss_len);
862 
863 		/* Try to remove multicast address on all ports */
864 		trunk_ioctl_allports(tr, SIOCDELMULTI, (caddr_t)ifr);
865 
866 		SLIST_REMOVE(&tr->tr_mc_head, mc, trunk_mc, mc_entries);
867 		free(mc, M_DEVBUF);
868 	}
869 }
870 
871 int
872 trunk_ether_cmdmulti(struct trunk_port *tp, u_long cmd)
873 {
874 	struct trunk_softc *tr = (struct trunk_softc *)tp->tp_trunk;
875 	struct trunk_mc *mc;
876 	struct trunk_ifreq ifs;
877 	struct ifreq *ifr = &ifs.ifreq.ifreq;
878 	int ret, error = 0;
879 
880 	bcopy(tp->tp_ifname, ifr->ifr_name, IFNAMSIZ);
881 	SLIST_FOREACH(mc, &tr->tr_mc_head, mc_entries) {
882 		bcopy(&mc->mc_addr, &ifr->ifr_addr, mc->mc_addr.ss_len);
883 
884 		if ((ret = tp->tp_ioctl(tp->tp_if, cmd, (caddr_t)ifr)) != 0) {
885 			if (tr->tr_ifflags & IFF_DEBUG) {
886 				printf("%s: ioctl %lu failed on %s: %d\n",
887 				    tr->tr_ifname, cmd, tp->tp_ifname, ret);
888 			}
889 			/* Store last known error and continue */
890 			error = ret;
891 		}
892 	}
893 
894 	return (error);
895 }
896 
897 int
898 trunk_ioctl_allports(struct trunk_softc *tr, u_long cmd, caddr_t data)
899 {
900 	struct ifreq *ifr = (struct ifreq *)data;
901 	struct trunk_port *tp;
902 	int ret, error = 0;
903 
904 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries) {
905 		bcopy(tp->tp_ifname, ifr->ifr_name, IFNAMSIZ);
906 		if ((ret = tp->tp_ioctl(tp->tp_if, cmd, data)) != 0) {
907 			if (tr->tr_ifflags & IFF_DEBUG) {
908 				printf("%s: ioctl %lu failed on %s: %d\n",
909 				    tr->tr_ifname, cmd, tp->tp_ifname, ret);
910 			}
911 			/* Store last known error and continue */
912 			error = ret;
913 		}
914 	}
915 
916 	return (error);
917 }
918 
919 void
920 trunk_start(struct ifnet *ifp)
921 {
922 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
923 	struct mbuf *m;
924 	int error = 0;
925 
926 	for (;; error = 0) {
927 		IFQ_DEQUEUE(&ifp->if_snd, m);
928 		if (m == NULL)
929 			break;
930 
931 #if NBPFILTER > 0
932 		if (ifp->if_bpf)
933 			bpf_mtap_ether(ifp->if_bpf, m, BPF_DIRECTION_OUT);
934 #endif
935 
936 		if (tr->tr_proto != TRUNK_PROTO_NONE && tr->tr_count) {
937 			error = (*tr->tr_start)(tr, m);
938 			if (error == 0)
939 				ifp->if_opackets++;
940 			else
941 				ifp->if_oerrors++;
942 		} else {
943 			m_freem(m);
944 			if (tr->tr_proto != TRUNK_PROTO_NONE)
945 				ifp->if_oerrors++;
946 		}
947 	}
948 }
949 
950 int
951 trunk_enqueue(struct ifnet *ifp, struct mbuf *m)
952 {
953 	int len, error = 0;
954 	u_short mflags;
955 
956 	splassert(IPL_NET);
957 
958 	/* Send mbuf */
959 	mflags = m->m_flags;
960 	len = m->m_pkthdr.len;
961 	IFQ_ENQUEUE(&ifp->if_snd, m, NULL, error);
962 	if (error)
963 		return (error);
964 	if_start(ifp);
965 
966 	ifp->if_obytes += len;
967 	if (mflags & M_MCAST)
968 		ifp->if_omcasts++;
969 
970 	return (error);
971 }
972 
973 u_int32_t
974 trunk_hashmbuf(struct mbuf *m, u_int32_t key)
975 {
976 	u_int16_t etype, ether_vtag;
977 	u_int32_t p = 0;
978 	u_int16_t *vlan, vlanbuf[2];
979 	int off;
980 	struct ether_header *eh;
981 #ifdef INET
982 	struct ip *ip, ipbuf;
983 #endif
984 #ifdef INET6
985 	u_int32_t flow;
986 	struct ip6_hdr *ip6, ip6buf;
987 #endif
988 
989 	off = sizeof(*eh);
990 	if (m->m_len < off)
991 		return (p);
992 	eh = mtod(m, struct ether_header *);
993 	etype = ntohs(eh->ether_type);
994 	p = hash32_buf(&eh->ether_shost, ETHER_ADDR_LEN, key);
995 	p = hash32_buf(&eh->ether_dhost, ETHER_ADDR_LEN, p);
996 
997 	/* Special handling for encapsulating VLAN frames */
998 	if (m->m_flags & M_VLANTAG) {
999 		ether_vtag = EVL_VLANOFTAG(m->m_pkthdr.ether_vtag);
1000 		p = hash32_buf(&ether_vtag, sizeof(ether_vtag), p);
1001 	} else if (etype == ETHERTYPE_VLAN) {
1002 		if ((vlan = (u_int16_t *)
1003 		    trunk_gethdr(m, off, EVL_ENCAPLEN, &vlanbuf)) == NULL)
1004 			return (p);
1005 		ether_vtag = EVL_VLANOFTAG(*vlan);
1006 		p = hash32_buf(&ether_vtag, sizeof(ether_vtag), p);
1007 		etype = ntohs(vlan[1]);
1008 		off += EVL_ENCAPLEN;
1009 	}
1010 
1011 	switch (etype) {
1012 #ifdef INET
1013 	case ETHERTYPE_IP:
1014 		if ((ip = (struct ip *)
1015 		    trunk_gethdr(m, off, sizeof(*ip), &ipbuf)) == NULL)
1016 			return (p);
1017 		p = hash32_buf(&ip->ip_src, sizeof(struct in_addr), p);
1018 		p = hash32_buf(&ip->ip_dst, sizeof(struct in_addr), p);
1019 		break;
1020 #endif
1021 #ifdef INET6
1022 	case ETHERTYPE_IPV6:
1023 		if ((ip6 = (struct ip6_hdr *)
1024 		    trunk_gethdr(m, off, sizeof(*ip6), &ip6buf)) == NULL)
1025 			return (p);
1026 		p = hash32_buf(&ip6->ip6_src, sizeof(struct in6_addr), p);
1027 		p = hash32_buf(&ip6->ip6_dst, sizeof(struct in6_addr), p);
1028 		flow = ip6->ip6_flow & IPV6_FLOWLABEL_MASK;
1029 		p = hash32_buf(&flow, sizeof(flow), p); /* IPv6 flow label */
1030 		break;
1031 #endif
1032 	}
1033 
1034 	return (p);
1035 }
1036 
1037 void
1038 trunk_init(struct ifnet *ifp)
1039 {
1040 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
1041 	int s;
1042 
1043 	s = splnet();
1044 
1045 	ifp->if_flags |= IFF_RUNNING;
1046 	ifp->if_flags &= ~IFF_OACTIVE;
1047 
1048 	if (tr->tr_init != NULL)
1049 		(*tr->tr_init)(tr);
1050 
1051 	splx(s);
1052 }
1053 
1054 void
1055 trunk_stop(struct ifnet *ifp)
1056 {
1057 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
1058 	int s;
1059 
1060 	s = splnet();
1061 
1062 	ifp->if_flags &= ~(IFF_RUNNING | IFF_OACTIVE);
1063 
1064 	if (tr->tr_stop != NULL)
1065 		(*tr->tr_stop)(tr);
1066 
1067 	splx(s);
1068 }
1069 
1070 void
1071 trunk_watchdog(struct ifnet *ifp)
1072 {
1073 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
1074 
1075 	if (tr->tr_proto != TRUNK_PROTO_NONE &&
1076 	    (*tr->tr_watchdog)(tr) != 0) {
1077 		ifp->if_oerrors++;
1078 	}
1079 
1080 }
1081 
1082 int
1083 trunk_input(struct ifnet *ifp, struct ether_header *eh, struct mbuf *m)
1084 {
1085 	struct trunk_softc *tr;
1086 	struct trunk_port *tp;
1087 	struct ifnet *trifp = NULL;
1088 	int error = 0;
1089 
1090 	/* Should be checked by the caller */
1091 	if (ifp->if_type != IFT_IEEE8023ADLAG) {
1092 		error = EPROTONOSUPPORT;
1093 		goto bad;
1094 	}
1095 	if ((tp = (struct trunk_port *)ifp->if_tp) == NULL ||
1096 	    (tr = (struct trunk_softc *)tp->tp_trunk) == NULL) {
1097 		error = ENOENT;
1098 		goto bad;
1099 	}
1100 	trifp = &tr->tr_ac.ac_if;
1101 	if (tr->tr_proto == TRUNK_PROTO_NONE) {
1102 		error = ENOENT;
1103 		goto bad;
1104 	}
1105 
1106 #if NBPFILTER > 0
1107 	if (trifp->if_bpf && tr->tr_proto != TRUNK_PROTO_FAILOVER)
1108 		bpf_mtap_hdr(trifp->if_bpf, (char *)eh, ETHER_HDR_LEN, m,
1109 		    BPF_DIRECTION_IN);
1110 #endif
1111 
1112 	error = (*tr->tr_input)(tr, tp, eh, m);
1113 	if (error != 0)
1114 		return (error);
1115 
1116 	trifp->if_ipackets++;
1117 	return (0);
1118 
1119  bad:
1120 	if (error > 0 && trifp != NULL)
1121 		trifp->if_ierrors++;
1122 	m_freem(m);
1123 	return (error);
1124 }
1125 
1126 int
1127 trunk_media_change(struct ifnet *ifp)
1128 {
1129 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
1130 
1131 	if (tr->tr_ifflags & IFF_DEBUG)
1132 		printf("%s\n", __func__);
1133 
1134 	/* Ignore */
1135 	return (0);
1136 }
1137 
1138 void
1139 trunk_media_status(struct ifnet *ifp, struct ifmediareq *imr)
1140 {
1141 	struct trunk_softc *tr = (struct trunk_softc *)ifp->if_softc;
1142 	struct trunk_port *tp;
1143 
1144 	imr->ifm_status = IFM_AVALID;
1145 	imr->ifm_active = IFM_ETHER | IFM_AUTO;
1146 
1147 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries) {
1148 		if (TRUNK_PORTACTIVE(tp))
1149 			imr->ifm_status |= IFM_ACTIVE;
1150 	}
1151 }
1152 
1153 void
1154 trunk_port_state(void *arg)
1155 {
1156 	struct trunk_port *tp = (struct trunk_port *)arg;
1157 	struct trunk_softc *tr = NULL;
1158 
1159 	if (tp != NULL)
1160 		tr = (struct trunk_softc *)tp->tp_trunk;
1161 	if (tr == NULL)
1162 		return;
1163 	if (tr->tr_linkstate != NULL)
1164 		(*tr->tr_linkstate)(tp);
1165 	trunk_link_active(tr, tp);
1166 }
1167 
1168 struct trunk_port *
1169 trunk_link_active(struct trunk_softc *tr, struct trunk_port *tp)
1170 {
1171 	struct trunk_port *tp_next, *rval = NULL;
1172 	int new_link = LINK_STATE_DOWN;
1173 
1174 	/*
1175 	 * Search a port which reports an active link state.
1176 	 */
1177 
1178 	if (tp == NULL)
1179 		goto search;
1180 	if (TRUNK_PORTACTIVE(tp)) {
1181 		rval = tp;
1182 		goto found;
1183 	}
1184 	if ((tp_next = SLIST_NEXT(tp, tp_entries)) != NULL &&
1185 	    TRUNK_PORTACTIVE(tp_next)) {
1186 		rval = tp_next;
1187 		goto found;
1188 	}
1189 
1190  search:
1191 	SLIST_FOREACH(tp_next, &tr->tr_ports, tp_entries) {
1192 		if (TRUNK_PORTACTIVE(tp_next)) {
1193 			rval = tp_next;
1194 			goto found;
1195 		}
1196 	}
1197 
1198  found:
1199 	if (rval != NULL) {
1200 		/*
1201 		 * The IEEE 802.1D standard assumes that a trunk with
1202 		 * multiple ports is always full duplex. This is valid
1203 		 * for load sharing trunks and if at least two links
1204 		 * are active. Unfortunately, checking the latter would
1205 		 * be too expensive at this point.
1206 		 */
1207 		if ((tr->tr_capabilities & IFCAP_TRUNK_FULLDUPLEX) &&
1208 		    (tr->tr_count > 1))
1209 			new_link = LINK_STATE_FULL_DUPLEX;
1210 		else
1211 			new_link = rval->tp_link_state;
1212 	}
1213 
1214 	if (tr->tr_ac.ac_if.if_link_state != new_link) {
1215 		tr->tr_ac.ac_if.if_link_state = new_link;
1216 		if_link_state_change(&tr->tr_ac.ac_if);
1217 	}
1218 
1219 	return (rval);
1220 }
1221 
1222 const void *
1223 trunk_gethdr(struct mbuf *m, u_int off, u_int len, void *buf)
1224 {
1225 	if (m->m_pkthdr.len < (off + len))
1226 		return (NULL);
1227 	else if (m->m_len < (off + len)) {
1228 		m_copydata(m, off, len, buf);
1229 		return (buf);
1230 	}
1231 	return (mtod(m, const void *) + off);
1232 }
1233 
1234 /*
1235  * Simple round robin trunking
1236  */
1237 
1238 int
1239 trunk_rr_attach(struct trunk_softc *tr)
1240 {
1241 	struct trunk_port *tp;
1242 
1243 	tr->tr_detach = trunk_rr_detach;
1244 	tr->tr_start = trunk_rr_start;
1245 	tr->tr_input = trunk_rr_input;
1246 	tr->tr_init = NULL;
1247 	tr->tr_stop = NULL;
1248 	tr->tr_port_create = NULL;
1249 	tr->tr_port_destroy = trunk_rr_port_destroy;
1250 	tr->tr_capabilities = IFCAP_TRUNK_FULLDUPLEX;
1251 	tr->tr_req = NULL;
1252 	tr->tr_portreq = NULL;
1253 
1254 	tp = SLIST_FIRST(&tr->tr_ports);
1255 	tr->tr_psc = (caddr_t)tp;
1256 
1257 	return (0);
1258 }
1259 
1260 int
1261 trunk_rr_detach(struct trunk_softc *tr)
1262 {
1263 	tr->tr_psc = NULL;
1264 	return (0);
1265 }
1266 
1267 void
1268 trunk_rr_port_destroy(struct trunk_port *tp)
1269 {
1270 	struct trunk_softc *tr = (struct trunk_softc *)tp->tp_trunk;
1271 
1272 	if (tp == (struct trunk_port *)tr->tr_psc)
1273 		tr->tr_psc = NULL;
1274 }
1275 
1276 int
1277 trunk_rr_start(struct trunk_softc *tr, struct mbuf *m)
1278 {
1279 	struct trunk_port *tp = (struct trunk_port *)tr->tr_psc, *tp_next;
1280 	int error = 0;
1281 
1282 	if (tp == NULL && (tp = trunk_link_active(tr, NULL)) == NULL) {
1283 		m_freem(m);
1284 		return (ENOENT);
1285 	}
1286 
1287 	/* Send mbuf */
1288 	if ((error = trunk_enqueue(tp->tp_if, m)) != 0)
1289 		return (error);
1290 
1291 	/* Get next active port */
1292 	tp_next = trunk_link_active(tr, SLIST_NEXT(tp, tp_entries));
1293 	tr->tr_psc = (caddr_t)tp_next;
1294 
1295 	return (0);
1296 }
1297 
1298 int
1299 trunk_rr_input(struct trunk_softc *tr, struct trunk_port *tp,
1300     struct ether_header *eh, struct mbuf *m)
1301 {
1302 	struct ifnet *ifp = &tr->tr_ac.ac_if;
1303 
1304 	/* Just pass in the packet to our trunk device */
1305 	m->m_pkthdr.rcvif = ifp;
1306 
1307 	return (0);
1308 }
1309 
1310 /*
1311  * Active failover
1312  */
1313 
1314 int
1315 trunk_fail_attach(struct trunk_softc *tr)
1316 {
1317 	tr->tr_detach = trunk_fail_detach;
1318 	tr->tr_start = trunk_fail_start;
1319 	tr->tr_input = trunk_fail_input;
1320 	tr->tr_init = NULL;
1321 	tr->tr_stop = NULL;
1322 	tr->tr_port_create = NULL;
1323 	tr->tr_port_destroy = NULL;
1324 	tr->tr_linkstate = NULL;
1325 	tr->tr_req = NULL;
1326 	tr->tr_portreq = NULL;
1327 
1328 	return (0);
1329 }
1330 
1331 int
1332 trunk_fail_detach(struct trunk_softc *tr)
1333 {
1334 	return (0);
1335 }
1336 
1337 int
1338 trunk_fail_start(struct trunk_softc *tr, struct mbuf *m)
1339 {
1340 	struct trunk_port *tp;
1341 
1342 	/* Use the master port if active or the next available port */
1343 	if ((tp = trunk_link_active(tr, tr->tr_primary)) == NULL) {
1344 		m_freem(m);
1345 		return (ENOENT);
1346 	}
1347 
1348 	/* Send mbuf */
1349 	return (trunk_enqueue(tp->tp_if, m));
1350 }
1351 
1352 int
1353 trunk_fail_input(struct trunk_softc *tr, struct trunk_port *tp,
1354     struct ether_header *eh, struct mbuf *m)
1355 {
1356 	struct ifnet *ifp = &tr->tr_ac.ac_if;
1357 	struct trunk_port *tmp_tp;
1358 	int accept = 0;
1359 
1360 	if (tp == tr->tr_primary) {
1361 		accept = 1;
1362 	} else if (tr->tr_primary->tp_link_state == LINK_STATE_DOWN) {
1363 		tmp_tp = trunk_link_active(tr, NULL);
1364 		/*
1365 		 * If tmp_tp is null, we've received a packet when all
1366 		 * our links are down. Weird, but process it anyways.
1367 		 */
1368 		if ((tmp_tp == NULL || tmp_tp == tp))
1369 			accept = 1;
1370 	}
1371 	if (!accept) {
1372 		m_freem(m);
1373 		return (-1);
1374 	}
1375 #if NBPFILTER > 0
1376 	if (ifp->if_bpf)
1377 		bpf_mtap_hdr(ifp->if_bpf, (char *)eh, ETHER_HDR_LEN, m,
1378 		    BPF_DIRECTION_IN);
1379 #endif
1380 
1381 	m->m_pkthdr.rcvif = ifp;
1382 	return (0);
1383 }
1384 
1385 /*
1386  * Loadbalancing
1387  */
1388 
1389 int
1390 trunk_lb_attach(struct trunk_softc *tr)
1391 {
1392 	struct trunk_lb *lb;
1393 
1394 	if ((lb = malloc(sizeof(*lb), M_DEVBUF, M_NOWAIT|M_ZERO)) == NULL)
1395 		return (ENOMEM);
1396 
1397 	tr->tr_detach = trunk_lb_detach;
1398 	tr->tr_start = trunk_lb_start;
1399 	tr->tr_input = trunk_lb_input;
1400 	tr->tr_port_create = trunk_lb_port_create;
1401 	tr->tr_port_destroy = trunk_lb_port_destroy;
1402 	tr->tr_linkstate = NULL;
1403 	tr->tr_capabilities = IFCAP_TRUNK_FULLDUPLEX;
1404 	tr->tr_req = NULL;
1405 	tr->tr_portreq = NULL;
1406 
1407 	lb->lb_key = arc4random();
1408 	tr->tr_psc = (caddr_t)lb;
1409 
1410 	return (0);
1411 }
1412 
1413 int
1414 trunk_lb_detach(struct trunk_softc *tr)
1415 {
1416 	struct trunk_lb *lb = (struct trunk_lb *)tr->tr_psc;
1417 	if (lb != NULL)
1418 		free(lb, M_DEVBUF);
1419 	return (0);
1420 }
1421 
1422 int
1423 trunk_lb_porttable(struct trunk_softc *tr, struct trunk_port *tp)
1424 {
1425 	struct trunk_lb *lb = (struct trunk_lb *)tr->tr_psc;
1426 	struct trunk_port *tp_next;
1427 	int i = 0;
1428 
1429 	bzero(&lb->lb_ports, sizeof(lb->lb_ports));
1430 	SLIST_FOREACH(tp_next, &tr->tr_ports, tp_entries) {
1431 		if (tp_next == tp)
1432 			continue;
1433 		if (i >= TRUNK_MAX_PORTS)
1434 			return (EINVAL);
1435 		if (tr->tr_ifflags & IFF_DEBUG)
1436 			printf("%s: port %s at index %d\n",
1437 			    tr->tr_ifname, tp_next->tp_ifname, i);
1438 		lb->lb_ports[i++] = tp_next;
1439 	}
1440 
1441 	return (0);
1442 }
1443 
1444 int
1445 trunk_lb_port_create(struct trunk_port *tp)
1446 {
1447 	struct trunk_softc *tr = (struct trunk_softc *)tp->tp_trunk;
1448 	return (trunk_lb_porttable(tr, NULL));
1449 }
1450 
1451 void
1452 trunk_lb_port_destroy(struct trunk_port *tp)
1453 {
1454 	struct trunk_softc *tr = (struct trunk_softc *)tp->tp_trunk;
1455 	trunk_lb_porttable(tr, tp);
1456 }
1457 
1458 int
1459 trunk_lb_start(struct trunk_softc *tr, struct mbuf *m)
1460 {
1461 	struct trunk_lb *lb = (struct trunk_lb *)tr->tr_psc;
1462 	struct trunk_port *tp = NULL;
1463 	u_int32_t p = 0;
1464 
1465 	p = trunk_hashmbuf(m, lb->lb_key);
1466 	p %= tr->tr_count;
1467 	tp = lb->lb_ports[p];
1468 
1469 	/*
1470 	 * Check the port's link state. This will return the next active
1471 	 * port if the link is down or the port is NULL.
1472 	 */
1473 	if ((tp = trunk_link_active(tr, tp)) == NULL) {
1474 		m_freem(m);
1475 		return (ENOENT);
1476 	}
1477 
1478 	/* Send mbuf */
1479 	return (trunk_enqueue(tp->tp_if, m));
1480 }
1481 
1482 int
1483 trunk_lb_input(struct trunk_softc *tr, struct trunk_port *tp,
1484     struct ether_header *eh, struct mbuf *m)
1485 {
1486 	struct ifnet *ifp = &tr->tr_ac.ac_if;
1487 
1488 	/* Just pass in the packet to our trunk device */
1489 	m->m_pkthdr.rcvif = ifp;
1490 
1491 	return (0);
1492 }
1493 
1494 /*
1495  * Broadcast mode
1496  */
1497 
1498 int
1499 trunk_bcast_attach(struct trunk_softc *tr)
1500 {
1501 	tr->tr_detach = trunk_bcast_detach;
1502 	tr->tr_start = trunk_bcast_start;
1503 	tr->tr_input = trunk_bcast_input;
1504 	tr->tr_init = NULL;
1505 	tr->tr_stop = NULL;
1506 	tr->tr_port_create = NULL;
1507 	tr->tr_port_destroy = NULL;
1508 	tr->tr_linkstate = NULL;
1509 	tr->tr_req = NULL;
1510 	tr->tr_portreq = NULL;
1511 
1512 	return (0);
1513 }
1514 
1515 int
1516 trunk_bcast_detach(struct trunk_softc *tr)
1517 {
1518 	return (0);
1519 }
1520 
1521 int
1522 trunk_bcast_start(struct trunk_softc *tr, struct mbuf *m)
1523 {
1524 	int			 active_ports = 0;
1525 	int			 errors = 0;
1526 	int			 ret;
1527 	struct trunk_port	*tp;
1528 	struct mbuf		*n;
1529 
1530 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries) {
1531 		if (TRUNK_PORTACTIVE(tp)) {
1532 			if (active_ports) {
1533 				n = m_copym(m, 0, M_COPYALL, M_DONTWAIT);
1534 				if (n == NULL) {
1535 					m_free(m);
1536 					return (ENOBUFS);
1537 				}
1538 			} else
1539 				n = m;
1540 			active_ports++;
1541 			if ((ret = trunk_enqueue(tp->tp_if, n)))
1542 				errors++;
1543 		}
1544 	}
1545 	if (active_ports == 0) {
1546 		m_free(m);
1547 		return (ENOENT);
1548 	}
1549 	if (errors == active_ports)
1550 		return (ret);
1551 	return (0);
1552 }
1553 
1554 int
1555 trunk_bcast_input(struct trunk_softc *tr, struct trunk_port *tp,
1556     struct ether_header *eh, struct mbuf *m)
1557 {
1558 	struct ifnet *ifp = &tr->tr_ac.ac_if;
1559 
1560 	m->m_pkthdr.rcvif = ifp;
1561 	return (0);
1562 }
1563 
1564 /*
1565  * 802.3ad LACP
1566  */
1567 
1568 int
1569 trunk_lacp_attach(struct trunk_softc *tr)
1570 {
1571 	struct trunk_port *tp;
1572 	int error;
1573 
1574 	tr->tr_detach = trunk_lacp_detach;
1575 	tr->tr_port_create = lacp_port_create;
1576 	tr->tr_port_destroy = lacp_port_destroy;
1577 	tr->tr_linkstate = lacp_linkstate;
1578 	tr->tr_start = trunk_lacp_start;
1579 	tr->tr_input = trunk_lacp_input;
1580 	tr->tr_init = lacp_init;
1581 	tr->tr_stop = lacp_stop;
1582 	tr->tr_req = lacp_req;
1583 	tr->tr_portreq = lacp_portreq;
1584 
1585 	error = lacp_attach(tr);
1586 	if (error)
1587 		return (error);
1588 
1589 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries)
1590 		lacp_port_create(tp);
1591 
1592 	return (error);
1593 }
1594 
1595 int
1596 trunk_lacp_detach(struct trunk_softc *tr)
1597 {
1598 	struct trunk_port *tp;
1599 	int error;
1600 
1601 	SLIST_FOREACH(tp, &tr->tr_ports, tp_entries)
1602 		lacp_port_destroy(tp);
1603 
1604 	/* unlocking is safe here */
1605 	error = lacp_detach(tr);
1606 
1607 	return (error);
1608 }
1609 
1610 int
1611 trunk_lacp_start(struct trunk_softc *tr, struct mbuf *m)
1612 {
1613 	struct trunk_port *tp;
1614 
1615 	tp = lacp_select_tx_port(tr, m);
1616 	if (tp == NULL) {
1617 		m_freem(m);
1618 		return (EBUSY);
1619 	}
1620 
1621 	/* Send mbuf */
1622 	return (trunk_enqueue(tp->tp_if, m));
1623 }
1624 
1625 int
1626 trunk_lacp_input(struct trunk_softc *tr, struct trunk_port *tp,
1627     struct ether_header *eh, struct mbuf *m)
1628 {
1629 	struct ifnet *ifp = &tr->tr_ac.ac_if;
1630 
1631 	m = lacp_input(tp, eh, m);
1632 	if (m == NULL)
1633 		return (-1);
1634 
1635 	m->m_pkthdr.rcvif = ifp;
1636 	return (0);
1637 }
1638