xref: /freebsd/sys/net/if_ethersubr.c (revision bc7512cc58af2e8bbe5bbf5ca0059b1daa1da897)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_netgraph.h"
38 #include "opt_mbuf_profiling.h"
39 #include "opt_rss.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/devctl.h>
44 #include <sys/eventhandler.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/mbuf.h>
50 #include <sys/module.h>
51 #include <sys/msan.h>
52 #include <sys/proc.h>
53 #include <sys/priv.h>
54 #include <sys/random.h>
55 #include <sys/socket.h>
56 #include <sys/sockio.h>
57 #include <sys/sysctl.h>
58 #include <sys/uuid.h>
59 
60 #include <net/ieee_oui.h>
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/if_arp.h>
64 #include <net/netisr.h>
65 #include <net/route.h>
66 #include <net/if_llc.h>
67 #include <net/if_dl.h>
68 #include <net/if_types.h>
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/if_bridgevar.h>
72 #include <net/if_vlan_var.h>
73 #include <net/if_llatbl.h>
74 #include <net/pfil.h>
75 #include <net/rss_config.h>
76 #include <net/vnet.h>
77 
78 #include <netpfil/pf/pf_mtag.h>
79 
80 #if defined(INET) || defined(INET6)
81 #include <netinet/in.h>
82 #include <netinet/in_var.h>
83 #include <netinet/if_ether.h>
84 #include <netinet/ip_carp.h>
85 #include <netinet/ip_var.h>
86 #endif
87 #ifdef INET6
88 #include <netinet6/nd6.h>
89 #endif
90 #include <security/mac/mac_framework.h>
91 
92 #include <crypto/sha1.h>
93 
94 #ifdef CTASSERT
95 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
96 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
97 #endif
98 
99 VNET_DEFINE(pfil_head_t, link_pfil_head);	/* Packet filter hooks */
100 
101 /* netgraph node hooks for ng_ether(4) */
102 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
103 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
104 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
105 void	(*ng_ether_attach_p)(struct ifnet *ifp);
106 void	(*ng_ether_detach_p)(struct ifnet *ifp);
107 
108 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
109 
110 /* if_bridge(4) support */
111 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
112 
113 /* if_lagg(4) support */
114 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
115 
116 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
117 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
118 
119 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
120 		struct sockaddr *);
121 static	int ether_requestencap(struct ifnet *, struct if_encap_req *);
122 
123 #define senderr(e) do { error = (e); goto bad;} while (0)
124 
125 static void
126 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
127 {
128 	int csum_flags = 0;
129 
130 	if (src->m_pkthdr.csum_flags & CSUM_IP)
131 		csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
132 	if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
133 		csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
134 	if (src->m_pkthdr.csum_flags & CSUM_SCTP)
135 		csum_flags |= CSUM_SCTP_VALID;
136 	dst->m_pkthdr.csum_flags |= csum_flags;
137 	if (csum_flags & CSUM_DATA_VALID)
138 		dst->m_pkthdr.csum_data = 0xffff;
139 }
140 
141 /*
142  * Handle link-layer encapsulation requests.
143  */
144 static int
145 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
146 {
147 	struct ether_header *eh;
148 	struct arphdr *ah;
149 	uint16_t etype;
150 	const u_char *lladdr;
151 
152 	if (req->rtype != IFENCAP_LL)
153 		return (EOPNOTSUPP);
154 
155 	if (req->bufsize < ETHER_HDR_LEN)
156 		return (ENOMEM);
157 
158 	eh = (struct ether_header *)req->buf;
159 	lladdr = req->lladdr;
160 	req->lladdr_off = 0;
161 
162 	switch (req->family) {
163 	case AF_INET:
164 		etype = htons(ETHERTYPE_IP);
165 		break;
166 	case AF_INET6:
167 		etype = htons(ETHERTYPE_IPV6);
168 		break;
169 	case AF_ARP:
170 		ah = (struct arphdr *)req->hdata;
171 		ah->ar_hrd = htons(ARPHRD_ETHER);
172 
173 		switch(ntohs(ah->ar_op)) {
174 		case ARPOP_REVREQUEST:
175 		case ARPOP_REVREPLY:
176 			etype = htons(ETHERTYPE_REVARP);
177 			break;
178 		case ARPOP_REQUEST:
179 		case ARPOP_REPLY:
180 		default:
181 			etype = htons(ETHERTYPE_ARP);
182 			break;
183 		}
184 
185 		if (req->flags & IFENCAP_FLAG_BROADCAST)
186 			lladdr = ifp->if_broadcastaddr;
187 		break;
188 	default:
189 		return (EAFNOSUPPORT);
190 	}
191 
192 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
193 	memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
194 	memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
195 	req->bufsize = sizeof(struct ether_header);
196 
197 	return (0);
198 }
199 
200 static int
201 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
202 	const struct sockaddr *dst, struct route *ro, u_char *phdr,
203 	uint32_t *pflags, struct llentry **plle)
204 {
205 	uint32_t lleflags = 0;
206 	int error = 0;
207 #if defined(INET) || defined(INET6)
208 	struct ether_header *eh = (struct ether_header *)phdr;
209 	uint16_t etype;
210 #endif
211 
212 	if (plle)
213 		*plle = NULL;
214 
215 	switch (dst->sa_family) {
216 #ifdef INET
217 	case AF_INET:
218 		if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
219 			error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
220 			    plle);
221 		else {
222 			if (m->m_flags & M_BCAST)
223 				memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
224 				    ETHER_ADDR_LEN);
225 			else {
226 				const struct in_addr *a;
227 				a = &(((const struct sockaddr_in *)dst)->sin_addr);
228 				ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
229 			}
230 			etype = htons(ETHERTYPE_IP);
231 			memcpy(&eh->ether_type, &etype, sizeof(etype));
232 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
233 		}
234 		break;
235 #endif
236 #ifdef INET6
237 	case AF_INET6:
238 		if ((m->m_flags & M_MCAST) == 0) {
239 			int af = RO_GET_FAMILY(ro, dst);
240 			error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
241 			    &lleflags, plle);
242 		} else {
243 			const struct in6_addr *a6;
244 			a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
245 			ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
246 			etype = htons(ETHERTYPE_IPV6);
247 			memcpy(&eh->ether_type, &etype, sizeof(etype));
248 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
249 		}
250 		break;
251 #endif
252 	default:
253 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
254 		if (m != NULL)
255 			m_freem(m);
256 		return (EAFNOSUPPORT);
257 	}
258 
259 	if (error == EHOSTDOWN) {
260 		if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
261 			error = EHOSTUNREACH;
262 	}
263 
264 	if (error != 0)
265 		return (error);
266 
267 	*pflags = RT_MAY_LOOP;
268 	if (lleflags & LLE_IFADDR)
269 		*pflags |= RT_L2_ME;
270 
271 	return (0);
272 }
273 
274 /*
275  * Ethernet output routine.
276  * Encapsulate a packet of type family for the local net.
277  * Use trailer local net encapsulation if enough data in first
278  * packet leaves a multiple of 512 bytes of data in remainder.
279  */
280 int
281 ether_output(struct ifnet *ifp, struct mbuf *m,
282 	const struct sockaddr *dst, struct route *ro)
283 {
284 	int error = 0;
285 	char linkhdr[ETHER_HDR_LEN], *phdr;
286 	struct ether_header *eh;
287 	struct pf_mtag *t;
288 	bool loop_copy;
289 	int hlen;	/* link layer header length */
290 	uint32_t pflags;
291 	struct llentry *lle = NULL;
292 	int addref = 0;
293 
294 	phdr = NULL;
295 	pflags = 0;
296 	if (ro != NULL) {
297 		/* XXX BPF uses ro_prepend */
298 		if (ro->ro_prepend != NULL) {
299 			phdr = ro->ro_prepend;
300 			hlen = ro->ro_plen;
301 		} else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
302 			if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
303 				lle = ro->ro_lle;
304 				if (lle != NULL &&
305 				    (lle->la_flags & LLE_VALID) == 0) {
306 					LLE_FREE(lle);
307 					lle = NULL;	/* redundant */
308 					ro->ro_lle = NULL;
309 				}
310 				if (lle == NULL) {
311 					/* if we lookup, keep cache */
312 					addref = 1;
313 				} else
314 					/*
315 					 * Notify LLE code that
316 					 * the entry was used
317 					 * by datapath.
318 					 */
319 					llentry_provide_feedback(lle);
320 			}
321 			if (lle != NULL) {
322 				phdr = lle->r_linkdata;
323 				hlen = lle->r_hdrlen;
324 				pflags = lle->r_flags;
325 			}
326 		}
327 	}
328 
329 #ifdef MAC
330 	error = mac_ifnet_check_transmit(ifp, m);
331 	if (error)
332 		senderr(error);
333 #endif
334 
335 	M_PROFILE(m);
336 	if (ifp->if_flags & IFF_MONITOR)
337 		senderr(ENETDOWN);
338 	if (!((ifp->if_flags & IFF_UP) &&
339 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
340 		senderr(ENETDOWN);
341 
342 	if (phdr == NULL) {
343 		/* No prepend data supplied. Try to calculate ourselves. */
344 		phdr = linkhdr;
345 		hlen = ETHER_HDR_LEN;
346 		error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
347 		    addref ? &lle : NULL);
348 		if (addref && lle != NULL)
349 			ro->ro_lle = lle;
350 		if (error != 0)
351 			return (error == EWOULDBLOCK ? 0 : error);
352 	}
353 
354 	if ((pflags & RT_L2_ME) != 0) {
355 		update_mbuf_csumflags(m, m);
356 		return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
357 	}
358 	loop_copy = (pflags & RT_MAY_LOOP) != 0;
359 
360 	/*
361 	 * Add local net header.  If no space in first mbuf,
362 	 * allocate another.
363 	 *
364 	 * Note that we do prepend regardless of RT_HAS_HEADER flag.
365 	 * This is done because BPF code shifts m_data pointer
366 	 * to the end of ethernet header prior to calling if_output().
367 	 */
368 	M_PREPEND(m, hlen, M_NOWAIT);
369 	if (m == NULL)
370 		senderr(ENOBUFS);
371 	if ((pflags & RT_HAS_HEADER) == 0) {
372 		eh = mtod(m, struct ether_header *);
373 		memcpy(eh, phdr, hlen);
374 	}
375 
376 	/*
377 	 * If a simplex interface, and the packet is being sent to our
378 	 * Ethernet address or a broadcast address, loopback a copy.
379 	 * XXX To make a simplex device behave exactly like a duplex
380 	 * device, we should copy in the case of sending to our own
381 	 * ethernet address (thus letting the original actually appear
382 	 * on the wire). However, we don't do that here for security
383 	 * reasons and compatibility with the original behavior.
384 	 */
385 	if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
386 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
387 		struct mbuf *n;
388 
389 		/*
390 		 * Because if_simloop() modifies the packet, we need a
391 		 * writable copy through m_dup() instead of a readonly
392 		 * one as m_copy[m] would give us. The alternative would
393 		 * be to modify if_simloop() to handle the readonly mbuf,
394 		 * but performancewise it is mostly equivalent (trading
395 		 * extra data copying vs. extra locking).
396 		 *
397 		 * XXX This is a local workaround.  A number of less
398 		 * often used kernel parts suffer from the same bug.
399 		 * See PR kern/105943 for a proposed general solution.
400 		 */
401 		if ((n = m_dup(m, M_NOWAIT)) != NULL) {
402 			update_mbuf_csumflags(m, n);
403 			(void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
404 		} else
405 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
406 	}
407 
408        /*
409 	* Bridges require special output handling.
410 	*/
411 	if (ifp->if_bridge) {
412 		BRIDGE_OUTPUT(ifp, m, error);
413 		return (error);
414 	}
415 
416 #if defined(INET) || defined(INET6)
417 	if (ifp->if_carp &&
418 	    (error = (*carp_output_p)(ifp, m, dst)))
419 		goto bad;
420 #endif
421 
422 	/* Handle ng_ether(4) processing, if any */
423 	if (ifp->if_l2com != NULL) {
424 		KASSERT(ng_ether_output_p != NULL,
425 		    ("ng_ether_output_p is NULL"));
426 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
427 bad:			if (m != NULL)
428 				m_freem(m);
429 			return (error);
430 		}
431 		if (m == NULL)
432 			return (0);
433 	}
434 
435 	/* Continue with link-layer output */
436 	return ether_output_frame(ifp, m);
437 }
438 
439 static bool
440 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
441 {
442 	struct ether_8021q_tag qtag;
443 	struct ether_header *eh;
444 
445 	eh = mtod(*mp, struct ether_header *);
446 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
447 	    ntohs(eh->ether_type) == ETHERTYPE_QINQ)
448 		return (true);
449 
450 	qtag.vid = 0;
451 	qtag.pcp = pcp;
452 	qtag.proto = ETHERTYPE_VLAN;
453 	if (ether_8021q_frame(mp, ifp, ifp, &qtag))
454 		return (true);
455 	if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
456 	return (false);
457 }
458 
459 /*
460  * Ethernet link layer output routine to send a raw frame to the device.
461  *
462  * This assumes that the 14 byte Ethernet header is present and contiguous
463  * in the first mbuf (if BRIDGE'ing).
464  */
465 int
466 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
467 {
468 	uint8_t pcp;
469 
470 	pcp = ifp->if_pcp;
471 	if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
472 	    !ether_set_pcp(&m, ifp, pcp))
473 		return (0);
474 
475 	if (PFIL_HOOKED_OUT(V_link_pfil_head))
476 		switch (pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_OUT,
477 		    NULL)) {
478 		case PFIL_DROPPED:
479 			return (EACCES);
480 		case PFIL_CONSUMED:
481 			return (0);
482 		}
483 
484 #ifdef EXPERIMENTAL
485 #if defined(INET6) && defined(INET)
486 	/* draft-ietf-6man-ipv6only-flag */
487 	/* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
488 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
489 		struct ether_header *eh;
490 
491 		eh = mtod(m, struct ether_header *);
492 		switch (ntohs(eh->ether_type)) {
493 		case ETHERTYPE_IP:
494 		case ETHERTYPE_ARP:
495 		case ETHERTYPE_REVARP:
496 			m_freem(m);
497 			return (EAFNOSUPPORT);
498 			/* NOTREACHED */
499 			break;
500 		};
501 	}
502 #endif
503 #endif
504 
505 	/*
506 	 * Queue message on interface, update output statistics if successful,
507 	 * and start output if interface not yet active.
508 	 *
509 	 * If KMSAN is enabled, use it to verify that the data does not contain
510 	 * any uninitialized bytes.
511 	 */
512 	kmsan_check_mbuf(m, "ether_output");
513 	return ((ifp->if_transmit)(ifp, m));
514 }
515 
516 /*
517  * Process a received Ethernet packet; the packet is in the
518  * mbuf chain m with the ethernet header at the front.
519  */
520 static void
521 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
522 {
523 	struct ether_header *eh;
524 	u_short etype;
525 
526 	if ((ifp->if_flags & IFF_UP) == 0) {
527 		m_freem(m);
528 		return;
529 	}
530 #ifdef DIAGNOSTIC
531 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
532 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
533 		m_freem(m);
534 		return;
535 	}
536 #endif
537 	if (m->m_len < ETHER_HDR_LEN) {
538 		/* XXX maybe should pullup? */
539 		if_printf(ifp, "discard frame w/o leading ethernet "
540 				"header (len %u pkt len %u)\n",
541 				m->m_len, m->m_pkthdr.len);
542 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
543 		m_freem(m);
544 		return;
545 	}
546 	eh = mtod(m, struct ether_header *);
547 	etype = ntohs(eh->ether_type);
548 	random_harvest_queue_ether(m, sizeof(*m));
549 
550 #ifdef EXPERIMENTAL
551 #if defined(INET6) && defined(INET)
552 	/* draft-ietf-6man-ipv6only-flag */
553 	/* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
554 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
555 		switch (etype) {
556 		case ETHERTYPE_IP:
557 		case ETHERTYPE_ARP:
558 		case ETHERTYPE_REVARP:
559 			m_freem(m);
560 			return;
561 			/* NOTREACHED */
562 			break;
563 		};
564 	}
565 #endif
566 #endif
567 
568 	CURVNET_SET_QUIET(ifp->if_vnet);
569 
570 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
571 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
572 			m->m_flags |= M_BCAST;
573 		else
574 			m->m_flags |= M_MCAST;
575 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
576 	}
577 
578 #ifdef MAC
579 	/*
580 	 * Tag the mbuf with an appropriate MAC label before any other
581 	 * consumers can get to it.
582 	 */
583 	mac_ifnet_create_mbuf(ifp, m);
584 #endif
585 
586 	/*
587 	 * Give bpf a chance at the packet.
588 	 */
589 	ETHER_BPF_MTAP(ifp, m);
590 
591 	/*
592 	 * If the CRC is still on the packet, trim it off. We do this once
593 	 * and once only in case we are re-entered. Nothing else on the
594 	 * Ethernet receive path expects to see the FCS.
595 	 */
596 	if (m->m_flags & M_HASFCS) {
597 		m_adj(m, -ETHER_CRC_LEN);
598 		m->m_flags &= ~M_HASFCS;
599 	}
600 
601 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
602 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
603 
604 	/* Allow monitor mode to claim this frame, after stats are updated. */
605 	if (ifp->if_flags & IFF_MONITOR) {
606 		m_freem(m);
607 		CURVNET_RESTORE();
608 		return;
609 	}
610 
611 	/* Handle input from a lagg(4) port */
612 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
613 		KASSERT(lagg_input_ethernet_p != NULL,
614 		    ("%s: if_lagg not loaded!", __func__));
615 		m = (*lagg_input_ethernet_p)(ifp, m);
616 		if (m != NULL)
617 			ifp = m->m_pkthdr.rcvif;
618 		else {
619 			CURVNET_RESTORE();
620 			return;
621 		}
622 	}
623 
624 	/*
625 	 * If the hardware did not process an 802.1Q tag, do this now,
626 	 * to allow 802.1P priority frames to be passed to the main input
627 	 * path correctly.
628 	 */
629 	if ((m->m_flags & M_VLANTAG) == 0 &&
630 	    ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
631 		struct ether_vlan_header *evl;
632 
633 		if (m->m_len < sizeof(*evl) &&
634 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
635 #ifdef DIAGNOSTIC
636 			if_printf(ifp, "cannot pullup VLAN header\n");
637 #endif
638 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
639 			CURVNET_RESTORE();
640 			return;
641 		}
642 
643 		evl = mtod(m, struct ether_vlan_header *);
644 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
645 		m->m_flags |= M_VLANTAG;
646 
647 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
648 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
649 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
650 		eh = mtod(m, struct ether_header *);
651 	}
652 
653 	M_SETFIB(m, ifp->if_fib);
654 
655 	/* Allow ng_ether(4) to claim this frame. */
656 	if (ifp->if_l2com != NULL) {
657 		KASSERT(ng_ether_input_p != NULL,
658 		    ("%s: ng_ether_input_p is NULL", __func__));
659 		m->m_flags &= ~M_PROMISC;
660 		(*ng_ether_input_p)(ifp, &m);
661 		if (m == NULL) {
662 			CURVNET_RESTORE();
663 			return;
664 		}
665 		eh = mtod(m, struct ether_header *);
666 	}
667 
668 	/*
669 	 * Allow if_bridge(4) to claim this frame.
670 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
671 	 * and the frame should be delivered locally.
672 	 */
673 	if (ifp->if_bridge != NULL) {
674 		m->m_flags &= ~M_PROMISC;
675 		BRIDGE_INPUT(ifp, m);
676 		if (m == NULL) {
677 			CURVNET_RESTORE();
678 			return;
679 		}
680 		eh = mtod(m, struct ether_header *);
681 	}
682 
683 #if defined(INET) || defined(INET6)
684 	/*
685 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
686 	 * mbuf flows up to Layer 3.
687 	 * FreeBSD's implementation of carp(4) uses the inprotosw
688 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
689 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
690 	 * is outside the scope of the M_PROMISC test below.
691 	 * TODO: Maintain a hash table of ethernet addresses other than
692 	 * ether_dhost which may be active on this ifp.
693 	 */
694 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
695 		m->m_flags &= ~M_PROMISC;
696 	} else
697 #endif
698 	{
699 		/*
700 		 * If the frame received was not for our MAC address, set the
701 		 * M_PROMISC flag on the mbuf chain. The frame may need to
702 		 * be seen by the rest of the Ethernet input path in case of
703 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
704 		 * seen by upper protocol layers.
705 		 */
706 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
707 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
708 			m->m_flags |= M_PROMISC;
709 	}
710 
711 	ether_demux(ifp, m);
712 	CURVNET_RESTORE();
713 }
714 
715 /*
716  * Ethernet input dispatch; by default, direct dispatch here regardless of
717  * global configuration.  However, if RSS is enabled, hook up RSS affinity
718  * so that when deferred or hybrid dispatch is enabled, we can redistribute
719  * load based on RSS.
720  *
721  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
722  * not it had already done work distribution via multi-queue.  Then we could
723  * direct dispatch in the event load balancing was already complete and
724  * handle the case of interfaces with different capabilities better.
725  *
726  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
727  * at multiple layers?
728  *
729  * XXXRW: For now, enable all this only if RSS is compiled in, although it
730  * works fine without RSS.  Need to characterise the performance overhead
731  * of the detour through the netisr code in the event the result is always
732  * direct dispatch.
733  */
734 static void
735 ether_nh_input(struct mbuf *m)
736 {
737 
738 	M_ASSERTPKTHDR(m);
739 	KASSERT(m->m_pkthdr.rcvif != NULL,
740 	    ("%s: NULL interface pointer", __func__));
741 	ether_input_internal(m->m_pkthdr.rcvif, m);
742 }
743 
744 static struct netisr_handler	ether_nh = {
745 	.nh_name = "ether",
746 	.nh_handler = ether_nh_input,
747 	.nh_proto = NETISR_ETHER,
748 #ifdef RSS
749 	.nh_policy = NETISR_POLICY_CPU,
750 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
751 	.nh_m2cpuid = rss_m2cpuid,
752 #else
753 	.nh_policy = NETISR_POLICY_SOURCE,
754 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
755 #endif
756 };
757 
758 static void
759 ether_init(__unused void *arg)
760 {
761 
762 	netisr_register(&ether_nh);
763 }
764 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
765 
766 static void
767 vnet_ether_init(__unused void *arg)
768 {
769 	struct pfil_head_args args;
770 
771 	args.pa_version = PFIL_VERSION;
772 	args.pa_flags = PFIL_IN | PFIL_OUT;
773 	args.pa_type = PFIL_TYPE_ETHERNET;
774 	args.pa_headname = PFIL_ETHER_NAME;
775 	V_link_pfil_head = pfil_head_register(&args);
776 
777 #ifdef VIMAGE
778 	netisr_register_vnet(&ether_nh);
779 #endif
780 }
781 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
782     vnet_ether_init, NULL);
783 
784 #ifdef VIMAGE
785 static void
786 vnet_ether_pfil_destroy(__unused void *arg)
787 {
788 
789 	pfil_head_unregister(V_link_pfil_head);
790 }
791 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
792     vnet_ether_pfil_destroy, NULL);
793 
794 static void
795 vnet_ether_destroy(__unused void *arg)
796 {
797 
798 	netisr_unregister_vnet(&ether_nh);
799 }
800 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
801     vnet_ether_destroy, NULL);
802 #endif
803 
804 static void
805 ether_input(struct ifnet *ifp, struct mbuf *m)
806 {
807 	struct epoch_tracker et;
808 	struct mbuf *mn;
809 	bool needs_epoch;
810 
811 	needs_epoch = !(ifp->if_flags & IFF_KNOWSEPOCH);
812 
813 	/*
814 	 * The drivers are allowed to pass in a chain of packets linked with
815 	 * m_nextpkt. We split them up into separate packets here and pass
816 	 * them up. This allows the drivers to amortize the receive lock.
817 	 */
818 	CURVNET_SET_QUIET(ifp->if_vnet);
819 	if (__predict_false(needs_epoch))
820 		NET_EPOCH_ENTER(et);
821 	while (m) {
822 		mn = m->m_nextpkt;
823 		m->m_nextpkt = NULL;
824 
825 		/*
826 		 * We will rely on rcvif being set properly in the deferred
827 		 * context, so assert it is correct here.
828 		 */
829 		MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
830 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
831 		    "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
832 		netisr_dispatch(NETISR_ETHER, m);
833 		m = mn;
834 	}
835 	if (__predict_false(needs_epoch))
836 		NET_EPOCH_EXIT(et);
837 	CURVNET_RESTORE();
838 }
839 
840 /*
841  * Upper layer processing for a received Ethernet packet.
842  */
843 void
844 ether_demux(struct ifnet *ifp, struct mbuf *m)
845 {
846 	struct ether_header *eh;
847 	int i, isr;
848 	u_short ether_type;
849 
850 	NET_EPOCH_ASSERT();
851 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
852 
853 	/* Do not grab PROMISC frames in case we are re-entered. */
854 	if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
855 		i = pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_IN, NULL);
856 		if (i != 0 || m == NULL)
857 			return;
858 	}
859 
860 	eh = mtod(m, struct ether_header *);
861 	ether_type = ntohs(eh->ether_type);
862 
863 	/*
864 	 * If this frame has a VLAN tag other than 0, call vlan_input()
865 	 * if its module is loaded. Otherwise, drop.
866 	 */
867 	if ((m->m_flags & M_VLANTAG) &&
868 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
869 		if (ifp->if_vlantrunk == NULL) {
870 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
871 			m_freem(m);
872 			return;
873 		}
874 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
875 		    __func__));
876 		/* Clear before possibly re-entering ether_input(). */
877 		m->m_flags &= ~M_PROMISC;
878 		(*vlan_input_p)(ifp, m);
879 		return;
880 	}
881 
882 	/*
883 	 * Pass promiscuously received frames to the upper layer if the user
884 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
885 	 */
886 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
887 		m_freem(m);
888 		return;
889 	}
890 
891 	/*
892 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
893 	 * Strip off Ethernet header.
894 	 */
895 	m->m_flags &= ~M_VLANTAG;
896 	m_clrprotoflags(m);
897 	m_adj(m, ETHER_HDR_LEN);
898 
899 	/*
900 	 * Dispatch frame to upper layer.
901 	 */
902 	switch (ether_type) {
903 #ifdef INET
904 	case ETHERTYPE_IP:
905 		isr = NETISR_IP;
906 		break;
907 
908 	case ETHERTYPE_ARP:
909 		if (ifp->if_flags & IFF_NOARP) {
910 			/* Discard packet if ARP is disabled on interface */
911 			m_freem(m);
912 			return;
913 		}
914 		isr = NETISR_ARP;
915 		break;
916 #endif
917 #ifdef INET6
918 	case ETHERTYPE_IPV6:
919 		isr = NETISR_IPV6;
920 		break;
921 #endif
922 	default:
923 		goto discard;
924 	}
925 	netisr_dispatch(isr, m);
926 	return;
927 
928 discard:
929 	/*
930 	 * Packet is to be discarded.  If netgraph is present,
931 	 * hand the packet to it for last chance processing;
932 	 * otherwise dispose of it.
933 	 */
934 	if (ifp->if_l2com != NULL) {
935 		KASSERT(ng_ether_input_orphan_p != NULL,
936 		    ("ng_ether_input_orphan_p is NULL"));
937 		/*
938 		 * Put back the ethernet header so netgraph has a
939 		 * consistent view of inbound packets.
940 		 */
941 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
942 		(*ng_ether_input_orphan_p)(ifp, m);
943 		return;
944 	}
945 	m_freem(m);
946 }
947 
948 /*
949  * Convert Ethernet address to printable (loggable) representation.
950  * This routine is for compatibility; it's better to just use
951  *
952  *	printf("%6D", <pointer to address>, ":");
953  *
954  * since there's no static buffer involved.
955  */
956 char *
957 ether_sprintf(const u_char *ap)
958 {
959 	static char etherbuf[18];
960 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
961 	return (etherbuf);
962 }
963 
964 /*
965  * Perform common duties while attaching to interface list
966  */
967 void
968 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
969 {
970 	int i;
971 	struct ifaddr *ifa;
972 	struct sockaddr_dl *sdl;
973 
974 	ifp->if_addrlen = ETHER_ADDR_LEN;
975 	ifp->if_hdrlen = ETHER_HDR_LEN;
976 	ifp->if_mtu = ETHERMTU;
977 	if_attach(ifp);
978 	ifp->if_output = ether_output;
979 	ifp->if_input = ether_input;
980 	ifp->if_resolvemulti = ether_resolvemulti;
981 	ifp->if_requestencap = ether_requestencap;
982 #ifdef VIMAGE
983 	ifp->if_reassign = ether_reassign;
984 #endif
985 	if (ifp->if_baudrate == 0)
986 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
987 	ifp->if_broadcastaddr = etherbroadcastaddr;
988 
989 	ifa = ifp->if_addr;
990 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
991 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
992 	sdl->sdl_type = IFT_ETHER;
993 	sdl->sdl_alen = ifp->if_addrlen;
994 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
995 
996 	if (ifp->if_hw_addr != NULL)
997 		bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
998 
999 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1000 	if (ng_ether_attach_p != NULL)
1001 		(*ng_ether_attach_p)(ifp);
1002 
1003 	/* Announce Ethernet MAC address if non-zero. */
1004 	for (i = 0; i < ifp->if_addrlen; i++)
1005 		if (lla[i] != 0)
1006 			break;
1007 	if (i != ifp->if_addrlen)
1008 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1009 
1010 	uuid_ether_add(LLADDR(sdl));
1011 
1012 	/* Add necessary bits are setup; announce it now. */
1013 	EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1014 	if (IS_DEFAULT_VNET(curvnet))
1015 		devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1016 }
1017 
1018 /*
1019  * Perform common duties while detaching an Ethernet interface
1020  */
1021 void
1022 ether_ifdetach(struct ifnet *ifp)
1023 {
1024 	struct sockaddr_dl *sdl;
1025 
1026 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1027 	uuid_ether_del(LLADDR(sdl));
1028 
1029 	if (ifp->if_l2com != NULL) {
1030 		KASSERT(ng_ether_detach_p != NULL,
1031 		    ("ng_ether_detach_p is NULL"));
1032 		(*ng_ether_detach_p)(ifp);
1033 	}
1034 
1035 	bpfdetach(ifp);
1036 	if_detach(ifp);
1037 }
1038 
1039 #ifdef VIMAGE
1040 void
1041 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1042 {
1043 
1044 	if (ifp->if_l2com != NULL) {
1045 		KASSERT(ng_ether_detach_p != NULL,
1046 		    ("ng_ether_detach_p is NULL"));
1047 		(*ng_ether_detach_p)(ifp);
1048 	}
1049 
1050 	if (ng_ether_attach_p != NULL) {
1051 		CURVNET_SET_QUIET(new_vnet);
1052 		(*ng_ether_attach_p)(ifp);
1053 		CURVNET_RESTORE();
1054 	}
1055 }
1056 #endif
1057 
1058 SYSCTL_DECL(_net_link);
1059 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1060     "Ethernet");
1061 
1062 #if 0
1063 /*
1064  * This is for reference.  We have a table-driven version
1065  * of the little-endian crc32 generator, which is faster
1066  * than the double-loop.
1067  */
1068 uint32_t
1069 ether_crc32_le(const uint8_t *buf, size_t len)
1070 {
1071 	size_t i;
1072 	uint32_t crc;
1073 	int bit;
1074 	uint8_t data;
1075 
1076 	crc = 0xffffffff;	/* initial value */
1077 
1078 	for (i = 0; i < len; i++) {
1079 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1080 			carry = (crc ^ data) & 1;
1081 			crc >>= 1;
1082 			if (carry)
1083 				crc = (crc ^ ETHER_CRC_POLY_LE);
1084 		}
1085 	}
1086 
1087 	return (crc);
1088 }
1089 #else
1090 uint32_t
1091 ether_crc32_le(const uint8_t *buf, size_t len)
1092 {
1093 	static const uint32_t crctab[] = {
1094 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1095 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1096 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1097 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1098 	};
1099 	size_t i;
1100 	uint32_t crc;
1101 
1102 	crc = 0xffffffff;	/* initial value */
1103 
1104 	for (i = 0; i < len; i++) {
1105 		crc ^= buf[i];
1106 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1107 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1108 	}
1109 
1110 	return (crc);
1111 }
1112 #endif
1113 
1114 uint32_t
1115 ether_crc32_be(const uint8_t *buf, size_t len)
1116 {
1117 	size_t i;
1118 	uint32_t crc, carry;
1119 	int bit;
1120 	uint8_t data;
1121 
1122 	crc = 0xffffffff;	/* initial value */
1123 
1124 	for (i = 0; i < len; i++) {
1125 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1126 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1127 			crc <<= 1;
1128 			if (carry)
1129 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1130 		}
1131 	}
1132 
1133 	return (crc);
1134 }
1135 
1136 int
1137 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1138 {
1139 	struct ifaddr *ifa = (struct ifaddr *) data;
1140 	struct ifreq *ifr = (struct ifreq *) data;
1141 	int error = 0;
1142 
1143 	switch (command) {
1144 	case SIOCSIFADDR:
1145 		ifp->if_flags |= IFF_UP;
1146 
1147 		switch (ifa->ifa_addr->sa_family) {
1148 #ifdef INET
1149 		case AF_INET:
1150 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1151 			arp_ifinit(ifp, ifa);
1152 			break;
1153 #endif
1154 		default:
1155 			ifp->if_init(ifp->if_softc);
1156 			break;
1157 		}
1158 		break;
1159 
1160 	case SIOCGIFADDR:
1161 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1162 		    ETHER_ADDR_LEN);
1163 		break;
1164 
1165 	case SIOCSIFMTU:
1166 		/*
1167 		 * Set the interface MTU.
1168 		 */
1169 		if (ifr->ifr_mtu > ETHERMTU) {
1170 			error = EINVAL;
1171 		} else {
1172 			ifp->if_mtu = ifr->ifr_mtu;
1173 		}
1174 		break;
1175 
1176 	case SIOCSLANPCP:
1177 		error = priv_check(curthread, PRIV_NET_SETLANPCP);
1178 		if (error != 0)
1179 			break;
1180 		if (ifr->ifr_lan_pcp > 7 &&
1181 		    ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1182 			error = EINVAL;
1183 		} else {
1184 			ifp->if_pcp = ifr->ifr_lan_pcp;
1185 			/* broadcast event about PCP change */
1186 			EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1187 		}
1188 		break;
1189 
1190 	case SIOCGLANPCP:
1191 		ifr->ifr_lan_pcp = ifp->if_pcp;
1192 		break;
1193 
1194 	default:
1195 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1196 		break;
1197 	}
1198 	return (error);
1199 }
1200 
1201 static int
1202 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1203 	struct sockaddr *sa)
1204 {
1205 	struct sockaddr_dl *sdl;
1206 #ifdef INET
1207 	struct sockaddr_in *sin;
1208 #endif
1209 #ifdef INET6
1210 	struct sockaddr_in6 *sin6;
1211 #endif
1212 	u_char *e_addr;
1213 
1214 	switch(sa->sa_family) {
1215 	case AF_LINK:
1216 		/*
1217 		 * No mapping needed. Just check that it's a valid MC address.
1218 		 */
1219 		sdl = (struct sockaddr_dl *)sa;
1220 		e_addr = LLADDR(sdl);
1221 		if (!ETHER_IS_MULTICAST(e_addr))
1222 			return EADDRNOTAVAIL;
1223 		*llsa = NULL;
1224 		return 0;
1225 
1226 #ifdef INET
1227 	case AF_INET:
1228 		sin = (struct sockaddr_in *)sa;
1229 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1230 			return EADDRNOTAVAIL;
1231 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1232 		sdl->sdl_alen = ETHER_ADDR_LEN;
1233 		e_addr = LLADDR(sdl);
1234 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1235 		*llsa = (struct sockaddr *)sdl;
1236 		return 0;
1237 #endif
1238 #ifdef INET6
1239 	case AF_INET6:
1240 		sin6 = (struct sockaddr_in6 *)sa;
1241 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1242 			/*
1243 			 * An IP6 address of 0 means listen to all
1244 			 * of the Ethernet multicast address used for IP6.
1245 			 * (This is used for multicast routers.)
1246 			 */
1247 			ifp->if_flags |= IFF_ALLMULTI;
1248 			*llsa = NULL;
1249 			return 0;
1250 		}
1251 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1252 			return EADDRNOTAVAIL;
1253 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1254 		sdl->sdl_alen = ETHER_ADDR_LEN;
1255 		e_addr = LLADDR(sdl);
1256 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1257 		*llsa = (struct sockaddr *)sdl;
1258 		return 0;
1259 #endif
1260 
1261 	default:
1262 		/*
1263 		 * Well, the text isn't quite right, but it's the name
1264 		 * that counts...
1265 		 */
1266 		return EAFNOSUPPORT;
1267 	}
1268 }
1269 
1270 static moduledata_t ether_mod = {
1271 	.name = "ether",
1272 };
1273 
1274 void
1275 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1276 {
1277 	struct ether_vlan_header vlan;
1278 	struct mbuf mv, mb;
1279 
1280 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1281 	    ("%s: vlan information not present", __func__));
1282 	KASSERT(m->m_len >= sizeof(struct ether_header),
1283 	    ("%s: mbuf not large enough for header", __func__));
1284 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1285 	vlan.evl_proto = vlan.evl_encap_proto;
1286 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1287 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1288 	m->m_len -= sizeof(struct ether_header);
1289 	m->m_data += sizeof(struct ether_header);
1290 	/*
1291 	 * If a data link has been supplied by the caller, then we will need to
1292 	 * re-create a stack allocated mbuf chain with the following structure:
1293 	 *
1294 	 * (1) mbuf #1 will contain the supplied data link
1295 	 * (2) mbuf #2 will contain the vlan header
1296 	 * (3) mbuf #3 will contain the original mbuf's packet data
1297 	 *
1298 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1299 	 */
1300 	if (data != NULL) {
1301 		mv.m_next = m;
1302 		mv.m_data = (caddr_t)&vlan;
1303 		mv.m_len = sizeof(vlan);
1304 		mb.m_next = &mv;
1305 		mb.m_data = data;
1306 		mb.m_len = dlen;
1307 		bpf_mtap(bp, &mb);
1308 	} else
1309 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1310 	m->m_len += sizeof(struct ether_header);
1311 	m->m_data -= sizeof(struct ether_header);
1312 }
1313 
1314 struct mbuf *
1315 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1316 {
1317 	struct ether_vlan_header *evl;
1318 
1319 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1320 	if (m == NULL)
1321 		return (NULL);
1322 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1323 
1324 	if (m->m_len < sizeof(*evl)) {
1325 		m = m_pullup(m, sizeof(*evl));
1326 		if (m == NULL)
1327 			return (NULL);
1328 	}
1329 
1330 	/*
1331 	 * Transform the Ethernet header into an Ethernet header
1332 	 * with 802.1Q encapsulation.
1333 	 */
1334 	evl = mtod(m, struct ether_vlan_header *);
1335 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1336 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1337 	evl->evl_encap_proto = htons(proto);
1338 	evl->evl_tag = htons(tag);
1339 	return (m);
1340 }
1341 
1342 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1343     "IEEE 802.1Q VLAN");
1344 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1345     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1346     "for consistency");
1347 
1348 VNET_DEFINE_STATIC(int, soft_pad);
1349 #define	V_soft_pad	VNET(soft_pad)
1350 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1351     &VNET_NAME(soft_pad), 0,
1352     "pad short frames before tagging");
1353 
1354 /*
1355  * For now, make preserving PCP via an mbuf tag optional, as it increases
1356  * per-packet memory allocations and frees.  In the future, it would be
1357  * preferable to reuse ether_vtag for this, or similar.
1358  */
1359 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1360 #define	V_vlan_mtag_pcp	VNET(vlan_mtag_pcp)
1361 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1362     &VNET_NAME(vlan_mtag_pcp), 0,
1363     "Retain VLAN PCP information as packets are passed up the stack");
1364 
1365 bool
1366 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1367     struct ether_8021q_tag *qtag)
1368 {
1369 	struct m_tag *mtag;
1370 	int n;
1371 	uint16_t tag;
1372 	static const char pad[8];	/* just zeros */
1373 
1374 	/*
1375 	 * Pad the frame to the minimum size allowed if told to.
1376 	 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1377 	 * paragraph C.4.4.3.b.  It can help to work around buggy
1378 	 * bridges that violate paragraph C.4.4.3.a from the same
1379 	 * document, i.e., fail to pad short frames after untagging.
1380 	 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1381 	 * untagging it will produce a 62-byte frame, which is a runt
1382 	 * and requires padding.  There are VLAN-enabled network
1383 	 * devices that just discard such runts instead or mishandle
1384 	 * them somehow.
1385 	 */
1386 	if (V_soft_pad && p->if_type == IFT_ETHER) {
1387 		for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1388 		     n > 0; n -= sizeof(pad)) {
1389 			if (!m_append(*mp, min(n, sizeof(pad)), pad))
1390 				break;
1391 		}
1392 		if (n > 0) {
1393 			m_freem(*mp);
1394 			*mp = NULL;
1395 			if_printf(ife, "cannot pad short frame");
1396 			return (false);
1397 		}
1398 	}
1399 
1400 	/*
1401 	 * If PCP is set in mbuf, use it
1402 	 */
1403 	if ((*mp)->m_flags & M_VLANTAG) {
1404 		qtag->pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1405 	}
1406 
1407 	/*
1408 	 * If underlying interface can do VLAN tag insertion itself,
1409 	 * just pass the packet along. However, we need some way to
1410 	 * tell the interface where the packet came from so that it
1411 	 * knows how to find the VLAN tag to use, so we attach a
1412 	 * packet tag that holds it.
1413 	 */
1414 	if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1415 	    MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1416 		tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1417 	else
1418 		tag = EVL_MAKETAG(qtag->vid, qtag->pcp, 0);
1419 	if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1420 	    (qtag->proto == ETHERTYPE_VLAN)) {
1421 		(*mp)->m_pkthdr.ether_vtag = tag;
1422 		(*mp)->m_flags |= M_VLANTAG;
1423 	} else {
1424 		*mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1425 		if (*mp == NULL) {
1426 			if_printf(ife, "unable to prepend 802.1Q header");
1427 			return (false);
1428 		}
1429 	}
1430 	return (true);
1431 }
1432 
1433 /*
1434  * Allocate an address from the FreeBSD Foundation OUI.  This uses a
1435  * cryptographic hash function on the containing jail's name, UUID and the
1436  * interface name to attempt to provide a unique but stable address.
1437  * Pseudo-interfaces which require a MAC address should use this function to
1438  * allocate non-locally-administered addresses.
1439  */
1440 void
1441 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1442 {
1443 	SHA1_CTX ctx;
1444 	char *buf;
1445 	char uuid[HOSTUUIDLEN + 1];
1446 	uint64_t addr;
1447 	int i, sz;
1448 	char digest[SHA1_RESULTLEN];
1449 	char jailname[MAXHOSTNAMELEN];
1450 
1451 	getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1452 	if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1453 		/* Fall back to a random mac address. */
1454 		goto rando;
1455 	}
1456 
1457 	/* If each (vnet) jail would also have a unique hostuuid this would not
1458 	 * be necessary. */
1459 	getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1460 	sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp),
1461 	    jailname);
1462 	if (sz < 0) {
1463 		/* Fall back to a random mac address. */
1464 		goto rando;
1465 	}
1466 
1467 	SHA1Init(&ctx);
1468 	SHA1Update(&ctx, buf, sz);
1469 	SHA1Final(digest, &ctx);
1470 	free(buf, M_TEMP);
1471 
1472 	addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1473 	    OUI_FREEBSD_GENERATED_MASK;
1474 	addr = OUI_FREEBSD(addr);
1475 	for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1476 		hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1477 		    0xFF;
1478 	}
1479 
1480 	return;
1481 rando:
1482 	arc4rand(hwaddr, sizeof(*hwaddr), 0);
1483 	/* Unicast */
1484 	hwaddr->octet[0] &= 0xFE;
1485 	/* Locally administered. */
1486 	hwaddr->octet[0] |= 0x02;
1487 }
1488 
1489 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1490 MODULE_VERSION(ether, 1);
1491