xref: /freebsd/sys/net/if_ethersubr.c (revision e5b786625f7f82a1fa91e41823332459ea5550f9)
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  */
33 
34 #include "opt_inet.h"
35 #include "opt_inet6.h"
36 #include "opt_netgraph.h"
37 #include "opt_mbuf_profiling.h"
38 #include "opt_rss.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/devctl.h>
43 #include <sys/eventhandler.h>
44 #include <sys/jail.h>
45 #include <sys/kernel.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/module.h>
50 #include <sys/msan.h>
51 #include <sys/proc.h>
52 #include <sys/priv.h>
53 #include <sys/random.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/uuid.h>
58 #ifdef KDB
59 #include <sys/kdb.h>
60 #endif
61 
62 #include <net/ieee_oui.h>
63 #include <net/if.h>
64 #include <net/if_var.h>
65 #include <net/if_private.h>
66 #include <net/if_arp.h>
67 #include <net/netisr.h>
68 #include <net/route.h>
69 #include <net/if_llc.h>
70 #include <net/if_dl.h>
71 #include <net/if_types.h>
72 #include <net/bpf.h>
73 #include <net/ethernet.h>
74 #include <net/if_bridgevar.h>
75 #include <net/if_vlan_var.h>
76 #include <net/if_llatbl.h>
77 #include <net/pfil.h>
78 #include <net/rss_config.h>
79 #include <net/vnet.h>
80 
81 #include <netpfil/pf/pf_mtag.h>
82 
83 #if defined(INET) || defined(INET6)
84 #include <netinet/in.h>
85 #include <netinet/in_var.h>
86 #include <netinet/if_ether.h>
87 #include <netinet/ip_carp.h>
88 #include <netinet/ip_var.h>
89 #endif
90 #ifdef INET6
91 #include <netinet6/nd6.h>
92 #endif
93 #include <security/mac/mac_framework.h>
94 
95 #include <crypto/sha1.h>
96 
97 #ifdef CTASSERT
98 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
99 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
100 #endif
101 
102 VNET_DEFINE(pfil_head_t, link_pfil_head);	/* Packet filter hooks */
103 
104 /* netgraph node hooks for ng_ether(4) */
105 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
106 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
107 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
108 void	(*ng_ether_attach_p)(struct ifnet *ifp);
109 void	(*ng_ether_detach_p)(struct ifnet *ifp);
110 
111 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
112 
113 /* if_bridge(4) support */
114 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
115 
116 /* if_lagg(4) support */
117 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
118 
119 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
120 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
121 
122 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
123 		struct sockaddr *);
124 static	int ether_requestencap(struct ifnet *, struct if_encap_req *);
125 
126 static inline bool ether_do_pcp(struct ifnet *, struct mbuf *);
127 
128 #define senderr(e) do { error = (e); goto bad;} while (0)
129 
130 static void
131 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
132 {
133 	int csum_flags = 0;
134 
135 	if (src->m_pkthdr.csum_flags & CSUM_IP)
136 		csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
137 	if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
138 		csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
139 	if (src->m_pkthdr.csum_flags & CSUM_SCTP)
140 		csum_flags |= CSUM_SCTP_VALID;
141 	dst->m_pkthdr.csum_flags |= csum_flags;
142 	if (csum_flags & CSUM_DATA_VALID)
143 		dst->m_pkthdr.csum_data = 0xffff;
144 }
145 
146 /*
147  * Handle link-layer encapsulation requests.
148  */
149 static int
150 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
151 {
152 	struct ether_header *eh;
153 	struct arphdr *ah;
154 	uint16_t etype;
155 	const u_char *lladdr;
156 
157 	if (req->rtype != IFENCAP_LL)
158 		return (EOPNOTSUPP);
159 
160 	if (req->bufsize < ETHER_HDR_LEN)
161 		return (ENOMEM);
162 
163 	eh = (struct ether_header *)req->buf;
164 	lladdr = req->lladdr;
165 	req->lladdr_off = 0;
166 
167 	switch (req->family) {
168 	case AF_INET:
169 		etype = htons(ETHERTYPE_IP);
170 		break;
171 	case AF_INET6:
172 		etype = htons(ETHERTYPE_IPV6);
173 		break;
174 	case AF_ARP:
175 		ah = (struct arphdr *)req->hdata;
176 		ah->ar_hrd = htons(ARPHRD_ETHER);
177 
178 		switch(ntohs(ah->ar_op)) {
179 		case ARPOP_REVREQUEST:
180 		case ARPOP_REVREPLY:
181 			etype = htons(ETHERTYPE_REVARP);
182 			break;
183 		case ARPOP_REQUEST:
184 		case ARPOP_REPLY:
185 		default:
186 			etype = htons(ETHERTYPE_ARP);
187 			break;
188 		}
189 
190 		if (req->flags & IFENCAP_FLAG_BROADCAST)
191 			lladdr = ifp->if_broadcastaddr;
192 		break;
193 	default:
194 		return (EAFNOSUPPORT);
195 	}
196 
197 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
198 	memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
199 	memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
200 	req->bufsize = sizeof(struct ether_header);
201 
202 	return (0);
203 }
204 
205 static int
206 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
207 	const struct sockaddr *dst, struct route *ro, u_char *phdr,
208 	uint32_t *pflags, struct llentry **plle)
209 {
210 	uint32_t lleflags = 0;
211 	int error = 0;
212 #if defined(INET) || defined(INET6)
213 	struct ether_header *eh = (struct ether_header *)phdr;
214 	uint16_t etype;
215 #endif
216 
217 	if (plle)
218 		*plle = NULL;
219 
220 	switch (dst->sa_family) {
221 #ifdef INET
222 	case AF_INET:
223 		if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
224 			error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
225 			    plle);
226 		else {
227 			if (m->m_flags & M_BCAST)
228 				memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
229 				    ETHER_ADDR_LEN);
230 			else {
231 				const struct in_addr *a;
232 				a = &(((const struct sockaddr_in *)dst)->sin_addr);
233 				ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
234 			}
235 			etype = htons(ETHERTYPE_IP);
236 			memcpy(&eh->ether_type, &etype, sizeof(etype));
237 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
238 		}
239 		break;
240 #endif
241 #ifdef INET6
242 	case AF_INET6:
243 		if ((m->m_flags & M_MCAST) == 0) {
244 			int af = RO_GET_FAMILY(ro, dst);
245 			error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
246 			    &lleflags, plle);
247 		} else {
248 			const struct in6_addr *a6;
249 			a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
250 			ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
251 			etype = htons(ETHERTYPE_IPV6);
252 			memcpy(&eh->ether_type, &etype, sizeof(etype));
253 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
254 		}
255 		break;
256 #endif
257 	default:
258 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
259 		if (m != NULL)
260 			m_freem(m);
261 		return (EAFNOSUPPORT);
262 	}
263 
264 	if (error == EHOSTDOWN) {
265 		if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
266 			error = EHOSTUNREACH;
267 	}
268 
269 	if (error != 0)
270 		return (error);
271 
272 	*pflags = RT_MAY_LOOP;
273 	if (lleflags & LLE_IFADDR)
274 		*pflags |= RT_L2_ME;
275 
276 	return (0);
277 }
278 
279 /*
280  * Ethernet output routine.
281  * Encapsulate a packet of type family for the local net.
282  * Use trailer local net encapsulation if enough data in first
283  * packet leaves a multiple of 512 bytes of data in remainder.
284  */
285 int
286 ether_output(struct ifnet *ifp, struct mbuf *m,
287 	const struct sockaddr *dst, struct route *ro)
288 {
289 	int error = 0;
290 	char linkhdr[ETHER_HDR_LEN], *phdr;
291 	struct ether_header *eh;
292 	struct pf_mtag *t;
293 	bool loop_copy;
294 	int hlen;	/* link layer header length */
295 	uint32_t pflags;
296 	struct llentry *lle = NULL;
297 	int addref = 0;
298 
299 	phdr = NULL;
300 	pflags = 0;
301 	if (ro != NULL) {
302 		/* XXX BPF uses ro_prepend */
303 		if (ro->ro_prepend != NULL) {
304 			phdr = ro->ro_prepend;
305 			hlen = ro->ro_plen;
306 		} else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
307 			if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
308 				lle = ro->ro_lle;
309 				if (lle != NULL &&
310 				    (lle->la_flags & LLE_VALID) == 0) {
311 					LLE_FREE(lle);
312 					lle = NULL;	/* redundant */
313 					ro->ro_lle = NULL;
314 				}
315 				if (lle == NULL) {
316 					/* if we lookup, keep cache */
317 					addref = 1;
318 				} else
319 					/*
320 					 * Notify LLE code that
321 					 * the entry was used
322 					 * by datapath.
323 					 */
324 					llentry_provide_feedback(lle);
325 			}
326 			if (lle != NULL) {
327 				phdr = lle->r_linkdata;
328 				hlen = lle->r_hdrlen;
329 				pflags = lle->r_flags;
330 			}
331 		}
332 	}
333 
334 #ifdef MAC
335 	error = mac_ifnet_check_transmit(ifp, m);
336 	if (error)
337 		senderr(error);
338 #endif
339 
340 	M_PROFILE(m);
341 	if (ifp->if_flags & IFF_MONITOR)
342 		senderr(ENETDOWN);
343 	if (!((ifp->if_flags & IFF_UP) &&
344 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
345 		senderr(ENETDOWN);
346 
347 	if (phdr == NULL) {
348 		/* No prepend data supplied. Try to calculate ourselves. */
349 		phdr = linkhdr;
350 		hlen = ETHER_HDR_LEN;
351 		error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
352 		    addref ? &lle : NULL);
353 		if (addref && lle != NULL)
354 			ro->ro_lle = lle;
355 		if (error != 0)
356 			return (error == EWOULDBLOCK ? 0 : error);
357 	}
358 
359 	if ((pflags & RT_L2_ME) != 0) {
360 		update_mbuf_csumflags(m, m);
361 		return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
362 	}
363 	loop_copy = (pflags & RT_MAY_LOOP) != 0;
364 
365 	/*
366 	 * Add local net header.  If no space in first mbuf,
367 	 * allocate another.
368 	 *
369 	 * Note that we do prepend regardless of RT_HAS_HEADER flag.
370 	 * This is done because BPF code shifts m_data pointer
371 	 * to the end of ethernet header prior to calling if_output().
372 	 */
373 	M_PREPEND(m, hlen, M_NOWAIT);
374 	if (m == NULL)
375 		senderr(ENOBUFS);
376 	if ((pflags & RT_HAS_HEADER) == 0) {
377 		eh = mtod(m, struct ether_header *);
378 		memcpy(eh, phdr, hlen);
379 	}
380 
381 	/*
382 	 * If a simplex interface, and the packet is being sent to our
383 	 * Ethernet address or a broadcast address, loopback a copy.
384 	 * XXX To make a simplex device behave exactly like a duplex
385 	 * device, we should copy in the case of sending to our own
386 	 * ethernet address (thus letting the original actually appear
387 	 * on the wire). However, we don't do that here for security
388 	 * reasons and compatibility with the original behavior.
389 	 */
390 	if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
391 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
392 		struct mbuf *n;
393 
394 		/*
395 		 * Because if_simloop() modifies the packet, we need a
396 		 * writable copy through m_dup() instead of a readonly
397 		 * one as m_copy[m] would give us. The alternative would
398 		 * be to modify if_simloop() to handle the readonly mbuf,
399 		 * but performancewise it is mostly equivalent (trading
400 		 * extra data copying vs. extra locking).
401 		 *
402 		 * XXX This is a local workaround.  A number of less
403 		 * often used kernel parts suffer from the same bug.
404 		 * See PR kern/105943 for a proposed general solution.
405 		 */
406 		if ((n = m_dup(m, M_NOWAIT)) != NULL) {
407 			update_mbuf_csumflags(m, n);
408 			(void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
409 		} else
410 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
411 	}
412 
413        /*
414 	* Bridges require special output handling.
415 	*/
416 	if (ifp->if_bridge) {
417 		BRIDGE_OUTPUT(ifp, m, error);
418 		return (error);
419 	}
420 
421 #if defined(INET) || defined(INET6)
422 	if (ifp->if_carp &&
423 	    (error = (*carp_output_p)(ifp, m, dst)))
424 		goto bad;
425 #endif
426 
427 	/* Handle ng_ether(4) processing, if any */
428 	if (ifp->if_l2com != NULL) {
429 		KASSERT(ng_ether_output_p != NULL,
430 		    ("ng_ether_output_p is NULL"));
431 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
432 bad:			if (m != NULL)
433 				m_freem(m);
434 			return (error);
435 		}
436 		if (m == NULL)
437 			return (0);
438 	}
439 
440 	/* Continue with link-layer output */
441 	return ether_output_frame(ifp, m);
442 }
443 
444 static bool
445 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
446 {
447 	struct ether_8021q_tag qtag;
448 	struct ether_header *eh;
449 
450 	eh = mtod(*mp, struct ether_header *);
451 	if (eh->ether_type == htons(ETHERTYPE_VLAN) ||
452 	    eh->ether_type == htons(ETHERTYPE_QINQ)) {
453 		(*mp)->m_flags &= ~M_VLANTAG;
454 		return (true);
455 	}
456 
457 	qtag.vid = 0;
458 	qtag.pcp = pcp;
459 	qtag.proto = ETHERTYPE_VLAN;
460 	if (ether_8021q_frame(mp, ifp, ifp, &qtag))
461 		return (true);
462 	if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
463 	return (false);
464 }
465 
466 /*
467  * Ethernet link layer output routine to send a raw frame to the device.
468  *
469  * This assumes that the 14 byte Ethernet header is present and contiguous
470  * in the first mbuf (if BRIDGE'ing).
471  */
472 int
473 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
474 {
475 	if (ether_do_pcp(ifp, m) && !ether_set_pcp(&m, ifp, ifp->if_pcp))
476 		return (0);
477 
478 	if (PFIL_HOOKED_OUT(V_link_pfil_head))
479 		switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, NULL)) {
480 		case PFIL_DROPPED:
481 			return (EACCES);
482 		case PFIL_CONSUMED:
483 			return (0);
484 		}
485 
486 #ifdef EXPERIMENTAL
487 #if defined(INET6) && defined(INET)
488 	/* draft-ietf-6man-ipv6only-flag */
489 	/* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
490 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
491 		struct ether_header *eh;
492 
493 		eh = mtod(m, struct ether_header *);
494 		switch (ntohs(eh->ether_type)) {
495 		case ETHERTYPE_IP:
496 		case ETHERTYPE_ARP:
497 		case ETHERTYPE_REVARP:
498 			m_freem(m);
499 			return (EAFNOSUPPORT);
500 			/* NOTREACHED */
501 			break;
502 		};
503 	}
504 #endif
505 #endif
506 
507 	/*
508 	 * Queue message on interface, update output statistics if successful,
509 	 * and start output if interface not yet active.
510 	 *
511 	 * If KMSAN is enabled, use it to verify that the data does not contain
512 	 * any uninitialized bytes.
513 	 */
514 	kmsan_check_mbuf(m, "ether_output");
515 	return ((ifp->if_transmit)(ifp, m));
516 }
517 
518 /*
519  * Process a received Ethernet packet; the packet is in the
520  * mbuf chain m with the ethernet header at the front.
521  */
522 static void
523 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
524 {
525 	struct ether_header *eh;
526 	u_short etype;
527 
528 	if ((ifp->if_flags & IFF_UP) == 0) {
529 		m_freem(m);
530 		return;
531 	}
532 #ifdef DIAGNOSTIC
533 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
534 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
535 		m_freem(m);
536 		return;
537 	}
538 #endif
539 	if (m->m_len < ETHER_HDR_LEN) {
540 		/* XXX maybe should pullup? */
541 		if_printf(ifp, "discard frame w/o leading ethernet "
542 				"header (len %u pkt len %u)\n",
543 				m->m_len, m->m_pkthdr.len);
544 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
545 		m_freem(m);
546 		return;
547 	}
548 	eh = mtod(m, struct ether_header *);
549 	etype = ntohs(eh->ether_type);
550 	random_harvest_queue_ether(m, sizeof(*m));
551 
552 #ifdef EXPERIMENTAL
553 #if defined(INET6) && defined(INET)
554 	/* draft-ietf-6man-ipv6only-flag */
555 	/* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
556 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
557 		switch (etype) {
558 		case ETHERTYPE_IP:
559 		case ETHERTYPE_ARP:
560 		case ETHERTYPE_REVARP:
561 			m_freem(m);
562 			return;
563 			/* NOTREACHED */
564 			break;
565 		};
566 	}
567 #endif
568 #endif
569 
570 	CURVNET_SET_QUIET(ifp->if_vnet);
571 
572 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
573 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
574 			m->m_flags |= M_BCAST;
575 		else
576 			m->m_flags |= M_MCAST;
577 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
578 	}
579 
580 #ifdef MAC
581 	/*
582 	 * Tag the mbuf with an appropriate MAC label before any other
583 	 * consumers can get to it.
584 	 */
585 	mac_ifnet_create_mbuf(ifp, m);
586 #endif
587 
588 	/*
589 	 * Give bpf a chance at the packet.
590 	 */
591 	ETHER_BPF_MTAP(ifp, m);
592 
593 	/*
594 	 * If the CRC is still on the packet, trim it off. We do this once
595 	 * and once only in case we are re-entered. Nothing else on the
596 	 * Ethernet receive path expects to see the FCS.
597 	 */
598 	if (m->m_flags & M_HASFCS) {
599 		m_adj(m, -ETHER_CRC_LEN);
600 		m->m_flags &= ~M_HASFCS;
601 	}
602 
603 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
604 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
605 
606 	/* Allow monitor mode to claim this frame, after stats are updated. */
607 	if (ifp->if_flags & IFF_MONITOR) {
608 		m_freem(m);
609 		CURVNET_RESTORE();
610 		return;
611 	}
612 
613 	/* Handle input from a lagg(4) port */
614 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
615 		KASSERT(lagg_input_ethernet_p != NULL,
616 		    ("%s: if_lagg not loaded!", __func__));
617 		m = (*lagg_input_ethernet_p)(ifp, m);
618 		if (m != NULL)
619 			ifp = m->m_pkthdr.rcvif;
620 		else {
621 			CURVNET_RESTORE();
622 			return;
623 		}
624 	}
625 
626 	/*
627 	 * If the hardware did not process an 802.1Q tag, do this now,
628 	 * to allow 802.1P priority frames to be passed to the main input
629 	 * path correctly.
630 	 */
631 	if ((m->m_flags & M_VLANTAG) == 0 &&
632 	    ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
633 		struct ether_vlan_header *evl;
634 
635 		if (m->m_len < sizeof(*evl) &&
636 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
637 #ifdef DIAGNOSTIC
638 			if_printf(ifp, "cannot pullup VLAN header\n");
639 #endif
640 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
641 			CURVNET_RESTORE();
642 			return;
643 		}
644 
645 		evl = mtod(m, struct ether_vlan_header *);
646 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
647 		m->m_flags |= M_VLANTAG;
648 
649 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
650 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
651 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
652 		eh = mtod(m, struct ether_header *);
653 	}
654 
655 	M_SETFIB(m, ifp->if_fib);
656 
657 	/* Allow ng_ether(4) to claim this frame. */
658 	if (ifp->if_l2com != NULL) {
659 		KASSERT(ng_ether_input_p != NULL,
660 		    ("%s: ng_ether_input_p is NULL", __func__));
661 		m->m_flags &= ~M_PROMISC;
662 		(*ng_ether_input_p)(ifp, &m);
663 		if (m == NULL) {
664 			CURVNET_RESTORE();
665 			return;
666 		}
667 		eh = mtod(m, struct ether_header *);
668 	}
669 
670 	/*
671 	 * Allow if_bridge(4) to claim this frame.
672 	 *
673 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
674 	 * and the frame should be delivered locally.
675 	 *
676 	 * If M_BRIDGE_INJECT is set, the packet was received directly by the
677 	 * bridge via netmap, so "ifp" is the bridge itself and the packet
678 	 * should be re-examined.
679 	 */
680 	if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) {
681 		m->m_flags &= ~M_PROMISC;
682 		BRIDGE_INPUT(ifp, m);
683 		if (m == NULL) {
684 			CURVNET_RESTORE();
685 			return;
686 		}
687 		eh = mtod(m, struct ether_header *);
688 	}
689 
690 #if defined(INET) || defined(INET6)
691 	/*
692 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
693 	 * mbuf flows up to Layer 3.
694 	 * FreeBSD's implementation of carp(4) uses the inprotosw
695 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
696 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
697 	 * is outside the scope of the M_PROMISC test below.
698 	 * TODO: Maintain a hash table of ethernet addresses other than
699 	 * ether_dhost which may be active on this ifp.
700 	 */
701 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
702 		m->m_flags &= ~M_PROMISC;
703 	} else
704 #endif
705 	{
706 		/*
707 		 * If the frame received was not for our MAC address, set the
708 		 * M_PROMISC flag on the mbuf chain. The frame may need to
709 		 * be seen by the rest of the Ethernet input path in case of
710 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
711 		 * seen by upper protocol layers.
712 		 */
713 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
714 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
715 			m->m_flags |= M_PROMISC;
716 	}
717 
718 	ether_demux(ifp, m);
719 	CURVNET_RESTORE();
720 }
721 
722 /*
723  * Ethernet input dispatch; by default, direct dispatch here regardless of
724  * global configuration.  However, if RSS is enabled, hook up RSS affinity
725  * so that when deferred or hybrid dispatch is enabled, we can redistribute
726  * load based on RSS.
727  *
728  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
729  * not it had already done work distribution via multi-queue.  Then we could
730  * direct dispatch in the event load balancing was already complete and
731  * handle the case of interfaces with different capabilities better.
732  *
733  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
734  * at multiple layers?
735  *
736  * XXXRW: For now, enable all this only if RSS is compiled in, although it
737  * works fine without RSS.  Need to characterise the performance overhead
738  * of the detour through the netisr code in the event the result is always
739  * direct dispatch.
740  */
741 static void
742 ether_nh_input(struct mbuf *m)
743 {
744 
745 	M_ASSERTPKTHDR(m);
746 	KASSERT(m->m_pkthdr.rcvif != NULL,
747 	    ("%s: NULL interface pointer", __func__));
748 	ether_input_internal(m->m_pkthdr.rcvif, m);
749 }
750 
751 static struct netisr_handler	ether_nh = {
752 	.nh_name = "ether",
753 	.nh_handler = ether_nh_input,
754 	.nh_proto = NETISR_ETHER,
755 #ifdef RSS
756 	.nh_policy = NETISR_POLICY_CPU,
757 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
758 	.nh_m2cpuid = rss_m2cpuid,
759 #else
760 	.nh_policy = NETISR_POLICY_SOURCE,
761 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
762 #endif
763 };
764 
765 static void
766 ether_init(__unused void *arg)
767 {
768 
769 	netisr_register(&ether_nh);
770 }
771 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
772 
773 static void
774 vnet_ether_init(__unused void *arg)
775 {
776 	struct pfil_head_args args;
777 
778 	args.pa_version = PFIL_VERSION;
779 	args.pa_flags = PFIL_IN | PFIL_OUT;
780 	args.pa_type = PFIL_TYPE_ETHERNET;
781 	args.pa_headname = PFIL_ETHER_NAME;
782 	V_link_pfil_head = pfil_head_register(&args);
783 
784 #ifdef VIMAGE
785 	netisr_register_vnet(&ether_nh);
786 #endif
787 }
788 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
789     vnet_ether_init, NULL);
790 
791 #ifdef VIMAGE
792 static void
793 vnet_ether_pfil_destroy(__unused void *arg)
794 {
795 
796 	pfil_head_unregister(V_link_pfil_head);
797 }
798 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
799     vnet_ether_pfil_destroy, NULL);
800 
801 static void
802 vnet_ether_destroy(__unused void *arg)
803 {
804 
805 	netisr_unregister_vnet(&ether_nh);
806 }
807 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
808     vnet_ether_destroy, NULL);
809 #endif
810 
811 static void
812 ether_input(struct ifnet *ifp, struct mbuf *m)
813 {
814 	struct epoch_tracker et;
815 	struct mbuf *mn;
816 	bool needs_epoch;
817 
818 	needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH);
819 #ifdef INVARIANTS
820 	/*
821 	 * This temporary code is here to prevent epoch unaware and unmarked
822 	 * drivers to panic the system.  Once all drivers are taken care of,
823 	 * the whole INVARIANTS block should go away.
824 	 */
825 	if (!needs_epoch && !in_epoch(net_epoch_preempt)) {
826 		static bool printedonce;
827 
828 		needs_epoch = true;
829 		if (!printedonce) {
830 			printedonce = true;
831 			if_printf(ifp, "called %s w/o net epoch! "
832 			    "PLEASE file a bug report.", __func__);
833 #ifdef KDB
834 			kdb_backtrace();
835 #endif
836 		}
837 	}
838 #endif
839 
840 	/*
841 	 * The drivers are allowed to pass in a chain of packets linked with
842 	 * m_nextpkt. We split them up into separate packets here and pass
843 	 * them up. This allows the drivers to amortize the receive lock.
844 	 */
845 	CURVNET_SET_QUIET(ifp->if_vnet);
846 	if (__predict_false(needs_epoch))
847 		NET_EPOCH_ENTER(et);
848 	while (m) {
849 		mn = m->m_nextpkt;
850 		m->m_nextpkt = NULL;
851 
852 		/*
853 		 * We will rely on rcvif being set properly in the deferred
854 		 * context, so assert it is correct here.
855 		 */
856 		MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
857 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
858 		    "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
859 		netisr_dispatch(NETISR_ETHER, m);
860 		m = mn;
861 	}
862 	if (__predict_false(needs_epoch))
863 		NET_EPOCH_EXIT(et);
864 	CURVNET_RESTORE();
865 }
866 
867 /*
868  * Upper layer processing for a received Ethernet packet.
869  */
870 void
871 ether_demux(struct ifnet *ifp, struct mbuf *m)
872 {
873 	struct ether_header *eh;
874 	int i, isr;
875 	u_short ether_type;
876 
877 	NET_EPOCH_ASSERT();
878 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
879 
880 	/* Do not grab PROMISC frames in case we are re-entered. */
881 	if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
882 		i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL);
883 		if (i != 0 || m == NULL)
884 			return;
885 	}
886 
887 	eh = mtod(m, struct ether_header *);
888 	ether_type = ntohs(eh->ether_type);
889 
890 	/*
891 	 * If this frame has a VLAN tag other than 0, call vlan_input()
892 	 * if its module is loaded. Otherwise, drop.
893 	 */
894 	if ((m->m_flags & M_VLANTAG) &&
895 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
896 		if (ifp->if_vlantrunk == NULL) {
897 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
898 			m_freem(m);
899 			return;
900 		}
901 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
902 		    __func__));
903 		/* Clear before possibly re-entering ether_input(). */
904 		m->m_flags &= ~M_PROMISC;
905 		(*vlan_input_p)(ifp, m);
906 		return;
907 	}
908 
909 	/*
910 	 * Pass promiscuously received frames to the upper layer if the user
911 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
912 	 */
913 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
914 		m_freem(m);
915 		return;
916 	}
917 
918 	/*
919 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
920 	 */
921 	m->m_flags &= ~M_VLANTAG;
922 	m_clrprotoflags(m);
923 
924 	/*
925 	 * Dispatch frame to upper layer.
926 	 */
927 	switch (ether_type) {
928 #ifdef INET
929 	case ETHERTYPE_IP:
930 		isr = NETISR_IP;
931 		break;
932 
933 	case ETHERTYPE_ARP:
934 		if (ifp->if_flags & IFF_NOARP) {
935 			/* Discard packet if ARP is disabled on interface */
936 			m_freem(m);
937 			return;
938 		}
939 		isr = NETISR_ARP;
940 		break;
941 #endif
942 #ifdef INET6
943 	case ETHERTYPE_IPV6:
944 		isr = NETISR_IPV6;
945 		break;
946 #endif
947 	default:
948 		goto discard;
949 	}
950 
951 	/* Strip off Ethernet header. */
952 	m_adj(m, ETHER_HDR_LEN);
953 
954 	netisr_dispatch(isr, m);
955 	return;
956 
957 discard:
958 	/*
959 	 * Packet is to be discarded.  If netgraph is present,
960 	 * hand the packet to it for last chance processing;
961 	 * otherwise dispose of it.
962 	 */
963 	if (ifp->if_l2com != NULL) {
964 		KASSERT(ng_ether_input_orphan_p != NULL,
965 		    ("ng_ether_input_orphan_p is NULL"));
966 		(*ng_ether_input_orphan_p)(ifp, m);
967 		return;
968 	}
969 	m_freem(m);
970 }
971 
972 /*
973  * Convert Ethernet address to printable (loggable) representation.
974  * This routine is for compatibility; it's better to just use
975  *
976  *	printf("%6D", <pointer to address>, ":");
977  *
978  * since there's no static buffer involved.
979  */
980 char *
981 ether_sprintf(const u_char *ap)
982 {
983 	static char etherbuf[18];
984 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
985 	return (etherbuf);
986 }
987 
988 /*
989  * Perform common duties while attaching to interface list
990  */
991 void
992 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
993 {
994 	int i;
995 	struct ifaddr *ifa;
996 	struct sockaddr_dl *sdl;
997 
998 	ifp->if_addrlen = ETHER_ADDR_LEN;
999 	ifp->if_hdrlen = ETHER_HDR_LEN;
1000 	ifp->if_mtu = ETHERMTU;
1001 	if_attach(ifp);
1002 	ifp->if_output = ether_output;
1003 	ifp->if_input = ether_input;
1004 	ifp->if_resolvemulti = ether_resolvemulti;
1005 	ifp->if_requestencap = ether_requestencap;
1006 #ifdef VIMAGE
1007 	ifp->if_reassign = ether_reassign;
1008 #endif
1009 	if (ifp->if_baudrate == 0)
1010 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
1011 	ifp->if_broadcastaddr = etherbroadcastaddr;
1012 
1013 	ifa = ifp->if_addr;
1014 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
1015 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1016 	sdl->sdl_type = IFT_ETHER;
1017 	sdl->sdl_alen = ifp->if_addrlen;
1018 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
1019 
1020 	if (ifp->if_hw_addr != NULL)
1021 		bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
1022 
1023 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1024 	if (ng_ether_attach_p != NULL)
1025 		(*ng_ether_attach_p)(ifp);
1026 
1027 	/* Announce Ethernet MAC address if non-zero. */
1028 	for (i = 0; i < ifp->if_addrlen; i++)
1029 		if (lla[i] != 0)
1030 			break;
1031 	if (i != ifp->if_addrlen)
1032 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1033 
1034 	uuid_ether_add(LLADDR(sdl));
1035 
1036 	/* Add necessary bits are setup; announce it now. */
1037 	EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1038 	if (IS_DEFAULT_VNET(curvnet))
1039 		devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1040 }
1041 
1042 /*
1043  * Perform common duties while detaching an Ethernet interface
1044  */
1045 void
1046 ether_ifdetach(struct ifnet *ifp)
1047 {
1048 	struct sockaddr_dl *sdl;
1049 
1050 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1051 	uuid_ether_del(LLADDR(sdl));
1052 
1053 	if (ifp->if_l2com != NULL) {
1054 		KASSERT(ng_ether_detach_p != NULL,
1055 		    ("ng_ether_detach_p is NULL"));
1056 		(*ng_ether_detach_p)(ifp);
1057 	}
1058 
1059 	bpfdetach(ifp);
1060 	if_detach(ifp);
1061 }
1062 
1063 #ifdef VIMAGE
1064 void
1065 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1066 {
1067 
1068 	if (ifp->if_l2com != NULL) {
1069 		KASSERT(ng_ether_detach_p != NULL,
1070 		    ("ng_ether_detach_p is NULL"));
1071 		(*ng_ether_detach_p)(ifp);
1072 	}
1073 
1074 	if (ng_ether_attach_p != NULL) {
1075 		CURVNET_SET_QUIET(new_vnet);
1076 		(*ng_ether_attach_p)(ifp);
1077 		CURVNET_RESTORE();
1078 	}
1079 }
1080 #endif
1081 
1082 SYSCTL_DECL(_net_link);
1083 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1084     "Ethernet");
1085 
1086 #if 0
1087 /*
1088  * This is for reference.  We have a table-driven version
1089  * of the little-endian crc32 generator, which is faster
1090  * than the double-loop.
1091  */
1092 uint32_t
1093 ether_crc32_le(const uint8_t *buf, size_t len)
1094 {
1095 	size_t i;
1096 	uint32_t crc;
1097 	int bit;
1098 	uint8_t data;
1099 
1100 	crc = 0xffffffff;	/* initial value */
1101 
1102 	for (i = 0; i < len; i++) {
1103 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1104 			carry = (crc ^ data) & 1;
1105 			crc >>= 1;
1106 			if (carry)
1107 				crc = (crc ^ ETHER_CRC_POLY_LE);
1108 		}
1109 	}
1110 
1111 	return (crc);
1112 }
1113 #else
1114 uint32_t
1115 ether_crc32_le(const uint8_t *buf, size_t len)
1116 {
1117 	static const uint32_t crctab[] = {
1118 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1119 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1120 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1121 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1122 	};
1123 	size_t i;
1124 	uint32_t crc;
1125 
1126 	crc = 0xffffffff;	/* initial value */
1127 
1128 	for (i = 0; i < len; i++) {
1129 		crc ^= buf[i];
1130 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1131 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1132 	}
1133 
1134 	return (crc);
1135 }
1136 #endif
1137 
1138 uint32_t
1139 ether_crc32_be(const uint8_t *buf, size_t len)
1140 {
1141 	size_t i;
1142 	uint32_t crc, carry;
1143 	int bit;
1144 	uint8_t data;
1145 
1146 	crc = 0xffffffff;	/* initial value */
1147 
1148 	for (i = 0; i < len; i++) {
1149 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1150 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1151 			crc <<= 1;
1152 			if (carry)
1153 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1154 		}
1155 	}
1156 
1157 	return (crc);
1158 }
1159 
1160 int
1161 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1162 {
1163 	struct ifaddr *ifa = (struct ifaddr *) data;
1164 	struct ifreq *ifr = (struct ifreq *) data;
1165 	int error = 0;
1166 
1167 	switch (command) {
1168 	case SIOCSIFADDR:
1169 		ifp->if_flags |= IFF_UP;
1170 
1171 		switch (ifa->ifa_addr->sa_family) {
1172 #ifdef INET
1173 		case AF_INET:
1174 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1175 			arp_ifinit(ifp, ifa);
1176 			break;
1177 #endif
1178 		default:
1179 			ifp->if_init(ifp->if_softc);
1180 			break;
1181 		}
1182 		break;
1183 
1184 	case SIOCGIFADDR:
1185 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1186 		    ETHER_ADDR_LEN);
1187 		break;
1188 
1189 	case SIOCSIFMTU:
1190 		/*
1191 		 * Set the interface MTU.
1192 		 */
1193 		if (ifr->ifr_mtu > ETHERMTU) {
1194 			error = EINVAL;
1195 		} else {
1196 			ifp->if_mtu = ifr->ifr_mtu;
1197 		}
1198 		break;
1199 
1200 	case SIOCSLANPCP:
1201 		error = priv_check(curthread, PRIV_NET_SETLANPCP);
1202 		if (error != 0)
1203 			break;
1204 		if (ifr->ifr_lan_pcp > 7 &&
1205 		    ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1206 			error = EINVAL;
1207 		} else {
1208 			ifp->if_pcp = ifr->ifr_lan_pcp;
1209 			/* broadcast event about PCP change */
1210 			EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1211 		}
1212 		break;
1213 
1214 	case SIOCGLANPCP:
1215 		ifr->ifr_lan_pcp = ifp->if_pcp;
1216 		break;
1217 
1218 	default:
1219 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1220 		break;
1221 	}
1222 	return (error);
1223 }
1224 
1225 static int
1226 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1227 	struct sockaddr *sa)
1228 {
1229 	struct sockaddr_dl *sdl;
1230 #ifdef INET
1231 	struct sockaddr_in *sin;
1232 #endif
1233 #ifdef INET6
1234 	struct sockaddr_in6 *sin6;
1235 #endif
1236 	u_char *e_addr;
1237 
1238 	switch(sa->sa_family) {
1239 	case AF_LINK:
1240 		/*
1241 		 * No mapping needed. Just check that it's a valid MC address.
1242 		 */
1243 		sdl = (struct sockaddr_dl *)sa;
1244 		e_addr = LLADDR(sdl);
1245 		if (!ETHER_IS_MULTICAST(e_addr))
1246 			return EADDRNOTAVAIL;
1247 		*llsa = NULL;
1248 		return 0;
1249 
1250 #ifdef INET
1251 	case AF_INET:
1252 		sin = (struct sockaddr_in *)sa;
1253 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1254 			return EADDRNOTAVAIL;
1255 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1256 		sdl->sdl_alen = ETHER_ADDR_LEN;
1257 		e_addr = LLADDR(sdl);
1258 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1259 		*llsa = (struct sockaddr *)sdl;
1260 		return 0;
1261 #endif
1262 #ifdef INET6
1263 	case AF_INET6:
1264 		sin6 = (struct sockaddr_in6 *)sa;
1265 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1266 			/*
1267 			 * An IP6 address of 0 means listen to all
1268 			 * of the Ethernet multicast address used for IP6.
1269 			 * (This is used for multicast routers.)
1270 			 */
1271 			ifp->if_flags |= IFF_ALLMULTI;
1272 			*llsa = NULL;
1273 			return 0;
1274 		}
1275 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1276 			return EADDRNOTAVAIL;
1277 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1278 		sdl->sdl_alen = ETHER_ADDR_LEN;
1279 		e_addr = LLADDR(sdl);
1280 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1281 		*llsa = (struct sockaddr *)sdl;
1282 		return 0;
1283 #endif
1284 
1285 	default:
1286 		/*
1287 		 * Well, the text isn't quite right, but it's the name
1288 		 * that counts...
1289 		 */
1290 		return EAFNOSUPPORT;
1291 	}
1292 }
1293 
1294 static moduledata_t ether_mod = {
1295 	.name = "ether",
1296 };
1297 
1298 void
1299 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1300 {
1301 	struct ether_vlan_header vlan;
1302 	struct mbuf mv, mb;
1303 
1304 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1305 	    ("%s: vlan information not present", __func__));
1306 	KASSERT(m->m_len >= sizeof(struct ether_header),
1307 	    ("%s: mbuf not large enough for header", __func__));
1308 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1309 	vlan.evl_proto = vlan.evl_encap_proto;
1310 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1311 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1312 	m->m_len -= sizeof(struct ether_header);
1313 	m->m_data += sizeof(struct ether_header);
1314 	/*
1315 	 * If a data link has been supplied by the caller, then we will need to
1316 	 * re-create a stack allocated mbuf chain with the following structure:
1317 	 *
1318 	 * (1) mbuf #1 will contain the supplied data link
1319 	 * (2) mbuf #2 will contain the vlan header
1320 	 * (3) mbuf #3 will contain the original mbuf's packet data
1321 	 *
1322 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1323 	 */
1324 	if (data != NULL) {
1325 		mv.m_next = m;
1326 		mv.m_data = (caddr_t)&vlan;
1327 		mv.m_len = sizeof(vlan);
1328 		mb.m_next = &mv;
1329 		mb.m_data = data;
1330 		mb.m_len = dlen;
1331 		bpf_mtap(bp, &mb);
1332 	} else
1333 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1334 	m->m_len += sizeof(struct ether_header);
1335 	m->m_data -= sizeof(struct ether_header);
1336 }
1337 
1338 struct mbuf *
1339 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1340 {
1341 	struct ether_vlan_header *evl;
1342 
1343 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1344 	if (m == NULL)
1345 		return (NULL);
1346 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1347 
1348 	if (m->m_len < sizeof(*evl)) {
1349 		m = m_pullup(m, sizeof(*evl));
1350 		if (m == NULL)
1351 			return (NULL);
1352 	}
1353 
1354 	/*
1355 	 * Transform the Ethernet header into an Ethernet header
1356 	 * with 802.1Q encapsulation.
1357 	 */
1358 	evl = mtod(m, struct ether_vlan_header *);
1359 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1360 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1361 	evl->evl_encap_proto = htons(proto);
1362 	evl->evl_tag = htons(tag);
1363 	return (m);
1364 }
1365 
1366 void
1367 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m)
1368 {
1369 	if (bpf_peers_present(ifp->if_bpf)) {
1370 		M_ASSERTVALID(m);
1371 		if ((m->m_flags & M_VLANTAG) != 0)
1372 			ether_vlan_mtap(ifp->if_bpf, m, NULL, 0);
1373 		else
1374 			bpf_mtap(ifp->if_bpf, m);
1375 	}
1376 }
1377 
1378 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1379     "IEEE 802.1Q VLAN");
1380 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1381     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1382     "for consistency");
1383 
1384 VNET_DEFINE_STATIC(int, soft_pad);
1385 #define	V_soft_pad	VNET(soft_pad)
1386 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1387     &VNET_NAME(soft_pad), 0,
1388     "pad short frames before tagging");
1389 
1390 /*
1391  * For now, make preserving PCP via an mbuf tag optional, as it increases
1392  * per-packet memory allocations and frees.  In the future, it would be
1393  * preferable to reuse ether_vtag for this, or similar.
1394  */
1395 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1396 #define	V_vlan_mtag_pcp	VNET(vlan_mtag_pcp)
1397 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1398     &VNET_NAME(vlan_mtag_pcp), 0,
1399     "Retain VLAN PCP information as packets are passed up the stack");
1400 
1401 static inline bool
1402 ether_do_pcp(struct ifnet *ifp, struct mbuf *m)
1403 {
1404 	if (ifp->if_type == IFT_L2VLAN)
1405 		return (false);
1406 	if (ifp->if_pcp != IFNET_PCP_NONE || (m->m_flags & M_VLANTAG) != 0)
1407 		return (true);
1408 	if (V_vlan_mtag_pcp &&
1409 	    m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_OUT, NULL) != NULL)
1410 		return (true);
1411 	return (false);
1412 }
1413 
1414 bool
1415 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1416     const struct ether_8021q_tag *qtag)
1417 {
1418 	struct m_tag *mtag;
1419 	int n;
1420 	uint16_t tag;
1421 	uint8_t pcp = qtag->pcp;
1422 	static const char pad[8];	/* just zeros */
1423 
1424 	/*
1425 	 * Pad the frame to the minimum size allowed if told to.
1426 	 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1427 	 * paragraph C.4.4.3.b.  It can help to work around buggy
1428 	 * bridges that violate paragraph C.4.4.3.a from the same
1429 	 * document, i.e., fail to pad short frames after untagging.
1430 	 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1431 	 * untagging it will produce a 62-byte frame, which is a runt
1432 	 * and requires padding.  There are VLAN-enabled network
1433 	 * devices that just discard such runts instead or mishandle
1434 	 * them somehow.
1435 	 */
1436 	if (V_soft_pad && p->if_type == IFT_ETHER) {
1437 		for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1438 		     n > 0; n -= sizeof(pad)) {
1439 			if (!m_append(*mp, min(n, sizeof(pad)), pad))
1440 				break;
1441 		}
1442 		if (n > 0) {
1443 			m_freem(*mp);
1444 			*mp = NULL;
1445 			if_printf(ife, "cannot pad short frame");
1446 			return (false);
1447 		}
1448 	}
1449 
1450 	/*
1451 	 * If PCP is set in mbuf, use it
1452 	 */
1453 	if ((*mp)->m_flags & M_VLANTAG) {
1454 		pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1455 	}
1456 
1457 	/*
1458 	 * If underlying interface can do VLAN tag insertion itself,
1459 	 * just pass the packet along. However, we need some way to
1460 	 * tell the interface where the packet came from so that it
1461 	 * knows how to find the VLAN tag to use, so we attach a
1462 	 * packet tag that holds it.
1463 	 */
1464 	if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1465 	    MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1466 		tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1467 	else
1468 		tag = EVL_MAKETAG(qtag->vid, pcp, 0);
1469 	if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1470 	    (qtag->proto == ETHERTYPE_VLAN)) {
1471 		(*mp)->m_pkthdr.ether_vtag = tag;
1472 		(*mp)->m_flags |= M_VLANTAG;
1473 	} else {
1474 		*mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1475 		if (*mp == NULL) {
1476 			if_printf(ife, "unable to prepend 802.1Q header");
1477 			return (false);
1478 		}
1479 		(*mp)->m_flags &= ~M_VLANTAG;
1480 	}
1481 	return (true);
1482 }
1483 
1484 /*
1485  * Allocate an address from the FreeBSD Foundation OUI.  This uses a
1486  * cryptographic hash function on the containing jail's name, UUID and the
1487  * interface name to attempt to provide a unique but stable address.
1488  * Pseudo-interfaces which require a MAC address should use this function to
1489  * allocate non-locally-administered addresses.
1490  */
1491 void
1492 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1493 {
1494 	SHA1_CTX ctx;
1495 	char *buf;
1496 	char uuid[HOSTUUIDLEN + 1];
1497 	uint64_t addr;
1498 	int i, sz;
1499 	char digest[SHA1_RESULTLEN];
1500 	char jailname[MAXHOSTNAMELEN];
1501 
1502 	getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1503 	if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1504 		/* Fall back to a random mac address. */
1505 		goto rando;
1506 	}
1507 
1508 	/* If each (vnet) jail would also have a unique hostuuid this would not
1509 	 * be necessary. */
1510 	getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1511 	sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, if_name(ifp),
1512 	    jailname);
1513 	if (sz < 0) {
1514 		/* Fall back to a random mac address. */
1515 		goto rando;
1516 	}
1517 
1518 	SHA1Init(&ctx);
1519 	SHA1Update(&ctx, buf, sz);
1520 	SHA1Final(digest, &ctx);
1521 	free(buf, M_TEMP);
1522 
1523 	addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1524 	    OUI_FREEBSD_GENERATED_MASK;
1525 	addr = OUI_FREEBSD(addr);
1526 	for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1527 		hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1528 		    0xFF;
1529 	}
1530 
1531 	return;
1532 rando:
1533 	arc4rand(hwaddr, sizeof(*hwaddr), 0);
1534 	/* Unicast */
1535 	hwaddr->octet[0] &= 0xFE;
1536 	/* Locally administered. */
1537 	hwaddr->octet[0] |= 0x02;
1538 }
1539 
1540 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1541 MODULE_VERSION(ether, 1);
1542