xref: /freebsd/sys/net/if_ethersubr.c (revision 09164454aa4c1868abf1ea724ec6160c4a556bdd)
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 
32 #include "opt_inet.h"
33 #include "opt_inet6.h"
34 #include "opt_netgraph.h"
35 #include "opt_mbuf_profiling.h"
36 #include "opt_rss.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/devctl.h>
41 #include <sys/eventhandler.h>
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/module.h>
48 #include <sys/msan.h>
49 #include <sys/proc.h>
50 #include <sys/priv.h>
51 #include <sys/random.h>
52 #include <sys/socket.h>
53 #include <sys/sockio.h>
54 #include <sys/sysctl.h>
55 #include <sys/uuid.h>
56 #ifdef KDB
57 #include <sys/kdb.h>
58 #endif
59 
60 #include <net/ieee_oui.h>
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/if_private.h>
64 #include <net/if_arp.h>
65 #include <net/netisr.h>
66 #include <net/route.h>
67 #include <net/if_llc.h>
68 #include <net/if_dl.h>
69 #include <net/if_types.h>
70 #include <net/bpf.h>
71 #include <net/ethernet.h>
72 #include <net/if_bridgevar.h>
73 #include <net/if_vlan_var.h>
74 #include <net/if_llatbl.h>
75 #include <net/pfil.h>
76 #include <net/rss_config.h>
77 #include <net/vnet.h>
78 
79 #include <netpfil/pf/pf_mtag.h>
80 
81 #if defined(INET) || defined(INET6)
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/if_ether.h>
85 #include <netinet/ip_carp.h>
86 #include <netinet/ip_var.h>
87 #endif
88 #ifdef INET6
89 #include <netinet6/nd6.h>
90 #endif
91 #include <security/mac/mac_framework.h>
92 
93 #include <crypto/sha1.h>
94 
95 #ifdef CTASSERT
96 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
97 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
98 #endif
99 
100 VNET_DEFINE(pfil_head_t, link_pfil_head);	/* Packet filter hooks */
101 
102 /* netgraph node hooks for ng_ether(4) */
103 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
104 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
105 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
106 void	(*ng_ether_attach_p)(struct ifnet *ifp);
107 void	(*ng_ether_detach_p)(struct ifnet *ifp);
108 
109 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
110 
111 /* if_bridge(4) support */
112 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
113 
114 /* if_lagg(4) support */
115 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
116 
117 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
118 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
119 
120 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
121 		struct sockaddr *);
122 static	int ether_requestencap(struct ifnet *, struct if_encap_req *);
123 
124 static inline bool ether_do_pcp(struct ifnet *, struct mbuf *);
125 
126 #define senderr(e) do { error = (e); goto bad;} while (0)
127 
128 static void
129 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
130 {
131 	int csum_flags = 0;
132 
133 	if (src->m_pkthdr.csum_flags & CSUM_IP)
134 		csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
135 	if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
136 		csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
137 	if (src->m_pkthdr.csum_flags & CSUM_SCTP)
138 		csum_flags |= CSUM_SCTP_VALID;
139 	dst->m_pkthdr.csum_flags |= csum_flags;
140 	if (csum_flags & CSUM_DATA_VALID)
141 		dst->m_pkthdr.csum_data = 0xffff;
142 }
143 
144 /*
145  * Handle link-layer encapsulation requests.
146  */
147 static int
148 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
149 {
150 	struct ether_header *eh;
151 	struct arphdr *ah;
152 	uint16_t etype;
153 	const u_char *lladdr;
154 
155 	if (req->rtype != IFENCAP_LL)
156 		return (EOPNOTSUPP);
157 
158 	if (req->bufsize < ETHER_HDR_LEN)
159 		return (ENOMEM);
160 
161 	eh = (struct ether_header *)req->buf;
162 	lladdr = req->lladdr;
163 	req->lladdr_off = 0;
164 
165 	switch (req->family) {
166 	case AF_INET:
167 		etype = htons(ETHERTYPE_IP);
168 		break;
169 	case AF_INET6:
170 		etype = htons(ETHERTYPE_IPV6);
171 		break;
172 	case AF_ARP:
173 		ah = (struct arphdr *)req->hdata;
174 		ah->ar_hrd = htons(ARPHRD_ETHER);
175 
176 		switch(ntohs(ah->ar_op)) {
177 		case ARPOP_REVREQUEST:
178 		case ARPOP_REVREPLY:
179 			etype = htons(ETHERTYPE_REVARP);
180 			break;
181 		case ARPOP_REQUEST:
182 		case ARPOP_REPLY:
183 		default:
184 			etype = htons(ETHERTYPE_ARP);
185 			break;
186 		}
187 
188 		if (req->flags & IFENCAP_FLAG_BROADCAST)
189 			lladdr = ifp->if_broadcastaddr;
190 		break;
191 	default:
192 		return (EAFNOSUPPORT);
193 	}
194 
195 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
196 	memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
197 	memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
198 	req->bufsize = sizeof(struct ether_header);
199 
200 	return (0);
201 }
202 
203 static int
204 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
205 	const struct sockaddr *dst, struct route *ro, u_char *phdr,
206 	uint32_t *pflags, struct llentry **plle)
207 {
208 	uint32_t lleflags = 0;
209 	int error = 0;
210 #if defined(INET) || defined(INET6)
211 	struct ether_header *eh = (struct ether_header *)phdr;
212 	uint16_t etype;
213 #endif
214 
215 	if (plle)
216 		*plle = NULL;
217 
218 	switch (dst->sa_family) {
219 #ifdef INET
220 	case AF_INET:
221 		if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
222 			error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
223 			    plle);
224 		else {
225 			if (m->m_flags & M_BCAST)
226 				memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
227 				    ETHER_ADDR_LEN);
228 			else {
229 				const struct in_addr *a;
230 				a = &(((const struct sockaddr_in *)dst)->sin_addr);
231 				ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
232 			}
233 			etype = htons(ETHERTYPE_IP);
234 			memcpy(&eh->ether_type, &etype, sizeof(etype));
235 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
236 		}
237 		break;
238 #endif
239 #ifdef INET6
240 	case AF_INET6:
241 		if ((m->m_flags & M_MCAST) == 0) {
242 			int af = RO_GET_FAMILY(ro, dst);
243 			error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
244 			    &lleflags, plle);
245 		} else {
246 			const struct in6_addr *a6;
247 			a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
248 			ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
249 			etype = htons(ETHERTYPE_IPV6);
250 			memcpy(&eh->ether_type, &etype, sizeof(etype));
251 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
252 		}
253 		break;
254 #endif
255 	default:
256 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
257 		if (m != NULL)
258 			m_freem(m);
259 		return (EAFNOSUPPORT);
260 	}
261 
262 	if (error == EHOSTDOWN) {
263 		if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
264 			error = EHOSTUNREACH;
265 	}
266 
267 	if (error != 0)
268 		return (error);
269 
270 	*pflags = RT_MAY_LOOP;
271 	if (lleflags & LLE_IFADDR)
272 		*pflags |= RT_L2_ME;
273 
274 	return (0);
275 }
276 
277 /*
278  * Ethernet output routine.
279  * Encapsulate a packet of type family for the local net.
280  * Use trailer local net encapsulation if enough data in first
281  * packet leaves a multiple of 512 bytes of data in remainder.
282  */
283 int
284 ether_output(struct ifnet *ifp, struct mbuf *m,
285 	const struct sockaddr *dst, struct route *ro)
286 {
287 	int error = 0;
288 	char linkhdr[ETHER_HDR_LEN], *phdr;
289 	struct ether_header *eh;
290 	struct pf_mtag *t;
291 	bool loop_copy;
292 	int hlen;	/* link layer header length */
293 	uint32_t pflags;
294 	struct llentry *lle = NULL;
295 	int addref = 0;
296 
297 	phdr = NULL;
298 	pflags = 0;
299 	if (ro != NULL) {
300 		/* XXX BPF uses ro_prepend */
301 		if (ro->ro_prepend != NULL) {
302 			phdr = ro->ro_prepend;
303 			hlen = ro->ro_plen;
304 		} else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
305 			if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
306 				lle = ro->ro_lle;
307 				if (lle != NULL &&
308 				    (lle->la_flags & LLE_VALID) == 0) {
309 					LLE_FREE(lle);
310 					lle = NULL;	/* redundant */
311 					ro->ro_lle = NULL;
312 				}
313 				if (lle == NULL) {
314 					/* if we lookup, keep cache */
315 					addref = 1;
316 				} else
317 					/*
318 					 * Notify LLE code that
319 					 * the entry was used
320 					 * by datapath.
321 					 */
322 					llentry_provide_feedback(lle);
323 			}
324 			if (lle != NULL) {
325 				phdr = lle->r_linkdata;
326 				hlen = lle->r_hdrlen;
327 				pflags = lle->r_flags;
328 			}
329 		}
330 	}
331 
332 #ifdef MAC
333 	error = mac_ifnet_check_transmit(ifp, m);
334 	if (error)
335 		senderr(error);
336 #endif
337 
338 	M_PROFILE(m);
339 	if (ifp->if_flags & IFF_MONITOR)
340 		senderr(ENETDOWN);
341 	if (!((ifp->if_flags & IFF_UP) &&
342 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
343 		senderr(ENETDOWN);
344 
345 	if (phdr == NULL) {
346 		/* No prepend data supplied. Try to calculate ourselves. */
347 		phdr = linkhdr;
348 		hlen = ETHER_HDR_LEN;
349 		error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
350 		    addref ? &lle : NULL);
351 		if (addref && lle != NULL)
352 			ro->ro_lle = lle;
353 		if (error != 0)
354 			return (error == EWOULDBLOCK ? 0 : error);
355 	}
356 
357 	if ((pflags & RT_L2_ME) != 0) {
358 		update_mbuf_csumflags(m, m);
359 		return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
360 	}
361 	loop_copy = (pflags & RT_MAY_LOOP) != 0;
362 
363 	/*
364 	 * Add local net header.  If no space in first mbuf,
365 	 * allocate another.
366 	 *
367 	 * Note that we do prepend regardless of RT_HAS_HEADER flag.
368 	 * This is done because BPF code shifts m_data pointer
369 	 * to the end of ethernet header prior to calling if_output().
370 	 */
371 	M_PREPEND(m, hlen, M_NOWAIT);
372 	if (m == NULL)
373 		senderr(ENOBUFS);
374 	if ((pflags & RT_HAS_HEADER) == 0) {
375 		eh = mtod(m, struct ether_header *);
376 		memcpy(eh, phdr, hlen);
377 	}
378 
379 	/*
380 	 * If a simplex interface, and the packet is being sent to our
381 	 * Ethernet address or a broadcast address, loopback a copy.
382 	 * XXX To make a simplex device behave exactly like a duplex
383 	 * device, we should copy in the case of sending to our own
384 	 * ethernet address (thus letting the original actually appear
385 	 * on the wire). However, we don't do that here for security
386 	 * reasons and compatibility with the original behavior.
387 	 */
388 	if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
389 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
390 		struct mbuf *n;
391 
392 		/*
393 		 * Because if_simloop() modifies the packet, we need a
394 		 * writable copy through m_dup() instead of a readonly
395 		 * one as m_copy[m] would give us. The alternative would
396 		 * be to modify if_simloop() to handle the readonly mbuf,
397 		 * but performancewise it is mostly equivalent (trading
398 		 * extra data copying vs. extra locking).
399 		 *
400 		 * XXX This is a local workaround.  A number of less
401 		 * often used kernel parts suffer from the same bug.
402 		 * See PR kern/105943 for a proposed general solution.
403 		 */
404 		if ((n = m_dup(m, M_NOWAIT)) != NULL) {
405 			update_mbuf_csumflags(m, n);
406 			(void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
407 		} else
408 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
409 	}
410 
411        /*
412 	* Bridges require special output handling.
413 	*/
414 	if (ifp->if_bridge) {
415 		BRIDGE_OUTPUT(ifp, m, error);
416 		return (error);
417 	}
418 
419 #if defined(INET) || defined(INET6)
420 	if (ifp->if_carp &&
421 	    (error = (*carp_output_p)(ifp, m, dst)))
422 		goto bad;
423 #endif
424 
425 	/* Handle ng_ether(4) processing, if any */
426 	if (ifp->if_l2com != NULL) {
427 		KASSERT(ng_ether_output_p != NULL,
428 		    ("ng_ether_output_p is NULL"));
429 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
430 bad:			if (m != NULL)
431 				m_freem(m);
432 			return (error);
433 		}
434 		if (m == NULL)
435 			return (0);
436 	}
437 
438 	/* Continue with link-layer output */
439 	return ether_output_frame(ifp, m);
440 }
441 
442 static bool
443 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
444 {
445 	struct ether_8021q_tag qtag;
446 	struct ether_header *eh;
447 
448 	eh = mtod(*mp, struct ether_header *);
449 	if (eh->ether_type == htons(ETHERTYPE_VLAN) ||
450 	    eh->ether_type == htons(ETHERTYPE_QINQ)) {
451 		(*mp)->m_flags &= ~M_VLANTAG;
452 		return (true);
453 	}
454 
455 	qtag.vid = 0;
456 	qtag.pcp = pcp;
457 	qtag.proto = ETHERTYPE_VLAN;
458 	if (ether_8021q_frame(mp, ifp, ifp, &qtag))
459 		return (true);
460 	if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
461 	return (false);
462 }
463 
464 /*
465  * Ethernet link layer output routine to send a raw frame to the device.
466  *
467  * This assumes that the 14 byte Ethernet header is present and contiguous
468  * in the first mbuf (if BRIDGE'ing).
469  */
470 int
471 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
472 {
473 	if (ether_do_pcp(ifp, m) && !ether_set_pcp(&m, ifp, ifp->if_pcp))
474 		return (0);
475 
476 	if (PFIL_HOOKED_OUT(V_link_pfil_head))
477 		switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, 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 (__predict_false(m->m_len < ETHER_HDR_LEN)) {
538 		/* Drivers should pullup and ensure the mbuf is valid */
539 		if_printf(ifp, "discard frame w/o leading ethernet "
540 				"header (len %d pkt len %d)\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 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
592 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
593 
594 	/* Allow monitor mode to claim this frame, after stats are updated. */
595 	if (ifp->if_flags & IFF_MONITOR) {
596 		m_freem(m);
597 		CURVNET_RESTORE();
598 		return;
599 	}
600 
601 	/* Handle input from a lagg(4) port */
602 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
603 		KASSERT(lagg_input_ethernet_p != NULL,
604 		    ("%s: if_lagg not loaded!", __func__));
605 		m = (*lagg_input_ethernet_p)(ifp, m);
606 		if (m != NULL)
607 			ifp = m->m_pkthdr.rcvif;
608 		else {
609 			CURVNET_RESTORE();
610 			return;
611 		}
612 	}
613 
614 	/*
615 	 * If the hardware did not process an 802.1Q tag, do this now,
616 	 * to allow 802.1P priority frames to be passed to the main input
617 	 * path correctly.
618 	 */
619 	if ((m->m_flags & M_VLANTAG) == 0 &&
620 	    ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
621 		struct ether_vlan_header *evl;
622 
623 		if (m->m_len < sizeof(*evl) &&
624 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
625 #ifdef DIAGNOSTIC
626 			if_printf(ifp, "cannot pullup VLAN header\n");
627 #endif
628 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
629 			CURVNET_RESTORE();
630 			return;
631 		}
632 
633 		evl = mtod(m, struct ether_vlan_header *);
634 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
635 		m->m_flags |= M_VLANTAG;
636 
637 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
638 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
639 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
640 		eh = mtod(m, struct ether_header *);
641 	}
642 
643 	M_SETFIB(m, ifp->if_fib);
644 
645 	/* Allow ng_ether(4) to claim this frame. */
646 	if (ifp->if_l2com != NULL) {
647 		KASSERT(ng_ether_input_p != NULL,
648 		    ("%s: ng_ether_input_p is NULL", __func__));
649 		m->m_flags &= ~M_PROMISC;
650 		(*ng_ether_input_p)(ifp, &m);
651 		if (m == NULL) {
652 			CURVNET_RESTORE();
653 			return;
654 		}
655 		eh = mtod(m, struct ether_header *);
656 	}
657 
658 	/*
659 	 * Allow if_bridge(4) to claim this frame.
660 	 *
661 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
662 	 * and the frame should be delivered locally.
663 	 *
664 	 * If M_BRIDGE_INJECT is set, the packet was received directly by the
665 	 * bridge via netmap, so "ifp" is the bridge itself and the packet
666 	 * should be re-examined.
667 	 */
668 	if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) {
669 		m->m_flags &= ~M_PROMISC;
670 		BRIDGE_INPUT(ifp, m);
671 		if (m == NULL) {
672 			CURVNET_RESTORE();
673 			return;
674 		}
675 		eh = mtod(m, struct ether_header *);
676 	}
677 
678 #if defined(INET) || defined(INET6)
679 	/*
680 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
681 	 * mbuf flows up to Layer 3.
682 	 * FreeBSD's implementation of carp(4) uses the inprotosw
683 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
684 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
685 	 * is outside the scope of the M_PROMISC test below.
686 	 * TODO: Maintain a hash table of ethernet addresses other than
687 	 * ether_dhost which may be active on this ifp.
688 	 */
689 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
690 		m->m_flags &= ~M_PROMISC;
691 	} else
692 #endif
693 	{
694 		/*
695 		 * If the frame received was not for our MAC address, set the
696 		 * M_PROMISC flag on the mbuf chain. The frame may need to
697 		 * be seen by the rest of the Ethernet input path in case of
698 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
699 		 * seen by upper protocol layers.
700 		 */
701 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
702 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
703 			m->m_flags |= M_PROMISC;
704 	}
705 
706 	ether_demux(ifp, m);
707 	CURVNET_RESTORE();
708 }
709 
710 /*
711  * Ethernet input dispatch; by default, direct dispatch here regardless of
712  * global configuration.  However, if RSS is enabled, hook up RSS affinity
713  * so that when deferred or hybrid dispatch is enabled, we can redistribute
714  * load based on RSS.
715  *
716  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
717  * not it had already done work distribution via multi-queue.  Then we could
718  * direct dispatch in the event load balancing was already complete and
719  * handle the case of interfaces with different capabilities better.
720  *
721  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
722  * at multiple layers?
723  *
724  * XXXRW: For now, enable all this only if RSS is compiled in, although it
725  * works fine without RSS.  Need to characterise the performance overhead
726  * of the detour through the netisr code in the event the result is always
727  * direct dispatch.
728  */
729 static void
730 ether_nh_input(struct mbuf *m)
731 {
732 
733 	M_ASSERTPKTHDR(m);
734 	KASSERT(m->m_pkthdr.rcvif != NULL,
735 	    ("%s: NULL interface pointer", __func__));
736 	ether_input_internal(m->m_pkthdr.rcvif, m);
737 }
738 
739 static struct netisr_handler	ether_nh = {
740 	.nh_name = "ether",
741 	.nh_handler = ether_nh_input,
742 	.nh_proto = NETISR_ETHER,
743 #ifdef RSS
744 	.nh_policy = NETISR_POLICY_CPU,
745 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
746 	.nh_m2cpuid = rss_m2cpuid,
747 #else
748 	.nh_policy = NETISR_POLICY_SOURCE,
749 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
750 #endif
751 };
752 
753 static void
754 ether_init(__unused void *arg)
755 {
756 
757 	netisr_register(&ether_nh);
758 }
759 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
760 
761 static void
762 vnet_ether_init(__unused void *arg)
763 {
764 	struct pfil_head_args args;
765 
766 	args.pa_version = PFIL_VERSION;
767 	args.pa_flags = PFIL_IN | PFIL_OUT;
768 	args.pa_type = PFIL_TYPE_ETHERNET;
769 	args.pa_headname = PFIL_ETHER_NAME;
770 	V_link_pfil_head = pfil_head_register(&args);
771 
772 #ifdef VIMAGE
773 	netisr_register_vnet(&ether_nh);
774 #endif
775 }
776 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
777     vnet_ether_init, NULL);
778 
779 #ifdef VIMAGE
780 static void
781 vnet_ether_pfil_destroy(__unused void *arg)
782 {
783 
784 	pfil_head_unregister(V_link_pfil_head);
785 }
786 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
787     vnet_ether_pfil_destroy, NULL);
788 
789 static void
790 vnet_ether_destroy(__unused void *arg)
791 {
792 
793 	netisr_unregister_vnet(&ether_nh);
794 }
795 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
796     vnet_ether_destroy, NULL);
797 #endif
798 
799 static void
800 ether_input(struct ifnet *ifp, struct mbuf *m)
801 {
802 	struct epoch_tracker et;
803 	struct mbuf *mn;
804 	bool needs_epoch;
805 
806 	needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH);
807 #ifdef INVARIANTS
808 	/*
809 	 * This temporary code is here to prevent epoch unaware and unmarked
810 	 * drivers to panic the system.  Once all drivers are taken care of,
811 	 * the whole INVARIANTS block should go away.
812 	 */
813 	if (!needs_epoch && !in_epoch(net_epoch_preempt)) {
814 		static bool printedonce;
815 
816 		needs_epoch = true;
817 		if (!printedonce) {
818 			printedonce = true;
819 			if_printf(ifp, "called %s w/o net epoch! "
820 			    "PLEASE file a bug report.", __func__);
821 #ifdef KDB
822 			kdb_backtrace();
823 #endif
824 		}
825 	}
826 #endif
827 
828 	/*
829 	 * The drivers are allowed to pass in a chain of packets linked with
830 	 * m_nextpkt. We split them up into separate packets here and pass
831 	 * them up. This allows the drivers to amortize the receive lock.
832 	 */
833 	CURVNET_SET_QUIET(ifp->if_vnet);
834 	if (__predict_false(needs_epoch))
835 		NET_EPOCH_ENTER(et);
836 	while (m) {
837 		mn = m->m_nextpkt;
838 		m->m_nextpkt = NULL;
839 
840 		/*
841 		 * We will rely on rcvif being set properly in the deferred
842 		 * context, so assert it is correct here.
843 		 */
844 		MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
845 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
846 		    "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
847 		netisr_dispatch(NETISR_ETHER, m);
848 		m = mn;
849 	}
850 	if (__predict_false(needs_epoch))
851 		NET_EPOCH_EXIT(et);
852 	CURVNET_RESTORE();
853 }
854 
855 /*
856  * Upper layer processing for a received Ethernet packet.
857  */
858 void
859 ether_demux(struct ifnet *ifp, struct mbuf *m)
860 {
861 	struct ether_header *eh;
862 	int i, isr;
863 	u_short ether_type;
864 
865 	NET_EPOCH_ASSERT();
866 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
867 
868 	/* Do not grab PROMISC frames in case we are re-entered. */
869 	if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
870 		i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL);
871 		if (i != PFIL_PASS)
872 			return;
873 	}
874 
875 	eh = mtod(m, struct ether_header *);
876 	ether_type = ntohs(eh->ether_type);
877 
878 	/*
879 	 * If this frame has a VLAN tag other than 0, call vlan_input()
880 	 * if its module is loaded. Otherwise, drop.
881 	 */
882 	if ((m->m_flags & M_VLANTAG) &&
883 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
884 		if (ifp->if_vlantrunk == NULL) {
885 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
886 			m_freem(m);
887 			return;
888 		}
889 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
890 		    __func__));
891 		/* Clear before possibly re-entering ether_input(). */
892 		m->m_flags &= ~M_PROMISC;
893 		(*vlan_input_p)(ifp, m);
894 		return;
895 	}
896 
897 	/*
898 	 * Pass promiscuously received frames to the upper layer if the user
899 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
900 	 */
901 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
902 		m_freem(m);
903 		return;
904 	}
905 
906 	/*
907 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
908 	 */
909 	m->m_flags &= ~M_VLANTAG;
910 	m_clrprotoflags(m);
911 
912 	/*
913 	 * Dispatch frame to upper layer.
914 	 */
915 	switch (ether_type) {
916 #ifdef INET
917 	case ETHERTYPE_IP:
918 		isr = NETISR_IP;
919 		break;
920 
921 	case ETHERTYPE_ARP:
922 		if (ifp->if_flags & IFF_NOARP) {
923 			/* Discard packet if ARP is disabled on interface */
924 			m_freem(m);
925 			return;
926 		}
927 		isr = NETISR_ARP;
928 		break;
929 #endif
930 #ifdef INET6
931 	case ETHERTYPE_IPV6:
932 		isr = NETISR_IPV6;
933 		break;
934 #endif
935 	default:
936 		goto discard;
937 	}
938 
939 	/* Strip off Ethernet header. */
940 	m_adj(m, ETHER_HDR_LEN);
941 
942 	netisr_dispatch(isr, m);
943 	return;
944 
945 discard:
946 	/*
947 	 * Packet is to be discarded.  If netgraph is present,
948 	 * hand the packet to it for last chance processing;
949 	 * otherwise dispose of it.
950 	 */
951 	if (ifp->if_l2com != NULL) {
952 		KASSERT(ng_ether_input_orphan_p != NULL,
953 		    ("ng_ether_input_orphan_p is NULL"));
954 		(*ng_ether_input_orphan_p)(ifp, m);
955 		return;
956 	}
957 	m_freem(m);
958 }
959 
960 /*
961  * Convert Ethernet address to printable (loggable) representation.
962  * This routine is for compatibility; it's better to just use
963  *
964  *	printf("%6D", <pointer to address>, ":");
965  *
966  * since there's no static buffer involved.
967  */
968 char *
969 ether_sprintf(const u_char *ap)
970 {
971 	static char etherbuf[18];
972 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
973 	return (etherbuf);
974 }
975 
976 /*
977  * Perform common duties while attaching to interface list
978  */
979 void
980 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
981 {
982 	int i;
983 	struct ifaddr *ifa;
984 	struct sockaddr_dl *sdl;
985 
986 	ifp->if_addrlen = ETHER_ADDR_LEN;
987 	ifp->if_hdrlen = ETHER_HDR_LEN;
988 	ifp->if_mtu = ETHERMTU;
989 	if_attach(ifp);
990 	ifp->if_output = ether_output;
991 	ifp->if_input = ether_input;
992 	ifp->if_resolvemulti = ether_resolvemulti;
993 	ifp->if_requestencap = ether_requestencap;
994 #ifdef VIMAGE
995 	ifp->if_reassign = ether_reassign;
996 #endif
997 	if (ifp->if_baudrate == 0)
998 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
999 	ifp->if_broadcastaddr = etherbroadcastaddr;
1000 
1001 	ifa = ifp->if_addr;
1002 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
1003 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1004 	sdl->sdl_type = IFT_ETHER;
1005 	sdl->sdl_alen = ifp->if_addrlen;
1006 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
1007 
1008 	if (ifp->if_hw_addr != NULL)
1009 		bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
1010 
1011 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1012 	if (ng_ether_attach_p != NULL)
1013 		(*ng_ether_attach_p)(ifp);
1014 
1015 	/* Announce Ethernet MAC address if non-zero. */
1016 	for (i = 0; i < ifp->if_addrlen; i++)
1017 		if (lla[i] != 0)
1018 			break;
1019 	if (i != ifp->if_addrlen)
1020 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1021 
1022 	uuid_ether_add(LLADDR(sdl));
1023 
1024 	/* Add necessary bits are setup; announce it now. */
1025 	EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1026 	if (IS_DEFAULT_VNET(curvnet))
1027 		devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1028 }
1029 
1030 /*
1031  * Perform common duties while detaching an Ethernet interface
1032  */
1033 void
1034 ether_ifdetach(struct ifnet *ifp)
1035 {
1036 	struct sockaddr_dl *sdl;
1037 
1038 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1039 	uuid_ether_del(LLADDR(sdl));
1040 
1041 	if (ifp->if_l2com != NULL) {
1042 		KASSERT(ng_ether_detach_p != NULL,
1043 		    ("ng_ether_detach_p is NULL"));
1044 		(*ng_ether_detach_p)(ifp);
1045 	}
1046 
1047 	bpfdetach(ifp);
1048 	if_detach(ifp);
1049 }
1050 
1051 #ifdef VIMAGE
1052 void
1053 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1054 {
1055 
1056 	if (ifp->if_l2com != NULL) {
1057 		KASSERT(ng_ether_detach_p != NULL,
1058 		    ("ng_ether_detach_p is NULL"));
1059 		(*ng_ether_detach_p)(ifp);
1060 	}
1061 
1062 	if (ng_ether_attach_p != NULL) {
1063 		CURVNET_SET_QUIET(new_vnet);
1064 		(*ng_ether_attach_p)(ifp);
1065 		CURVNET_RESTORE();
1066 	}
1067 }
1068 #endif
1069 
1070 SYSCTL_DECL(_net_link);
1071 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1072     "Ethernet");
1073 
1074 #if 0
1075 /*
1076  * This is for reference.  We have a table-driven version
1077  * of the little-endian crc32 generator, which is faster
1078  * than the double-loop.
1079  */
1080 uint32_t
1081 ether_crc32_le(const uint8_t *buf, size_t len)
1082 {
1083 	size_t i;
1084 	uint32_t crc;
1085 	int bit;
1086 	uint8_t data;
1087 
1088 	crc = 0xffffffff;	/* initial value */
1089 
1090 	for (i = 0; i < len; i++) {
1091 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1092 			carry = (crc ^ data) & 1;
1093 			crc >>= 1;
1094 			if (carry)
1095 				crc = (crc ^ ETHER_CRC_POLY_LE);
1096 		}
1097 	}
1098 
1099 	return (crc);
1100 }
1101 #else
1102 uint32_t
1103 ether_crc32_le(const uint8_t *buf, size_t len)
1104 {
1105 	static const uint32_t crctab[] = {
1106 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1107 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1108 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1109 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1110 	};
1111 	size_t i;
1112 	uint32_t crc;
1113 
1114 	crc = 0xffffffff;	/* initial value */
1115 
1116 	for (i = 0; i < len; i++) {
1117 		crc ^= buf[i];
1118 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1119 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1120 	}
1121 
1122 	return (crc);
1123 }
1124 #endif
1125 
1126 uint32_t
1127 ether_crc32_be(const uint8_t *buf, size_t len)
1128 {
1129 	size_t i;
1130 	uint32_t crc, carry;
1131 	int bit;
1132 	uint8_t data;
1133 
1134 	crc = 0xffffffff;	/* initial value */
1135 
1136 	for (i = 0; i < len; i++) {
1137 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1138 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1139 			crc <<= 1;
1140 			if (carry)
1141 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1142 		}
1143 	}
1144 
1145 	return (crc);
1146 }
1147 
1148 int
1149 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1150 {
1151 	struct ifaddr *ifa = (struct ifaddr *) data;
1152 	struct ifreq *ifr = (struct ifreq *) data;
1153 	int error = 0;
1154 
1155 	switch (command) {
1156 	case SIOCSIFADDR:
1157 		ifp->if_flags |= IFF_UP;
1158 
1159 		switch (ifa->ifa_addr->sa_family) {
1160 #ifdef INET
1161 		case AF_INET:
1162 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1163 			arp_ifinit(ifp, ifa);
1164 			break;
1165 #endif
1166 		default:
1167 			ifp->if_init(ifp->if_softc);
1168 			break;
1169 		}
1170 		break;
1171 
1172 	case SIOCGIFADDR:
1173 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1174 		    ETHER_ADDR_LEN);
1175 		break;
1176 
1177 	case SIOCSIFMTU:
1178 		/*
1179 		 * Set the interface MTU.
1180 		 */
1181 		if (ifr->ifr_mtu > ETHERMTU) {
1182 			error = EINVAL;
1183 		} else {
1184 			ifp->if_mtu = ifr->ifr_mtu;
1185 		}
1186 		break;
1187 
1188 	case SIOCSLANPCP:
1189 		error = priv_check(curthread, PRIV_NET_SETLANPCP);
1190 		if (error != 0)
1191 			break;
1192 		if (ifr->ifr_lan_pcp > 7 &&
1193 		    ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1194 			error = EINVAL;
1195 		} else {
1196 			ifp->if_pcp = ifr->ifr_lan_pcp;
1197 			/* broadcast event about PCP change */
1198 			EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1199 		}
1200 		break;
1201 
1202 	case SIOCGLANPCP:
1203 		ifr->ifr_lan_pcp = ifp->if_pcp;
1204 		break;
1205 
1206 	default:
1207 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1208 		break;
1209 	}
1210 	return (error);
1211 }
1212 
1213 static int
1214 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1215 	struct sockaddr *sa)
1216 {
1217 	struct sockaddr_dl *sdl;
1218 #ifdef INET
1219 	struct sockaddr_in *sin;
1220 #endif
1221 #ifdef INET6
1222 	struct sockaddr_in6 *sin6;
1223 #endif
1224 	u_char *e_addr;
1225 
1226 	switch(sa->sa_family) {
1227 	case AF_LINK:
1228 		/*
1229 		 * No mapping needed. Just check that it's a valid MC address.
1230 		 */
1231 		sdl = (struct sockaddr_dl *)sa;
1232 		e_addr = LLADDR(sdl);
1233 		if (!ETHER_IS_MULTICAST(e_addr))
1234 			return EADDRNOTAVAIL;
1235 		*llsa = NULL;
1236 		return 0;
1237 
1238 #ifdef INET
1239 	case AF_INET:
1240 		sin = (struct sockaddr_in *)sa;
1241 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1242 			return EADDRNOTAVAIL;
1243 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1244 		sdl->sdl_alen = ETHER_ADDR_LEN;
1245 		e_addr = LLADDR(sdl);
1246 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1247 		*llsa = (struct sockaddr *)sdl;
1248 		return 0;
1249 #endif
1250 #ifdef INET6
1251 	case AF_INET6:
1252 		sin6 = (struct sockaddr_in6 *)sa;
1253 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1254 			/*
1255 			 * An IP6 address of 0 means listen to all
1256 			 * of the Ethernet multicast address used for IP6.
1257 			 * (This is used for multicast routers.)
1258 			 */
1259 			ifp->if_flags |= IFF_ALLMULTI;
1260 			*llsa = NULL;
1261 			return 0;
1262 		}
1263 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1264 			return EADDRNOTAVAIL;
1265 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1266 		sdl->sdl_alen = ETHER_ADDR_LEN;
1267 		e_addr = LLADDR(sdl);
1268 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1269 		*llsa = (struct sockaddr *)sdl;
1270 		return 0;
1271 #endif
1272 
1273 	default:
1274 		/*
1275 		 * Well, the text isn't quite right, but it's the name
1276 		 * that counts...
1277 		 */
1278 		return EAFNOSUPPORT;
1279 	}
1280 }
1281 
1282 static moduledata_t ether_mod = {
1283 	.name = "ether",
1284 };
1285 
1286 void
1287 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1288 {
1289 	struct ether_vlan_header vlan;
1290 	struct mbuf mv, mb;
1291 
1292 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1293 	    ("%s: vlan information not present", __func__));
1294 	KASSERT(m->m_len >= sizeof(struct ether_header),
1295 	    ("%s: mbuf not large enough for header", __func__));
1296 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1297 	vlan.evl_proto = vlan.evl_encap_proto;
1298 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1299 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1300 	m->m_len -= sizeof(struct ether_header);
1301 	m->m_data += sizeof(struct ether_header);
1302 	/*
1303 	 * If a data link has been supplied by the caller, then we will need to
1304 	 * re-create a stack allocated mbuf chain with the following structure:
1305 	 *
1306 	 * (1) mbuf #1 will contain the supplied data link
1307 	 * (2) mbuf #2 will contain the vlan header
1308 	 * (3) mbuf #3 will contain the original mbuf's packet data
1309 	 *
1310 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1311 	 */
1312 	if (data != NULL) {
1313 		mv.m_next = m;
1314 		mv.m_data = (caddr_t)&vlan;
1315 		mv.m_len = sizeof(vlan);
1316 		mb.m_next = &mv;
1317 		mb.m_data = data;
1318 		mb.m_len = dlen;
1319 		bpf_mtap(bp, &mb);
1320 	} else
1321 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1322 	m->m_len += sizeof(struct ether_header);
1323 	m->m_data -= sizeof(struct ether_header);
1324 }
1325 
1326 struct mbuf *
1327 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1328 {
1329 	struct ether_vlan_header *evl;
1330 
1331 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1332 	if (m == NULL)
1333 		return (NULL);
1334 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1335 
1336 	if (m->m_len < sizeof(*evl)) {
1337 		m = m_pullup(m, sizeof(*evl));
1338 		if (m == NULL)
1339 			return (NULL);
1340 	}
1341 
1342 	/*
1343 	 * Transform the Ethernet header into an Ethernet header
1344 	 * with 802.1Q encapsulation.
1345 	 */
1346 	evl = mtod(m, struct ether_vlan_header *);
1347 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1348 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1349 	evl->evl_encap_proto = htons(proto);
1350 	evl->evl_tag = htons(tag);
1351 	return (m);
1352 }
1353 
1354 void
1355 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m)
1356 {
1357 	if (bpf_peers_present(ifp->if_bpf)) {
1358 		M_ASSERTVALID(m);
1359 		if ((m->m_flags & M_VLANTAG) != 0)
1360 			ether_vlan_mtap(ifp->if_bpf, m, NULL, 0);
1361 		else
1362 			bpf_mtap(ifp->if_bpf, m);
1363 	}
1364 }
1365 
1366 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1367     "IEEE 802.1Q VLAN");
1368 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1369     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1370     "for consistency");
1371 
1372 VNET_DEFINE_STATIC(int, soft_pad);
1373 #define	V_soft_pad	VNET(soft_pad)
1374 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1375     &VNET_NAME(soft_pad), 0,
1376     "pad short frames before tagging");
1377 
1378 /*
1379  * For now, make preserving PCP via an mbuf tag optional, as it increases
1380  * per-packet memory allocations and frees.  In the future, it would be
1381  * preferable to reuse ether_vtag for this, or similar.
1382  */
1383 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1384 #define	V_vlan_mtag_pcp	VNET(vlan_mtag_pcp)
1385 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1386     &VNET_NAME(vlan_mtag_pcp), 0,
1387     "Retain VLAN PCP information as packets are passed up the stack");
1388 
1389 static inline bool
1390 ether_do_pcp(struct ifnet *ifp, struct mbuf *m)
1391 {
1392 	if (ifp->if_type == IFT_L2VLAN)
1393 		return (false);
1394 	if (ifp->if_pcp != IFNET_PCP_NONE || (m->m_flags & M_VLANTAG) != 0)
1395 		return (true);
1396 	if (V_vlan_mtag_pcp &&
1397 	    m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_OUT, NULL) != NULL)
1398 		return (true);
1399 	return (false);
1400 }
1401 
1402 bool
1403 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1404     const struct ether_8021q_tag *qtag)
1405 {
1406 	struct m_tag *mtag;
1407 	int n;
1408 	uint16_t tag;
1409 	uint8_t pcp = qtag->pcp;
1410 	static const char pad[8];	/* just zeros */
1411 
1412 	/*
1413 	 * Pad the frame to the minimum size allowed if told to.
1414 	 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1415 	 * paragraph C.4.4.3.b.  It can help to work around buggy
1416 	 * bridges that violate paragraph C.4.4.3.a from the same
1417 	 * document, i.e., fail to pad short frames after untagging.
1418 	 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1419 	 * untagging it will produce a 62-byte frame, which is a runt
1420 	 * and requires padding.  There are VLAN-enabled network
1421 	 * devices that just discard such runts instead or mishandle
1422 	 * them somehow.
1423 	 */
1424 	if (V_soft_pad && p->if_type == IFT_ETHER) {
1425 		for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1426 		     n > 0; n -= sizeof(pad)) {
1427 			if (!m_append(*mp, min(n, sizeof(pad)), pad))
1428 				break;
1429 		}
1430 		if (n > 0) {
1431 			m_freem(*mp);
1432 			*mp = NULL;
1433 			if_printf(ife, "cannot pad short frame");
1434 			return (false);
1435 		}
1436 	}
1437 
1438 	/*
1439 	 * If PCP is set in mbuf, use it
1440 	 */
1441 	if ((*mp)->m_flags & M_VLANTAG) {
1442 		pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1443 	}
1444 
1445 	/*
1446 	 * If underlying interface can do VLAN tag insertion itself,
1447 	 * just pass the packet along. However, we need some way to
1448 	 * tell the interface where the packet came from so that it
1449 	 * knows how to find the VLAN tag to use, so we attach a
1450 	 * packet tag that holds it.
1451 	 */
1452 	if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1453 	    MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1454 		tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1455 	else
1456 		tag = EVL_MAKETAG(qtag->vid, pcp, 0);
1457 	if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1458 	    (qtag->proto == ETHERTYPE_VLAN)) {
1459 		(*mp)->m_pkthdr.ether_vtag = tag;
1460 		(*mp)->m_flags |= M_VLANTAG;
1461 	} else {
1462 		*mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1463 		if (*mp == NULL) {
1464 			if_printf(ife, "unable to prepend 802.1Q header");
1465 			return (false);
1466 		}
1467 		(*mp)->m_flags &= ~M_VLANTAG;
1468 	}
1469 	return (true);
1470 }
1471 
1472 /*
1473  * Allocate an address from the FreeBSD Foundation OUI.  This uses a
1474  * cryptographic hash function on the containing jail's name, UUID and the
1475  * interface name to attempt to provide a unique but stable address.
1476  * Pseudo-interfaces which require a MAC address should use this function to
1477  * allocate non-locally-administered addresses.
1478  */
1479 void
1480 ether_gen_addr_byname(const char *nameunit, struct ether_addr *hwaddr)
1481 {
1482 	SHA1_CTX ctx;
1483 	char *buf;
1484 	char uuid[HOSTUUIDLEN + 1];
1485 	uint64_t addr;
1486 	int i, sz;
1487 	char digest[SHA1_RESULTLEN];
1488 	char jailname[MAXHOSTNAMELEN];
1489 
1490 	getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1491 	if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1492 		/* Fall back to a random mac address. */
1493 		goto rando;
1494 	}
1495 
1496 	/* If each (vnet) jail would also have a unique hostuuid this would not
1497 	 * be necessary. */
1498 	getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1499 	sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, nameunit,
1500 	    jailname);
1501 	if (sz < 0) {
1502 		/* Fall back to a random mac address. */
1503 		goto rando;
1504 	}
1505 
1506 	SHA1Init(&ctx);
1507 	SHA1Update(&ctx, buf, sz);
1508 	SHA1Final(digest, &ctx);
1509 	free(buf, M_TEMP);
1510 
1511 	addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1512 	    OUI_FREEBSD_GENERATED_MASK;
1513 	addr = OUI_FREEBSD(addr);
1514 	for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1515 		hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1516 		    0xFF;
1517 	}
1518 
1519 	return;
1520 rando:
1521 	arc4rand(hwaddr, sizeof(*hwaddr), 0);
1522 	/* Unicast */
1523 	hwaddr->octet[0] &= 0xFE;
1524 	/* Locally administered. */
1525 	hwaddr->octet[0] |= 0x02;
1526 }
1527 
1528 void
1529 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1530 {
1531 	ether_gen_addr_byname(if_name(ifp), hwaddr);
1532 }
1533 
1534 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1535 MODULE_VERSION(ether, 1);
1536