xref: /freebsd/sys/net/if_ethersubr.c (revision 312809fe7fefbc8d5caa2b59089a5d9266378057)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  *
31  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
32  * $FreeBSD$
33  */
34 
35 #include "opt_inet.h"
36 #include "opt_inet6.h"
37 #include "opt_netgraph.h"
38 #include "opt_mbuf_profiling.h"
39 #include "opt_rss.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/bus.h>
44 #include <sys/eventhandler.h>
45 #include <sys/jail.h>
46 #include <sys/kernel.h>
47 #include <sys/lock.h>
48 #include <sys/malloc.h>
49 #include <sys/module.h>
50 #include <sys/mbuf.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 
59 #include <net/ieee_oui.h>
60 #include <net/if.h>
61 #include <net/if_var.h>
62 #include <net/if_arp.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/if_llc.h>
66 #include <net/if_dl.h>
67 #include <net/if_types.h>
68 #include <net/bpf.h>
69 #include <net/ethernet.h>
70 #include <net/if_bridgevar.h>
71 #include <net/if_vlan_var.h>
72 #include <net/if_llatbl.h>
73 #include <net/pfil.h>
74 #include <net/rss_config.h>
75 #include <net/vnet.h>
76 
77 #include <netpfil/pf/pf_mtag.h>
78 
79 #if defined(INET) || defined(INET6)
80 #include <netinet/in.h>
81 #include <netinet/in_var.h>
82 #include <netinet/if_ether.h>
83 #include <netinet/ip_carp.h>
84 #include <netinet/ip_var.h>
85 #endif
86 #ifdef INET6
87 #include <netinet6/nd6.h>
88 #endif
89 #include <security/mac/mac_framework.h>
90 
91 #include <crypto/sha1.h>
92 
93 #ifdef CTASSERT
94 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
95 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
96 #endif
97 
98 VNET_DEFINE(pfil_head_t, link_pfil_head);	/* Packet filter hooks */
99 
100 /* netgraph node hooks for ng_ether(4) */
101 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
102 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
103 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
104 void	(*ng_ether_attach_p)(struct ifnet *ifp);
105 void	(*ng_ether_detach_p)(struct ifnet *ifp);
106 
107 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
108 
109 /* if_bridge(4) support */
110 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
111 
112 /* if_lagg(4) support */
113 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
114 
115 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
116 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
117 
118 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
119 		struct sockaddr *);
120 #ifdef VIMAGE
121 static	void ether_reassign(struct ifnet *, struct vnet *, char *);
122 #endif
123 static	int ether_requestencap(struct ifnet *, struct if_encap_req *);
124 
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 
204 static int
205 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
206 	const struct sockaddr *dst, struct route *ro, u_char *phdr,
207 	uint32_t *pflags, struct llentry **plle)
208 {
209 	struct ether_header *eh;
210 	uint32_t lleflags = 0;
211 	int error = 0;
212 #if defined(INET) || defined(INET6)
213 	uint16_t etype;
214 #endif
215 
216 	if (plle)
217 		*plle = NULL;
218 	eh = (struct ether_header *)phdr;
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 			error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags,
245 			    plle);
246 		else {
247 			const struct in6_addr *a6;
248 			a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
249 			ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
250 			etype = htons(ETHERTYPE_IPV6);
251 			memcpy(&eh->ether_type, &etype, sizeof(etype));
252 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
253 		}
254 		break;
255 #endif
256 	default:
257 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
258 		if (m != NULL)
259 			m_freem(m);
260 		return (EAFNOSUPPORT);
261 	}
262 
263 	if (error == EHOSTDOWN) {
264 		if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
265 			error = EHOSTUNREACH;
266 	}
267 
268 	if (error != 0)
269 		return (error);
270 
271 	*pflags = RT_MAY_LOOP;
272 	if (lleflags & LLE_IFADDR)
273 		*pflags |= RT_L2_ME;
274 
275 	return (0);
276 }
277 
278 /*
279  * Ethernet output routine.
280  * Encapsulate a packet of type family for the local net.
281  * Use trailer local net encapsulation if enough data in first
282  * packet leaves a multiple of 512 bytes of data in remainder.
283  */
284 int
285 ether_output(struct ifnet *ifp, struct mbuf *m,
286 	const struct sockaddr *dst, struct route *ro)
287 {
288 	int error = 0;
289 	char linkhdr[ETHER_HDR_LEN], *phdr;
290 	struct ether_header *eh;
291 	struct pf_mtag *t;
292 	int loop_copy = 1;
293 	int hlen;	/* link layer header length */
294 	uint32_t pflags;
295 	struct llentry *lle = NULL;
296 	int addref = 0;
297 
298 	phdr = NULL;
299 	pflags = 0;
300 	if (ro != NULL) {
301 		/* XXX BPF uses ro_prepend */
302 		if (ro->ro_prepend != NULL) {
303 			phdr = ro->ro_prepend;
304 			hlen = ro->ro_plen;
305 		} else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
306 			if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
307 				lle = ro->ro_lle;
308 				if (lle != NULL &&
309 				    (lle->la_flags & LLE_VALID) == 0) {
310 					LLE_FREE(lle);
311 					lle = NULL;	/* redundant */
312 					ro->ro_lle = NULL;
313 				}
314 				if (lle == NULL) {
315 					/* if we lookup, keep cache */
316 					addref = 1;
317 				} else
318 					/*
319 					 * Notify LLE code that
320 					 * the entry was used
321 					 * by datapath.
322 					 */
323 					llentry_mark_used(lle);
324 			}
325 			if (lle != NULL) {
326 				phdr = lle->r_linkdata;
327 				hlen = lle->r_hdrlen;
328 				pflags = lle->r_flags;
329 			}
330 		}
331 	}
332 
333 #ifdef MAC
334 	error = mac_ifnet_check_transmit(ifp, m);
335 	if (error)
336 		senderr(error);
337 #endif
338 
339 	M_PROFILE(m);
340 	if (ifp->if_flags & IFF_MONITOR)
341 		senderr(ENETDOWN);
342 	if (!((ifp->if_flags & IFF_UP) &&
343 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
344 		senderr(ENETDOWN);
345 
346 	if (phdr == NULL) {
347 		/* No prepend data supplied. Try to calculate ourselves. */
348 		phdr = linkhdr;
349 		hlen = ETHER_HDR_LEN;
350 		error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
351 		    addref ? &lle : NULL);
352 		if (addref && lle != NULL)
353 			ro->ro_lle = lle;
354 		if (error != 0)
355 			return (error == EWOULDBLOCK ? 0 : error);
356 	}
357 
358 	if ((pflags & RT_L2_ME) != 0) {
359 		update_mbuf_csumflags(m, m);
360 		return (if_simloop(ifp, m, dst->sa_family, 0));
361 	}
362 	loop_copy = pflags & RT_MAY_LOOP;
363 
364 	/*
365 	 * Add local net header.  If no space in first mbuf,
366 	 * allocate another.
367 	 *
368 	 * Note that we do prepend regardless of RT_HAS_HEADER flag.
369 	 * This is done because BPF code shifts m_data pointer
370 	 * to the end of ethernet header prior to calling if_output().
371 	 */
372 	M_PREPEND(m, hlen, M_NOWAIT);
373 	if (m == NULL)
374 		senderr(ENOBUFS);
375 	if ((pflags & RT_HAS_HEADER) == 0) {
376 		eh = mtod(m, struct ether_header *);
377 		memcpy(eh, phdr, hlen);
378 	}
379 
380 	/*
381 	 * If a simplex interface, and the packet is being sent to our
382 	 * Ethernet address or a broadcast address, loopback a copy.
383 	 * XXX To make a simplex device behave exactly like a duplex
384 	 * device, we should copy in the case of sending to our own
385 	 * ethernet address (thus letting the original actually appear
386 	 * on the wire). However, we don't do that here for security
387 	 * reasons and compatibility with the original behavior.
388 	 */
389 	if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
390 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
391 		struct mbuf *n;
392 
393 		/*
394 		 * Because if_simloop() modifies the packet, we need a
395 		 * writable copy through m_dup() instead of a readonly
396 		 * one as m_copy[m] would give us. The alternative would
397 		 * be to modify if_simloop() to handle the readonly mbuf,
398 		 * but performancewise it is mostly equivalent (trading
399 		 * extra data copying vs. extra locking).
400 		 *
401 		 * XXX This is a local workaround.  A number of less
402 		 * often used kernel parts suffer from the same bug.
403 		 * See PR kern/105943 for a proposed general solution.
404 		 */
405 		if ((n = m_dup(m, M_NOWAIT)) != NULL) {
406 			update_mbuf_csumflags(m, n);
407 			(void)if_simloop(ifp, n, dst->sa_family, hlen);
408 		} else
409 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
410 	}
411 
412        /*
413 	* Bridges require special output handling.
414 	*/
415 	if (ifp->if_bridge) {
416 		BRIDGE_OUTPUT(ifp, m, error);
417 		return (error);
418 	}
419 
420 #if defined(INET) || defined(INET6)
421 	if (ifp->if_carp &&
422 	    (error = (*carp_output_p)(ifp, m, dst)))
423 		goto bad;
424 #endif
425 
426 	/* Handle ng_ether(4) processing, if any */
427 	if (ifp->if_l2com != NULL) {
428 		KASSERT(ng_ether_output_p != NULL,
429 		    ("ng_ether_output_p is NULL"));
430 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
431 bad:			if (m != NULL)
432 				m_freem(m);
433 			return (error);
434 		}
435 		if (m == NULL)
436 			return (0);
437 	}
438 
439 	/* Continue with link-layer output */
440 	return ether_output_frame(ifp, m);
441 }
442 
443 static bool
444 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
445 {
446 	struct ether_header *eh;
447 
448 	eh = mtod(*mp, struct ether_header *);
449 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN ||
450 	    ether_8021q_frame(mp, ifp, ifp, 0, pcp))
451 		return (true);
452 	if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
453 	return (false);
454 }
455 
456 /*
457  * Ethernet link layer output routine to send a raw frame to the device.
458  *
459  * This assumes that the 14 byte Ethernet header is present and contiguous
460  * in the first mbuf (if BRIDGE'ing).
461  */
462 int
463 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
464 {
465 	uint8_t pcp;
466 
467 	pcp = ifp->if_pcp;
468 	if (pcp != IFNET_PCP_NONE && ifp->if_type != IFT_L2VLAN &&
469 	    !ether_set_pcp(&m, ifp, pcp))
470 		return (0);
471 
472 	if (PFIL_HOOKED_OUT(V_link_pfil_head))
473 		switch (pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_OUT,
474 		    NULL)) {
475 		case PFIL_DROPPED:
476 			return (EACCES);
477 		case PFIL_CONSUMED:
478 			return (0);
479 		}
480 
481 #ifdef EXPERIMENTAL
482 #if defined(INET6) && defined(INET)
483 	/* draft-ietf-6man-ipv6only-flag */
484 	/* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
485 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
486 		struct ether_header *eh;
487 
488 		eh = mtod(m, struct ether_header *);
489 		switch (ntohs(eh->ether_type)) {
490 		case ETHERTYPE_IP:
491 		case ETHERTYPE_ARP:
492 		case ETHERTYPE_REVARP:
493 			m_freem(m);
494 			return (EAFNOSUPPORT);
495 			/* NOTREACHED */
496 			break;
497 		};
498 	}
499 #endif
500 #endif
501 
502 	/*
503 	 * Queue message on interface, update output statistics if
504 	 * successful, and start output if interface not yet active.
505 	 */
506 	return ((ifp->if_transmit)(ifp, m));
507 }
508 
509 /*
510  * Process a received Ethernet packet; the packet is in the
511  * mbuf chain m with the ethernet header at the front.
512  */
513 static void
514 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
515 {
516 	struct ether_header *eh;
517 	u_short etype;
518 
519 	if ((ifp->if_flags & IFF_UP) == 0) {
520 		m_freem(m);
521 		return;
522 	}
523 #ifdef DIAGNOSTIC
524 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
525 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
526 		m_freem(m);
527 		return;
528 	}
529 #endif
530 	if (m->m_len < ETHER_HDR_LEN) {
531 		/* XXX maybe should pullup? */
532 		if_printf(ifp, "discard frame w/o leading ethernet "
533 				"header (len %u pkt len %u)\n",
534 				m->m_len, m->m_pkthdr.len);
535 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
536 		m_freem(m);
537 		return;
538 	}
539 	eh = mtod(m, struct ether_header *);
540 	etype = ntohs(eh->ether_type);
541 	random_harvest_queue_ether(m, sizeof(*m));
542 
543 #ifdef EXPERIMENTAL
544 #if defined(INET6) && defined(INET)
545 	/* draft-ietf-6man-ipv6only-flag */
546 	/* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
547 	if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
548 
549 		switch (etype) {
550 		case ETHERTYPE_IP:
551 		case ETHERTYPE_ARP:
552 		case ETHERTYPE_REVARP:
553 			m_freem(m);
554 			return;
555 			/* NOTREACHED */
556 			break;
557 		};
558 	}
559 #endif
560 #endif
561 
562 	CURVNET_SET_QUIET(ifp->if_vnet);
563 
564 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
565 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
566 			m->m_flags |= M_BCAST;
567 		else
568 			m->m_flags |= M_MCAST;
569 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
570 	}
571 
572 #ifdef MAC
573 	/*
574 	 * Tag the mbuf with an appropriate MAC label before any other
575 	 * consumers can get to it.
576 	 */
577 	mac_ifnet_create_mbuf(ifp, m);
578 #endif
579 
580 	/*
581 	 * Give bpf a chance at the packet.
582 	 */
583 	ETHER_BPF_MTAP(ifp, m);
584 
585 	/*
586 	 * If the CRC is still on the packet, trim it off. We do this once
587 	 * and once only in case we are re-entered. Nothing else on the
588 	 * Ethernet receive path expects to see the FCS.
589 	 */
590 	if (m->m_flags & M_HASFCS) {
591 		m_adj(m, -ETHER_CRC_LEN);
592 		m->m_flags &= ~M_HASFCS;
593 	}
594 
595 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
596 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
597 
598 	/* Allow monitor mode to claim this frame, after stats are updated. */
599 	if (ifp->if_flags & IFF_MONITOR) {
600 		m_freem(m);
601 		CURVNET_RESTORE();
602 		return;
603 	}
604 
605 	/* Handle input from a lagg(4) port */
606 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
607 		KASSERT(lagg_input_p != NULL,
608 		    ("%s: if_lagg not loaded!", __func__));
609 		m = (*lagg_input_p)(ifp, m);
610 		if (m != NULL)
611 			ifp = m->m_pkthdr.rcvif;
612 		else {
613 			CURVNET_RESTORE();
614 			return;
615 		}
616 	}
617 
618 	/*
619 	 * If the hardware did not process an 802.1Q tag, do this now,
620 	 * to allow 802.1P priority frames to be passed to the main input
621 	 * path correctly.
622 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
623 	 */
624 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
625 		struct ether_vlan_header *evl;
626 
627 		if (m->m_len < sizeof(*evl) &&
628 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
629 #ifdef DIAGNOSTIC
630 			if_printf(ifp, "cannot pullup VLAN header\n");
631 #endif
632 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
633 			CURVNET_RESTORE();
634 			return;
635 		}
636 
637 		evl = mtod(m, struct ether_vlan_header *);
638 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
639 		m->m_flags |= M_VLANTAG;
640 
641 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
642 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
643 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
644 		eh = mtod(m, struct ether_header *);
645 	}
646 
647 	M_SETFIB(m, ifp->if_fib);
648 
649 	/* Allow ng_ether(4) to claim this frame. */
650 	if (ifp->if_l2com != NULL) {
651 		KASSERT(ng_ether_input_p != NULL,
652 		    ("%s: ng_ether_input_p is NULL", __func__));
653 		m->m_flags &= ~M_PROMISC;
654 		(*ng_ether_input_p)(ifp, &m);
655 		if (m == NULL) {
656 			CURVNET_RESTORE();
657 			return;
658 		}
659 		eh = mtod(m, struct ether_header *);
660 	}
661 
662 	/*
663 	 * Allow if_bridge(4) to claim this frame.
664 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
665 	 * and the frame should be delivered locally.
666 	 */
667 	if (ifp->if_bridge != NULL) {
668 		m->m_flags &= ~M_PROMISC;
669 		BRIDGE_INPUT(ifp, m);
670 		if (m == NULL) {
671 			CURVNET_RESTORE();
672 			return;
673 		}
674 		eh = mtod(m, struct ether_header *);
675 	}
676 
677 #if defined(INET) || defined(INET6)
678 	/*
679 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
680 	 * mbuf flows up to Layer 3.
681 	 * FreeBSD's implementation of carp(4) uses the inprotosw
682 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
683 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
684 	 * is outside the scope of the M_PROMISC test below.
685 	 * TODO: Maintain a hash table of ethernet addresses other than
686 	 * ether_dhost which may be active on this ifp.
687 	 */
688 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
689 		m->m_flags &= ~M_PROMISC;
690 	} else
691 #endif
692 	{
693 		/*
694 		 * If the frame received was not for our MAC address, set the
695 		 * M_PROMISC flag on the mbuf chain. The frame may need to
696 		 * be seen by the rest of the Ethernet input path in case of
697 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
698 		 * seen by upper protocol layers.
699 		 */
700 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
701 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
702 			m->m_flags |= M_PROMISC;
703 	}
704 
705 	ether_demux(ifp, m);
706 	CURVNET_RESTORE();
707 }
708 
709 /*
710  * Ethernet input dispatch; by default, direct dispatch here regardless of
711  * global configuration.  However, if RSS is enabled, hook up RSS affinity
712  * so that when deferred or hybrid dispatch is enabled, we can redistribute
713  * load based on RSS.
714  *
715  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
716  * not it had already done work distribution via multi-queue.  Then we could
717  * direct dispatch in the event load balancing was already complete and
718  * handle the case of interfaces with different capabilities better.
719  *
720  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
721  * at multiple layers?
722  *
723  * XXXRW: For now, enable all this only if RSS is compiled in, although it
724  * works fine without RSS.  Need to characterise the performance overhead
725  * of the detour through the netisr code in the event the result is always
726  * direct dispatch.
727  */
728 static void
729 ether_nh_input(struct mbuf *m)
730 {
731 
732 	M_ASSERTPKTHDR(m);
733 	KASSERT(m->m_pkthdr.rcvif != NULL,
734 	    ("%s: NULL interface pointer", __func__));
735 	ether_input_internal(m->m_pkthdr.rcvif, m);
736 }
737 
738 static struct netisr_handler	ether_nh = {
739 	.nh_name = "ether",
740 	.nh_handler = ether_nh_input,
741 	.nh_proto = NETISR_ETHER,
742 #ifdef RSS
743 	.nh_policy = NETISR_POLICY_CPU,
744 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
745 	.nh_m2cpuid = rss_m2cpuid,
746 #else
747 	.nh_policy = NETISR_POLICY_SOURCE,
748 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
749 #endif
750 };
751 
752 static void
753 ether_init(__unused void *arg)
754 {
755 
756 	netisr_register(&ether_nh);
757 }
758 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
759 
760 static void
761 vnet_ether_init(__unused void *arg)
762 {
763 	struct pfil_head_args args;
764 
765 	args.pa_version = PFIL_VERSION;
766 	args.pa_flags = PFIL_IN | PFIL_OUT;
767 	args.pa_type = PFIL_TYPE_ETHERNET;
768 	args.pa_headname = PFIL_ETHER_NAME;
769 	V_link_pfil_head = pfil_head_register(&args);
770 
771 #ifdef VIMAGE
772 	netisr_register_vnet(&ether_nh);
773 #endif
774 }
775 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
776     vnet_ether_init, NULL);
777 
778 #ifdef VIMAGE
779 static void
780 vnet_ether_pfil_destroy(__unused void *arg)
781 {
782 
783 	pfil_head_unregister(V_link_pfil_head);
784 }
785 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
786     vnet_ether_pfil_destroy, NULL);
787 
788 static void
789 vnet_ether_destroy(__unused void *arg)
790 {
791 
792 	netisr_unregister_vnet(&ether_nh);
793 }
794 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
795     vnet_ether_destroy, NULL);
796 #endif
797 
798 
799 
800 static void
801 ether_input(struct ifnet *ifp, struct mbuf *m)
802 {
803 	struct epoch_tracker et;
804 	struct mbuf *mn;
805 
806 	/*
807 	 * The drivers are allowed to pass in a chain of packets linked with
808 	 * m_nextpkt. We split them up into separate packets here and pass
809 	 * them up. This allows the drivers to amortize the receive lock.
810 	 */
811 	CURVNET_SET_QUIET(ifp->if_vnet);
812 	NET_EPOCH_ENTER(et);
813 	while (m) {
814 		mn = m->m_nextpkt;
815 		m->m_nextpkt = NULL;
816 
817 		/*
818 		 * We will rely on rcvif being set properly in the deferred
819 		 * context, so assert it is correct here.
820 		 */
821 		MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
822 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
823 		    "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
824 		netisr_dispatch(NETISR_ETHER, m);
825 		m = mn;
826 	}
827 	NET_EPOCH_EXIT(et);
828 	CURVNET_RESTORE();
829 }
830 
831 /*
832  * Upper layer processing for a received Ethernet packet.
833  */
834 void
835 ether_demux(struct ifnet *ifp, struct mbuf *m)
836 {
837 	struct ether_header *eh;
838 	int i, isr;
839 	u_short ether_type;
840 
841 	NET_EPOCH_ASSERT();
842 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
843 
844 	/* Do not grab PROMISC frames in case we are re-entered. */
845 	if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
846 		i = pfil_run_hooks(V_link_pfil_head, &m, ifp, PFIL_IN, NULL);
847 		if (i != 0 || m == NULL)
848 			return;
849 	}
850 
851 	eh = mtod(m, struct ether_header *);
852 	ether_type = ntohs(eh->ether_type);
853 
854 	/*
855 	 * If this frame has a VLAN tag other than 0, call vlan_input()
856 	 * if its module is loaded. Otherwise, drop.
857 	 */
858 	if ((m->m_flags & M_VLANTAG) &&
859 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
860 		if (ifp->if_vlantrunk == NULL) {
861 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
862 			m_freem(m);
863 			return;
864 		}
865 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
866 		    __func__));
867 		/* Clear before possibly re-entering ether_input(). */
868 		m->m_flags &= ~M_PROMISC;
869 		(*vlan_input_p)(ifp, m);
870 		return;
871 	}
872 
873 	/*
874 	 * Pass promiscuously received frames to the upper layer if the user
875 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
876 	 */
877 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
878 		m_freem(m);
879 		return;
880 	}
881 
882 	/*
883 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
884 	 * Strip off Ethernet header.
885 	 */
886 	m->m_flags &= ~M_VLANTAG;
887 	m_clrprotoflags(m);
888 	m_adj(m, ETHER_HDR_LEN);
889 
890 	/*
891 	 * Dispatch frame to upper layer.
892 	 */
893 	switch (ether_type) {
894 #ifdef INET
895 	case ETHERTYPE_IP:
896 		isr = NETISR_IP;
897 		break;
898 
899 	case ETHERTYPE_ARP:
900 		if (ifp->if_flags & IFF_NOARP) {
901 			/* Discard packet if ARP is disabled on interface */
902 			m_freem(m);
903 			return;
904 		}
905 		isr = NETISR_ARP;
906 		break;
907 #endif
908 #ifdef INET6
909 	case ETHERTYPE_IPV6:
910 		isr = NETISR_IPV6;
911 		break;
912 #endif
913 	default:
914 		goto discard;
915 	}
916 	netisr_dispatch(isr, m);
917 	return;
918 
919 discard:
920 	/*
921 	 * Packet is to be discarded.  If netgraph is present,
922 	 * hand the packet to it for last chance processing;
923 	 * otherwise dispose of it.
924 	 */
925 	if (ifp->if_l2com != NULL) {
926 		KASSERT(ng_ether_input_orphan_p != NULL,
927 		    ("ng_ether_input_orphan_p is NULL"));
928 		/*
929 		 * Put back the ethernet header so netgraph has a
930 		 * consistent view of inbound packets.
931 		 */
932 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
933 		(*ng_ether_input_orphan_p)(ifp, m);
934 		return;
935 	}
936 	m_freem(m);
937 }
938 
939 /*
940  * Convert Ethernet address to printable (loggable) representation.
941  * This routine is for compatibility; it's better to just use
942  *
943  *	printf("%6D", <pointer to address>, ":");
944  *
945  * since there's no static buffer involved.
946  */
947 char *
948 ether_sprintf(const u_char *ap)
949 {
950 	static char etherbuf[18];
951 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
952 	return (etherbuf);
953 }
954 
955 /*
956  * Perform common duties while attaching to interface list
957  */
958 void
959 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
960 {
961 	int i;
962 	struct ifaddr *ifa;
963 	struct sockaddr_dl *sdl;
964 
965 	ifp->if_addrlen = ETHER_ADDR_LEN;
966 	ifp->if_hdrlen = ETHER_HDR_LEN;
967 	if_attach(ifp);
968 	ifp->if_mtu = ETHERMTU;
969 	ifp->if_output = ether_output;
970 	ifp->if_input = ether_input;
971 	ifp->if_resolvemulti = ether_resolvemulti;
972 	ifp->if_requestencap = ether_requestencap;
973 #ifdef VIMAGE
974 	ifp->if_reassign = ether_reassign;
975 #endif
976 	if (ifp->if_baudrate == 0)
977 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
978 	ifp->if_broadcastaddr = etherbroadcastaddr;
979 
980 	ifa = ifp->if_addr;
981 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
982 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
983 	sdl->sdl_type = IFT_ETHER;
984 	sdl->sdl_alen = ifp->if_addrlen;
985 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
986 
987 	if (ifp->if_hw_addr != NULL)
988 		bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
989 
990 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
991 	if (ng_ether_attach_p != NULL)
992 		(*ng_ether_attach_p)(ifp);
993 
994 	/* Announce Ethernet MAC address if non-zero. */
995 	for (i = 0; i < ifp->if_addrlen; i++)
996 		if (lla[i] != 0)
997 			break;
998 	if (i != ifp->if_addrlen)
999 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1000 
1001 	uuid_ether_add(LLADDR(sdl));
1002 
1003 	/* Add necessary bits are setup; announce it now. */
1004 	EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1005 	if (IS_DEFAULT_VNET(curvnet))
1006 		devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1007 }
1008 
1009 /*
1010  * Perform common duties while detaching an Ethernet interface
1011  */
1012 void
1013 ether_ifdetach(struct ifnet *ifp)
1014 {
1015 	struct sockaddr_dl *sdl;
1016 
1017 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1018 	uuid_ether_del(LLADDR(sdl));
1019 
1020 	if (ifp->if_l2com != NULL) {
1021 		KASSERT(ng_ether_detach_p != NULL,
1022 		    ("ng_ether_detach_p is NULL"));
1023 		(*ng_ether_detach_p)(ifp);
1024 	}
1025 
1026 	bpfdetach(ifp);
1027 	if_detach(ifp);
1028 }
1029 
1030 #ifdef VIMAGE
1031 void
1032 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1033 {
1034 
1035 	if (ifp->if_l2com != NULL) {
1036 		KASSERT(ng_ether_detach_p != NULL,
1037 		    ("ng_ether_detach_p is NULL"));
1038 		(*ng_ether_detach_p)(ifp);
1039 	}
1040 
1041 	if (ng_ether_attach_p != NULL) {
1042 		CURVNET_SET_QUIET(new_vnet);
1043 		(*ng_ether_attach_p)(ifp);
1044 		CURVNET_RESTORE();
1045 	}
1046 }
1047 #endif
1048 
1049 SYSCTL_DECL(_net_link);
1050 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
1051 
1052 #if 0
1053 /*
1054  * This is for reference.  We have a table-driven version
1055  * of the little-endian crc32 generator, which is faster
1056  * than the double-loop.
1057  */
1058 uint32_t
1059 ether_crc32_le(const uint8_t *buf, size_t len)
1060 {
1061 	size_t i;
1062 	uint32_t crc;
1063 	int bit;
1064 	uint8_t data;
1065 
1066 	crc = 0xffffffff;	/* initial value */
1067 
1068 	for (i = 0; i < len; i++) {
1069 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1070 			carry = (crc ^ data) & 1;
1071 			crc >>= 1;
1072 			if (carry)
1073 				crc = (crc ^ ETHER_CRC_POLY_LE);
1074 		}
1075 	}
1076 
1077 	return (crc);
1078 }
1079 #else
1080 uint32_t
1081 ether_crc32_le(const uint8_t *buf, size_t len)
1082 {
1083 	static const uint32_t crctab[] = {
1084 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1085 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1086 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1087 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1088 	};
1089 	size_t i;
1090 	uint32_t crc;
1091 
1092 	crc = 0xffffffff;	/* initial value */
1093 
1094 	for (i = 0; i < len; i++) {
1095 		crc ^= buf[i];
1096 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1097 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1098 	}
1099 
1100 	return (crc);
1101 }
1102 #endif
1103 
1104 uint32_t
1105 ether_crc32_be(const uint8_t *buf, size_t len)
1106 {
1107 	size_t i;
1108 	uint32_t crc, carry;
1109 	int bit;
1110 	uint8_t data;
1111 
1112 	crc = 0xffffffff;	/* initial value */
1113 
1114 	for (i = 0; i < len; i++) {
1115 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1116 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1117 			crc <<= 1;
1118 			if (carry)
1119 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1120 		}
1121 	}
1122 
1123 	return (crc);
1124 }
1125 
1126 int
1127 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1128 {
1129 	struct ifaddr *ifa = (struct ifaddr *) data;
1130 	struct ifreq *ifr = (struct ifreq *) data;
1131 	int error = 0;
1132 
1133 	switch (command) {
1134 	case SIOCSIFADDR:
1135 		ifp->if_flags |= IFF_UP;
1136 
1137 		switch (ifa->ifa_addr->sa_family) {
1138 #ifdef INET
1139 		case AF_INET:
1140 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1141 			arp_ifinit(ifp, ifa);
1142 			break;
1143 #endif
1144 		default:
1145 			ifp->if_init(ifp->if_softc);
1146 			break;
1147 		}
1148 		break;
1149 
1150 	case SIOCGIFADDR:
1151 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1152 		    ETHER_ADDR_LEN);
1153 		break;
1154 
1155 	case SIOCSIFMTU:
1156 		/*
1157 		 * Set the interface MTU.
1158 		 */
1159 		if (ifr->ifr_mtu > ETHERMTU) {
1160 			error = EINVAL;
1161 		} else {
1162 			ifp->if_mtu = ifr->ifr_mtu;
1163 		}
1164 		break;
1165 
1166 	case SIOCSLANPCP:
1167 		error = priv_check(curthread, PRIV_NET_SETLANPCP);
1168 		if (error != 0)
1169 			break;
1170 		if (ifr->ifr_lan_pcp > 7 &&
1171 		    ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1172 			error = EINVAL;
1173 		} else {
1174 			ifp->if_pcp = ifr->ifr_lan_pcp;
1175 			/* broadcast event about PCP change */
1176 			EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1177 		}
1178 		break;
1179 
1180 	case SIOCGLANPCP:
1181 		ifr->ifr_lan_pcp = ifp->if_pcp;
1182 		break;
1183 
1184 	default:
1185 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1186 		break;
1187 	}
1188 	return (error);
1189 }
1190 
1191 static int
1192 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1193 	struct sockaddr *sa)
1194 {
1195 	struct sockaddr_dl *sdl;
1196 #ifdef INET
1197 	struct sockaddr_in *sin;
1198 #endif
1199 #ifdef INET6
1200 	struct sockaddr_in6 *sin6;
1201 #endif
1202 	u_char *e_addr;
1203 
1204 	switch(sa->sa_family) {
1205 	case AF_LINK:
1206 		/*
1207 		 * No mapping needed. Just check that it's a valid MC address.
1208 		 */
1209 		sdl = (struct sockaddr_dl *)sa;
1210 		e_addr = LLADDR(sdl);
1211 		if (!ETHER_IS_MULTICAST(e_addr))
1212 			return EADDRNOTAVAIL;
1213 		*llsa = NULL;
1214 		return 0;
1215 
1216 #ifdef INET
1217 	case AF_INET:
1218 		sin = (struct sockaddr_in *)sa;
1219 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1220 			return EADDRNOTAVAIL;
1221 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1222 		sdl->sdl_alen = ETHER_ADDR_LEN;
1223 		e_addr = LLADDR(sdl);
1224 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1225 		*llsa = (struct sockaddr *)sdl;
1226 		return 0;
1227 #endif
1228 #ifdef INET6
1229 	case AF_INET6:
1230 		sin6 = (struct sockaddr_in6 *)sa;
1231 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1232 			/*
1233 			 * An IP6 address of 0 means listen to all
1234 			 * of the Ethernet multicast address used for IP6.
1235 			 * (This is used for multicast routers.)
1236 			 */
1237 			ifp->if_flags |= IFF_ALLMULTI;
1238 			*llsa = NULL;
1239 			return 0;
1240 		}
1241 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_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_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1247 		*llsa = (struct sockaddr *)sdl;
1248 		return 0;
1249 #endif
1250 
1251 	default:
1252 		/*
1253 		 * Well, the text isn't quite right, but it's the name
1254 		 * that counts...
1255 		 */
1256 		return EAFNOSUPPORT;
1257 	}
1258 }
1259 
1260 static moduledata_t ether_mod = {
1261 	.name = "ether",
1262 };
1263 
1264 void
1265 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1266 {
1267 	struct ether_vlan_header vlan;
1268 	struct mbuf mv, mb;
1269 
1270 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1271 	    ("%s: vlan information not present", __func__));
1272 	KASSERT(m->m_len >= sizeof(struct ether_header),
1273 	    ("%s: mbuf not large enough for header", __func__));
1274 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1275 	vlan.evl_proto = vlan.evl_encap_proto;
1276 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1277 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1278 	m->m_len -= sizeof(struct ether_header);
1279 	m->m_data += sizeof(struct ether_header);
1280 	/*
1281 	 * If a data link has been supplied by the caller, then we will need to
1282 	 * re-create a stack allocated mbuf chain with the following structure:
1283 	 *
1284 	 * (1) mbuf #1 will contain the supplied data link
1285 	 * (2) mbuf #2 will contain the vlan header
1286 	 * (3) mbuf #3 will contain the original mbuf's packet data
1287 	 *
1288 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1289 	 */
1290 	if (data != NULL) {
1291 		mv.m_next = m;
1292 		mv.m_data = (caddr_t)&vlan;
1293 		mv.m_len = sizeof(vlan);
1294 		mb.m_next = &mv;
1295 		mb.m_data = data;
1296 		mb.m_len = dlen;
1297 		bpf_mtap(bp, &mb);
1298 	} else
1299 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1300 	m->m_len += sizeof(struct ether_header);
1301 	m->m_data -= sizeof(struct ether_header);
1302 }
1303 
1304 struct mbuf *
1305 ether_vlanencap(struct mbuf *m, uint16_t tag)
1306 {
1307 	struct ether_vlan_header *evl;
1308 
1309 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1310 	if (m == NULL)
1311 		return (NULL);
1312 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1313 
1314 	if (m->m_len < sizeof(*evl)) {
1315 		m = m_pullup(m, sizeof(*evl));
1316 		if (m == NULL)
1317 			return (NULL);
1318 	}
1319 
1320 	/*
1321 	 * Transform the Ethernet header into an Ethernet header
1322 	 * with 802.1Q encapsulation.
1323 	 */
1324 	evl = mtod(m, struct ether_vlan_header *);
1325 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1326 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1327 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1328 	evl->evl_tag = htons(tag);
1329 	return (m);
1330 }
1331 
1332 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW, 0,
1333     "IEEE 802.1Q VLAN");
1334 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link, CTLFLAG_RW, 0,
1335     "for consistency");
1336 
1337 VNET_DEFINE_STATIC(int, soft_pad);
1338 #define	V_soft_pad	VNET(soft_pad)
1339 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1340     &VNET_NAME(soft_pad), 0,
1341     "pad short frames before tagging");
1342 
1343 /*
1344  * For now, make preserving PCP via an mbuf tag optional, as it increases
1345  * per-packet memory allocations and frees.  In the future, it would be
1346  * preferable to reuse ether_vtag for this, or similar.
1347  */
1348 int vlan_mtag_pcp = 0;
1349 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW,
1350     &vlan_mtag_pcp, 0,
1351     "Retain VLAN PCP information as packets are passed up the stack");
1352 
1353 bool
1354 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1355     uint16_t vid, uint8_t pcp)
1356 {
1357 	struct m_tag *mtag;
1358 	int n;
1359 	uint16_t tag;
1360 	static const char pad[8];	/* just zeros */
1361 
1362 	/*
1363 	 * Pad the frame to the minimum size allowed if told to.
1364 	 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1365 	 * paragraph C.4.4.3.b.  It can help to work around buggy
1366 	 * bridges that violate paragraph C.4.4.3.a from the same
1367 	 * document, i.e., fail to pad short frames after untagging.
1368 	 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1369 	 * untagging it will produce a 62-byte frame, which is a runt
1370 	 * and requires padding.  There are VLAN-enabled network
1371 	 * devices that just discard such runts instead or mishandle
1372 	 * them somehow.
1373 	 */
1374 	if (V_soft_pad && p->if_type == IFT_ETHER) {
1375 		for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1376 		     n > 0; n -= sizeof(pad)) {
1377 			if (!m_append(*mp, min(n, sizeof(pad)), pad))
1378 				break;
1379 		}
1380 		if (n > 0) {
1381 			m_freem(*mp);
1382 			*mp = NULL;
1383 			if_printf(ife, "cannot pad short frame");
1384 			return (false);
1385 		}
1386 	}
1387 
1388 	/*
1389 	 * If underlying interface can do VLAN tag insertion itself,
1390 	 * just pass the packet along. However, we need some way to
1391 	 * tell the interface where the packet came from so that it
1392 	 * knows how to find the VLAN tag to use, so we attach a
1393 	 * packet tag that holds it.
1394 	 */
1395 	if (vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1396 	    MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1397 		tag = EVL_MAKETAG(vid, *(uint8_t *)(mtag + 1), 0);
1398 	else
1399 		tag = EVL_MAKETAG(vid, pcp, 0);
1400 	if (p->if_capenable & IFCAP_VLAN_HWTAGGING) {
1401 		(*mp)->m_pkthdr.ether_vtag = tag;
1402 		(*mp)->m_flags |= M_VLANTAG;
1403 	} else {
1404 		*mp = ether_vlanencap(*mp, tag);
1405 		if (*mp == NULL) {
1406 			if_printf(ife, "unable to prepend 802.1Q header");
1407 			return (false);
1408 		}
1409 	}
1410 	return (true);
1411 }
1412 
1413 /*
1414  * Allocate an address from the FreeBSD Foundation OUI.  This uses a
1415  * cryptographic hash function on the containing jail's UUID and the interface
1416  * name to attempt to provide a unique but stable address.  Pseudo-interfaces
1417  * which require a MAC address should use this function to allocate
1418  * non-locally-administered addresses.
1419  */
1420 void
1421 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1422 {
1423 #define	ETHER_GEN_ADDR_BUFSIZ	HOSTUUIDLEN + IFNAMSIZ + 2
1424 	SHA1_CTX ctx;
1425 	char buf[ETHER_GEN_ADDR_BUFSIZ];
1426 	char uuid[HOSTUUIDLEN + 1];
1427 	uint64_t addr;
1428 	int i, sz;
1429 	char digest[SHA1_RESULTLEN];
1430 
1431 	getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1432 	sz = snprintf(buf, ETHER_GEN_ADDR_BUFSIZ, "%s-%s", uuid, ifp->if_xname);
1433 	SHA1Init(&ctx);
1434 	SHA1Update(&ctx, buf, sz);
1435 	SHA1Final(digest, &ctx);
1436 
1437 	addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1438 	    OUI_FREEBSD_GENERATED_MASK;
1439 	addr = OUI_FREEBSD(addr);
1440 	for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1441 		hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1442 		    0xFF;
1443 	}
1444 }
1445 
1446 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1447 MODULE_VERSION(ether, 1);
1448