xref: /freebsd/sys/net/if_ethersubr.c (revision d9f0ce31900a48d1a2bfc1c8c86f79d1e831451a)
1 /*-
2  * Copyright (c) 1982, 1989, 1993
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
30  * $FreeBSD$
31  */
32 
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 #include "opt_netgraph.h"
36 #include "opt_mbuf_profiling.h"
37 #include "opt_rss.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/module.h>
45 #include <sys/mbuf.h>
46 #include <sys/random.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/sysctl.h>
50 #include <sys/uuid.h>
51 
52 #include <net/if.h>
53 #include <net/if_var.h>
54 #include <net/if_arp.h>
55 #include <net/netisr.h>
56 #include <net/route.h>
57 #include <net/if_llc.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/bpf.h>
61 #include <net/ethernet.h>
62 #include <net/if_bridgevar.h>
63 #include <net/if_vlan_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/pfil.h>
66 #include <net/rss_config.h>
67 #include <net/vnet.h>
68 
69 #include <netpfil/pf/pf_mtag.h>
70 
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/if_ether.h>
75 #include <netinet/ip_carp.h>
76 #include <netinet/ip_var.h>
77 #endif
78 #ifdef INET6
79 #include <netinet6/nd6.h>
80 #endif
81 #include <security/mac/mac_framework.h>
82 
83 #ifdef CTASSERT
84 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
85 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
86 #endif
87 
88 VNET_DEFINE(struct pfil_head, link_pfil_hook);	/* Packet filter hooks */
89 
90 /* netgraph node hooks for ng_ether(4) */
91 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
92 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
93 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
94 void	(*ng_ether_attach_p)(struct ifnet *ifp);
95 void	(*ng_ether_detach_p)(struct ifnet *ifp);
96 
97 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
98 
99 /* if_bridge(4) support */
100 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
101 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
102 		struct sockaddr *, struct rtentry *);
103 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
104 
105 /* if_lagg(4) support */
106 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
107 
108 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
109 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
110 
111 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
112 		struct sockaddr *);
113 #ifdef VIMAGE
114 static	void ether_reassign(struct ifnet *, struct vnet *, char *);
115 #endif
116 static	int ether_requestencap(struct ifnet *, struct if_encap_req *);
117 
118 #define	ETHER_IS_BROADCAST(addr) \
119 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
120 
121 #define senderr(e) do { error = (e); goto bad;} while (0)
122 
123 static void
124 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
125 {
126 	int csum_flags = 0;
127 
128 	if (src->m_pkthdr.csum_flags & CSUM_IP)
129 		csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
130 	if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
131 		csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
132 	if (src->m_pkthdr.csum_flags & CSUM_SCTP)
133 		csum_flags |= CSUM_SCTP_VALID;
134 	dst->m_pkthdr.csum_flags |= csum_flags;
135 	if (csum_flags & CSUM_DATA_VALID)
136 		dst->m_pkthdr.csum_data = 0xffff;
137 }
138 
139 /*
140  * Handle link-layer encapsulation requests.
141  */
142 static int
143 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
144 {
145 	struct ether_header *eh;
146 	struct arphdr *ah;
147 	uint16_t etype;
148 	const u_char *lladdr;
149 
150 	if (req->rtype != IFENCAP_LL)
151 		return (EOPNOTSUPP);
152 
153 	if (req->bufsize < ETHER_HDR_LEN)
154 		return (ENOMEM);
155 
156 	eh = (struct ether_header *)req->buf;
157 	lladdr = req->lladdr;
158 	req->lladdr_off = 0;
159 
160 	switch (req->family) {
161 	case AF_INET:
162 		etype = htons(ETHERTYPE_IP);
163 		break;
164 	case AF_INET6:
165 		etype = htons(ETHERTYPE_IPV6);
166 		break;
167 	case AF_ARP:
168 		ah = (struct arphdr *)req->hdata;
169 		ah->ar_hrd = htons(ARPHRD_ETHER);
170 
171 		switch(ntohs(ah->ar_op)) {
172 		case ARPOP_REVREQUEST:
173 		case ARPOP_REVREPLY:
174 			etype = htons(ETHERTYPE_REVARP);
175 			break;
176 		case ARPOP_REQUEST:
177 		case ARPOP_REPLY:
178 		default:
179 			etype = htons(ETHERTYPE_ARP);
180 			break;
181 		}
182 
183 		if (req->flags & IFENCAP_FLAG_BROADCAST)
184 			lladdr = ifp->if_broadcastaddr;
185 		break;
186 	default:
187 		return (EAFNOSUPPORT);
188 	}
189 
190 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
191 	memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
192 	memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
193 	req->bufsize = sizeof(struct ether_header);
194 
195 	return (0);
196 }
197 
198 
199 static int
200 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
201 	const struct sockaddr *dst, struct route *ro, u_char *phdr,
202 	uint32_t *pflags)
203 {
204 	struct ether_header *eh;
205 	uint32_t lleflags = 0;
206 	int error = 0;
207 #if defined(INET) || defined(INET6)
208 	uint16_t etype;
209 #endif
210 
211 	eh = (struct ether_header *)phdr;
212 
213 	switch (dst->sa_family) {
214 #ifdef INET
215 	case AF_INET:
216 		if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
217 			error = arpresolve(ifp, 0, m, dst, phdr, &lleflags);
218 		else {
219 			if (m->m_flags & M_BCAST)
220 				memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
221 				    ETHER_ADDR_LEN);
222 			else {
223 				const struct in_addr *a;
224 				a = &(((const struct sockaddr_in *)dst)->sin_addr);
225 				ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
226 			}
227 			etype = htons(ETHERTYPE_IP);
228 			memcpy(&eh->ether_type, &etype, sizeof(etype));
229 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
230 		}
231 		break;
232 #endif
233 #ifdef INET6
234 	case AF_INET6:
235 		if ((m->m_flags & M_MCAST) == 0)
236 			error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags);
237 		else {
238 			const struct in6_addr *a6;
239 			a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
240 			ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
241 			etype = htons(ETHERTYPE_IPV6);
242 			memcpy(&eh->ether_type, &etype, sizeof(etype));
243 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
244 		}
245 		break;
246 #endif
247 	default:
248 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
249 		if (m != NULL)
250 			m_freem(m);
251 		return (EAFNOSUPPORT);
252 	}
253 
254 	if (error == EHOSTDOWN) {
255 		if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
256 			error = EHOSTUNREACH;
257 	}
258 
259 	if (error != 0)
260 		return (error);
261 
262 	*pflags = RT_MAY_LOOP;
263 	if (lleflags & LLE_IFADDR)
264 		*pflags |= RT_L2_ME;
265 
266 	return (0);
267 }
268 
269 /*
270  * Ethernet output routine.
271  * Encapsulate a packet of type family for the local net.
272  * Use trailer local net encapsulation if enough data in first
273  * packet leaves a multiple of 512 bytes of data in remainder.
274  */
275 int
276 ether_output(struct ifnet *ifp, struct mbuf *m,
277 	const struct sockaddr *dst, struct route *ro)
278 {
279 	int error = 0;
280 	char linkhdr[ETHER_HDR_LEN], *phdr;
281 	struct ether_header *eh;
282 	struct pf_mtag *t;
283 	int loop_copy = 1;
284 	int hlen;	/* link layer header length */
285 	uint32_t pflags;
286 
287 	phdr = NULL;
288 	pflags = 0;
289 	if (ro != NULL) {
290 		phdr = ro->ro_prepend;
291 		hlen = ro->ro_plen;
292 		pflags = ro->ro_flags;
293 	}
294 #ifdef MAC
295 	error = mac_ifnet_check_transmit(ifp, m);
296 	if (error)
297 		senderr(error);
298 #endif
299 
300 	M_PROFILE(m);
301 	if (ifp->if_flags & IFF_MONITOR)
302 		senderr(ENETDOWN);
303 	if (!((ifp->if_flags & IFF_UP) &&
304 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
305 		senderr(ENETDOWN);
306 
307 	if (phdr == NULL) {
308 		/* No prepend data supplied. Try to calculate ourselves. */
309 		phdr = linkhdr;
310 		hlen = ETHER_HDR_LEN;
311 		error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags);
312 		if (error != 0)
313 			return (error == EWOULDBLOCK ? 0 : error);
314 	}
315 
316 	if ((pflags & RT_L2_ME) != 0) {
317 		update_mbuf_csumflags(m, m);
318 		return (if_simloop(ifp, m, dst->sa_family, 0));
319 	}
320 	loop_copy = pflags & RT_MAY_LOOP;
321 
322 	/*
323 	 * Add local net header.  If no space in first mbuf,
324 	 * allocate another.
325 	 *
326 	 * Note that we do prepend regardless of RT_HAS_HEADER flag.
327 	 * This is done because BPF code shifts m_data pointer
328 	 * to the end of ethernet header prior to calling if_output().
329 	 */
330 	M_PREPEND(m, hlen, M_NOWAIT);
331 	if (m == NULL)
332 		senderr(ENOBUFS);
333 	if ((pflags & RT_HAS_HEADER) == 0) {
334 		eh = mtod(m, struct ether_header *);
335 		memcpy(eh, phdr, hlen);
336 	}
337 
338 	/*
339 	 * If a simplex interface, and the packet is being sent to our
340 	 * Ethernet address or a broadcast address, loopback a copy.
341 	 * XXX To make a simplex device behave exactly like a duplex
342 	 * device, we should copy in the case of sending to our own
343 	 * ethernet address (thus letting the original actually appear
344 	 * on the wire). However, we don't do that here for security
345 	 * reasons and compatibility with the original behavior.
346 	 */
347 	if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
348 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
349 		struct mbuf *n;
350 
351 		/*
352 		 * Because if_simloop() modifies the packet, we need a
353 		 * writable copy through m_dup() instead of a readonly
354 		 * one as m_copy[m] would give us. The alternative would
355 		 * be to modify if_simloop() to handle the readonly mbuf,
356 		 * but performancewise it is mostly equivalent (trading
357 		 * extra data copying vs. extra locking).
358 		 *
359 		 * XXX This is a local workaround.  A number of less
360 		 * often used kernel parts suffer from the same bug.
361 		 * See PR kern/105943 for a proposed general solution.
362 		 */
363 		if ((n = m_dup(m, M_NOWAIT)) != NULL) {
364 			update_mbuf_csumflags(m, n);
365 			(void)if_simloop(ifp, n, dst->sa_family, hlen);
366 		} else
367 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
368 	}
369 
370        /*
371 	* Bridges require special output handling.
372 	*/
373 	if (ifp->if_bridge) {
374 		BRIDGE_OUTPUT(ifp, m, error);
375 		return (error);
376 	}
377 
378 #if defined(INET) || defined(INET6)
379 	if (ifp->if_carp &&
380 	    (error = (*carp_output_p)(ifp, m, dst)))
381 		goto bad;
382 #endif
383 
384 	/* Handle ng_ether(4) processing, if any */
385 	if (ifp->if_l2com != NULL) {
386 		KASSERT(ng_ether_output_p != NULL,
387 		    ("ng_ether_output_p is NULL"));
388 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
389 bad:			if (m != NULL)
390 				m_freem(m);
391 			return (error);
392 		}
393 		if (m == NULL)
394 			return (0);
395 	}
396 
397 	/* Continue with link-layer output */
398 	return ether_output_frame(ifp, m);
399 }
400 
401 /*
402  * Ethernet link layer output routine to send a raw frame to the device.
403  *
404  * This assumes that the 14 byte Ethernet header is present and contiguous
405  * in the first mbuf (if BRIDGE'ing).
406  */
407 int
408 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
409 {
410 	int i;
411 
412 	if (PFIL_HOOKED(&V_link_pfil_hook)) {
413 		i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL);
414 
415 		if (i != 0)
416 			return (EACCES);
417 
418 		if (m == NULL)
419 			return (0);
420 	}
421 
422 	/*
423 	 * Queue message on interface, update output statistics if
424 	 * successful, and start output if interface not yet active.
425 	 */
426 	return ((ifp->if_transmit)(ifp, m));
427 }
428 
429 #if defined(INET) || defined(INET6)
430 #endif
431 
432 /*
433  * Process a received Ethernet packet; the packet is in the
434  * mbuf chain m with the ethernet header at the front.
435  */
436 static void
437 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
438 {
439 	struct ether_header *eh;
440 	u_short etype;
441 
442 	if ((ifp->if_flags & IFF_UP) == 0) {
443 		m_freem(m);
444 		return;
445 	}
446 #ifdef DIAGNOSTIC
447 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
448 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
449 		m_freem(m);
450 		return;
451 	}
452 #endif
453 	if (m->m_len < ETHER_HDR_LEN) {
454 		/* XXX maybe should pullup? */
455 		if_printf(ifp, "discard frame w/o leading ethernet "
456 				"header (len %u pkt len %u)\n",
457 				m->m_len, m->m_pkthdr.len);
458 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
459 		m_freem(m);
460 		return;
461 	}
462 	eh = mtod(m, struct ether_header *);
463 	etype = ntohs(eh->ether_type);
464 	random_harvest_queue(m, sizeof(*m), 2, RANDOM_NET_ETHER);
465 
466 	CURVNET_SET_QUIET(ifp->if_vnet);
467 
468 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
469 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
470 			m->m_flags |= M_BCAST;
471 		else
472 			m->m_flags |= M_MCAST;
473 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
474 	}
475 
476 #ifdef MAC
477 	/*
478 	 * Tag the mbuf with an appropriate MAC label before any other
479 	 * consumers can get to it.
480 	 */
481 	mac_ifnet_create_mbuf(ifp, m);
482 #endif
483 
484 	/*
485 	 * Give bpf a chance at the packet.
486 	 */
487 	ETHER_BPF_MTAP(ifp, m);
488 
489 	/*
490 	 * If the CRC is still on the packet, trim it off. We do this once
491 	 * and once only in case we are re-entered. Nothing else on the
492 	 * Ethernet receive path expects to see the FCS.
493 	 */
494 	if (m->m_flags & M_HASFCS) {
495 		m_adj(m, -ETHER_CRC_LEN);
496 		m->m_flags &= ~M_HASFCS;
497 	}
498 
499 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
500 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
501 
502 	/* Allow monitor mode to claim this frame, after stats are updated. */
503 	if (ifp->if_flags & IFF_MONITOR) {
504 		m_freem(m);
505 		CURVNET_RESTORE();
506 		return;
507 	}
508 
509 	/* Handle input from a lagg(4) port */
510 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
511 		KASSERT(lagg_input_p != NULL,
512 		    ("%s: if_lagg not loaded!", __func__));
513 		m = (*lagg_input_p)(ifp, m);
514 		if (m != NULL)
515 			ifp = m->m_pkthdr.rcvif;
516 		else {
517 			CURVNET_RESTORE();
518 			return;
519 		}
520 	}
521 
522 	/*
523 	 * If the hardware did not process an 802.1Q tag, do this now,
524 	 * to allow 802.1P priority frames to be passed to the main input
525 	 * path correctly.
526 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
527 	 */
528 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
529 		struct ether_vlan_header *evl;
530 
531 		if (m->m_len < sizeof(*evl) &&
532 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
533 #ifdef DIAGNOSTIC
534 			if_printf(ifp, "cannot pullup VLAN header\n");
535 #endif
536 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
537 			CURVNET_RESTORE();
538 			return;
539 		}
540 
541 		evl = mtod(m, struct ether_vlan_header *);
542 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
543 		m->m_flags |= M_VLANTAG;
544 
545 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
546 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
547 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
548 		eh = mtod(m, struct ether_header *);
549 	}
550 
551 	M_SETFIB(m, ifp->if_fib);
552 
553 	/* Allow ng_ether(4) to claim this frame. */
554 	if (ifp->if_l2com != NULL) {
555 		KASSERT(ng_ether_input_p != NULL,
556 		    ("%s: ng_ether_input_p is NULL", __func__));
557 		m->m_flags &= ~M_PROMISC;
558 		(*ng_ether_input_p)(ifp, &m);
559 		if (m == NULL) {
560 			CURVNET_RESTORE();
561 			return;
562 		}
563 		eh = mtod(m, struct ether_header *);
564 	}
565 
566 	/*
567 	 * Allow if_bridge(4) to claim this frame.
568 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
569 	 * and the frame should be delivered locally.
570 	 */
571 	if (ifp->if_bridge != NULL) {
572 		m->m_flags &= ~M_PROMISC;
573 		BRIDGE_INPUT(ifp, m);
574 		if (m == NULL) {
575 			CURVNET_RESTORE();
576 			return;
577 		}
578 		eh = mtod(m, struct ether_header *);
579 	}
580 
581 #if defined(INET) || defined(INET6)
582 	/*
583 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
584 	 * mbuf flows up to Layer 3.
585 	 * FreeBSD's implementation of carp(4) uses the inprotosw
586 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
587 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
588 	 * is outside the scope of the M_PROMISC test below.
589 	 * TODO: Maintain a hash table of ethernet addresses other than
590 	 * ether_dhost which may be active on this ifp.
591 	 */
592 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
593 		m->m_flags &= ~M_PROMISC;
594 	} else
595 #endif
596 	{
597 		/*
598 		 * If the frame received was not for our MAC address, set the
599 		 * M_PROMISC flag on the mbuf chain. The frame may need to
600 		 * be seen by the rest of the Ethernet input path in case of
601 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
602 		 * seen by upper protocol layers.
603 		 */
604 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
605 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
606 			m->m_flags |= M_PROMISC;
607 	}
608 
609 	ether_demux(ifp, m);
610 	CURVNET_RESTORE();
611 }
612 
613 /*
614  * Ethernet input dispatch; by default, direct dispatch here regardless of
615  * global configuration.  However, if RSS is enabled, hook up RSS affinity
616  * so that when deferred or hybrid dispatch is enabled, we can redistribute
617  * load based on RSS.
618  *
619  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
620  * not it had already done work distribution via multi-queue.  Then we could
621  * direct dispatch in the event load balancing was already complete and
622  * handle the case of interfaces with different capabilities better.
623  *
624  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
625  * at multiple layers?
626  *
627  * XXXRW: For now, enable all this only if RSS is compiled in, although it
628  * works fine without RSS.  Need to characterise the performance overhead
629  * of the detour through the netisr code in the event the result is always
630  * direct dispatch.
631  */
632 static void
633 ether_nh_input(struct mbuf *m)
634 {
635 
636 	M_ASSERTPKTHDR(m);
637 	KASSERT(m->m_pkthdr.rcvif != NULL,
638 	    ("%s: NULL interface pointer", __func__));
639 	ether_input_internal(m->m_pkthdr.rcvif, m);
640 }
641 
642 static struct netisr_handler	ether_nh = {
643 	.nh_name = "ether",
644 	.nh_handler = ether_nh_input,
645 	.nh_proto = NETISR_ETHER,
646 #ifdef RSS
647 	.nh_policy = NETISR_POLICY_CPU,
648 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
649 	.nh_m2cpuid = rss_m2cpuid,
650 #else
651 	.nh_policy = NETISR_POLICY_SOURCE,
652 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
653 #endif
654 };
655 
656 static void
657 ether_init(__unused void *arg)
658 {
659 
660 	netisr_register(&ether_nh);
661 }
662 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
663 
664 static void
665 vnet_ether_init(__unused void *arg)
666 {
667 	int i;
668 
669 	/* Initialize packet filter hooks. */
670 	V_link_pfil_hook.ph_type = PFIL_TYPE_AF;
671 	V_link_pfil_hook.ph_af = AF_LINK;
672 	if ((i = pfil_head_register(&V_link_pfil_hook)) != 0)
673 		printf("%s: WARNING: unable to register pfil link hook, "
674 			"error %d\n", __func__, i);
675 }
676 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
677     vnet_ether_init, NULL);
678 
679 static void
680 vnet_ether_destroy(__unused void *arg)
681 {
682 	int i;
683 
684 	if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0)
685 		printf("%s: WARNING: unable to unregister pfil link hook, "
686 			"error %d\n", __func__, i);
687 }
688 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
689     vnet_ether_destroy, NULL);
690 
691 
692 
693 static void
694 ether_input(struct ifnet *ifp, struct mbuf *m)
695 {
696 
697 	struct mbuf *mn;
698 
699 	/*
700 	 * The drivers are allowed to pass in a chain of packets linked with
701 	 * m_nextpkt. We split them up into separate packets here and pass
702 	 * them up. This allows the drivers to amortize the receive lock.
703 	 */
704 	while (m) {
705 		mn = m->m_nextpkt;
706 		m->m_nextpkt = NULL;
707 
708 		/*
709 		 * We will rely on rcvif being set properly in the deferred context,
710 		 * so assert it is correct here.
711 		 */
712 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__));
713 		netisr_dispatch(NETISR_ETHER, m);
714 		m = mn;
715 	}
716 }
717 
718 /*
719  * Upper layer processing for a received Ethernet packet.
720  */
721 void
722 ether_demux(struct ifnet *ifp, struct mbuf *m)
723 {
724 	struct ether_header *eh;
725 	int i, isr;
726 	u_short ether_type;
727 
728 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
729 
730 	/* Do not grab PROMISC frames in case we are re-entered. */
731 	if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) {
732 		i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL);
733 
734 		if (i != 0 || m == NULL)
735 			return;
736 	}
737 
738 	eh = mtod(m, struct ether_header *);
739 	ether_type = ntohs(eh->ether_type);
740 
741 	/*
742 	 * If this frame has a VLAN tag other than 0, call vlan_input()
743 	 * if its module is loaded. Otherwise, drop.
744 	 */
745 	if ((m->m_flags & M_VLANTAG) &&
746 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
747 		if (ifp->if_vlantrunk == NULL) {
748 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
749 			m_freem(m);
750 			return;
751 		}
752 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
753 		    __func__));
754 		/* Clear before possibly re-entering ether_input(). */
755 		m->m_flags &= ~M_PROMISC;
756 		(*vlan_input_p)(ifp, m);
757 		return;
758 	}
759 
760 	/*
761 	 * Pass promiscuously received frames to the upper layer if the user
762 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
763 	 */
764 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
765 		m_freem(m);
766 		return;
767 	}
768 
769 	/*
770 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
771 	 * Strip off Ethernet header.
772 	 */
773 	m->m_flags &= ~M_VLANTAG;
774 	m_clrprotoflags(m);
775 	m_adj(m, ETHER_HDR_LEN);
776 
777 	/*
778 	 * Dispatch frame to upper layer.
779 	 */
780 	switch (ether_type) {
781 #ifdef INET
782 	case ETHERTYPE_IP:
783 		isr = NETISR_IP;
784 		break;
785 
786 	case ETHERTYPE_ARP:
787 		if (ifp->if_flags & IFF_NOARP) {
788 			/* Discard packet if ARP is disabled on interface */
789 			m_freem(m);
790 			return;
791 		}
792 		isr = NETISR_ARP;
793 		break;
794 #endif
795 #ifdef INET6
796 	case ETHERTYPE_IPV6:
797 		isr = NETISR_IPV6;
798 		break;
799 #endif
800 	default:
801 		goto discard;
802 	}
803 	netisr_dispatch(isr, m);
804 	return;
805 
806 discard:
807 	/*
808 	 * Packet is to be discarded.  If netgraph is present,
809 	 * hand the packet to it for last chance processing;
810 	 * otherwise dispose of it.
811 	 */
812 	if (ifp->if_l2com != NULL) {
813 		KASSERT(ng_ether_input_orphan_p != NULL,
814 		    ("ng_ether_input_orphan_p is NULL"));
815 		/*
816 		 * Put back the ethernet header so netgraph has a
817 		 * consistent view of inbound packets.
818 		 */
819 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
820 		(*ng_ether_input_orphan_p)(ifp, m);
821 		return;
822 	}
823 	m_freem(m);
824 }
825 
826 /*
827  * Convert Ethernet address to printable (loggable) representation.
828  * This routine is for compatibility; it's better to just use
829  *
830  *	printf("%6D", <pointer to address>, ":");
831  *
832  * since there's no static buffer involved.
833  */
834 char *
835 ether_sprintf(const u_char *ap)
836 {
837 	static char etherbuf[18];
838 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
839 	return (etherbuf);
840 }
841 
842 /*
843  * Perform common duties while attaching to interface list
844  */
845 void
846 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
847 {
848 	int i;
849 	struct ifaddr *ifa;
850 	struct sockaddr_dl *sdl;
851 
852 	ifp->if_addrlen = ETHER_ADDR_LEN;
853 	ifp->if_hdrlen = ETHER_HDR_LEN;
854 	if_attach(ifp);
855 	ifp->if_mtu = ETHERMTU;
856 	ifp->if_output = ether_output;
857 	ifp->if_input = ether_input;
858 	ifp->if_resolvemulti = ether_resolvemulti;
859 	ifp->if_requestencap = ether_requestencap;
860 #ifdef VIMAGE
861 	ifp->if_reassign = ether_reassign;
862 #endif
863 	if (ifp->if_baudrate == 0)
864 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
865 	ifp->if_broadcastaddr = etherbroadcastaddr;
866 
867 	ifa = ifp->if_addr;
868 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
869 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
870 	sdl->sdl_type = IFT_ETHER;
871 	sdl->sdl_alen = ifp->if_addrlen;
872 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
873 
874 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
875 	if (ng_ether_attach_p != NULL)
876 		(*ng_ether_attach_p)(ifp);
877 
878 	/* Announce Ethernet MAC address if non-zero. */
879 	for (i = 0; i < ifp->if_addrlen; i++)
880 		if (lla[i] != 0)
881 			break;
882 	if (i != ifp->if_addrlen)
883 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
884 
885 	uuid_ether_add(LLADDR(sdl));
886 }
887 
888 /*
889  * Perform common duties while detaching an Ethernet interface
890  */
891 void
892 ether_ifdetach(struct ifnet *ifp)
893 {
894 	struct sockaddr_dl *sdl;
895 
896 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
897 	uuid_ether_del(LLADDR(sdl));
898 
899 	if (ifp->if_l2com != NULL) {
900 		KASSERT(ng_ether_detach_p != NULL,
901 		    ("ng_ether_detach_p is NULL"));
902 		(*ng_ether_detach_p)(ifp);
903 	}
904 
905 	bpfdetach(ifp);
906 	if_detach(ifp);
907 }
908 
909 #ifdef VIMAGE
910 void
911 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
912 {
913 
914 	if (ifp->if_l2com != NULL) {
915 		KASSERT(ng_ether_detach_p != NULL,
916 		    ("ng_ether_detach_p is NULL"));
917 		(*ng_ether_detach_p)(ifp);
918 	}
919 
920 	if (ng_ether_attach_p != NULL) {
921 		CURVNET_SET_QUIET(new_vnet);
922 		(*ng_ether_attach_p)(ifp);
923 		CURVNET_RESTORE();
924 	}
925 }
926 #endif
927 
928 SYSCTL_DECL(_net_link);
929 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
930 
931 #if 0
932 /*
933  * This is for reference.  We have a table-driven version
934  * of the little-endian crc32 generator, which is faster
935  * than the double-loop.
936  */
937 uint32_t
938 ether_crc32_le(const uint8_t *buf, size_t len)
939 {
940 	size_t i;
941 	uint32_t crc;
942 	int bit;
943 	uint8_t data;
944 
945 	crc = 0xffffffff;	/* initial value */
946 
947 	for (i = 0; i < len; i++) {
948 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
949 			carry = (crc ^ data) & 1;
950 			crc >>= 1;
951 			if (carry)
952 				crc = (crc ^ ETHER_CRC_POLY_LE);
953 		}
954 	}
955 
956 	return (crc);
957 }
958 #else
959 uint32_t
960 ether_crc32_le(const uint8_t *buf, size_t len)
961 {
962 	static const uint32_t crctab[] = {
963 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
964 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
965 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
966 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
967 	};
968 	size_t i;
969 	uint32_t crc;
970 
971 	crc = 0xffffffff;	/* initial value */
972 
973 	for (i = 0; i < len; i++) {
974 		crc ^= buf[i];
975 		crc = (crc >> 4) ^ crctab[crc & 0xf];
976 		crc = (crc >> 4) ^ crctab[crc & 0xf];
977 	}
978 
979 	return (crc);
980 }
981 #endif
982 
983 uint32_t
984 ether_crc32_be(const uint8_t *buf, size_t len)
985 {
986 	size_t i;
987 	uint32_t crc, carry;
988 	int bit;
989 	uint8_t data;
990 
991 	crc = 0xffffffff;	/* initial value */
992 
993 	for (i = 0; i < len; i++) {
994 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
995 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
996 			crc <<= 1;
997 			if (carry)
998 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
999 		}
1000 	}
1001 
1002 	return (crc);
1003 }
1004 
1005 int
1006 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1007 {
1008 	struct ifaddr *ifa = (struct ifaddr *) data;
1009 	struct ifreq *ifr = (struct ifreq *) data;
1010 	int error = 0;
1011 
1012 	switch (command) {
1013 	case SIOCSIFADDR:
1014 		ifp->if_flags |= IFF_UP;
1015 
1016 		switch (ifa->ifa_addr->sa_family) {
1017 #ifdef INET
1018 		case AF_INET:
1019 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1020 			arp_ifinit(ifp, ifa);
1021 			break;
1022 #endif
1023 		default:
1024 			ifp->if_init(ifp->if_softc);
1025 			break;
1026 		}
1027 		break;
1028 
1029 	case SIOCGIFADDR:
1030 		{
1031 			struct sockaddr *sa;
1032 
1033 			sa = (struct sockaddr *) & ifr->ifr_data;
1034 			bcopy(IF_LLADDR(ifp),
1035 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1036 		}
1037 		break;
1038 
1039 	case SIOCSIFMTU:
1040 		/*
1041 		 * Set the interface MTU.
1042 		 */
1043 		if (ifr->ifr_mtu > ETHERMTU) {
1044 			error = EINVAL;
1045 		} else {
1046 			ifp->if_mtu = ifr->ifr_mtu;
1047 		}
1048 		break;
1049 	default:
1050 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1051 		break;
1052 	}
1053 	return (error);
1054 }
1055 
1056 static int
1057 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1058 	struct sockaddr *sa)
1059 {
1060 	struct sockaddr_dl *sdl;
1061 #ifdef INET
1062 	struct sockaddr_in *sin;
1063 #endif
1064 #ifdef INET6
1065 	struct sockaddr_in6 *sin6;
1066 #endif
1067 	u_char *e_addr;
1068 
1069 	switch(sa->sa_family) {
1070 	case AF_LINK:
1071 		/*
1072 		 * No mapping needed. Just check that it's a valid MC address.
1073 		 */
1074 		sdl = (struct sockaddr_dl *)sa;
1075 		e_addr = LLADDR(sdl);
1076 		if (!ETHER_IS_MULTICAST(e_addr))
1077 			return EADDRNOTAVAIL;
1078 		*llsa = NULL;
1079 		return 0;
1080 
1081 #ifdef INET
1082 	case AF_INET:
1083 		sin = (struct sockaddr_in *)sa;
1084 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1085 			return EADDRNOTAVAIL;
1086 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1087 		sdl->sdl_alen = ETHER_ADDR_LEN;
1088 		e_addr = LLADDR(sdl);
1089 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1090 		*llsa = (struct sockaddr *)sdl;
1091 		return 0;
1092 #endif
1093 #ifdef INET6
1094 	case AF_INET6:
1095 		sin6 = (struct sockaddr_in6 *)sa;
1096 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1097 			/*
1098 			 * An IP6 address of 0 means listen to all
1099 			 * of the Ethernet multicast address used for IP6.
1100 			 * (This is used for multicast routers.)
1101 			 */
1102 			ifp->if_flags |= IFF_ALLMULTI;
1103 			*llsa = NULL;
1104 			return 0;
1105 		}
1106 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1107 			return EADDRNOTAVAIL;
1108 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1109 		sdl->sdl_alen = ETHER_ADDR_LEN;
1110 		e_addr = LLADDR(sdl);
1111 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1112 		*llsa = (struct sockaddr *)sdl;
1113 		return 0;
1114 #endif
1115 
1116 	default:
1117 		/*
1118 		 * Well, the text isn't quite right, but it's the name
1119 		 * that counts...
1120 		 */
1121 		return EAFNOSUPPORT;
1122 	}
1123 }
1124 
1125 static moduledata_t ether_mod = {
1126 	.name = "ether",
1127 };
1128 
1129 void
1130 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1131 {
1132 	struct ether_vlan_header vlan;
1133 	struct mbuf mv, mb;
1134 
1135 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1136 	    ("%s: vlan information not present", __func__));
1137 	KASSERT(m->m_len >= sizeof(struct ether_header),
1138 	    ("%s: mbuf not large enough for header", __func__));
1139 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1140 	vlan.evl_proto = vlan.evl_encap_proto;
1141 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1142 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1143 	m->m_len -= sizeof(struct ether_header);
1144 	m->m_data += sizeof(struct ether_header);
1145 	/*
1146 	 * If a data link has been supplied by the caller, then we will need to
1147 	 * re-create a stack allocated mbuf chain with the following structure:
1148 	 *
1149 	 * (1) mbuf #1 will contain the supplied data link
1150 	 * (2) mbuf #2 will contain the vlan header
1151 	 * (3) mbuf #3 will contain the original mbuf's packet data
1152 	 *
1153 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1154 	 */
1155 	if (data != NULL) {
1156 		mv.m_next = m;
1157 		mv.m_data = (caddr_t)&vlan;
1158 		mv.m_len = sizeof(vlan);
1159 		mb.m_next = &mv;
1160 		mb.m_data = data;
1161 		mb.m_len = dlen;
1162 		bpf_mtap(bp, &mb);
1163 	} else
1164 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1165 	m->m_len += sizeof(struct ether_header);
1166 	m->m_data -= sizeof(struct ether_header);
1167 }
1168 
1169 struct mbuf *
1170 ether_vlanencap(struct mbuf *m, uint16_t tag)
1171 {
1172 	struct ether_vlan_header *evl;
1173 
1174 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1175 	if (m == NULL)
1176 		return (NULL);
1177 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1178 
1179 	if (m->m_len < sizeof(*evl)) {
1180 		m = m_pullup(m, sizeof(*evl));
1181 		if (m == NULL)
1182 			return (NULL);
1183 	}
1184 
1185 	/*
1186 	 * Transform the Ethernet header into an Ethernet header
1187 	 * with 802.1Q encapsulation.
1188 	 */
1189 	evl = mtod(m, struct ether_vlan_header *);
1190 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1191 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1192 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1193 	evl->evl_tag = htons(tag);
1194 	return (m);
1195 }
1196 
1197 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1198 MODULE_VERSION(ether, 1);
1199