xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 73029d086fecb97607aec6269511b65ab8a3131d)
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  * 3. 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: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
31  */
32 
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 #include "opt_mpls.h"
36 #include "opt_netgraph.h"
37 #include "opt_carp.h"
38 #include "opt_rss.h"
39 
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/globaldata.h>
43 #include <sys/kernel.h>
44 #include <sys/ktr.h>
45 #include <sys/lock.h>
46 #include <sys/malloc.h>
47 #include <sys/mbuf.h>
48 #include <sys/msgport.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 #include <sys/thread.h>
53 
54 #include <sys/thread2.h>
55 #include <sys/mplock2.h>
56 
57 #include <net/if.h>
58 #include <net/netisr.h>
59 #include <net/route.h>
60 #include <net/if_llc.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/ifq_var.h>
64 #include <net/bpf.h>
65 #include <net/ethernet.h>
66 #include <net/vlan/if_vlan_ether.h>
67 #include <net/vlan/if_vlan_var.h>
68 #include <net/netmsg2.h>
69 #include <net/netisr2.h>
70 
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/tcp_var.h>
75 #include <netinet/if_ether.h>
76 #include <netinet/ip_flow.h>
77 #include <net/ipfw/ip_fw.h>
78 #include <net/dummynet/ip_dummynet.h>
79 #endif
80 #ifdef INET6
81 #include <netinet6/nd6.h>
82 #endif
83 
84 #ifdef CARP
85 #include <netinet/ip_carp.h>
86 #endif
87 
88 #ifdef MPLS
89 #include <netproto/mpls/mpls.h>
90 #endif
91 
92 /* netgraph node hooks for ng_ether(4) */
93 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
94 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
95 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
96 void	(*ng_ether_attach_p)(struct ifnet *ifp);
97 void	(*ng_ether_detach_p)(struct ifnet *ifp);
98 
99 void	(*vlan_input_p)(struct mbuf *);
100 
101 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
102 			struct rtentry *);
103 static void ether_restore_header(struct mbuf **, const struct ether_header *,
104 				 const struct ether_header *);
105 static int ether_characterize(struct mbuf **);
106 static void ether_dispatch(int, struct mbuf *, int);
107 
108 /*
109  * if_bridge support
110  */
111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
112 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
113 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
114 struct ifnet *(*bridge_interface_p)(void *if_bridge);
115 
116 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
117 			      struct sockaddr *);
118 
119 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
120 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
121 };
122 
123 #define gotoerr(e) do { error = (e); goto bad; } while (0)
124 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
125 
126 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
127 				struct ip_fw **rule,
128 				const struct ether_header *eh);
129 
130 static int ether_ipfw;
131 static u_long ether_restore_hdr;
132 static u_long ether_prepend_hdr;
133 static u_long ether_input_wronghash;
134 static int ether_debug;
135 
136 #ifdef RSS_DEBUG
137 static u_long ether_pktinfo_try;
138 static u_long ether_pktinfo_hit;
139 static u_long ether_rss_nopi;
140 static u_long ether_rss_nohash;
141 static u_long ether_input_requeue;
142 #endif
143 static u_long ether_input_wronghwhash;
144 static int ether_input_ckhash;
145 
146 #define ETHER_TSOLEN_DEFAULT	(4 * ETHERMTU)
147 
148 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
149 TUNABLE_INT("net.link.ether.tsolen", &ether_tsolen_default);
150 
151 SYSCTL_DECL(_net_link);
152 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
153 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
154     &ether_debug, 0, "Ether debug");
155 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
156     &ether_ipfw, 0, "Pass ether pkts through firewall");
157 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
158     &ether_restore_hdr, 0, "# of ether header restoration");
159 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
160     &ether_prepend_hdr, 0,
161     "# of ether header restoration which prepends mbuf");
162 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
163     &ether_input_wronghash, 0, "# of input packets with wrong hash");
164 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
165     &ether_tsolen_default, 0, "Default max TSO length");
166 
167 #ifdef RSS_DEBUG
168 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
169     &ether_rss_nopi, 0, "# of packets do not have pktinfo");
170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
171     &ether_rss_nohash, 0, "# of packets do not have hash");
172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
173     &ether_pktinfo_try, 0,
174     "# of tries to find packets' msgport using pktinfo");
175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
176     &ether_pktinfo_hit, 0,
177     "# of packets whose msgport are found using pktinfo");
178 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
179     &ether_input_requeue, 0, "# of input packets gets requeued");
180 #endif
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
182     &ether_input_wronghwhash, 0, "# of input packets with wrong hw hash");
183 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
184     &ether_input_ckhash, 0, "always check hash");
185 
186 #define ETHER_KTR_STR		"ifp=%p"
187 #define ETHER_KTR_ARGS	struct ifnet *ifp
188 #ifndef KTR_ETHERNET
189 #define KTR_ETHERNET		KTR_ALL
190 #endif
191 KTR_INFO_MASTER(ether);
192 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
193 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
194 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
195 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
196 #define logether(name, arg)	KTR_LOG(ether_ ## name, arg)
197 
198 /*
199  * Ethernet output routine.
200  * Encapsulate a packet of type family for the local net.
201  * Use trailer local net encapsulation if enough data in first
202  * packet leaves a multiple of 512 bytes of data in remainder.
203  * Assumes that ifp is actually pointer to arpcom structure.
204  */
205 static int
206 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
207 	     struct rtentry *rt)
208 {
209 	struct ether_header *eh, *deh;
210 	u_char *edst;
211 	int loop_copy = 0;
212 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
213 	struct arpcom *ac = IFP2AC(ifp);
214 	int error;
215 
216 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
217 
218 	if (ifp->if_flags & IFF_MONITOR)
219 		gotoerr(ENETDOWN);
220 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
221 		gotoerr(ENETDOWN);
222 
223 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
224 	if (m == NULL)
225 		return (ENOBUFS);
226 	m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
227 	eh = mtod(m, struct ether_header *);
228 	edst = eh->ether_dhost;
229 
230 	/*
231 	 * Fill in the destination ethernet address and frame type.
232 	 */
233 	switch (dst->sa_family) {
234 #ifdef INET
235 	case AF_INET:
236 		if (!arpresolve(ifp, rt, m, dst, edst))
237 			return (0);	/* if not yet resolved */
238 #ifdef MPLS
239 		if (m->m_flags & M_MPLSLABELED)
240 			eh->ether_type = htons(ETHERTYPE_MPLS);
241 		else
242 #endif
243 			eh->ether_type = htons(ETHERTYPE_IP);
244 		break;
245 #endif
246 #ifdef INET6
247 	case AF_INET6:
248 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
249 			return (0);		/* Something bad happenned. */
250 		eh->ether_type = htons(ETHERTYPE_IPV6);
251 		break;
252 #endif
253 	case pseudo_AF_HDRCMPLT:
254 	case AF_UNSPEC:
255 		loop_copy = -1; /* if this is for us, don't do it */
256 		deh = (struct ether_header *)dst->sa_data;
257 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
258 		eh->ether_type = deh->ether_type;
259 		break;
260 
261 	default:
262 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
263 		gotoerr(EAFNOSUPPORT);
264 	}
265 
266 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
267 		memcpy(eh->ether_shost,
268 		       ((struct ether_header *)dst->sa_data)->ether_shost,
269 		       ETHER_ADDR_LEN);
270 	else
271 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
272 
273 	/*
274 	 * Bridges require special output handling.
275 	 */
276 	if (ifp->if_bridge) {
277 		KASSERT(bridge_output_p != NULL,
278 			("%s: if_bridge not loaded!", __func__));
279 		return bridge_output_p(ifp, m);
280 	}
281 
282 	/*
283 	 * If a simplex interface, and the packet is being sent to our
284 	 * Ethernet address or a broadcast address, loopback a copy.
285 	 * XXX To make a simplex device behave exactly like a duplex
286 	 * device, we should copy in the case of sending to our own
287 	 * ethernet address (thus letting the original actually appear
288 	 * on the wire). However, we don't do that here for security
289 	 * reasons and compatibility with the original behavior.
290 	 */
291 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
292 		int csum_flags = 0;
293 
294 		if (m->m_pkthdr.csum_flags & CSUM_IP)
295 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
296 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
297 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
298 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
299 			struct mbuf *n;
300 
301 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
302 				n->m_pkthdr.csum_flags |= csum_flags;
303 				if (csum_flags & CSUM_DATA_VALID)
304 					n->m_pkthdr.csum_data = 0xffff;
305 				if_simloop(ifp, n, dst->sa_family, hlen);
306 			} else
307 				IFNET_STAT_INC(ifp, iqdrops, 1);
308 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
309 				ETHER_ADDR_LEN) == 0) {
310 			m->m_pkthdr.csum_flags |= csum_flags;
311 			if (csum_flags & CSUM_DATA_VALID)
312 				m->m_pkthdr.csum_data = 0xffff;
313 			if_simloop(ifp, m, dst->sa_family, hlen);
314 			return (0);	/* XXX */
315 		}
316 	}
317 
318 #ifdef CARP
319 	if (ifp->if_type == IFT_CARP) {
320 		ifp = carp_parent(ifp);
321 		if (ifp == NULL)
322 			gotoerr(ENETUNREACH);
323 
324 		ac = IFP2AC(ifp);
325 
326 		/*
327 		 * Check precondition again
328 		 */
329 		ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
330 
331 		if (ifp->if_flags & IFF_MONITOR)
332 			gotoerr(ENETDOWN);
333 		if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
334 		    (IFF_UP | IFF_RUNNING))
335 			gotoerr(ENETDOWN);
336 	}
337 #endif
338 
339 	/* Handle ng_ether(4) processing, if any */
340 	if (ng_ether_output_p != NULL) {
341 		/*
342 		 * Hold BGL and recheck ng_ether_output_p
343 		 */
344 		get_mplock();
345 		if (ng_ether_output_p != NULL) {
346 			if ((error = ng_ether_output_p(ifp, &m)) != 0) {
347 				rel_mplock();
348 				goto bad;
349 			}
350 			if (m == NULL) {
351 				rel_mplock();
352 				return (0);
353 			}
354 		}
355 		rel_mplock();
356 	}
357 
358 	/* Continue with link-layer output */
359 	return ether_output_frame(ifp, m);
360 
361 bad:
362 	m_freem(m);
363 	return (error);
364 }
365 
366 /*
367  * Returns the bridge interface an ifp is associated
368  * with.
369  *
370  * Only call if ifp->if_bridge != NULL.
371  */
372 struct ifnet *
373 ether_bridge_interface(struct ifnet *ifp)
374 {
375 	if (bridge_interface_p)
376 		return(bridge_interface_p(ifp->if_bridge));
377 	return (ifp);
378 }
379 
380 /*
381  * Ethernet link layer output routine to send a raw frame to the device.
382  *
383  * This assumes that the 14 byte Ethernet header is present and contiguous
384  * in the first mbuf.
385  */
386 int
387 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
388 {
389 	struct ip_fw *rule = NULL;
390 	int error = 0;
391 	struct altq_pktattr pktattr;
392 
393 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
394 
395 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
396 		struct m_tag *mtag;
397 
398 		/* Extract info from dummynet tag */
399 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
400 		KKASSERT(mtag != NULL);
401 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
402 		KKASSERT(rule != NULL);
403 
404 		m_tag_delete(m, mtag);
405 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
406 	}
407 
408 	if (ifq_is_enabled(&ifp->if_snd))
409 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
410 	crit_enter();
411 	if (IPFW_LOADED && ether_ipfw != 0) {
412 		struct ether_header save_eh, *eh;
413 
414 		eh = mtod(m, struct ether_header *);
415 		save_eh = *eh;
416 		m_adj(m, ETHER_HDR_LEN);
417 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
418 			crit_exit();
419 			if (m != NULL) {
420 				m_freem(m);
421 				return ENOBUFS; /* pkt dropped */
422 			} else
423 				return 0;	/* consumed e.g. in a pipe */
424 		}
425 
426 		/* packet was ok, restore the ethernet header */
427 		ether_restore_header(&m, eh, &save_eh);
428 		if (m == NULL) {
429 			crit_exit();
430 			return ENOBUFS;
431 		}
432 	}
433 	crit_exit();
434 
435 	/*
436 	 * Queue message on interface, update output statistics if
437 	 * successful, and start output if interface not yet active.
438 	 */
439 	error = ifq_dispatch(ifp, m, &pktattr);
440 	return (error);
441 }
442 
443 /*
444  * ipfw processing for ethernet packets (in and out).
445  * The second parameter is NULL from ether_demux(), and ifp from
446  * ether_output_frame().
447  */
448 static boolean_t
449 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
450 	       const struct ether_header *eh)
451 {
452 	struct ether_header save_eh = *eh;	/* might be a ptr in *m0 */
453 	struct ip_fw_args args;
454 	struct m_tag *mtag;
455 	struct mbuf *m;
456 	int i;
457 
458 	if (*rule != NULL && fw_one_pass)
459 		return TRUE; /* dummynet packet, already partially processed */
460 
461 	/*
462 	 * I need some amount of data to be contiguous.
463 	 */
464 	i = min((*m0)->m_pkthdr.len, max_protohdr);
465 	if ((*m0)->m_len < i) {
466 		*m0 = m_pullup(*m0, i);
467 		if (*m0 == NULL)
468 			return FALSE;
469 	}
470 
471 	/*
472 	 * Clean up tags
473 	 */
474 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
475 		m_tag_delete(*m0, mtag);
476 	if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
477 		mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
478 		KKASSERT(mtag != NULL);
479 		m_tag_delete(*m0, mtag);
480 		(*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
481 	}
482 
483 	args.m = *m0;		/* the packet we are looking at		*/
484 	args.oif = dst;		/* destination, if any			*/
485 	args.rule = *rule;	/* matching rule to restart		*/
486 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
487 	i = ip_fw_chk_ptr(&args);
488 	*m0 = args.m;
489 	*rule = args.rule;
490 
491 	if (*m0 == NULL)
492 		return FALSE;
493 
494 	switch (i) {
495 	case IP_FW_PASS:
496 		return TRUE;
497 
498 	case IP_FW_DIVERT:
499 	case IP_FW_TEE:
500 	case IP_FW_DENY:
501 		/*
502 		 * XXX at some point add support for divert/forward actions.
503 		 * If none of the above matches, we have to drop the pkt.
504 		 */
505 		return FALSE;
506 
507 	case IP_FW_DUMMYNET:
508 		/*
509 		 * Pass the pkt to dummynet, which consumes it.
510 		 */
511 		m = *m0;	/* pass the original to dummynet */
512 		*m0 = NULL;	/* and nothing back to the caller */
513 
514 		ether_restore_header(&m, eh, &save_eh);
515 		if (m == NULL)
516 			return FALSE;
517 
518 		ip_fw_dn_io_ptr(m, args.cookie,
519 				dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
520 		ip_dn_queue(m);
521 		return FALSE;
522 
523 	default:
524 		panic("unknown ipfw return value: %d", i);
525 	}
526 }
527 
528 static void
529 ether_input(struct ifnet *ifp, struct mbuf *m,
530     const struct pktinfo *pi, int cpuid)
531 {
532 	ether_input_pkt(ifp, m, pi, cpuid);
533 }
534 
535 /*
536  * Perform common duties while attaching to interface list
537  */
538 void
539 ether_ifattach(struct ifnet *ifp, const uint8_t *lla,
540     lwkt_serialize_t serializer)
541 {
542 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
543 	    serializer);
544 }
545 
546 void
547 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla,
548     u_int dlt, u_int hdrlen, lwkt_serialize_t serializer)
549 {
550 	struct sockaddr_dl *sdl;
551 	char ethstr[ETHER_ADDRSTRLEN + 1];
552 	struct ifaltq *ifq;
553 	int i;
554 
555 	ifp->if_type = IFT_ETHER;
556 	ifp->if_addrlen = ETHER_ADDR_LEN;
557 	ifp->if_hdrlen = ETHER_HDR_LEN;
558 	if_attach(ifp, serializer);
559 	ifq = &ifp->if_snd;
560 	for (i = 0; i < ifq->altq_subq_cnt; ++i) {
561 		struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i);
562 
563 		ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen *
564 		    (ETHER_MAX_LEN - ETHER_CRC_LEN);
565 	}
566 	ifp->if_mtu = ETHERMTU;
567 	if (ifp->if_tsolen <= 0) {
568 		if ((ether_tsolen_default / ETHERMTU) < 2) {
569 			kprintf("ether TSO maxlen %d -> %d\n",
570 			    ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
571 			ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
572 		}
573 		ifp->if_tsolen = ether_tsolen_default;
574 	}
575 	if (ifp->if_baudrate == 0)
576 		ifp->if_baudrate = 10000000;
577 	ifp->if_output = ether_output;
578 	ifp->if_input = ether_input;
579 	ifp->if_resolvemulti = ether_resolvemulti;
580 	ifp->if_broadcastaddr = etherbroadcastaddr;
581 	sdl = IF_LLSOCKADDR(ifp);
582 	sdl->sdl_type = IFT_ETHER;
583 	sdl->sdl_alen = ifp->if_addrlen;
584 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
585 	/*
586 	 * XXX Keep the current drivers happy.
587 	 * XXX Remove once all drivers have been cleaned up
588 	 */
589 	if (lla != IFP2AC(ifp)->ac_enaddr)
590 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
591 	bpfattach(ifp, dlt, hdrlen);
592 	if (ng_ether_attach_p != NULL)
593 		(*ng_ether_attach_p)(ifp);
594 
595 	if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
596 }
597 
598 /*
599  * Perform common duties while detaching an Ethernet interface
600  */
601 void
602 ether_ifdetach(struct ifnet *ifp)
603 {
604 	if_down(ifp);
605 
606 	if (ng_ether_detach_p != NULL)
607 		(*ng_ether_detach_p)(ifp);
608 	bpfdetach(ifp);
609 	if_detach(ifp);
610 }
611 
612 int
613 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
614 {
615 	struct ifaddr *ifa = (struct ifaddr *) data;
616 	struct ifreq *ifr = (struct ifreq *) data;
617 	int error = 0;
618 
619 #define IF_INIT(ifp) \
620 do { \
621 	if (((ifp)->if_flags & IFF_UP) == 0) { \
622 		(ifp)->if_flags |= IFF_UP; \
623 		(ifp)->if_init((ifp)->if_softc); \
624 	} \
625 } while (0)
626 
627 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
628 
629 	switch (command) {
630 	case SIOCSIFADDR:
631 		switch (ifa->ifa_addr->sa_family) {
632 #ifdef INET
633 		case AF_INET:
634 			IF_INIT(ifp);	/* before arpwhohas */
635 			arp_ifinit(ifp, ifa);
636 			break;
637 #endif
638 		default:
639 			IF_INIT(ifp);
640 			break;
641 		}
642 		break;
643 
644 	case SIOCGIFADDR:
645 		bcopy(IFP2AC(ifp)->ac_enaddr,
646 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
647 		      ETHER_ADDR_LEN);
648 		break;
649 
650 	case SIOCSIFMTU:
651 		/*
652 		 * Set the interface MTU.
653 		 */
654 		if (ifr->ifr_mtu > ETHERMTU) {
655 			error = EINVAL;
656 		} else {
657 			ifp->if_mtu = ifr->ifr_mtu;
658 		}
659 		break;
660 	default:
661 		error = EINVAL;
662 		break;
663 	}
664 	return (error);
665 
666 #undef IF_INIT
667 }
668 
669 int
670 ether_resolvemulti(
671 	struct ifnet *ifp,
672 	struct sockaddr **llsa,
673 	struct sockaddr *sa)
674 {
675 	struct sockaddr_dl *sdl;
676 #ifdef INET
677 	struct sockaddr_in *sin;
678 #endif
679 #ifdef INET6
680 	struct sockaddr_in6 *sin6;
681 #endif
682 	u_char *e_addr;
683 
684 	switch(sa->sa_family) {
685 	case AF_LINK:
686 		/*
687 		 * No mapping needed. Just check that it's a valid MC address.
688 		 */
689 		sdl = (struct sockaddr_dl *)sa;
690 		e_addr = LLADDR(sdl);
691 		if ((e_addr[0] & 1) != 1)
692 			return EADDRNOTAVAIL;
693 		*llsa = NULL;
694 		return 0;
695 
696 #ifdef INET
697 	case AF_INET:
698 		sin = (struct sockaddr_in *)sa;
699 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
700 			return EADDRNOTAVAIL;
701 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
702 		sdl->sdl_len = sizeof *sdl;
703 		sdl->sdl_family = AF_LINK;
704 		sdl->sdl_index = ifp->if_index;
705 		sdl->sdl_type = IFT_ETHER;
706 		sdl->sdl_alen = ETHER_ADDR_LEN;
707 		e_addr = LLADDR(sdl);
708 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
709 		*llsa = (struct sockaddr *)sdl;
710 		return 0;
711 #endif
712 #ifdef INET6
713 	case AF_INET6:
714 		sin6 = (struct sockaddr_in6 *)sa;
715 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
716 			/*
717 			 * An IP6 address of 0 means listen to all
718 			 * of the Ethernet multicast address used for IP6.
719 			 * (This is used for multicast routers.)
720 			 */
721 			ifp->if_flags |= IFF_ALLMULTI;
722 			*llsa = NULL;
723 			return 0;
724 		}
725 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
726 			return EADDRNOTAVAIL;
727 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
728 		sdl->sdl_len = sizeof *sdl;
729 		sdl->sdl_family = AF_LINK;
730 		sdl->sdl_index = ifp->if_index;
731 		sdl->sdl_type = IFT_ETHER;
732 		sdl->sdl_alen = ETHER_ADDR_LEN;
733 		e_addr = LLADDR(sdl);
734 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
735 		*llsa = (struct sockaddr *)sdl;
736 		return 0;
737 #endif
738 
739 	default:
740 		/*
741 		 * Well, the text isn't quite right, but it's the name
742 		 * that counts...
743 		 */
744 		return EAFNOSUPPORT;
745 	}
746 }
747 
748 #if 0
749 /*
750  * This is for reference.  We have a table-driven version
751  * of the little-endian crc32 generator, which is faster
752  * than the double-loop.
753  */
754 uint32_t
755 ether_crc32_le(const uint8_t *buf, size_t len)
756 {
757 	uint32_t c, crc, carry;
758 	size_t i, j;
759 
760 	crc = 0xffffffffU;	/* initial value */
761 
762 	for (i = 0; i < len; i++) {
763 		c = buf[i];
764 		for (j = 0; j < 8; j++) {
765 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
766 			crc >>= 1;
767 			c >>= 1;
768 			if (carry)
769 				crc = (crc ^ ETHER_CRC_POLY_LE);
770 		}
771 	}
772 
773 	return (crc);
774 }
775 #else
776 uint32_t
777 ether_crc32_le(const uint8_t *buf, size_t len)
778 {
779 	static const uint32_t crctab[] = {
780 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
781 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
782 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
783 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
784 	};
785 	uint32_t crc;
786 	size_t i;
787 
788 	crc = 0xffffffffU;	/* initial value */
789 
790 	for (i = 0; i < len; i++) {
791 		crc ^= buf[i];
792 		crc = (crc >> 4) ^ crctab[crc & 0xf];
793 		crc = (crc >> 4) ^ crctab[crc & 0xf];
794 	}
795 
796 	return (crc);
797 }
798 #endif
799 
800 uint32_t
801 ether_crc32_be(const uint8_t *buf, size_t len)
802 {
803 	uint32_t c, crc, carry;
804 	size_t i, j;
805 
806 	crc = 0xffffffffU;	/* initial value */
807 
808 	for (i = 0; i < len; i++) {
809 		c = buf[i];
810 		for (j = 0; j < 8; j++) {
811 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
812 			crc <<= 1;
813 			c >>= 1;
814 			if (carry)
815 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
816 		}
817 	}
818 
819 	return (crc);
820 }
821 
822 /*
823  * find the size of ethernet header, and call classifier
824  */
825 void
826 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
827 		   struct altq_pktattr *pktattr)
828 {
829 	struct ether_header *eh;
830 	uint16_t ether_type;
831 	int hlen, af, hdrsize;
832 
833 	hlen = sizeof(struct ether_header);
834 	eh = mtod(m, struct ether_header *);
835 
836 	ether_type = ntohs(eh->ether_type);
837 	if (ether_type < ETHERMTU) {
838 		/* ick! LLC/SNAP */
839 		struct llc *llc = (struct llc *)(eh + 1);
840 		hlen += 8;
841 
842 		if (m->m_len < hlen ||
843 		    llc->llc_dsap != LLC_SNAP_LSAP ||
844 		    llc->llc_ssap != LLC_SNAP_LSAP ||
845 		    llc->llc_control != LLC_UI)
846 			goto bad;  /* not snap! */
847 
848 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
849 	}
850 
851 	if (ether_type == ETHERTYPE_IP) {
852 		af = AF_INET;
853 		hdrsize = 20;  /* sizeof(struct ip) */
854 #ifdef INET6
855 	} else if (ether_type == ETHERTYPE_IPV6) {
856 		af = AF_INET6;
857 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
858 #endif
859 	} else
860 		goto bad;
861 
862 	while (m->m_len <= hlen) {
863 		hlen -= m->m_len;
864 		m = m->m_next;
865 	}
866 	if (m->m_len < hlen + hdrsize) {
867 		/*
868 		 * ip header is not in a single mbuf.  this should not
869 		 * happen in the current code.
870 		 * (todo: use m_pulldown in the future)
871 		 */
872 		goto bad;
873 	}
874 	m->m_data += hlen;
875 	m->m_len -= hlen;
876 	ifq_classify(ifq, m, af, pktattr);
877 	m->m_data -= hlen;
878 	m->m_len += hlen;
879 
880 	return;
881 
882 bad:
883 	pktattr->pattr_class = NULL;
884 	pktattr->pattr_hdr = NULL;
885 	pktattr->pattr_af = AF_UNSPEC;
886 }
887 
888 static void
889 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
890 		     const struct ether_header *save_eh)
891 {
892 	struct mbuf *m = *m0;
893 
894 	ether_restore_hdr++;
895 
896 	/*
897 	 * Prepend the header, optimize for the common case of
898 	 * eh pointing into the mbuf.
899 	 */
900 	if ((const void *)(eh + 1) == (void *)m->m_data) {
901 		m->m_data -= ETHER_HDR_LEN;
902 		m->m_len += ETHER_HDR_LEN;
903 		m->m_pkthdr.len += ETHER_HDR_LEN;
904 	} else {
905 		ether_prepend_hdr++;
906 
907 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
908 		if (m != NULL) {
909 			bcopy(save_eh, mtod(m, struct ether_header *),
910 			      ETHER_HDR_LEN);
911 		}
912 	}
913 	*m0 = m;
914 }
915 
916 /*
917  * Upper layer processing for a received Ethernet packet.
918  */
919 void
920 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
921 {
922 	struct ether_header *eh;
923 	int isr, discard = 0;
924 	u_short ether_type;
925 	struct ip_fw *rule = NULL;
926 
927 	M_ASSERTPKTHDR(m);
928 	KASSERT(m->m_len >= ETHER_HDR_LEN,
929 		("ether header is not contiguous!"));
930 
931 	eh = mtod(m, struct ether_header *);
932 
933 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
934 		struct m_tag *mtag;
935 
936 		/* Extract info from dummynet tag */
937 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
938 		KKASSERT(mtag != NULL);
939 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
940 		KKASSERT(rule != NULL);
941 
942 		m_tag_delete(m, mtag);
943 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
944 
945 		/* packet is passing the second time */
946 		goto post_stats;
947 	}
948 
949 	/*
950 	 * We got a packet which was unicast to a different Ethernet
951 	 * address.  If the driver is working properly, then this
952 	 * situation can only happen when the interface is in
953 	 * promiscuous mode.  We defer the packet discarding until the
954 	 * vlan processing is done, so that vlan/bridge or vlan/netgraph
955 	 * could work.
956 	 */
957 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
958 	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
959 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
960 		if (ether_debug & 1) {
961 			kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
962 				"%02x:%02x:%02x:%02x:%02x:%02x "
963 				"%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
964 				eh->ether_dhost[0],
965 				eh->ether_dhost[1],
966 				eh->ether_dhost[2],
967 				eh->ether_dhost[3],
968 				eh->ether_dhost[4],
969 				eh->ether_dhost[5],
970 				eh->ether_shost[0],
971 				eh->ether_shost[1],
972 				eh->ether_shost[2],
973 				eh->ether_shost[3],
974 				eh->ether_shost[4],
975 				eh->ether_shost[5],
976 				eh->ether_type,
977 				((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
978 				((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
979 				((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
980 				((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
981 				((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
982 				((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
983 			);
984 		}
985 		if ((ether_debug & 2) == 0)
986 			discard = 1;
987 	}
988 
989 post_stats:
990 	if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
991 		struct ether_header save_eh = *eh;
992 
993 		/* XXX old crufty stuff, needs to be removed */
994 		m_adj(m, sizeof(struct ether_header));
995 
996 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
997 			m_freem(m);
998 			return;
999 		}
1000 
1001 		ether_restore_header(&m, eh, &save_eh);
1002 		if (m == NULL)
1003 			return;
1004 		eh = mtod(m, struct ether_header *);
1005 	}
1006 
1007 	ether_type = ntohs(eh->ether_type);
1008 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1009 
1010 	if (m->m_flags & M_VLANTAG) {
1011 		void (*vlan_input_func)(struct mbuf *);
1012 
1013 		vlan_input_func = vlan_input_p;
1014 		if (vlan_input_func != NULL) {
1015 			vlan_input_func(m);
1016 		} else {
1017 			IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1018 			m_freem(m);
1019 		}
1020 		return;
1021 	}
1022 
1023 	/*
1024 	 * If we have been asked to discard this packet
1025 	 * (e.g. not for us), drop it before entering
1026 	 * the upper layer.
1027 	 */
1028 	if (discard) {
1029 		m_freem(m);
1030 		return;
1031 	}
1032 
1033 	/*
1034 	 * Clear protocol specific flags,
1035 	 * before entering the upper layer.
1036 	 */
1037 	m->m_flags &= ~M_ETHER_FLAGS;
1038 
1039 	/* Strip ethernet header. */
1040 	m_adj(m, sizeof(struct ether_header));
1041 
1042 	switch (ether_type) {
1043 #ifdef INET
1044 	case ETHERTYPE_IP:
1045 		if ((m->m_flags & M_LENCHECKED) == 0) {
1046 			if (!ip_lengthcheck(&m, 0))
1047 				return;
1048 		}
1049 		if (ipflow_fastforward(m))
1050 			return;
1051 		isr = NETISR_IP;
1052 		break;
1053 
1054 	case ETHERTYPE_ARP:
1055 		if (ifp->if_flags & IFF_NOARP) {
1056 			/* Discard packet if ARP is disabled on interface */
1057 			m_freem(m);
1058 			return;
1059 		}
1060 		isr = NETISR_ARP;
1061 		break;
1062 #endif
1063 
1064 #ifdef INET6
1065 	case ETHERTYPE_IPV6:
1066 		isr = NETISR_IPV6;
1067 		break;
1068 #endif
1069 
1070 #ifdef MPLS
1071 	case ETHERTYPE_MPLS:
1072 	case ETHERTYPE_MPLS_MCAST:
1073 		/* Should have been set by ether_input_pkt(). */
1074 		KKASSERT(m->m_flags & M_MPLSLABELED);
1075 		isr = NETISR_MPLS;
1076 		break;
1077 #endif
1078 
1079 	default:
1080 		/*
1081 		 * The accurate msgport is not determined before
1082 		 * we reach here, so recharacterize packet.
1083 		 */
1084 		m->m_flags &= ~M_HASH;
1085 		if (ng_ether_input_orphan_p != NULL) {
1086 			/*
1087 			 * Put back the ethernet header so netgraph has a
1088 			 * consistent view of inbound packets.
1089 			 */
1090 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1091 			if (m == NULL) {
1092 				/*
1093 				 * M_PREPEND frees the mbuf in case of failure.
1094 				 */
1095 				return;
1096 			}
1097 			/*
1098 			 * Hold BGL and recheck ng_ether_input_orphan_p
1099 			 */
1100 			get_mplock();
1101 			if (ng_ether_input_orphan_p != NULL) {
1102 				ng_ether_input_orphan_p(ifp, m);
1103 				rel_mplock();
1104 				return;
1105 			}
1106 			rel_mplock();
1107 		}
1108 		m_freem(m);
1109 		return;
1110 	}
1111 
1112 	if (m->m_flags & M_HASH) {
1113 		if (&curthread->td_msgport ==
1114 		    netisr_hashport(m->m_pkthdr.hash)) {
1115 			netisr_handle(isr, m);
1116 			return;
1117 		} else {
1118 			/*
1119 			 * XXX Something is wrong,
1120 			 * we probably should panic here!
1121 			 */
1122 			m->m_flags &= ~M_HASH;
1123 			atomic_add_long(&ether_input_wronghash, 1);
1124 		}
1125 	}
1126 #ifdef RSS_DEBUG
1127 	atomic_add_long(&ether_input_requeue, 1);
1128 #endif
1129 	netisr_queue(isr, m);
1130 }
1131 
1132 /*
1133  * First we perform any link layer operations, then continue to the
1134  * upper layers with ether_demux_oncpu().
1135  */
1136 static void
1137 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1138 {
1139 #ifdef CARP
1140 	void *carp;
1141 #endif
1142 
1143 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1144 		/*
1145 		 * Receiving interface's flags are changed, when this
1146 		 * packet is waiting for processing; discard it.
1147 		 */
1148 		m_freem(m);
1149 		return;
1150 	}
1151 
1152 	/*
1153 	 * Tap the packet off here for a bridge.  bridge_input()
1154 	 * will return NULL if it has consumed the packet, otherwise
1155 	 * it gets processed as normal.  Note that bridge_input()
1156 	 * will always return the original packet if we need to
1157 	 * process it locally.
1158 	 */
1159 	if (ifp->if_bridge) {
1160 		KASSERT(bridge_input_p != NULL,
1161 			("%s: if_bridge not loaded!", __func__));
1162 
1163 		if(m->m_flags & M_ETHER_BRIDGED) {
1164 			m->m_flags &= ~M_ETHER_BRIDGED;
1165 		} else {
1166 			m = bridge_input_p(ifp, m);
1167 			if (m == NULL)
1168 				return;
1169 
1170 			KASSERT(ifp == m->m_pkthdr.rcvif,
1171 				("bridge_input_p changed rcvif"));
1172 		}
1173 	}
1174 
1175 #ifdef CARP
1176 	carp = ifp->if_carp;
1177 	if (carp) {
1178 		m = carp_input(carp, m);
1179 		if (m == NULL)
1180 			return;
1181 		KASSERT(ifp == m->m_pkthdr.rcvif,
1182 		    ("carp_input changed rcvif"));
1183 	}
1184 #endif
1185 
1186 	/* Handle ng_ether(4) processing, if any */
1187 	if (ng_ether_input_p != NULL) {
1188 		/*
1189 		 * Hold BGL and recheck ng_ether_input_p
1190 		 */
1191 		get_mplock();
1192 		if (ng_ether_input_p != NULL)
1193 			ng_ether_input_p(ifp, &m);
1194 		rel_mplock();
1195 
1196 		if (m == NULL)
1197 			return;
1198 	}
1199 
1200 	/* Continue with upper layer processing */
1201 	ether_demux_oncpu(ifp, m);
1202 }
1203 
1204 /*
1205  * Perform certain functions of ether_input_pkt():
1206  * - Test IFF_UP
1207  * - Update statistics
1208  * - Run bpf(4) tap if requested
1209  * Then pass the packet to ether_input_oncpu().
1210  *
1211  * This function should be used by pseudo interface (e.g. vlan(4)),
1212  * when it tries to claim that the packet is received by it.
1213  *
1214  * REINPUT_KEEPRCVIF
1215  * REINPUT_RUNBPF
1216  */
1217 void
1218 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1219 {
1220 	/* Discard packet if interface is not up */
1221 	if (!(ifp->if_flags & IFF_UP)) {
1222 		m_freem(m);
1223 		return;
1224 	}
1225 
1226 	/*
1227 	 * Change receiving interface.  The bridge will often pass a flag to
1228 	 * ask that this not be done so ARPs get applied to the correct
1229 	 * side.
1230 	 */
1231 	if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1232 	    m->m_pkthdr.rcvif == NULL) {
1233 		m->m_pkthdr.rcvif = ifp;
1234 	}
1235 
1236 	/* Update statistics */
1237 	IFNET_STAT_INC(ifp, ipackets, 1);
1238 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1239 	if (m->m_flags & (M_MCAST | M_BCAST))
1240 		IFNET_STAT_INC(ifp, imcasts, 1);
1241 
1242 	if (reinput_flags & REINPUT_RUNBPF)
1243 		BPF_MTAP(ifp, m);
1244 
1245 	ether_input_oncpu(ifp, m);
1246 }
1247 
1248 static __inline boolean_t
1249 ether_vlancheck(struct mbuf **m0)
1250 {
1251 	struct mbuf *m = *m0;
1252 	struct ether_header *eh;
1253 	uint16_t ether_type;
1254 
1255 	eh = mtod(m, struct ether_header *);
1256 	ether_type = ntohs(eh->ether_type);
1257 
1258 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1259 		/*
1260 		 * Extract vlan tag if hardware does not do it for us
1261 		 */
1262 		vlan_ether_decap(&m);
1263 		if (m == NULL)
1264 			goto failed;
1265 
1266 		eh = mtod(m, struct ether_header *);
1267 		ether_type = ntohs(eh->ether_type);
1268 	}
1269 
1270 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1271 		/*
1272 		 * To prevent possible dangerous recursion,
1273 		 * we don't do vlan-in-vlan
1274 		 */
1275 		IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1276 		goto failed;
1277 	}
1278 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1279 
1280 	m->m_flags |= M_ETHER_VLANCHECKED;
1281 	*m0 = m;
1282 	return TRUE;
1283 failed:
1284 	if (m != NULL)
1285 		m_freem(m);
1286 	*m0 = NULL;
1287 	return FALSE;
1288 }
1289 
1290 static void
1291 ether_input_handler(netmsg_t nmsg)
1292 {
1293 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1294 	struct ether_header *eh;
1295 	struct ifnet *ifp;
1296 	struct mbuf *m;
1297 
1298 	m = nmp->nm_packet;
1299 	M_ASSERTPKTHDR(m);
1300 
1301 	if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1302 		if (!ether_vlancheck(&m)) {
1303 			KKASSERT(m == NULL);
1304 			return;
1305 		}
1306 	}
1307 	if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1308 	    __predict_false(ether_input_ckhash)) {
1309 		int isr;
1310 
1311 		/*
1312 		 * Need to verify the hash supplied by the hardware
1313 		 * which could be wrong.
1314 		 */
1315 		m->m_flags &= ~(M_HASH | M_CKHASH);
1316 		isr = ether_characterize(&m);
1317 		if (m == NULL)
1318 			return;
1319 		KKASSERT(m->m_flags & M_HASH);
1320 
1321 		if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1322 			/*
1323 			 * Wrong hardware supplied hash; redispatch
1324 			 */
1325 			ether_dispatch(isr, m, -1);
1326 			if (__predict_false(ether_input_ckhash))
1327 				atomic_add_long(&ether_input_wronghwhash, 1);
1328 			return;
1329 		}
1330 	}
1331 	ifp = m->m_pkthdr.rcvif;
1332 
1333 	eh = mtod(m, struct ether_header *);
1334 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1335 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1336 			 ifp->if_addrlen) == 0)
1337 			m->m_flags |= M_BCAST;
1338 		else
1339 			m->m_flags |= M_MCAST;
1340 		IFNET_STAT_INC(ifp, imcasts, 1);
1341 	}
1342 
1343 	ether_input_oncpu(ifp, m);
1344 }
1345 
1346 /*
1347  * Send the packet to the target netisr msgport
1348  *
1349  * At this point the packet must be characterized (M_HASH set),
1350  * so we know which netisr to send it to.
1351  */
1352 static void
1353 ether_dispatch(int isr, struct mbuf *m, int cpuid)
1354 {
1355 	struct netmsg_packet *pmsg;
1356 	int target_cpuid;
1357 
1358 	KKASSERT(m->m_flags & M_HASH);
1359 	target_cpuid = netisr_hashcpu(m->m_pkthdr.hash);
1360 
1361 	pmsg = &m->m_hdr.mh_netmsg;
1362 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1363 		    0, ether_input_handler);
1364 	pmsg->nm_packet = m;
1365 	pmsg->base.lmsg.u.ms_result = isr;
1366 
1367 	logether(disp_beg, NULL);
1368 	if (target_cpuid == cpuid) {
1369 		lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid),
1370 		    &pmsg->base.lmsg);
1371 	} else {
1372 		lwkt_sendmsg(netisr_cpuport(target_cpuid),
1373 		    &pmsg->base.lmsg);
1374 	}
1375 	logether(disp_end, NULL);
1376 }
1377 
1378 /*
1379  * Process a received Ethernet packet.
1380  *
1381  * The ethernet header is assumed to be in the mbuf so the caller
1382  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1383  * bytes in the first mbuf.
1384  *
1385  * If the caller knows that the current thread is stick to the current
1386  * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid
1387  * (mycpuid) should be passed through 'cpuid' argument.  Else -1 should
1388  * be passed as 'cpuid' argument.
1389  */
1390 void
1391 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi,
1392     int cpuid)
1393 {
1394 	int isr;
1395 
1396 	M_ASSERTPKTHDR(m);
1397 
1398 	/* Discard packet if interface is not up */
1399 	if (!(ifp->if_flags & IFF_UP)) {
1400 		m_freem(m);
1401 		return;
1402 	}
1403 
1404 	if (m->m_len < sizeof(struct ether_header)) {
1405 		/* XXX error in the caller. */
1406 		m_freem(m);
1407 		return;
1408 	}
1409 
1410 	m->m_pkthdr.rcvif = ifp;
1411 
1412 	logether(pkt_beg, ifp);
1413 
1414 	ETHER_BPF_MTAP(ifp, m);
1415 
1416 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1417 
1418 	if (ifp->if_flags & IFF_MONITOR) {
1419 		struct ether_header *eh;
1420 
1421 		eh = mtod(m, struct ether_header *);
1422 		if (ETHER_IS_MULTICAST(eh->ether_dhost))
1423 			IFNET_STAT_INC(ifp, imcasts, 1);
1424 
1425 		/*
1426 		 * Interface marked for monitoring; discard packet.
1427 		 */
1428 		m_freem(m);
1429 
1430 		logether(pkt_end, ifp);
1431 		return;
1432 	}
1433 
1434 	/*
1435 	 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1436 	 * we can dispatch it immediately with trivial checks.
1437 	 */
1438 	if (pi != NULL && (m->m_flags & M_HASH)) {
1439 #ifdef RSS_DEBUG
1440 		atomic_add_long(&ether_pktinfo_try, 1);
1441 #endif
1442 		netisr_hashcheck(pi->pi_netisr, m, pi);
1443 		if (m->m_flags & M_HASH) {
1444 			ether_dispatch(pi->pi_netisr, m, cpuid);
1445 #ifdef RSS_DEBUG
1446 			atomic_add_long(&ether_pktinfo_hit, 1);
1447 #endif
1448 			logether(pkt_end, ifp);
1449 			return;
1450 		}
1451 	}
1452 #ifdef RSS_DEBUG
1453 	else if (ifp->if_capenable & IFCAP_RSS) {
1454 		if (pi == NULL)
1455 			atomic_add_long(&ether_rss_nopi, 1);
1456 		else
1457 			atomic_add_long(&ether_rss_nohash, 1);
1458 	}
1459 #endif
1460 
1461 	/*
1462 	 * Packet hash will be recalculated by software, so clear
1463 	 * the M_HASH and M_CKHASH flag set by the driver; the hash
1464 	 * value calculated by the hardware may not be exactly what
1465 	 * we want.
1466 	 */
1467 	m->m_flags &= ~(M_HASH | M_CKHASH);
1468 
1469 	if (!ether_vlancheck(&m)) {
1470 		KKASSERT(m == NULL);
1471 		logether(pkt_end, ifp);
1472 		return;
1473 	}
1474 
1475 	isr = ether_characterize(&m);
1476 	if (m == NULL) {
1477 		logether(pkt_end, ifp);
1478 		return;
1479 	}
1480 
1481 	/*
1482 	 * Finally dispatch it
1483 	 */
1484 	ether_dispatch(isr, m, cpuid);
1485 
1486 	logether(pkt_end, ifp);
1487 }
1488 
1489 static int
1490 ether_characterize(struct mbuf **m0)
1491 {
1492 	struct mbuf *m = *m0;
1493 	struct ether_header *eh;
1494 	uint16_t ether_type;
1495 	int isr;
1496 
1497 	eh = mtod(m, struct ether_header *);
1498 	ether_type = ntohs(eh->ether_type);
1499 
1500 	/*
1501 	 * Map ether type to netisr id.
1502 	 */
1503 	switch (ether_type) {
1504 #ifdef INET
1505 	case ETHERTYPE_IP:
1506 		isr = NETISR_IP;
1507 		break;
1508 
1509 	case ETHERTYPE_ARP:
1510 		isr = NETISR_ARP;
1511 		break;
1512 #endif
1513 
1514 #ifdef INET6
1515 	case ETHERTYPE_IPV6:
1516 		isr = NETISR_IPV6;
1517 		break;
1518 #endif
1519 
1520 #ifdef MPLS
1521 	case ETHERTYPE_MPLS:
1522 	case ETHERTYPE_MPLS_MCAST:
1523 		m->m_flags |= M_MPLSLABELED;
1524 		isr = NETISR_MPLS;
1525 		break;
1526 #endif
1527 
1528 	default:
1529 		/*
1530 		 * NETISR_MAX is an invalid value; it is chosen to let
1531 		 * netisr_characterize() know that we have no clear
1532 		 * idea where this packet should go.
1533 		 */
1534 		isr = NETISR_MAX;
1535 		break;
1536 	}
1537 
1538 	/*
1539 	 * Ask the isr to characterize the packet since we couldn't.
1540 	 * This is an attempt to optimally get us onto the correct protocol
1541 	 * thread.
1542 	 */
1543 	netisr_characterize(isr, &m, sizeof(struct ether_header));
1544 
1545 	*m0 = m;
1546 	return isr;
1547 }
1548 
1549 static void
1550 ether_demux_handler(netmsg_t nmsg)
1551 {
1552 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1553 	struct ifnet *ifp;
1554 	struct mbuf *m;
1555 
1556 	m = nmp->nm_packet;
1557 	M_ASSERTPKTHDR(m);
1558 	ifp = m->m_pkthdr.rcvif;
1559 
1560 	ether_demux_oncpu(ifp, m);
1561 }
1562 
1563 void
1564 ether_demux(struct mbuf *m)
1565 {
1566 	struct netmsg_packet *pmsg;
1567 	int isr;
1568 
1569 	isr = ether_characterize(&m);
1570 	if (m == NULL)
1571 		return;
1572 
1573 	KKASSERT(m->m_flags & M_HASH);
1574 	pmsg = &m->m_hdr.mh_netmsg;
1575 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1576 	    0, ether_demux_handler);
1577 	pmsg->nm_packet = m;
1578 	pmsg->base.lmsg.u.ms_result = isr;
1579 
1580 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1581 }
1582 
1583 u_char *
1584 kether_aton(const char *macstr, u_char *addr)
1585 {
1586         unsigned int o0, o1, o2, o3, o4, o5;
1587         int n;
1588 
1589         if (macstr == NULL || addr == NULL)
1590                 return NULL;
1591 
1592         n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1593             &o3, &o4, &o5);
1594         if (n != 6)
1595                 return NULL;
1596 
1597         addr[0] = o0;
1598         addr[1] = o1;
1599         addr[2] = o2;
1600         addr[3] = o3;
1601         addr[4] = o4;
1602         addr[5] = o5;
1603 
1604         return addr;
1605 }
1606 
1607 char *
1608 kether_ntoa(const u_char *addr, char *buf)
1609 {
1610         int len = ETHER_ADDRSTRLEN + 1;
1611         int n;
1612 
1613         n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1614             addr[1], addr[2], addr[3], addr[4], addr[5]);
1615 
1616         if (n < 17)
1617                 return NULL;
1618 
1619         return buf;
1620 }
1621 
1622 MODULE_VERSION(ether, 1);
1623