xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 7f8c68295613ce24cc71827cf210cb3d1e3bc69b)
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_ipx.h"
36 #include "opt_mpls.h"
37 #include "opt_netgraph.h"
38 #include "opt_carp.h"
39 #include "opt_rss.h"
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/globaldata.h>
44 #include <sys/kernel.h>
45 #include <sys/ktr.h>
46 #include <sys/lock.h>
47 #include <sys/malloc.h>
48 #include <sys/mbuf.h>
49 #include <sys/msgport.h>
50 #include <sys/socket.h>
51 #include <sys/sockio.h>
52 #include <sys/sysctl.h>
53 #include <sys/thread.h>
54 
55 #include <sys/thread2.h>
56 #include <sys/mplock2.h>
57 
58 #include <net/if.h>
59 #include <net/netisr.h>
60 #include <net/route.h>
61 #include <net/if_llc.h>
62 #include <net/if_dl.h>
63 #include <net/if_types.h>
64 #include <net/ifq_var.h>
65 #include <net/bpf.h>
66 #include <net/ethernet.h>
67 #include <net/vlan/if_vlan_ether.h>
68 #include <net/vlan/if_vlan_var.h>
69 #include <net/netmsg2.h>
70 #include <net/netisr2.h>
71 
72 #if defined(INET) || defined(INET6)
73 #include <netinet/in.h>
74 #include <netinet/ip_var.h>
75 #include <netinet/tcp_var.h>
76 #include <netinet/if_ether.h>
77 #include <netinet/ip_flow.h>
78 #include <net/ipfw/ip_fw.h>
79 #include <net/dummynet/ip_dummynet.h>
80 #endif
81 #ifdef INET6
82 #include <netinet6/nd6.h>
83 #endif
84 
85 #ifdef CARP
86 #include <netinet/ip_carp.h>
87 #endif
88 
89 #ifdef IPX
90 #include <netproto/ipx/ipx.h>
91 #include <netproto/ipx/ipx_if.h>
92 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
93 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
94 		  short *tp, int *hlen);
95 #endif
96 
97 #ifdef MPLS
98 #include <netproto/mpls/mpls.h>
99 #endif
100 
101 /* netgraph node hooks for ng_ether(4) */
102 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
103 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
104 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
105 void	(*ng_ether_attach_p)(struct ifnet *ifp);
106 void	(*ng_ether_detach_p)(struct ifnet *ifp);
107 
108 void	(*vlan_input_p)(struct mbuf *);
109 
110 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
111 			struct rtentry *);
112 static void ether_restore_header(struct mbuf **, const struct ether_header *,
113 				 const struct ether_header *);
114 static int ether_characterize(struct mbuf **);
115 static void ether_dispatch(int, struct mbuf *);
116 
117 /*
118  * if_bridge support
119  */
120 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
121 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
122 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
123 struct ifnet *(*bridge_interface_p)(void *if_bridge);
124 
125 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
126 			      struct sockaddr *);
127 
128 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
129 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
130 };
131 
132 #define gotoerr(e) do { error = (e); goto bad; } while (0)
133 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
134 
135 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
136 				struct ip_fw **rule,
137 				const struct ether_header *eh);
138 
139 static int ether_ipfw;
140 static u_long ether_restore_hdr;
141 static u_long ether_prepend_hdr;
142 static u_long ether_input_wronghash;
143 static int ether_debug;
144 
145 #ifdef RSS_DEBUG
146 static u_long ether_pktinfo_try;
147 static u_long ether_pktinfo_hit;
148 static u_long ether_rss_nopi;
149 static u_long ether_rss_nohash;
150 static u_long ether_input_requeue;
151 #endif
152 static u_long ether_input_wronghwhash;
153 static int ether_input_ckhash;
154 
155 #define ETHER_TSOLEN_DEFAULT	(4 * ETHERMTU)
156 
157 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
158 TUNABLE_INT("net.link.ether.tsolen", &ether_tsolen_default);
159 
160 SYSCTL_DECL(_net_link);
161 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
162 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
163     &ether_debug, 0, "Ether debug");
164 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
165     &ether_ipfw, 0, "Pass ether pkts through firewall");
166 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
167     &ether_restore_hdr, 0, "# of ether header restoration");
168 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
169     &ether_prepend_hdr, 0,
170     "# of ether header restoration which prepends mbuf");
171 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
172     &ether_input_wronghash, 0, "# of input packets with wrong hash");
173 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
174     &ether_tsolen_default, 0, "Default max TSO length");
175 
176 #ifdef RSS_DEBUG
177 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
178     &ether_rss_nopi, 0, "# of packets do not have pktinfo");
179 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
180     &ether_rss_nohash, 0, "# of packets do not have hash");
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
182     &ether_pktinfo_try, 0,
183     "# of tries to find packets' msgport using pktinfo");
184 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
185     &ether_pktinfo_hit, 0,
186     "# of packets whose msgport are found using pktinfo");
187 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
188     &ether_input_requeue, 0, "# of input packets gets requeued");
189 #endif
190 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
191     &ether_input_wronghwhash, 0, "# of input packets with wrong hw hash");
192 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
193     &ether_input_ckhash, 0, "always check hash");
194 
195 #define ETHER_KTR_STR		"ifp=%p"
196 #define ETHER_KTR_ARGS	struct ifnet *ifp
197 #ifndef KTR_ETHERNET
198 #define KTR_ETHERNET		KTR_ALL
199 #endif
200 KTR_INFO_MASTER(ether);
201 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
202 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
203 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
204 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
205 #define logether(name, arg)	KTR_LOG(ether_ ## name, arg)
206 
207 /*
208  * Ethernet output routine.
209  * Encapsulate a packet of type family for the local net.
210  * Use trailer local net encapsulation if enough data in first
211  * packet leaves a multiple of 512 bytes of data in remainder.
212  * Assumes that ifp is actually pointer to arpcom structure.
213  */
214 static int
215 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
216 	     struct rtentry *rt)
217 {
218 	struct ether_header *eh, *deh;
219 	u_char *edst;
220 	int loop_copy = 0;
221 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
222 	struct arpcom *ac = IFP2AC(ifp);
223 	int error;
224 
225 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
226 
227 	if (ifp->if_flags & IFF_MONITOR)
228 		gotoerr(ENETDOWN);
229 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
230 		gotoerr(ENETDOWN);
231 
232 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
233 	if (m == NULL)
234 		return (ENOBUFS);
235 	m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
236 	eh = mtod(m, struct ether_header *);
237 	edst = eh->ether_dhost;
238 
239 	/*
240 	 * Fill in the destination ethernet address and frame type.
241 	 */
242 	switch (dst->sa_family) {
243 #ifdef INET
244 	case AF_INET:
245 		if (!arpresolve(ifp, rt, m, dst, edst))
246 			return (0);	/* if not yet resolved */
247 #ifdef MPLS
248 		if (m->m_flags & M_MPLSLABELED)
249 			eh->ether_type = htons(ETHERTYPE_MPLS);
250 		else
251 #endif
252 			eh->ether_type = htons(ETHERTYPE_IP);
253 		break;
254 #endif
255 #ifdef INET6
256 	case AF_INET6:
257 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
258 			return (0);		/* Something bad happenned. */
259 		eh->ether_type = htons(ETHERTYPE_IPV6);
260 		break;
261 #endif
262 #ifdef IPX
263 	case AF_IPX:
264 		if (ef_outputp != NULL) {
265 			/*
266 			 * Hold BGL and recheck ef_outputp
267 			 */
268 			get_mplock();
269 			if (ef_outputp != NULL) {
270 				error = ef_outputp(ifp, &m, dst,
271 						   &eh->ether_type, &hlen);
272 				rel_mplock();
273 				if (error)
274 					goto bad;
275 				else
276 					break;
277 			}
278 			rel_mplock();
279 		}
280 		eh->ether_type = htons(ETHERTYPE_IPX);
281 		bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
282 		      edst, ETHER_ADDR_LEN);
283 		break;
284 #endif
285 	case pseudo_AF_HDRCMPLT:
286 	case AF_UNSPEC:
287 		loop_copy = -1; /* if this is for us, don't do it */
288 		deh = (struct ether_header *)dst->sa_data;
289 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
290 		eh->ether_type = deh->ether_type;
291 		break;
292 
293 	default:
294 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
295 		gotoerr(EAFNOSUPPORT);
296 	}
297 
298 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
299 		memcpy(eh->ether_shost,
300 		       ((struct ether_header *)dst->sa_data)->ether_shost,
301 		       ETHER_ADDR_LEN);
302 	else
303 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
304 
305 	/*
306 	 * Bridges require special output handling.
307 	 */
308 	if (ifp->if_bridge) {
309 		KASSERT(bridge_output_p != NULL,
310 			("%s: if_bridge not loaded!", __func__));
311 		return bridge_output_p(ifp, m);
312 	}
313 
314 	/*
315 	 * If a simplex interface, and the packet is being sent to our
316 	 * Ethernet address or a broadcast address, loopback a copy.
317 	 * XXX To make a simplex device behave exactly like a duplex
318 	 * device, we should copy in the case of sending to our own
319 	 * ethernet address (thus letting the original actually appear
320 	 * on the wire). However, we don't do that here for security
321 	 * reasons and compatibility with the original behavior.
322 	 */
323 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
324 		int csum_flags = 0;
325 
326 		if (m->m_pkthdr.csum_flags & CSUM_IP)
327 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
328 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
329 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
330 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
331 			struct mbuf *n;
332 
333 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
334 				n->m_pkthdr.csum_flags |= csum_flags;
335 				if (csum_flags & CSUM_DATA_VALID)
336 					n->m_pkthdr.csum_data = 0xffff;
337 				if_simloop(ifp, n, dst->sa_family, hlen);
338 			} else
339 				IFNET_STAT_INC(ifp, iqdrops, 1);
340 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
341 				ETHER_ADDR_LEN) == 0) {
342 			m->m_pkthdr.csum_flags |= csum_flags;
343 			if (csum_flags & CSUM_DATA_VALID)
344 				m->m_pkthdr.csum_data = 0xffff;
345 			if_simloop(ifp, m, dst->sa_family, hlen);
346 			return (0);	/* XXX */
347 		}
348 	}
349 
350 #ifdef CARP
351 	if (ifp->if_type == IFT_CARP) {
352 		ifp = carp_parent(ifp);
353 		if (ifp == NULL)
354 			gotoerr(ENETUNREACH);
355 
356 		ac = IFP2AC(ifp);
357 
358 		/*
359 		 * Check precondition again
360 		 */
361 		ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
362 
363 		if (ifp->if_flags & IFF_MONITOR)
364 			gotoerr(ENETDOWN);
365 		if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
366 		    (IFF_UP | IFF_RUNNING))
367 			gotoerr(ENETDOWN);
368 	}
369 #endif
370 
371 	/* Handle ng_ether(4) processing, if any */
372 	if (ng_ether_output_p != NULL) {
373 		/*
374 		 * Hold BGL and recheck ng_ether_output_p
375 		 */
376 		get_mplock();
377 		if (ng_ether_output_p != NULL) {
378 			if ((error = ng_ether_output_p(ifp, &m)) != 0) {
379 				rel_mplock();
380 				goto bad;
381 			}
382 			if (m == NULL) {
383 				rel_mplock();
384 				return (0);
385 			}
386 		}
387 		rel_mplock();
388 	}
389 
390 	/* Continue with link-layer output */
391 	return ether_output_frame(ifp, m);
392 
393 bad:
394 	m_freem(m);
395 	return (error);
396 }
397 
398 /*
399  * Returns the bridge interface an ifp is associated
400  * with.
401  *
402  * Only call if ifp->if_bridge != NULL.
403  */
404 struct ifnet *
405 ether_bridge_interface(struct ifnet *ifp)
406 {
407 	if (bridge_interface_p)
408 		return(bridge_interface_p(ifp->if_bridge));
409 	return (ifp);
410 }
411 
412 /*
413  * Ethernet link layer output routine to send a raw frame to the device.
414  *
415  * This assumes that the 14 byte Ethernet header is present and contiguous
416  * in the first mbuf.
417  */
418 int
419 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
420 {
421 	struct ip_fw *rule = NULL;
422 	int error = 0;
423 	struct altq_pktattr pktattr;
424 
425 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
426 
427 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
428 		struct m_tag *mtag;
429 
430 		/* Extract info from dummynet tag */
431 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
432 		KKASSERT(mtag != NULL);
433 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
434 		KKASSERT(rule != NULL);
435 
436 		m_tag_delete(m, mtag);
437 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
438 	}
439 
440 	if (ifq_is_enabled(&ifp->if_snd))
441 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
442 	crit_enter();
443 	if (IPFW_LOADED && ether_ipfw != 0) {
444 		struct ether_header save_eh, *eh;
445 
446 		eh = mtod(m, struct ether_header *);
447 		save_eh = *eh;
448 		m_adj(m, ETHER_HDR_LEN);
449 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
450 			crit_exit();
451 			if (m != NULL) {
452 				m_freem(m);
453 				return ENOBUFS; /* pkt dropped */
454 			} else
455 				return 0;	/* consumed e.g. in a pipe */
456 		}
457 
458 		/* packet was ok, restore the ethernet header */
459 		ether_restore_header(&m, eh, &save_eh);
460 		if (m == NULL) {
461 			crit_exit();
462 			return ENOBUFS;
463 		}
464 	}
465 	crit_exit();
466 
467 	/*
468 	 * Queue message on interface, update output statistics if
469 	 * successful, and start output if interface not yet active.
470 	 */
471 	error = ifq_dispatch(ifp, m, &pktattr);
472 	return (error);
473 }
474 
475 /*
476  * ipfw processing for ethernet packets (in and out).
477  * The second parameter is NULL from ether_demux(), and ifp from
478  * ether_output_frame().
479  */
480 static boolean_t
481 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
482 	       const struct ether_header *eh)
483 {
484 	struct ether_header save_eh = *eh;	/* might be a ptr in *m0 */
485 	struct ip_fw_args args;
486 	struct m_tag *mtag;
487 	struct mbuf *m;
488 	int i;
489 
490 	if (*rule != NULL && fw_one_pass)
491 		return TRUE; /* dummynet packet, already partially processed */
492 
493 	/*
494 	 * I need some amount of data to be contiguous.
495 	 */
496 	i = min((*m0)->m_pkthdr.len, max_protohdr);
497 	if ((*m0)->m_len < i) {
498 		*m0 = m_pullup(*m0, i);
499 		if (*m0 == NULL)
500 			return FALSE;
501 	}
502 
503 	/*
504 	 * Clean up tags
505 	 */
506 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
507 		m_tag_delete(*m0, mtag);
508 	if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
509 		mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
510 		KKASSERT(mtag != NULL);
511 		m_tag_delete(*m0, mtag);
512 		(*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
513 	}
514 
515 	args.m = *m0;		/* the packet we are looking at		*/
516 	args.oif = dst;		/* destination, if any			*/
517 	args.rule = *rule;	/* matching rule to restart		*/
518 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
519 	i = ip_fw_chk_ptr(&args);
520 	*m0 = args.m;
521 	*rule = args.rule;
522 
523 	if (*m0 == NULL)
524 		return FALSE;
525 
526 	switch (i) {
527 	case IP_FW_PASS:
528 		return TRUE;
529 
530 	case IP_FW_DIVERT:
531 	case IP_FW_TEE:
532 	case IP_FW_DENY:
533 		/*
534 		 * XXX at some point add support for divert/forward actions.
535 		 * If none of the above matches, we have to drop the pkt.
536 		 */
537 		return FALSE;
538 
539 	case IP_FW_DUMMYNET:
540 		/*
541 		 * Pass the pkt to dummynet, which consumes it.
542 		 */
543 		m = *m0;	/* pass the original to dummynet */
544 		*m0 = NULL;	/* and nothing back to the caller */
545 
546 		ether_restore_header(&m, eh, &save_eh);
547 		if (m == NULL)
548 			return FALSE;
549 
550 		ip_fw_dn_io_ptr(m, args.cookie,
551 				dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
552 		ip_dn_queue(m);
553 		return FALSE;
554 
555 	default:
556 		panic("unknown ipfw return value: %d", i);
557 	}
558 }
559 
560 static void
561 ether_input(struct ifnet *ifp, struct mbuf *m)
562 {
563 	ether_input_pkt(ifp, m, NULL);
564 }
565 
566 /*
567  * Perform common duties while attaching to interface list
568  */
569 void
570 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
571 {
572 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
573 			   serializer);
574 }
575 
576 void
577 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
578 		   lwkt_serialize_t serializer)
579 {
580 	struct sockaddr_dl *sdl;
581 	char ethstr[ETHER_ADDRSTRLEN + 1];
582 
583 	ifp->if_type = IFT_ETHER;
584 	ifp->if_addrlen = ETHER_ADDR_LEN;
585 	ifp->if_hdrlen = ETHER_HDR_LEN;
586 	if_attach(ifp, serializer);
587 	ifp->if_mtu = ETHERMTU;
588 	if (ifp->if_tsolen <= 0) {
589 		if ((ether_tsolen_default / ETHERMTU) < 2) {
590 			kprintf("ether TSO maxlen %d -> %d\n",
591 			    ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
592 			ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
593 		}
594 		ifp->if_tsolen = ether_tsolen_default;
595 	}
596 	if (ifp->if_baudrate == 0)
597 		ifp->if_baudrate = 10000000;
598 	ifp->if_output = ether_output;
599 	ifp->if_input = ether_input;
600 	ifp->if_resolvemulti = ether_resolvemulti;
601 	ifp->if_broadcastaddr = etherbroadcastaddr;
602 	sdl = IF_LLSOCKADDR(ifp);
603 	sdl->sdl_type = IFT_ETHER;
604 	sdl->sdl_alen = ifp->if_addrlen;
605 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
606 	/*
607 	 * XXX Keep the current drivers happy.
608 	 * XXX Remove once all drivers have been cleaned up
609 	 */
610 	if (lla != IFP2AC(ifp)->ac_enaddr)
611 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
612 	bpfattach(ifp, dlt, hdrlen);
613 	if (ng_ether_attach_p != NULL)
614 		(*ng_ether_attach_p)(ifp);
615 
616 	if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
617 }
618 
619 /*
620  * Perform common duties while detaching an Ethernet interface
621  */
622 void
623 ether_ifdetach(struct ifnet *ifp)
624 {
625 	if_down(ifp);
626 
627 	if (ng_ether_detach_p != NULL)
628 		(*ng_ether_detach_p)(ifp);
629 	bpfdetach(ifp);
630 	if_detach(ifp);
631 }
632 
633 int
634 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
635 {
636 	struct ifaddr *ifa = (struct ifaddr *) data;
637 	struct ifreq *ifr = (struct ifreq *) data;
638 	int error = 0;
639 
640 #define IF_INIT(ifp) \
641 do { \
642 	if (((ifp)->if_flags & IFF_UP) == 0) { \
643 		(ifp)->if_flags |= IFF_UP; \
644 		(ifp)->if_init((ifp)->if_softc); \
645 	} \
646 } while (0)
647 
648 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
649 
650 	switch (command) {
651 	case SIOCSIFADDR:
652 		switch (ifa->ifa_addr->sa_family) {
653 #ifdef INET
654 		case AF_INET:
655 			IF_INIT(ifp);	/* before arpwhohas */
656 			arp_ifinit(ifp, ifa);
657 			break;
658 #endif
659 #ifdef IPX
660 		/*
661 		 * XXX - This code is probably wrong
662 		 */
663 		case AF_IPX:
664 			{
665 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
666 			struct arpcom *ac = IFP2AC(ifp);
667 
668 			if (ipx_nullhost(*ina))
669 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
670 			else
671 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
672 				      sizeof ac->ac_enaddr);
673 
674 			IF_INIT(ifp);	/* Set new address. */
675 			break;
676 			}
677 #endif
678 		default:
679 			IF_INIT(ifp);
680 			break;
681 		}
682 		break;
683 
684 	case SIOCGIFADDR:
685 		bcopy(IFP2AC(ifp)->ac_enaddr,
686 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
687 		      ETHER_ADDR_LEN);
688 		break;
689 
690 	case SIOCSIFMTU:
691 		/*
692 		 * Set the interface MTU.
693 		 */
694 		if (ifr->ifr_mtu > ETHERMTU) {
695 			error = EINVAL;
696 		} else {
697 			ifp->if_mtu = ifr->ifr_mtu;
698 		}
699 		break;
700 	default:
701 		error = EINVAL;
702 		break;
703 	}
704 	return (error);
705 
706 #undef IF_INIT
707 }
708 
709 int
710 ether_resolvemulti(
711 	struct ifnet *ifp,
712 	struct sockaddr **llsa,
713 	struct sockaddr *sa)
714 {
715 	struct sockaddr_dl *sdl;
716 #ifdef INET
717 	struct sockaddr_in *sin;
718 #endif
719 #ifdef INET6
720 	struct sockaddr_in6 *sin6;
721 #endif
722 	u_char *e_addr;
723 
724 	switch(sa->sa_family) {
725 	case AF_LINK:
726 		/*
727 		 * No mapping needed. Just check that it's a valid MC address.
728 		 */
729 		sdl = (struct sockaddr_dl *)sa;
730 		e_addr = LLADDR(sdl);
731 		if ((e_addr[0] & 1) != 1)
732 			return EADDRNOTAVAIL;
733 		*llsa = NULL;
734 		return 0;
735 
736 #ifdef INET
737 	case AF_INET:
738 		sin = (struct sockaddr_in *)sa;
739 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
740 			return EADDRNOTAVAIL;
741 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
742 		sdl->sdl_len = sizeof *sdl;
743 		sdl->sdl_family = AF_LINK;
744 		sdl->sdl_index = ifp->if_index;
745 		sdl->sdl_type = IFT_ETHER;
746 		sdl->sdl_alen = ETHER_ADDR_LEN;
747 		e_addr = LLADDR(sdl);
748 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
749 		*llsa = (struct sockaddr *)sdl;
750 		return 0;
751 #endif
752 #ifdef INET6
753 	case AF_INET6:
754 		sin6 = (struct sockaddr_in6 *)sa;
755 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
756 			/*
757 			 * An IP6 address of 0 means listen to all
758 			 * of the Ethernet multicast address used for IP6.
759 			 * (This is used for multicast routers.)
760 			 */
761 			ifp->if_flags |= IFF_ALLMULTI;
762 			*llsa = NULL;
763 			return 0;
764 		}
765 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
766 			return EADDRNOTAVAIL;
767 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
768 		sdl->sdl_len = sizeof *sdl;
769 		sdl->sdl_family = AF_LINK;
770 		sdl->sdl_index = ifp->if_index;
771 		sdl->sdl_type = IFT_ETHER;
772 		sdl->sdl_alen = ETHER_ADDR_LEN;
773 		e_addr = LLADDR(sdl);
774 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
775 		*llsa = (struct sockaddr *)sdl;
776 		return 0;
777 #endif
778 
779 	default:
780 		/*
781 		 * Well, the text isn't quite right, but it's the name
782 		 * that counts...
783 		 */
784 		return EAFNOSUPPORT;
785 	}
786 }
787 
788 #if 0
789 /*
790  * This is for reference.  We have a table-driven version
791  * of the little-endian crc32 generator, which is faster
792  * than the double-loop.
793  */
794 uint32_t
795 ether_crc32_le(const uint8_t *buf, size_t len)
796 {
797 	uint32_t c, crc, carry;
798 	size_t i, j;
799 
800 	crc = 0xffffffffU;	/* initial value */
801 
802 	for (i = 0; i < len; i++) {
803 		c = buf[i];
804 		for (j = 0; j < 8; j++) {
805 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
806 			crc >>= 1;
807 			c >>= 1;
808 			if (carry)
809 				crc = (crc ^ ETHER_CRC_POLY_LE);
810 		}
811 	}
812 
813 	return (crc);
814 }
815 #else
816 uint32_t
817 ether_crc32_le(const uint8_t *buf, size_t len)
818 {
819 	static const uint32_t crctab[] = {
820 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
821 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
822 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
823 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
824 	};
825 	uint32_t crc;
826 	size_t i;
827 
828 	crc = 0xffffffffU;	/* initial value */
829 
830 	for (i = 0; i < len; i++) {
831 		crc ^= buf[i];
832 		crc = (crc >> 4) ^ crctab[crc & 0xf];
833 		crc = (crc >> 4) ^ crctab[crc & 0xf];
834 	}
835 
836 	return (crc);
837 }
838 #endif
839 
840 uint32_t
841 ether_crc32_be(const uint8_t *buf, size_t len)
842 {
843 	uint32_t c, crc, carry;
844 	size_t i, j;
845 
846 	crc = 0xffffffffU;	/* initial value */
847 
848 	for (i = 0; i < len; i++) {
849 		c = buf[i];
850 		for (j = 0; j < 8; j++) {
851 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
852 			crc <<= 1;
853 			c >>= 1;
854 			if (carry)
855 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
856 		}
857 	}
858 
859 	return (crc);
860 }
861 
862 /*
863  * find the size of ethernet header, and call classifier
864  */
865 void
866 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
867 		   struct altq_pktattr *pktattr)
868 {
869 	struct ether_header *eh;
870 	uint16_t ether_type;
871 	int hlen, af, hdrsize;
872 
873 	hlen = sizeof(struct ether_header);
874 	eh = mtod(m, struct ether_header *);
875 
876 	ether_type = ntohs(eh->ether_type);
877 	if (ether_type < ETHERMTU) {
878 		/* ick! LLC/SNAP */
879 		struct llc *llc = (struct llc *)(eh + 1);
880 		hlen += 8;
881 
882 		if (m->m_len < hlen ||
883 		    llc->llc_dsap != LLC_SNAP_LSAP ||
884 		    llc->llc_ssap != LLC_SNAP_LSAP ||
885 		    llc->llc_control != LLC_UI)
886 			goto bad;  /* not snap! */
887 
888 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
889 	}
890 
891 	if (ether_type == ETHERTYPE_IP) {
892 		af = AF_INET;
893 		hdrsize = 20;  /* sizeof(struct ip) */
894 #ifdef INET6
895 	} else if (ether_type == ETHERTYPE_IPV6) {
896 		af = AF_INET6;
897 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
898 #endif
899 	} else
900 		goto bad;
901 
902 	while (m->m_len <= hlen) {
903 		hlen -= m->m_len;
904 		m = m->m_next;
905 	}
906 	if (m->m_len < hlen + hdrsize) {
907 		/*
908 		 * ip header is not in a single mbuf.  this should not
909 		 * happen in the current code.
910 		 * (todo: use m_pulldown in the future)
911 		 */
912 		goto bad;
913 	}
914 	m->m_data += hlen;
915 	m->m_len -= hlen;
916 	ifq_classify(ifq, m, af, pktattr);
917 	m->m_data -= hlen;
918 	m->m_len += hlen;
919 
920 	return;
921 
922 bad:
923 	pktattr->pattr_class = NULL;
924 	pktattr->pattr_hdr = NULL;
925 	pktattr->pattr_af = AF_UNSPEC;
926 }
927 
928 static void
929 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
930 		     const struct ether_header *save_eh)
931 {
932 	struct mbuf *m = *m0;
933 
934 	ether_restore_hdr++;
935 
936 	/*
937 	 * Prepend the header, optimize for the common case of
938 	 * eh pointing into the mbuf.
939 	 */
940 	if ((const void *)(eh + 1) == (void *)m->m_data) {
941 		m->m_data -= ETHER_HDR_LEN;
942 		m->m_len += ETHER_HDR_LEN;
943 		m->m_pkthdr.len += ETHER_HDR_LEN;
944 	} else {
945 		ether_prepend_hdr++;
946 
947 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
948 		if (m != NULL) {
949 			bcopy(save_eh, mtod(m, struct ether_header *),
950 			      ETHER_HDR_LEN);
951 		}
952 	}
953 	*m0 = m;
954 }
955 
956 /*
957  * Upper layer processing for a received Ethernet packet.
958  */
959 void
960 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
961 {
962 	struct ether_header *eh;
963 	int isr, discard = 0;
964 	u_short ether_type;
965 	struct ip_fw *rule = NULL;
966 
967 	M_ASSERTPKTHDR(m);
968 	KASSERT(m->m_len >= ETHER_HDR_LEN,
969 		("ether header is not contiguous!"));
970 
971 	eh = mtod(m, struct ether_header *);
972 
973 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
974 		struct m_tag *mtag;
975 
976 		/* Extract info from dummynet tag */
977 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
978 		KKASSERT(mtag != NULL);
979 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
980 		KKASSERT(rule != NULL);
981 
982 		m_tag_delete(m, mtag);
983 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
984 
985 		/* packet is passing the second time */
986 		goto post_stats;
987 	}
988 
989 	/*
990 	 * We got a packet which was unicast to a different Ethernet
991 	 * address.  If the driver is working properly, then this
992 	 * situation can only happen when the interface is in
993 	 * promiscuous mode.  We defer the packet discarding until the
994 	 * vlan processing is done, so that vlan/bridge or vlan/netgraph
995 	 * could work.
996 	 */
997 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
998 	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
999 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1000 		if (ether_debug & 1) {
1001 			kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
1002 				"%02x:%02x:%02x:%02x:%02x:%02x "
1003 				"%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
1004 				eh->ether_dhost[0],
1005 				eh->ether_dhost[1],
1006 				eh->ether_dhost[2],
1007 				eh->ether_dhost[3],
1008 				eh->ether_dhost[4],
1009 				eh->ether_dhost[5],
1010 				eh->ether_shost[0],
1011 				eh->ether_shost[1],
1012 				eh->ether_shost[2],
1013 				eh->ether_shost[3],
1014 				eh->ether_shost[4],
1015 				eh->ether_shost[5],
1016 				eh->ether_type,
1017 				((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1018 				((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1019 				((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1020 				((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1021 				((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1022 				((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1023 			);
1024 		}
1025 		if ((ether_debug & 2) == 0)
1026 			discard = 1;
1027 	}
1028 
1029 post_stats:
1030 	if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
1031 		struct ether_header save_eh = *eh;
1032 
1033 		/* XXX old crufty stuff, needs to be removed */
1034 		m_adj(m, sizeof(struct ether_header));
1035 
1036 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1037 			m_freem(m);
1038 			return;
1039 		}
1040 
1041 		ether_restore_header(&m, eh, &save_eh);
1042 		if (m == NULL)
1043 			return;
1044 		eh = mtod(m, struct ether_header *);
1045 	}
1046 
1047 	ether_type = ntohs(eh->ether_type);
1048 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1049 
1050 	if (m->m_flags & M_VLANTAG) {
1051 		void (*vlan_input_func)(struct mbuf *);
1052 
1053 		vlan_input_func = vlan_input_p;
1054 		if (vlan_input_func != NULL) {
1055 			vlan_input_func(m);
1056 		} else {
1057 			IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1058 			m_freem(m);
1059 		}
1060 		return;
1061 	}
1062 
1063 	/*
1064 	 * If we have been asked to discard this packet
1065 	 * (e.g. not for us), drop it before entering
1066 	 * the upper layer.
1067 	 */
1068 	if (discard) {
1069 		m_freem(m);
1070 		return;
1071 	}
1072 
1073 	/*
1074 	 * Clear protocol specific flags,
1075 	 * before entering the upper layer.
1076 	 */
1077 	m->m_flags &= ~M_ETHER_FLAGS;
1078 
1079 	/* Strip ethernet header. */
1080 	m_adj(m, sizeof(struct ether_header));
1081 
1082 	switch (ether_type) {
1083 #ifdef INET
1084 	case ETHERTYPE_IP:
1085 		if ((m->m_flags & M_LENCHECKED) == 0) {
1086 			if (!ip_lengthcheck(&m, 0))
1087 				return;
1088 		}
1089 		if (ipflow_fastforward(m))
1090 			return;
1091 		isr = NETISR_IP;
1092 		break;
1093 
1094 	case ETHERTYPE_ARP:
1095 		if (ifp->if_flags & IFF_NOARP) {
1096 			/* Discard packet if ARP is disabled on interface */
1097 			m_freem(m);
1098 			return;
1099 		}
1100 		isr = NETISR_ARP;
1101 		break;
1102 #endif
1103 
1104 #ifdef INET6
1105 	case ETHERTYPE_IPV6:
1106 		isr = NETISR_IPV6;
1107 		break;
1108 #endif
1109 
1110 #ifdef IPX
1111 	case ETHERTYPE_IPX:
1112 		if (ef_inputp) {
1113 			/*
1114 			 * Hold BGL and recheck ef_inputp
1115 			 */
1116 			get_mplock();
1117 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1118 				rel_mplock();
1119 				return;
1120 			}
1121 			rel_mplock();
1122 		}
1123 		isr = NETISR_IPX;
1124 		break;
1125 #endif
1126 
1127 #ifdef MPLS
1128 	case ETHERTYPE_MPLS:
1129 	case ETHERTYPE_MPLS_MCAST:
1130 		/* Should have been set by ether_input_pkt(). */
1131 		KKASSERT(m->m_flags & M_MPLSLABELED);
1132 		isr = NETISR_MPLS;
1133 		break;
1134 #endif
1135 
1136 	default:
1137 		/*
1138 		 * The accurate msgport is not determined before
1139 		 * we reach here, so recharacterize packet.
1140 		 */
1141 		m->m_flags &= ~M_HASH;
1142 #ifdef IPX
1143 		if (ef_inputp) {
1144 			/*
1145 			 * Hold BGL and recheck ef_inputp
1146 			 */
1147 			get_mplock();
1148 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1149 				rel_mplock();
1150 				return;
1151 			}
1152 			rel_mplock();
1153 		}
1154 #endif
1155 		if (ng_ether_input_orphan_p != NULL) {
1156 			/*
1157 			 * Put back the ethernet header so netgraph has a
1158 			 * consistent view of inbound packets.
1159 			 */
1160 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1161 			if (m == NULL) {
1162 				/*
1163 				 * M_PREPEND frees the mbuf in case of failure.
1164 				 */
1165 				return;
1166 			}
1167 			/*
1168 			 * Hold BGL and recheck ng_ether_input_orphan_p
1169 			 */
1170 			get_mplock();
1171 			if (ng_ether_input_orphan_p != NULL) {
1172 				ng_ether_input_orphan_p(ifp, m);
1173 				rel_mplock();
1174 				return;
1175 			}
1176 			rel_mplock();
1177 		}
1178 		m_freem(m);
1179 		return;
1180 	}
1181 
1182 	if (m->m_flags & M_HASH) {
1183 		if (&curthread->td_msgport ==
1184 		    netisr_hashport(m->m_pkthdr.hash)) {
1185 			netisr_handle(isr, m);
1186 			return;
1187 		} else {
1188 			/*
1189 			 * XXX Something is wrong,
1190 			 * we probably should panic here!
1191 			 */
1192 			m->m_flags &= ~M_HASH;
1193 			atomic_add_long(&ether_input_wronghash, 1);
1194 		}
1195 	}
1196 #ifdef RSS_DEBUG
1197 	atomic_add_long(&ether_input_requeue, 1);
1198 #endif
1199 	netisr_queue(isr, m);
1200 }
1201 
1202 /*
1203  * First we perform any link layer operations, then continue to the
1204  * upper layers with ether_demux_oncpu().
1205  */
1206 static void
1207 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1208 {
1209 #ifdef CARP
1210 	void *carp;
1211 #endif
1212 
1213 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1214 		/*
1215 		 * Receiving interface's flags are changed, when this
1216 		 * packet is waiting for processing; discard it.
1217 		 */
1218 		m_freem(m);
1219 		return;
1220 	}
1221 
1222 	/*
1223 	 * Tap the packet off here for a bridge.  bridge_input()
1224 	 * will return NULL if it has consumed the packet, otherwise
1225 	 * it gets processed as normal.  Note that bridge_input()
1226 	 * will always return the original packet if we need to
1227 	 * process it locally.
1228 	 */
1229 	if (ifp->if_bridge) {
1230 		KASSERT(bridge_input_p != NULL,
1231 			("%s: if_bridge not loaded!", __func__));
1232 
1233 		if(m->m_flags & M_ETHER_BRIDGED) {
1234 			m->m_flags &= ~M_ETHER_BRIDGED;
1235 		} else {
1236 			m = bridge_input_p(ifp, m);
1237 			if (m == NULL)
1238 				return;
1239 
1240 			KASSERT(ifp == m->m_pkthdr.rcvif,
1241 				("bridge_input_p changed rcvif"));
1242 		}
1243 	}
1244 
1245 #ifdef CARP
1246 	carp = ifp->if_carp;
1247 	if (carp) {
1248 		m = carp_input(carp, m);
1249 		if (m == NULL)
1250 			return;
1251 		KASSERT(ifp == m->m_pkthdr.rcvif,
1252 		    ("carp_input changed rcvif"));
1253 	}
1254 #endif
1255 
1256 	/* Handle ng_ether(4) processing, if any */
1257 	if (ng_ether_input_p != NULL) {
1258 		/*
1259 		 * Hold BGL and recheck ng_ether_input_p
1260 		 */
1261 		get_mplock();
1262 		if (ng_ether_input_p != NULL)
1263 			ng_ether_input_p(ifp, &m);
1264 		rel_mplock();
1265 
1266 		if (m == NULL)
1267 			return;
1268 	}
1269 
1270 	/* Continue with upper layer processing */
1271 	ether_demux_oncpu(ifp, m);
1272 }
1273 
1274 /*
1275  * Perform certain functions of ether_input_pkt():
1276  * - Test IFF_UP
1277  * - Update statistics
1278  * - Run bpf(4) tap if requested
1279  * Then pass the packet to ether_input_oncpu().
1280  *
1281  * This function should be used by pseudo interface (e.g. vlan(4)),
1282  * when it tries to claim that the packet is received by it.
1283  *
1284  * REINPUT_KEEPRCVIF
1285  * REINPUT_RUNBPF
1286  */
1287 void
1288 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1289 {
1290 	/* Discard packet if interface is not up */
1291 	if (!(ifp->if_flags & IFF_UP)) {
1292 		m_freem(m);
1293 		return;
1294 	}
1295 
1296 	/*
1297 	 * Change receiving interface.  The bridge will often pass a flag to
1298 	 * ask that this not be done so ARPs get applied to the correct
1299 	 * side.
1300 	 */
1301 	if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1302 	    m->m_pkthdr.rcvif == NULL) {
1303 		m->m_pkthdr.rcvif = ifp;
1304 	}
1305 
1306 	/* Update statistics */
1307 	IFNET_STAT_INC(ifp, ipackets, 1);
1308 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1309 	if (m->m_flags & (M_MCAST | M_BCAST))
1310 		IFNET_STAT_INC(ifp, imcasts, 1);
1311 
1312 	if (reinput_flags & REINPUT_RUNBPF)
1313 		BPF_MTAP(ifp, m);
1314 
1315 	ether_input_oncpu(ifp, m);
1316 }
1317 
1318 static __inline boolean_t
1319 ether_vlancheck(struct mbuf **m0)
1320 {
1321 	struct mbuf *m = *m0;
1322 	struct ether_header *eh;
1323 	uint16_t ether_type;
1324 
1325 	eh = mtod(m, struct ether_header *);
1326 	ether_type = ntohs(eh->ether_type);
1327 
1328 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1329 		/*
1330 		 * Extract vlan tag if hardware does not do it for us
1331 		 */
1332 		vlan_ether_decap(&m);
1333 		if (m == NULL)
1334 			goto failed;
1335 
1336 		eh = mtod(m, struct ether_header *);
1337 		ether_type = ntohs(eh->ether_type);
1338 	}
1339 
1340 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1341 		/*
1342 		 * To prevent possible dangerous recursion,
1343 		 * we don't do vlan-in-vlan
1344 		 */
1345 		IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1346 		goto failed;
1347 	}
1348 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1349 
1350 	m->m_flags |= M_ETHER_VLANCHECKED;
1351 	*m0 = m;
1352 	return TRUE;
1353 failed:
1354 	if (m != NULL)
1355 		m_freem(m);
1356 	*m0 = NULL;
1357 	return FALSE;
1358 }
1359 
1360 static void
1361 ether_input_handler(netmsg_t nmsg)
1362 {
1363 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1364 	struct ether_header *eh;
1365 	struct ifnet *ifp;
1366 	struct mbuf *m;
1367 
1368 	m = nmp->nm_packet;
1369 	M_ASSERTPKTHDR(m);
1370 
1371 	if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1372 		if (!ether_vlancheck(&m)) {
1373 			KKASSERT(m == NULL);
1374 			return;
1375 		}
1376 	}
1377 	if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1378 	    __predict_false(ether_input_ckhash)) {
1379 		int isr;
1380 
1381 		/*
1382 		 * Need to verify the hash supplied by the hardware
1383 		 * which could be wrong.
1384 		 */
1385 		m->m_flags &= ~(M_HASH | M_CKHASH);
1386 		isr = ether_characterize(&m);
1387 		if (m == NULL)
1388 			return;
1389 		KKASSERT(m->m_flags & M_HASH);
1390 
1391 		if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1392 			/*
1393 			 * Wrong hardware supplied hash; redispatch
1394 			 */
1395 			ether_dispatch(isr, m);
1396 			if (__predict_false(ether_input_ckhash))
1397 				atomic_add_long(&ether_input_wronghwhash, 1);
1398 			return;
1399 		}
1400 	}
1401 	ifp = m->m_pkthdr.rcvif;
1402 
1403 	eh = mtod(m, struct ether_header *);
1404 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1405 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1406 			 ifp->if_addrlen) == 0)
1407 			m->m_flags |= M_BCAST;
1408 		else
1409 			m->m_flags |= M_MCAST;
1410 		IFNET_STAT_INC(ifp, imcasts, 1);
1411 	}
1412 
1413 	ether_input_oncpu(ifp, m);
1414 }
1415 
1416 /*
1417  * Send the packet to the target netisr msgport
1418  *
1419  * At this point the packet must be characterized (M_HASH set),
1420  * so we know which netisr to send it to.
1421  */
1422 static void
1423 ether_dispatch(int isr, struct mbuf *m)
1424 {
1425 	struct netmsg_packet *pmsg;
1426 
1427 	KKASSERT(m->m_flags & M_HASH);
1428 	pmsg = &m->m_hdr.mh_netmsg;
1429 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1430 		    0, ether_input_handler);
1431 	pmsg->nm_packet = m;
1432 	pmsg->base.lmsg.u.ms_result = isr;
1433 
1434 	logether(disp_beg, NULL);
1435 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1436 	logether(disp_end, NULL);
1437 }
1438 
1439 /*
1440  * Process a received Ethernet packet.
1441  *
1442  * The ethernet header is assumed to be in the mbuf so the caller
1443  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1444  * bytes in the first mbuf.
1445  */
1446 void
1447 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi)
1448 {
1449 	int isr;
1450 
1451 	M_ASSERTPKTHDR(m);
1452 
1453 	/* Discard packet if interface is not up */
1454 	if (!(ifp->if_flags & IFF_UP)) {
1455 		m_freem(m);
1456 		return;
1457 	}
1458 
1459 	if (m->m_len < sizeof(struct ether_header)) {
1460 		/* XXX error in the caller. */
1461 		m_freem(m);
1462 		return;
1463 	}
1464 
1465 	m->m_pkthdr.rcvif = ifp;
1466 
1467 	logether(pkt_beg, ifp);
1468 
1469 	ETHER_BPF_MTAP(ifp, m);
1470 
1471 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1472 
1473 	if (ifp->if_flags & IFF_MONITOR) {
1474 		struct ether_header *eh;
1475 
1476 		eh = mtod(m, struct ether_header *);
1477 		if (ETHER_IS_MULTICAST(eh->ether_dhost))
1478 			IFNET_STAT_INC(ifp, imcasts, 1);
1479 
1480 		/*
1481 		 * Interface marked for monitoring; discard packet.
1482 		 */
1483 		m_freem(m);
1484 
1485 		logether(pkt_end, ifp);
1486 		return;
1487 	}
1488 
1489 	/*
1490 	 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1491 	 * we can dispatch it immediately with trivial checks.
1492 	 */
1493 	if (pi != NULL && (m->m_flags & M_HASH)) {
1494 #ifdef RSS_DEBUG
1495 		atomic_add_long(&ether_pktinfo_try, 1);
1496 #endif
1497 		netisr_hashcheck(pi->pi_netisr, m, pi);
1498 		if (m->m_flags & M_HASH) {
1499 			ether_dispatch(pi->pi_netisr, m);
1500 #ifdef RSS_DEBUG
1501 			atomic_add_long(&ether_pktinfo_hit, 1);
1502 #endif
1503 			logether(pkt_end, ifp);
1504 			return;
1505 		}
1506 	}
1507 #ifdef RSS_DEBUG
1508 	else if (ifp->if_capenable & IFCAP_RSS) {
1509 		if (pi == NULL)
1510 			atomic_add_long(&ether_rss_nopi, 1);
1511 		else
1512 			atomic_add_long(&ether_rss_nohash, 1);
1513 	}
1514 #endif
1515 
1516 	/*
1517 	 * Packet hash will be recalculated by software, so clear
1518 	 * the M_HASH and M_CKHASH flag set by the driver; the hash
1519 	 * value calculated by the hardware may not be exactly what
1520 	 * we want.
1521 	 */
1522 	m->m_flags &= ~(M_HASH | M_CKHASH);
1523 
1524 	if (!ether_vlancheck(&m)) {
1525 		KKASSERT(m == NULL);
1526 		logether(pkt_end, ifp);
1527 		return;
1528 	}
1529 
1530 	isr = ether_characterize(&m);
1531 	if (m == NULL) {
1532 		logether(pkt_end, ifp);
1533 		return;
1534 	}
1535 
1536 	/*
1537 	 * Finally dispatch it
1538 	 */
1539 	ether_dispatch(isr, m);
1540 
1541 	logether(pkt_end, ifp);
1542 }
1543 
1544 static int
1545 ether_characterize(struct mbuf **m0)
1546 {
1547 	struct mbuf *m = *m0;
1548 	struct ether_header *eh;
1549 	uint16_t ether_type;
1550 	int isr;
1551 
1552 	eh = mtod(m, struct ether_header *);
1553 	ether_type = ntohs(eh->ether_type);
1554 
1555 	/*
1556 	 * Map ether type to netisr id.
1557 	 */
1558 	switch (ether_type) {
1559 #ifdef INET
1560 	case ETHERTYPE_IP:
1561 		isr = NETISR_IP;
1562 		break;
1563 
1564 	case ETHERTYPE_ARP:
1565 		isr = NETISR_ARP;
1566 		break;
1567 #endif
1568 
1569 #ifdef INET6
1570 	case ETHERTYPE_IPV6:
1571 		isr = NETISR_IPV6;
1572 		break;
1573 #endif
1574 
1575 #ifdef IPX
1576 	case ETHERTYPE_IPX:
1577 		isr = NETISR_IPX;
1578 		break;
1579 #endif
1580 
1581 #ifdef MPLS
1582 	case ETHERTYPE_MPLS:
1583 	case ETHERTYPE_MPLS_MCAST:
1584 		m->m_flags |= M_MPLSLABELED;
1585 		isr = NETISR_MPLS;
1586 		break;
1587 #endif
1588 
1589 	default:
1590 		/*
1591 		 * NETISR_MAX is an invalid value; it is chosen to let
1592 		 * netisr_characterize() know that we have no clear
1593 		 * idea where this packet should go.
1594 		 */
1595 		isr = NETISR_MAX;
1596 		break;
1597 	}
1598 
1599 	/*
1600 	 * Ask the isr to characterize the packet since we couldn't.
1601 	 * This is an attempt to optimally get us onto the correct protocol
1602 	 * thread.
1603 	 */
1604 	netisr_characterize(isr, &m, sizeof(struct ether_header));
1605 
1606 	*m0 = m;
1607 	return isr;
1608 }
1609 
1610 static void
1611 ether_demux_handler(netmsg_t nmsg)
1612 {
1613 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1614 	struct ifnet *ifp;
1615 	struct mbuf *m;
1616 
1617 	m = nmp->nm_packet;
1618 	M_ASSERTPKTHDR(m);
1619 	ifp = m->m_pkthdr.rcvif;
1620 
1621 	ether_demux_oncpu(ifp, m);
1622 }
1623 
1624 void
1625 ether_demux(struct mbuf *m)
1626 {
1627 	struct netmsg_packet *pmsg;
1628 	int isr;
1629 
1630 	isr = ether_characterize(&m);
1631 	if (m == NULL)
1632 		return;
1633 
1634 	KKASSERT(m->m_flags & M_HASH);
1635 	pmsg = &m->m_hdr.mh_netmsg;
1636 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1637 	    0, ether_demux_handler);
1638 	pmsg->nm_packet = m;
1639 	pmsg->base.lmsg.u.ms_result = isr;
1640 
1641 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1642 }
1643 
1644 u_char *
1645 kether_aton(const char *macstr, u_char *addr)
1646 {
1647         unsigned int o0, o1, o2, o3, o4, o5;
1648         int n;
1649 
1650         if (macstr == NULL || addr == NULL)
1651                 return NULL;
1652 
1653         n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1654             &o3, &o4, &o5);
1655         if (n != 6)
1656                 return NULL;
1657 
1658         addr[0] = o0;
1659         addr[1] = o1;
1660         addr[2] = o2;
1661         addr[3] = o3;
1662         addr[4] = o4;
1663         addr[5] = o5;
1664 
1665         return addr;
1666 }
1667 
1668 char *
1669 kether_ntoa(const u_char *addr, char *buf)
1670 {
1671         int len = ETHER_ADDRSTRLEN + 1;
1672         int n;
1673 
1674         n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1675             addr[1], addr[2], addr[3], addr[4], addr[5]);
1676 
1677         if (n < 17)
1678                 return NULL;
1679 
1680         return buf;
1681 }
1682 
1683 MODULE_VERSION(ether, 1);
1684