xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 2d0700913d3c55b6181d2b703dd69aae2179ce8c)
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 	struct ifaltq *ifq;
583 	int i;
584 
585 	ifp->if_type = IFT_ETHER;
586 	ifp->if_addrlen = ETHER_ADDR_LEN;
587 	ifp->if_hdrlen = ETHER_HDR_LEN;
588 	if_attach(ifp, serializer);
589 	ifq = &ifp->if_snd;
590 	for (i = 0; i < ifq->altq_subq_cnt; ++i) {
591 		struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i);
592 
593 		ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen *
594 		    (ETHER_MAX_LEN - ETHER_CRC_LEN);
595 	}
596 	ifp->if_mtu = ETHERMTU;
597 	if (ifp->if_tsolen <= 0) {
598 		if ((ether_tsolen_default / ETHERMTU) < 2) {
599 			kprintf("ether TSO maxlen %d -> %d\n",
600 			    ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
601 			ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
602 		}
603 		ifp->if_tsolen = ether_tsolen_default;
604 	}
605 	if (ifp->if_baudrate == 0)
606 		ifp->if_baudrate = 10000000;
607 	ifp->if_output = ether_output;
608 	ifp->if_input = ether_input;
609 	ifp->if_resolvemulti = ether_resolvemulti;
610 	ifp->if_broadcastaddr = etherbroadcastaddr;
611 	sdl = IF_LLSOCKADDR(ifp);
612 	sdl->sdl_type = IFT_ETHER;
613 	sdl->sdl_alen = ifp->if_addrlen;
614 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
615 	/*
616 	 * XXX Keep the current drivers happy.
617 	 * XXX Remove once all drivers have been cleaned up
618 	 */
619 	if (lla != IFP2AC(ifp)->ac_enaddr)
620 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
621 	bpfattach(ifp, dlt, hdrlen);
622 	if (ng_ether_attach_p != NULL)
623 		(*ng_ether_attach_p)(ifp);
624 
625 	if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
626 }
627 
628 /*
629  * Perform common duties while detaching an Ethernet interface
630  */
631 void
632 ether_ifdetach(struct ifnet *ifp)
633 {
634 	if_down(ifp);
635 
636 	if (ng_ether_detach_p != NULL)
637 		(*ng_ether_detach_p)(ifp);
638 	bpfdetach(ifp);
639 	if_detach(ifp);
640 }
641 
642 int
643 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
644 {
645 	struct ifaddr *ifa = (struct ifaddr *) data;
646 	struct ifreq *ifr = (struct ifreq *) data;
647 	int error = 0;
648 
649 #define IF_INIT(ifp) \
650 do { \
651 	if (((ifp)->if_flags & IFF_UP) == 0) { \
652 		(ifp)->if_flags |= IFF_UP; \
653 		(ifp)->if_init((ifp)->if_softc); \
654 	} \
655 } while (0)
656 
657 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
658 
659 	switch (command) {
660 	case SIOCSIFADDR:
661 		switch (ifa->ifa_addr->sa_family) {
662 #ifdef INET
663 		case AF_INET:
664 			IF_INIT(ifp);	/* before arpwhohas */
665 			arp_ifinit(ifp, ifa);
666 			break;
667 #endif
668 #ifdef IPX
669 		/*
670 		 * XXX - This code is probably wrong
671 		 */
672 		case AF_IPX:
673 			{
674 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
675 			struct arpcom *ac = IFP2AC(ifp);
676 
677 			if (ipx_nullhost(*ina))
678 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
679 			else
680 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
681 				      sizeof ac->ac_enaddr);
682 
683 			IF_INIT(ifp);	/* Set new address. */
684 			break;
685 			}
686 #endif
687 		default:
688 			IF_INIT(ifp);
689 			break;
690 		}
691 		break;
692 
693 	case SIOCGIFADDR:
694 		bcopy(IFP2AC(ifp)->ac_enaddr,
695 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
696 		      ETHER_ADDR_LEN);
697 		break;
698 
699 	case SIOCSIFMTU:
700 		/*
701 		 * Set the interface MTU.
702 		 */
703 		if (ifr->ifr_mtu > ETHERMTU) {
704 			error = EINVAL;
705 		} else {
706 			ifp->if_mtu = ifr->ifr_mtu;
707 		}
708 		break;
709 	default:
710 		error = EINVAL;
711 		break;
712 	}
713 	return (error);
714 
715 #undef IF_INIT
716 }
717 
718 int
719 ether_resolvemulti(
720 	struct ifnet *ifp,
721 	struct sockaddr **llsa,
722 	struct sockaddr *sa)
723 {
724 	struct sockaddr_dl *sdl;
725 #ifdef INET
726 	struct sockaddr_in *sin;
727 #endif
728 #ifdef INET6
729 	struct sockaddr_in6 *sin6;
730 #endif
731 	u_char *e_addr;
732 
733 	switch(sa->sa_family) {
734 	case AF_LINK:
735 		/*
736 		 * No mapping needed. Just check that it's a valid MC address.
737 		 */
738 		sdl = (struct sockaddr_dl *)sa;
739 		e_addr = LLADDR(sdl);
740 		if ((e_addr[0] & 1) != 1)
741 			return EADDRNOTAVAIL;
742 		*llsa = NULL;
743 		return 0;
744 
745 #ifdef INET
746 	case AF_INET:
747 		sin = (struct sockaddr_in *)sa;
748 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
749 			return EADDRNOTAVAIL;
750 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
751 		sdl->sdl_len = sizeof *sdl;
752 		sdl->sdl_family = AF_LINK;
753 		sdl->sdl_index = ifp->if_index;
754 		sdl->sdl_type = IFT_ETHER;
755 		sdl->sdl_alen = ETHER_ADDR_LEN;
756 		e_addr = LLADDR(sdl);
757 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
758 		*llsa = (struct sockaddr *)sdl;
759 		return 0;
760 #endif
761 #ifdef INET6
762 	case AF_INET6:
763 		sin6 = (struct sockaddr_in6 *)sa;
764 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
765 			/*
766 			 * An IP6 address of 0 means listen to all
767 			 * of the Ethernet multicast address used for IP6.
768 			 * (This is used for multicast routers.)
769 			 */
770 			ifp->if_flags |= IFF_ALLMULTI;
771 			*llsa = NULL;
772 			return 0;
773 		}
774 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
775 			return EADDRNOTAVAIL;
776 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
777 		sdl->sdl_len = sizeof *sdl;
778 		sdl->sdl_family = AF_LINK;
779 		sdl->sdl_index = ifp->if_index;
780 		sdl->sdl_type = IFT_ETHER;
781 		sdl->sdl_alen = ETHER_ADDR_LEN;
782 		e_addr = LLADDR(sdl);
783 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
784 		*llsa = (struct sockaddr *)sdl;
785 		return 0;
786 #endif
787 
788 	default:
789 		/*
790 		 * Well, the text isn't quite right, but it's the name
791 		 * that counts...
792 		 */
793 		return EAFNOSUPPORT;
794 	}
795 }
796 
797 #if 0
798 /*
799  * This is for reference.  We have a table-driven version
800  * of the little-endian crc32 generator, which is faster
801  * than the double-loop.
802  */
803 uint32_t
804 ether_crc32_le(const uint8_t *buf, size_t len)
805 {
806 	uint32_t c, crc, carry;
807 	size_t i, j;
808 
809 	crc = 0xffffffffU;	/* initial value */
810 
811 	for (i = 0; i < len; i++) {
812 		c = buf[i];
813 		for (j = 0; j < 8; j++) {
814 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
815 			crc >>= 1;
816 			c >>= 1;
817 			if (carry)
818 				crc = (crc ^ ETHER_CRC_POLY_LE);
819 		}
820 	}
821 
822 	return (crc);
823 }
824 #else
825 uint32_t
826 ether_crc32_le(const uint8_t *buf, size_t len)
827 {
828 	static const uint32_t crctab[] = {
829 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
830 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
831 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
832 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
833 	};
834 	uint32_t crc;
835 	size_t i;
836 
837 	crc = 0xffffffffU;	/* initial value */
838 
839 	for (i = 0; i < len; i++) {
840 		crc ^= buf[i];
841 		crc = (crc >> 4) ^ crctab[crc & 0xf];
842 		crc = (crc >> 4) ^ crctab[crc & 0xf];
843 	}
844 
845 	return (crc);
846 }
847 #endif
848 
849 uint32_t
850 ether_crc32_be(const uint8_t *buf, size_t len)
851 {
852 	uint32_t c, crc, carry;
853 	size_t i, j;
854 
855 	crc = 0xffffffffU;	/* initial value */
856 
857 	for (i = 0; i < len; i++) {
858 		c = buf[i];
859 		for (j = 0; j < 8; j++) {
860 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
861 			crc <<= 1;
862 			c >>= 1;
863 			if (carry)
864 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
865 		}
866 	}
867 
868 	return (crc);
869 }
870 
871 /*
872  * find the size of ethernet header, and call classifier
873  */
874 void
875 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
876 		   struct altq_pktattr *pktattr)
877 {
878 	struct ether_header *eh;
879 	uint16_t ether_type;
880 	int hlen, af, hdrsize;
881 
882 	hlen = sizeof(struct ether_header);
883 	eh = mtod(m, struct ether_header *);
884 
885 	ether_type = ntohs(eh->ether_type);
886 	if (ether_type < ETHERMTU) {
887 		/* ick! LLC/SNAP */
888 		struct llc *llc = (struct llc *)(eh + 1);
889 		hlen += 8;
890 
891 		if (m->m_len < hlen ||
892 		    llc->llc_dsap != LLC_SNAP_LSAP ||
893 		    llc->llc_ssap != LLC_SNAP_LSAP ||
894 		    llc->llc_control != LLC_UI)
895 			goto bad;  /* not snap! */
896 
897 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
898 	}
899 
900 	if (ether_type == ETHERTYPE_IP) {
901 		af = AF_INET;
902 		hdrsize = 20;  /* sizeof(struct ip) */
903 #ifdef INET6
904 	} else if (ether_type == ETHERTYPE_IPV6) {
905 		af = AF_INET6;
906 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
907 #endif
908 	} else
909 		goto bad;
910 
911 	while (m->m_len <= hlen) {
912 		hlen -= m->m_len;
913 		m = m->m_next;
914 	}
915 	if (m->m_len < hlen + hdrsize) {
916 		/*
917 		 * ip header is not in a single mbuf.  this should not
918 		 * happen in the current code.
919 		 * (todo: use m_pulldown in the future)
920 		 */
921 		goto bad;
922 	}
923 	m->m_data += hlen;
924 	m->m_len -= hlen;
925 	ifq_classify(ifq, m, af, pktattr);
926 	m->m_data -= hlen;
927 	m->m_len += hlen;
928 
929 	return;
930 
931 bad:
932 	pktattr->pattr_class = NULL;
933 	pktattr->pattr_hdr = NULL;
934 	pktattr->pattr_af = AF_UNSPEC;
935 }
936 
937 static void
938 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
939 		     const struct ether_header *save_eh)
940 {
941 	struct mbuf *m = *m0;
942 
943 	ether_restore_hdr++;
944 
945 	/*
946 	 * Prepend the header, optimize for the common case of
947 	 * eh pointing into the mbuf.
948 	 */
949 	if ((const void *)(eh + 1) == (void *)m->m_data) {
950 		m->m_data -= ETHER_HDR_LEN;
951 		m->m_len += ETHER_HDR_LEN;
952 		m->m_pkthdr.len += ETHER_HDR_LEN;
953 	} else {
954 		ether_prepend_hdr++;
955 
956 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
957 		if (m != NULL) {
958 			bcopy(save_eh, mtod(m, struct ether_header *),
959 			      ETHER_HDR_LEN);
960 		}
961 	}
962 	*m0 = m;
963 }
964 
965 /*
966  * Upper layer processing for a received Ethernet packet.
967  */
968 void
969 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
970 {
971 	struct ether_header *eh;
972 	int isr, discard = 0;
973 	u_short ether_type;
974 	struct ip_fw *rule = NULL;
975 
976 	M_ASSERTPKTHDR(m);
977 	KASSERT(m->m_len >= ETHER_HDR_LEN,
978 		("ether header is not contiguous!"));
979 
980 	eh = mtod(m, struct ether_header *);
981 
982 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
983 		struct m_tag *mtag;
984 
985 		/* Extract info from dummynet tag */
986 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
987 		KKASSERT(mtag != NULL);
988 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
989 		KKASSERT(rule != NULL);
990 
991 		m_tag_delete(m, mtag);
992 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
993 
994 		/* packet is passing the second time */
995 		goto post_stats;
996 	}
997 
998 	/*
999 	 * We got a packet which was unicast to a different Ethernet
1000 	 * address.  If the driver is working properly, then this
1001 	 * situation can only happen when the interface is in
1002 	 * promiscuous mode.  We defer the packet discarding until the
1003 	 * vlan processing is done, so that vlan/bridge or vlan/netgraph
1004 	 * could work.
1005 	 */
1006 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1007 	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
1008 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1009 		if (ether_debug & 1) {
1010 			kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
1011 				"%02x:%02x:%02x:%02x:%02x:%02x "
1012 				"%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
1013 				eh->ether_dhost[0],
1014 				eh->ether_dhost[1],
1015 				eh->ether_dhost[2],
1016 				eh->ether_dhost[3],
1017 				eh->ether_dhost[4],
1018 				eh->ether_dhost[5],
1019 				eh->ether_shost[0],
1020 				eh->ether_shost[1],
1021 				eh->ether_shost[2],
1022 				eh->ether_shost[3],
1023 				eh->ether_shost[4],
1024 				eh->ether_shost[5],
1025 				eh->ether_type,
1026 				((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1027 				((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1028 				((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1029 				((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1030 				((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1031 				((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1032 			);
1033 		}
1034 		if ((ether_debug & 2) == 0)
1035 			discard = 1;
1036 	}
1037 
1038 post_stats:
1039 	if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
1040 		struct ether_header save_eh = *eh;
1041 
1042 		/* XXX old crufty stuff, needs to be removed */
1043 		m_adj(m, sizeof(struct ether_header));
1044 
1045 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1046 			m_freem(m);
1047 			return;
1048 		}
1049 
1050 		ether_restore_header(&m, eh, &save_eh);
1051 		if (m == NULL)
1052 			return;
1053 		eh = mtod(m, struct ether_header *);
1054 	}
1055 
1056 	ether_type = ntohs(eh->ether_type);
1057 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1058 
1059 	if (m->m_flags & M_VLANTAG) {
1060 		void (*vlan_input_func)(struct mbuf *);
1061 
1062 		vlan_input_func = vlan_input_p;
1063 		if (vlan_input_func != NULL) {
1064 			vlan_input_func(m);
1065 		} else {
1066 			IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1067 			m_freem(m);
1068 		}
1069 		return;
1070 	}
1071 
1072 	/*
1073 	 * If we have been asked to discard this packet
1074 	 * (e.g. not for us), drop it before entering
1075 	 * the upper layer.
1076 	 */
1077 	if (discard) {
1078 		m_freem(m);
1079 		return;
1080 	}
1081 
1082 	/*
1083 	 * Clear protocol specific flags,
1084 	 * before entering the upper layer.
1085 	 */
1086 	m->m_flags &= ~M_ETHER_FLAGS;
1087 
1088 	/* Strip ethernet header. */
1089 	m_adj(m, sizeof(struct ether_header));
1090 
1091 	switch (ether_type) {
1092 #ifdef INET
1093 	case ETHERTYPE_IP:
1094 		if ((m->m_flags & M_LENCHECKED) == 0) {
1095 			if (!ip_lengthcheck(&m, 0))
1096 				return;
1097 		}
1098 		if (ipflow_fastforward(m))
1099 			return;
1100 		isr = NETISR_IP;
1101 		break;
1102 
1103 	case ETHERTYPE_ARP:
1104 		if (ifp->if_flags & IFF_NOARP) {
1105 			/* Discard packet if ARP is disabled on interface */
1106 			m_freem(m);
1107 			return;
1108 		}
1109 		isr = NETISR_ARP;
1110 		break;
1111 #endif
1112 
1113 #ifdef INET6
1114 	case ETHERTYPE_IPV6:
1115 		isr = NETISR_IPV6;
1116 		break;
1117 #endif
1118 
1119 #ifdef IPX
1120 	case ETHERTYPE_IPX:
1121 		if (ef_inputp) {
1122 			/*
1123 			 * Hold BGL and recheck ef_inputp
1124 			 */
1125 			get_mplock();
1126 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1127 				rel_mplock();
1128 				return;
1129 			}
1130 			rel_mplock();
1131 		}
1132 		isr = NETISR_IPX;
1133 		break;
1134 #endif
1135 
1136 #ifdef MPLS
1137 	case ETHERTYPE_MPLS:
1138 	case ETHERTYPE_MPLS_MCAST:
1139 		/* Should have been set by ether_input_pkt(). */
1140 		KKASSERT(m->m_flags & M_MPLSLABELED);
1141 		isr = NETISR_MPLS;
1142 		break;
1143 #endif
1144 
1145 	default:
1146 		/*
1147 		 * The accurate msgport is not determined before
1148 		 * we reach here, so recharacterize packet.
1149 		 */
1150 		m->m_flags &= ~M_HASH;
1151 #ifdef IPX
1152 		if (ef_inputp) {
1153 			/*
1154 			 * Hold BGL and recheck ef_inputp
1155 			 */
1156 			get_mplock();
1157 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1158 				rel_mplock();
1159 				return;
1160 			}
1161 			rel_mplock();
1162 		}
1163 #endif
1164 		if (ng_ether_input_orphan_p != NULL) {
1165 			/*
1166 			 * Put back the ethernet header so netgraph has a
1167 			 * consistent view of inbound packets.
1168 			 */
1169 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1170 			if (m == NULL) {
1171 				/*
1172 				 * M_PREPEND frees the mbuf in case of failure.
1173 				 */
1174 				return;
1175 			}
1176 			/*
1177 			 * Hold BGL and recheck ng_ether_input_orphan_p
1178 			 */
1179 			get_mplock();
1180 			if (ng_ether_input_orphan_p != NULL) {
1181 				ng_ether_input_orphan_p(ifp, m);
1182 				rel_mplock();
1183 				return;
1184 			}
1185 			rel_mplock();
1186 		}
1187 		m_freem(m);
1188 		return;
1189 	}
1190 
1191 	if (m->m_flags & M_HASH) {
1192 		if (&curthread->td_msgport ==
1193 		    netisr_hashport(m->m_pkthdr.hash)) {
1194 			netisr_handle(isr, m);
1195 			return;
1196 		} else {
1197 			/*
1198 			 * XXX Something is wrong,
1199 			 * we probably should panic here!
1200 			 */
1201 			m->m_flags &= ~M_HASH;
1202 			atomic_add_long(&ether_input_wronghash, 1);
1203 		}
1204 	}
1205 #ifdef RSS_DEBUG
1206 	atomic_add_long(&ether_input_requeue, 1);
1207 #endif
1208 	netisr_queue(isr, m);
1209 }
1210 
1211 /*
1212  * First we perform any link layer operations, then continue to the
1213  * upper layers with ether_demux_oncpu().
1214  */
1215 static void
1216 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1217 {
1218 #ifdef CARP
1219 	void *carp;
1220 #endif
1221 
1222 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1223 		/*
1224 		 * Receiving interface's flags are changed, when this
1225 		 * packet is waiting for processing; discard it.
1226 		 */
1227 		m_freem(m);
1228 		return;
1229 	}
1230 
1231 	/*
1232 	 * Tap the packet off here for a bridge.  bridge_input()
1233 	 * will return NULL if it has consumed the packet, otherwise
1234 	 * it gets processed as normal.  Note that bridge_input()
1235 	 * will always return the original packet if we need to
1236 	 * process it locally.
1237 	 */
1238 	if (ifp->if_bridge) {
1239 		KASSERT(bridge_input_p != NULL,
1240 			("%s: if_bridge not loaded!", __func__));
1241 
1242 		if(m->m_flags & M_ETHER_BRIDGED) {
1243 			m->m_flags &= ~M_ETHER_BRIDGED;
1244 		} else {
1245 			m = bridge_input_p(ifp, m);
1246 			if (m == NULL)
1247 				return;
1248 
1249 			KASSERT(ifp == m->m_pkthdr.rcvif,
1250 				("bridge_input_p changed rcvif"));
1251 		}
1252 	}
1253 
1254 #ifdef CARP
1255 	carp = ifp->if_carp;
1256 	if (carp) {
1257 		m = carp_input(carp, m);
1258 		if (m == NULL)
1259 			return;
1260 		KASSERT(ifp == m->m_pkthdr.rcvif,
1261 		    ("carp_input changed rcvif"));
1262 	}
1263 #endif
1264 
1265 	/* Handle ng_ether(4) processing, if any */
1266 	if (ng_ether_input_p != NULL) {
1267 		/*
1268 		 * Hold BGL and recheck ng_ether_input_p
1269 		 */
1270 		get_mplock();
1271 		if (ng_ether_input_p != NULL)
1272 			ng_ether_input_p(ifp, &m);
1273 		rel_mplock();
1274 
1275 		if (m == NULL)
1276 			return;
1277 	}
1278 
1279 	/* Continue with upper layer processing */
1280 	ether_demux_oncpu(ifp, m);
1281 }
1282 
1283 /*
1284  * Perform certain functions of ether_input_pkt():
1285  * - Test IFF_UP
1286  * - Update statistics
1287  * - Run bpf(4) tap if requested
1288  * Then pass the packet to ether_input_oncpu().
1289  *
1290  * This function should be used by pseudo interface (e.g. vlan(4)),
1291  * when it tries to claim that the packet is received by it.
1292  *
1293  * REINPUT_KEEPRCVIF
1294  * REINPUT_RUNBPF
1295  */
1296 void
1297 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1298 {
1299 	/* Discard packet if interface is not up */
1300 	if (!(ifp->if_flags & IFF_UP)) {
1301 		m_freem(m);
1302 		return;
1303 	}
1304 
1305 	/*
1306 	 * Change receiving interface.  The bridge will often pass a flag to
1307 	 * ask that this not be done so ARPs get applied to the correct
1308 	 * side.
1309 	 */
1310 	if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1311 	    m->m_pkthdr.rcvif == NULL) {
1312 		m->m_pkthdr.rcvif = ifp;
1313 	}
1314 
1315 	/* Update statistics */
1316 	IFNET_STAT_INC(ifp, ipackets, 1);
1317 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1318 	if (m->m_flags & (M_MCAST | M_BCAST))
1319 		IFNET_STAT_INC(ifp, imcasts, 1);
1320 
1321 	if (reinput_flags & REINPUT_RUNBPF)
1322 		BPF_MTAP(ifp, m);
1323 
1324 	ether_input_oncpu(ifp, m);
1325 }
1326 
1327 static __inline boolean_t
1328 ether_vlancheck(struct mbuf **m0)
1329 {
1330 	struct mbuf *m = *m0;
1331 	struct ether_header *eh;
1332 	uint16_t ether_type;
1333 
1334 	eh = mtod(m, struct ether_header *);
1335 	ether_type = ntohs(eh->ether_type);
1336 
1337 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1338 		/*
1339 		 * Extract vlan tag if hardware does not do it for us
1340 		 */
1341 		vlan_ether_decap(&m);
1342 		if (m == NULL)
1343 			goto failed;
1344 
1345 		eh = mtod(m, struct ether_header *);
1346 		ether_type = ntohs(eh->ether_type);
1347 	}
1348 
1349 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1350 		/*
1351 		 * To prevent possible dangerous recursion,
1352 		 * we don't do vlan-in-vlan
1353 		 */
1354 		IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1355 		goto failed;
1356 	}
1357 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1358 
1359 	m->m_flags |= M_ETHER_VLANCHECKED;
1360 	*m0 = m;
1361 	return TRUE;
1362 failed:
1363 	if (m != NULL)
1364 		m_freem(m);
1365 	*m0 = NULL;
1366 	return FALSE;
1367 }
1368 
1369 static void
1370 ether_input_handler(netmsg_t nmsg)
1371 {
1372 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1373 	struct ether_header *eh;
1374 	struct ifnet *ifp;
1375 	struct mbuf *m;
1376 
1377 	m = nmp->nm_packet;
1378 	M_ASSERTPKTHDR(m);
1379 
1380 	if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1381 		if (!ether_vlancheck(&m)) {
1382 			KKASSERT(m == NULL);
1383 			return;
1384 		}
1385 	}
1386 	if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1387 	    __predict_false(ether_input_ckhash)) {
1388 		int isr;
1389 
1390 		/*
1391 		 * Need to verify the hash supplied by the hardware
1392 		 * which could be wrong.
1393 		 */
1394 		m->m_flags &= ~(M_HASH | M_CKHASH);
1395 		isr = ether_characterize(&m);
1396 		if (m == NULL)
1397 			return;
1398 		KKASSERT(m->m_flags & M_HASH);
1399 
1400 		if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1401 			/*
1402 			 * Wrong hardware supplied hash; redispatch
1403 			 */
1404 			ether_dispatch(isr, m);
1405 			if (__predict_false(ether_input_ckhash))
1406 				atomic_add_long(&ether_input_wronghwhash, 1);
1407 			return;
1408 		}
1409 	}
1410 	ifp = m->m_pkthdr.rcvif;
1411 
1412 	eh = mtod(m, struct ether_header *);
1413 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1414 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1415 			 ifp->if_addrlen) == 0)
1416 			m->m_flags |= M_BCAST;
1417 		else
1418 			m->m_flags |= M_MCAST;
1419 		IFNET_STAT_INC(ifp, imcasts, 1);
1420 	}
1421 
1422 	ether_input_oncpu(ifp, m);
1423 }
1424 
1425 /*
1426  * Send the packet to the target netisr msgport
1427  *
1428  * At this point the packet must be characterized (M_HASH set),
1429  * so we know which netisr to send it to.
1430  */
1431 static void
1432 ether_dispatch(int isr, struct mbuf *m)
1433 {
1434 	struct netmsg_packet *pmsg;
1435 
1436 	KKASSERT(m->m_flags & M_HASH);
1437 	pmsg = &m->m_hdr.mh_netmsg;
1438 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1439 		    0, ether_input_handler);
1440 	pmsg->nm_packet = m;
1441 	pmsg->base.lmsg.u.ms_result = isr;
1442 
1443 	logether(disp_beg, NULL);
1444 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1445 	logether(disp_end, NULL);
1446 }
1447 
1448 /*
1449  * Process a received Ethernet packet.
1450  *
1451  * The ethernet header is assumed to be in the mbuf so the caller
1452  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1453  * bytes in the first mbuf.
1454  */
1455 void
1456 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi)
1457 {
1458 	int isr;
1459 
1460 	M_ASSERTPKTHDR(m);
1461 
1462 	/* Discard packet if interface is not up */
1463 	if (!(ifp->if_flags & IFF_UP)) {
1464 		m_freem(m);
1465 		return;
1466 	}
1467 
1468 	if (m->m_len < sizeof(struct ether_header)) {
1469 		/* XXX error in the caller. */
1470 		m_freem(m);
1471 		return;
1472 	}
1473 
1474 	m->m_pkthdr.rcvif = ifp;
1475 
1476 	logether(pkt_beg, ifp);
1477 
1478 	ETHER_BPF_MTAP(ifp, m);
1479 
1480 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1481 
1482 	if (ifp->if_flags & IFF_MONITOR) {
1483 		struct ether_header *eh;
1484 
1485 		eh = mtod(m, struct ether_header *);
1486 		if (ETHER_IS_MULTICAST(eh->ether_dhost))
1487 			IFNET_STAT_INC(ifp, imcasts, 1);
1488 
1489 		/*
1490 		 * Interface marked for monitoring; discard packet.
1491 		 */
1492 		m_freem(m);
1493 
1494 		logether(pkt_end, ifp);
1495 		return;
1496 	}
1497 
1498 	/*
1499 	 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1500 	 * we can dispatch it immediately with trivial checks.
1501 	 */
1502 	if (pi != NULL && (m->m_flags & M_HASH)) {
1503 #ifdef RSS_DEBUG
1504 		atomic_add_long(&ether_pktinfo_try, 1);
1505 #endif
1506 		netisr_hashcheck(pi->pi_netisr, m, pi);
1507 		if (m->m_flags & M_HASH) {
1508 			ether_dispatch(pi->pi_netisr, m);
1509 #ifdef RSS_DEBUG
1510 			atomic_add_long(&ether_pktinfo_hit, 1);
1511 #endif
1512 			logether(pkt_end, ifp);
1513 			return;
1514 		}
1515 	}
1516 #ifdef RSS_DEBUG
1517 	else if (ifp->if_capenable & IFCAP_RSS) {
1518 		if (pi == NULL)
1519 			atomic_add_long(&ether_rss_nopi, 1);
1520 		else
1521 			atomic_add_long(&ether_rss_nohash, 1);
1522 	}
1523 #endif
1524 
1525 	/*
1526 	 * Packet hash will be recalculated by software, so clear
1527 	 * the M_HASH and M_CKHASH flag set by the driver; the hash
1528 	 * value calculated by the hardware may not be exactly what
1529 	 * we want.
1530 	 */
1531 	m->m_flags &= ~(M_HASH | M_CKHASH);
1532 
1533 	if (!ether_vlancheck(&m)) {
1534 		KKASSERT(m == NULL);
1535 		logether(pkt_end, ifp);
1536 		return;
1537 	}
1538 
1539 	isr = ether_characterize(&m);
1540 	if (m == NULL) {
1541 		logether(pkt_end, ifp);
1542 		return;
1543 	}
1544 
1545 	/*
1546 	 * Finally dispatch it
1547 	 */
1548 	ether_dispatch(isr, m);
1549 
1550 	logether(pkt_end, ifp);
1551 }
1552 
1553 static int
1554 ether_characterize(struct mbuf **m0)
1555 {
1556 	struct mbuf *m = *m0;
1557 	struct ether_header *eh;
1558 	uint16_t ether_type;
1559 	int isr;
1560 
1561 	eh = mtod(m, struct ether_header *);
1562 	ether_type = ntohs(eh->ether_type);
1563 
1564 	/*
1565 	 * Map ether type to netisr id.
1566 	 */
1567 	switch (ether_type) {
1568 #ifdef INET
1569 	case ETHERTYPE_IP:
1570 		isr = NETISR_IP;
1571 		break;
1572 
1573 	case ETHERTYPE_ARP:
1574 		isr = NETISR_ARP;
1575 		break;
1576 #endif
1577 
1578 #ifdef INET6
1579 	case ETHERTYPE_IPV6:
1580 		isr = NETISR_IPV6;
1581 		break;
1582 #endif
1583 
1584 #ifdef IPX
1585 	case ETHERTYPE_IPX:
1586 		isr = NETISR_IPX;
1587 		break;
1588 #endif
1589 
1590 #ifdef MPLS
1591 	case ETHERTYPE_MPLS:
1592 	case ETHERTYPE_MPLS_MCAST:
1593 		m->m_flags |= M_MPLSLABELED;
1594 		isr = NETISR_MPLS;
1595 		break;
1596 #endif
1597 
1598 	default:
1599 		/*
1600 		 * NETISR_MAX is an invalid value; it is chosen to let
1601 		 * netisr_characterize() know that we have no clear
1602 		 * idea where this packet should go.
1603 		 */
1604 		isr = NETISR_MAX;
1605 		break;
1606 	}
1607 
1608 	/*
1609 	 * Ask the isr to characterize the packet since we couldn't.
1610 	 * This is an attempt to optimally get us onto the correct protocol
1611 	 * thread.
1612 	 */
1613 	netisr_characterize(isr, &m, sizeof(struct ether_header));
1614 
1615 	*m0 = m;
1616 	return isr;
1617 }
1618 
1619 static void
1620 ether_demux_handler(netmsg_t nmsg)
1621 {
1622 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1623 	struct ifnet *ifp;
1624 	struct mbuf *m;
1625 
1626 	m = nmp->nm_packet;
1627 	M_ASSERTPKTHDR(m);
1628 	ifp = m->m_pkthdr.rcvif;
1629 
1630 	ether_demux_oncpu(ifp, m);
1631 }
1632 
1633 void
1634 ether_demux(struct mbuf *m)
1635 {
1636 	struct netmsg_packet *pmsg;
1637 	int isr;
1638 
1639 	isr = ether_characterize(&m);
1640 	if (m == NULL)
1641 		return;
1642 
1643 	KKASSERT(m->m_flags & M_HASH);
1644 	pmsg = &m->m_hdr.mh_netmsg;
1645 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1646 	    0, ether_demux_handler);
1647 	pmsg->nm_packet = m;
1648 	pmsg->base.lmsg.u.ms_result = isr;
1649 
1650 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1651 }
1652 
1653 u_char *
1654 kether_aton(const char *macstr, u_char *addr)
1655 {
1656         unsigned int o0, o1, o2, o3, o4, o5;
1657         int n;
1658 
1659         if (macstr == NULL || addr == NULL)
1660                 return NULL;
1661 
1662         n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1663             &o3, &o4, &o5);
1664         if (n != 6)
1665                 return NULL;
1666 
1667         addr[0] = o0;
1668         addr[1] = o1;
1669         addr[2] = o2;
1670         addr[3] = o3;
1671         addr[4] = o4;
1672         addr[5] = o5;
1673 
1674         return addr;
1675 }
1676 
1677 char *
1678 kether_ntoa(const u_char *addr, char *buf)
1679 {
1680         int len = ETHER_ADDRSTRLEN + 1;
1681         int n;
1682 
1683         n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1684             addr[1], addr[2], addr[3], addr[4], addr[5]);
1685 
1686         if (n < 17)
1687                 return NULL;
1688 
1689         return buf;
1690 }
1691 
1692 MODULE_VERSION(ether, 1);
1693