xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 767a43118674e3039d1f3dd7d3f2662a2eccf599)
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. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
34  * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
35  */
36 
37 #include "opt_inet.h"
38 #include "opt_inet6.h"
39 #include "opt_ipx.h"
40 #include "opt_mpls.h"
41 #include "opt_netgraph.h"
42 #include "opt_carp.h"
43 #include "opt_rss.h"
44 
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/globaldata.h>
48 #include <sys/kernel.h>
49 #include <sys/ktr.h>
50 #include <sys/lock.h>
51 #include <sys/malloc.h>
52 #include <sys/mbuf.h>
53 #include <sys/msgport.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/thread.h>
58 
59 #include <sys/thread2.h>
60 #include <sys/mplock2.h>
61 
62 #include <net/if.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/if_llc.h>
66 #include <net/if_dl.h>
67 #include <net/if_types.h>
68 #include <net/ifq_var.h>
69 #include <net/bpf.h>
70 #include <net/ethernet.h>
71 #include <net/vlan/if_vlan_ether.h>
72 #include <net/vlan/if_vlan_var.h>
73 #include <net/netmsg2.h>
74 
75 #if defined(INET) || defined(INET6)
76 #include <netinet/in.h>
77 #include <netinet/ip_var.h>
78 #include <netinet/tcp_var.h>
79 #include <netinet/if_ether.h>
80 #include <netinet/ip_flow.h>
81 #include <net/ipfw/ip_fw.h>
82 #include <net/dummynet/ip_dummynet.h>
83 #endif
84 #ifdef INET6
85 #include <netinet6/nd6.h>
86 #endif
87 
88 #ifdef CARP
89 #include <netinet/ip_carp.h>
90 #endif
91 
92 #ifdef IPX
93 #include <netproto/ipx/ipx.h>
94 #include <netproto/ipx/ipx_if.h>
95 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
96 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
97 		  short *tp, int *hlen);
98 #endif
99 
100 #ifdef MPLS
101 #include <netproto/mpls/mpls.h>
102 #endif
103 
104 /* netgraph node hooks for ng_ether(4) */
105 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
106 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
107 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
108 void	(*ng_ether_attach_p)(struct ifnet *ifp);
109 void	(*ng_ether_detach_p)(struct ifnet *ifp);
110 
111 void	(*vlan_input_p)(struct mbuf *);
112 
113 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
114 			struct rtentry *);
115 static void ether_restore_header(struct mbuf **, const struct ether_header *,
116 				 const struct ether_header *);
117 static int ether_characterize(struct mbuf **);
118 static void ether_dispatch(int, struct mbuf *);
119 
120 /*
121  * if_bridge support
122  */
123 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
124 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
125 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
126 struct ifnet *(*bridge_interface_p)(void *if_bridge);
127 
128 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
129 			      struct sockaddr *);
130 
131 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
132 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
133 };
134 
135 #define gotoerr(e) do { error = (e); goto bad; } while (0)
136 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
137 
138 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
139 				struct ip_fw **rule,
140 				const struct ether_header *eh);
141 
142 static int ether_ipfw;
143 static u_long ether_restore_hdr;
144 static u_long ether_prepend_hdr;
145 static u_long ether_input_wronghash;
146 static int ether_debug;
147 
148 #ifdef RSS_DEBUG
149 static u_long ether_pktinfo_try;
150 static u_long ether_pktinfo_hit;
151 static u_long ether_rss_nopi;
152 static u_long ether_rss_nohash;
153 static u_long ether_input_requeue;
154 static u_long ether_input_wronghwhash;
155 static int ether_input_ckhash;
156 #endif
157 
158 SYSCTL_DECL(_net_link);
159 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
160 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
161     &ether_debug, 0, "Ether debug");
162 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
163     &ether_ipfw, 0, "Pass ether pkts through firewall");
164 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
165     &ether_restore_hdr, 0, "# of ether header restoration");
166 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
167     &ether_prepend_hdr, 0,
168     "# of ether header restoration which prepends mbuf");
169 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
170     &ether_input_wronghash, 0, "# of input packets with wrong hash");
171 #ifdef RSS_DEBUG
172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
173     &ether_rss_nopi, 0, "# of packets do not have pktinfo");
174 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
175     &ether_rss_nohash, 0, "# of packets do not have hash");
176 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
177     &ether_pktinfo_try, 0,
178     "# of tries to find packets' msgport using pktinfo");
179 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
180     &ether_pktinfo_hit, 0,
181     "# of packets whose msgport are found using pktinfo");
182 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
183     &ether_input_requeue, 0, "# of input packets gets requeued");
184 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
185     &ether_input_wronghwhash, 0, "# of input packets with wrong hw hash");
186 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
187     &ether_input_ckhash, 0, "always check hash");
188 #endif
189 
190 #define ETHER_KTR_STR		"ifp=%p"
191 #define ETHER_KTR_ARGS	struct ifnet *ifp
192 #ifndef KTR_ETHERNET
193 #define KTR_ETHERNET		KTR_ALL
194 #endif
195 KTR_INFO_MASTER(ether);
196 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
197 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
198 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
199 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
200 #define logether(name, arg)	KTR_LOG(ether_ ## name, arg)
201 
202 /*
203  * Ethernet output routine.
204  * Encapsulate a packet of type family for the local net.
205  * Use trailer local net encapsulation if enough data in first
206  * packet leaves a multiple of 512 bytes of data in remainder.
207  * Assumes that ifp is actually pointer to arpcom structure.
208  */
209 static int
210 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
211 	     struct rtentry *rt)
212 {
213 	struct ether_header *eh, *deh;
214 	u_char *edst;
215 	int loop_copy = 0;
216 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
217 	struct arpcom *ac = IFP2AC(ifp);
218 	int error;
219 
220 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
221 
222 	if (ifp->if_flags & IFF_MONITOR)
223 		gotoerr(ENETDOWN);
224 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
225 		gotoerr(ENETDOWN);
226 
227 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
228 	if (m == NULL)
229 		return (ENOBUFS);
230 	m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
231 	eh = mtod(m, struct ether_header *);
232 	edst = eh->ether_dhost;
233 
234 	/*
235 	 * Fill in the destination ethernet address and frame type.
236 	 */
237 	switch (dst->sa_family) {
238 #ifdef INET
239 	case AF_INET:
240 		if (!arpresolve(ifp, rt, m, dst, edst))
241 			return (0);	/* if not yet resolved */
242 #ifdef MPLS
243 		if (m->m_flags & M_MPLSLABELED)
244 			eh->ether_type = htons(ETHERTYPE_MPLS);
245 		else
246 #endif
247 			eh->ether_type = htons(ETHERTYPE_IP);
248 		break;
249 #endif
250 #ifdef INET6
251 	case AF_INET6:
252 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
253 			return (0);		/* Something bad happenned. */
254 		eh->ether_type = htons(ETHERTYPE_IPV6);
255 		break;
256 #endif
257 #ifdef IPX
258 	case AF_IPX:
259 		if (ef_outputp != NULL) {
260 			/*
261 			 * Hold BGL and recheck ef_outputp
262 			 */
263 			get_mplock();
264 			if (ef_outputp != NULL) {
265 				error = ef_outputp(ifp, &m, dst,
266 						   &eh->ether_type, &hlen);
267 				rel_mplock();
268 				if (error)
269 					goto bad;
270 				else
271 					break;
272 			}
273 			rel_mplock();
274 		}
275 		eh->ether_type = htons(ETHERTYPE_IPX);
276 		bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
277 		      edst, ETHER_ADDR_LEN);
278 		break;
279 #endif
280 	case pseudo_AF_HDRCMPLT:
281 	case AF_UNSPEC:
282 		loop_copy = -1; /* if this is for us, don't do it */
283 		deh = (struct ether_header *)dst->sa_data;
284 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
285 		eh->ether_type = deh->ether_type;
286 		break;
287 
288 	default:
289 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
290 		gotoerr(EAFNOSUPPORT);
291 	}
292 
293 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
294 		memcpy(eh->ether_shost,
295 		       ((struct ether_header *)dst->sa_data)->ether_shost,
296 		       ETHER_ADDR_LEN);
297 	else
298 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
299 
300 	/*
301 	 * Bridges require special output handling.
302 	 */
303 	if (ifp->if_bridge) {
304 		KASSERT(bridge_output_p != NULL,
305 			("%s: if_bridge not loaded!", __func__));
306 		return bridge_output_p(ifp, m);
307 	}
308 
309 	/*
310 	 * If a simplex interface, and the packet is being sent to our
311 	 * Ethernet address or a broadcast address, loopback a copy.
312 	 * XXX To make a simplex device behave exactly like a duplex
313 	 * device, we should copy in the case of sending to our own
314 	 * ethernet address (thus letting the original actually appear
315 	 * on the wire). However, we don't do that here for security
316 	 * reasons and compatibility with the original behavior.
317 	 */
318 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
319 		int csum_flags = 0;
320 
321 		if (m->m_pkthdr.csum_flags & CSUM_IP)
322 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
323 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
324 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
325 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
326 			struct mbuf *n;
327 
328 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
329 				n->m_pkthdr.csum_flags |= csum_flags;
330 				if (csum_flags & CSUM_DATA_VALID)
331 					n->m_pkthdr.csum_data = 0xffff;
332 				if_simloop(ifp, n, dst->sa_family, hlen);
333 			} else
334 				ifp->if_iqdrops++;
335 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
336 				ETHER_ADDR_LEN) == 0) {
337 			m->m_pkthdr.csum_flags |= csum_flags;
338 			if (csum_flags & CSUM_DATA_VALID)
339 				m->m_pkthdr.csum_data = 0xffff;
340 			if_simloop(ifp, m, dst->sa_family, hlen);
341 			return (0);	/* XXX */
342 		}
343 	}
344 
345 #ifdef CARP
346 	if (ifp->if_type == IFT_CARP) {
347 		ifp = carp_parent(ifp);
348 		if (ifp == NULL)
349 			gotoerr(ENETUNREACH);
350 
351 		ac = IFP2AC(ifp);
352 
353 		/*
354 		 * Check precondition again
355 		 */
356 		ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
357 
358 		if (ifp->if_flags & IFF_MONITOR)
359 			gotoerr(ENETDOWN);
360 		if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
361 		    (IFF_UP | IFF_RUNNING))
362 			gotoerr(ENETDOWN);
363 	}
364 #endif
365 
366 	/* Handle ng_ether(4) processing, if any */
367 	if (ng_ether_output_p != NULL) {
368 		/*
369 		 * Hold BGL and recheck ng_ether_output_p
370 		 */
371 		get_mplock();
372 		if (ng_ether_output_p != NULL) {
373 			if ((error = ng_ether_output_p(ifp, &m)) != 0) {
374 				rel_mplock();
375 				goto bad;
376 			}
377 			if (m == NULL) {
378 				rel_mplock();
379 				return (0);
380 			}
381 		}
382 		rel_mplock();
383 	}
384 
385 	/* Continue with link-layer output */
386 	return ether_output_frame(ifp, m);
387 
388 bad:
389 	m_freem(m);
390 	return (error);
391 }
392 
393 /*
394  * Returns the bridge interface an ifp is associated
395  * with.
396  *
397  * Only call if ifp->if_bridge != NULL.
398  */
399 struct ifnet *
400 ether_bridge_interface(struct ifnet *ifp)
401 {
402 	if (bridge_interface_p)
403 		return(bridge_interface_p(ifp->if_bridge));
404 	return (ifp);
405 }
406 
407 /*
408  * Ethernet link layer output routine to send a raw frame to the device.
409  *
410  * This assumes that the 14 byte Ethernet header is present and contiguous
411  * in the first mbuf.
412  */
413 int
414 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
415 {
416 	struct ip_fw *rule = NULL;
417 	int error = 0;
418 	struct altq_pktattr pktattr;
419 
420 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
421 
422 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
423 		struct m_tag *mtag;
424 
425 		/* Extract info from dummynet tag */
426 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
427 		KKASSERT(mtag != NULL);
428 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
429 		KKASSERT(rule != NULL);
430 
431 		m_tag_delete(m, mtag);
432 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
433 	}
434 
435 	if (ifq_is_enabled(&ifp->if_snd))
436 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
437 	crit_enter();
438 	if (IPFW_LOADED && ether_ipfw != 0) {
439 		struct ether_header save_eh, *eh;
440 
441 		eh = mtod(m, struct ether_header *);
442 		save_eh = *eh;
443 		m_adj(m, ETHER_HDR_LEN);
444 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
445 			crit_exit();
446 			if (m != NULL) {
447 				m_freem(m);
448 				return ENOBUFS; /* pkt dropped */
449 			} else
450 				return 0;	/* consumed e.g. in a pipe */
451 		}
452 
453 		/* packet was ok, restore the ethernet header */
454 		ether_restore_header(&m, eh, &save_eh);
455 		if (m == NULL) {
456 			crit_exit();
457 			return ENOBUFS;
458 		}
459 	}
460 	crit_exit();
461 
462 	/*
463 	 * Queue message on interface, update output statistics if
464 	 * successful, and start output if interface not yet active.
465 	 */
466 	error = ifq_dispatch(ifp, m, &pktattr);
467 	return (error);
468 }
469 
470 /*
471  * ipfw processing for ethernet packets (in and out).
472  * The second parameter is NULL from ether_demux(), and ifp from
473  * ether_output_frame().
474  */
475 static boolean_t
476 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
477 	       const struct ether_header *eh)
478 {
479 	struct ether_header save_eh = *eh;	/* might be a ptr in *m0 */
480 	struct ip_fw_args args;
481 	struct m_tag *mtag;
482 	struct mbuf *m;
483 	int i;
484 
485 	if (*rule != NULL && fw_one_pass)
486 		return TRUE; /* dummynet packet, already partially processed */
487 
488 	/*
489 	 * I need some amount of data to be contiguous.
490 	 */
491 	i = min((*m0)->m_pkthdr.len, max_protohdr);
492 	if ((*m0)->m_len < i) {
493 		*m0 = m_pullup(*m0, i);
494 		if (*m0 == NULL)
495 			return FALSE;
496 	}
497 
498 	/*
499 	 * Clean up tags
500 	 */
501 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
502 		m_tag_delete(*m0, mtag);
503 	if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
504 		mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
505 		KKASSERT(mtag != NULL);
506 		m_tag_delete(*m0, mtag);
507 		(*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
508 	}
509 
510 	args.m = *m0;		/* the packet we are looking at		*/
511 	args.oif = dst;		/* destination, if any			*/
512 	args.rule = *rule;	/* matching rule to restart		*/
513 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
514 	i = ip_fw_chk_ptr(&args);
515 	*m0 = args.m;
516 	*rule = args.rule;
517 
518 	if (*m0 == NULL)
519 		return FALSE;
520 
521 	switch (i) {
522 	case IP_FW_PASS:
523 		return TRUE;
524 
525 	case IP_FW_DIVERT:
526 	case IP_FW_TEE:
527 	case IP_FW_DENY:
528 		/*
529 		 * XXX at some point add support for divert/forward actions.
530 		 * If none of the above matches, we have to drop the pkt.
531 		 */
532 		return FALSE;
533 
534 	case IP_FW_DUMMYNET:
535 		/*
536 		 * Pass the pkt to dummynet, which consumes it.
537 		 */
538 		m = *m0;	/* pass the original to dummynet */
539 		*m0 = NULL;	/* and nothing back to the caller */
540 
541 		ether_restore_header(&m, eh, &save_eh);
542 		if (m == NULL)
543 			return FALSE;
544 
545 		ip_fw_dn_io_ptr(m, args.cookie,
546 				dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
547 		ip_dn_queue(m);
548 		return FALSE;
549 
550 	default:
551 		panic("unknown ipfw return value: %d", i);
552 	}
553 }
554 
555 static void
556 ether_input(struct ifnet *ifp, struct mbuf *m)
557 {
558 	ether_input_pkt(ifp, m, NULL);
559 }
560 
561 /*
562  * Perform common duties while attaching to interface list
563  */
564 void
565 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
566 {
567 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
568 			   serializer);
569 }
570 
571 void
572 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
573 		   lwkt_serialize_t serializer)
574 {
575 	struct sockaddr_dl *sdl;
576 
577 	ifp->if_type = IFT_ETHER;
578 	ifp->if_addrlen = ETHER_ADDR_LEN;
579 	ifp->if_hdrlen = ETHER_HDR_LEN;
580 	if_attach(ifp, serializer);
581 	ifp->if_mtu = ETHERMTU;
582 	if (ifp->if_baudrate == 0)
583 		ifp->if_baudrate = 10000000;
584 	ifp->if_output = ether_output;
585 	ifp->if_input = ether_input;
586 	ifp->if_resolvemulti = ether_resolvemulti;
587 	ifp->if_broadcastaddr = etherbroadcastaddr;
588 	sdl = IF_LLSOCKADDR(ifp);
589 	sdl->sdl_type = IFT_ETHER;
590 	sdl->sdl_alen = ifp->if_addrlen;
591 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
592 	/*
593 	 * XXX Keep the current drivers happy.
594 	 * XXX Remove once all drivers have been cleaned up
595 	 */
596 	if (lla != IFP2AC(ifp)->ac_enaddr)
597 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
598 	bpfattach(ifp, dlt, hdrlen);
599 	if (ng_ether_attach_p != NULL)
600 		(*ng_ether_attach_p)(ifp);
601 
602 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
603 }
604 
605 /*
606  * Perform common duties while detaching an Ethernet interface
607  */
608 void
609 ether_ifdetach(struct ifnet *ifp)
610 {
611 	if_down(ifp);
612 
613 	if (ng_ether_detach_p != NULL)
614 		(*ng_ether_detach_p)(ifp);
615 	bpfdetach(ifp);
616 	if_detach(ifp);
617 }
618 
619 int
620 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
621 {
622 	struct ifaddr *ifa = (struct ifaddr *) data;
623 	struct ifreq *ifr = (struct ifreq *) data;
624 	int error = 0;
625 
626 #define IF_INIT(ifp) \
627 do { \
628 	if (((ifp)->if_flags & IFF_UP) == 0) { \
629 		(ifp)->if_flags |= IFF_UP; \
630 		(ifp)->if_init((ifp)->if_softc); \
631 	} \
632 } while (0)
633 
634 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
635 
636 	switch (command) {
637 	case SIOCSIFADDR:
638 		switch (ifa->ifa_addr->sa_family) {
639 #ifdef INET
640 		case AF_INET:
641 			IF_INIT(ifp);	/* before arpwhohas */
642 			arp_ifinit(ifp, ifa);
643 			break;
644 #endif
645 #ifdef IPX
646 		/*
647 		 * XXX - This code is probably wrong
648 		 */
649 		case AF_IPX:
650 			{
651 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
652 			struct arpcom *ac = IFP2AC(ifp);
653 
654 			if (ipx_nullhost(*ina))
655 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
656 			else
657 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
658 				      sizeof ac->ac_enaddr);
659 
660 			IF_INIT(ifp);	/* Set new address. */
661 			break;
662 			}
663 #endif
664 		default:
665 			IF_INIT(ifp);
666 			break;
667 		}
668 		break;
669 
670 	case SIOCGIFADDR:
671 		bcopy(IFP2AC(ifp)->ac_enaddr,
672 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
673 		      ETHER_ADDR_LEN);
674 		break;
675 
676 	case SIOCSIFMTU:
677 		/*
678 		 * Set the interface MTU.
679 		 */
680 		if (ifr->ifr_mtu > ETHERMTU) {
681 			error = EINVAL;
682 		} else {
683 			ifp->if_mtu = ifr->ifr_mtu;
684 		}
685 		break;
686 	default:
687 		error = EINVAL;
688 		break;
689 	}
690 	return (error);
691 
692 #undef IF_INIT
693 }
694 
695 int
696 ether_resolvemulti(
697 	struct ifnet *ifp,
698 	struct sockaddr **llsa,
699 	struct sockaddr *sa)
700 {
701 	struct sockaddr_dl *sdl;
702 #ifdef INET
703 	struct sockaddr_in *sin;
704 #endif
705 #ifdef INET6
706 	struct sockaddr_in6 *sin6;
707 #endif
708 	u_char *e_addr;
709 
710 	switch(sa->sa_family) {
711 	case AF_LINK:
712 		/*
713 		 * No mapping needed. Just check that it's a valid MC address.
714 		 */
715 		sdl = (struct sockaddr_dl *)sa;
716 		e_addr = LLADDR(sdl);
717 		if ((e_addr[0] & 1) != 1)
718 			return EADDRNOTAVAIL;
719 		*llsa = NULL;
720 		return 0;
721 
722 #ifdef INET
723 	case AF_INET:
724 		sin = (struct sockaddr_in *)sa;
725 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
726 			return EADDRNOTAVAIL;
727 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
728 		sdl->sdl_len = sizeof *sdl;
729 		sdl->sdl_family = AF_LINK;
730 		sdl->sdl_index = ifp->if_index;
731 		sdl->sdl_type = IFT_ETHER;
732 		sdl->sdl_alen = ETHER_ADDR_LEN;
733 		e_addr = LLADDR(sdl);
734 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
735 		*llsa = (struct sockaddr *)sdl;
736 		return 0;
737 #endif
738 #ifdef INET6
739 	case AF_INET6:
740 		sin6 = (struct sockaddr_in6 *)sa;
741 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
742 			/*
743 			 * An IP6 address of 0 means listen to all
744 			 * of the Ethernet multicast address used for IP6.
745 			 * (This is used for multicast routers.)
746 			 */
747 			ifp->if_flags |= IFF_ALLMULTI;
748 			*llsa = NULL;
749 			return 0;
750 		}
751 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
752 			return EADDRNOTAVAIL;
753 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
754 		sdl->sdl_len = sizeof *sdl;
755 		sdl->sdl_family = AF_LINK;
756 		sdl->sdl_index = ifp->if_index;
757 		sdl->sdl_type = IFT_ETHER;
758 		sdl->sdl_alen = ETHER_ADDR_LEN;
759 		e_addr = LLADDR(sdl);
760 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
761 		*llsa = (struct sockaddr *)sdl;
762 		return 0;
763 #endif
764 
765 	default:
766 		/*
767 		 * Well, the text isn't quite right, but it's the name
768 		 * that counts...
769 		 */
770 		return EAFNOSUPPORT;
771 	}
772 }
773 
774 #if 0
775 /*
776  * This is for reference.  We have a table-driven version
777  * of the little-endian crc32 generator, which is faster
778  * than the double-loop.
779  */
780 uint32_t
781 ether_crc32_le(const uint8_t *buf, size_t len)
782 {
783 	uint32_t c, crc, carry;
784 	size_t i, j;
785 
786 	crc = 0xffffffffU;	/* initial value */
787 
788 	for (i = 0; i < len; i++) {
789 		c = buf[i];
790 		for (j = 0; j < 8; j++) {
791 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
792 			crc >>= 1;
793 			c >>= 1;
794 			if (carry)
795 				crc = (crc ^ ETHER_CRC_POLY_LE);
796 		}
797 	}
798 
799 	return (crc);
800 }
801 #else
802 uint32_t
803 ether_crc32_le(const uint8_t *buf, size_t len)
804 {
805 	static const uint32_t crctab[] = {
806 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
807 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
808 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
809 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
810 	};
811 	uint32_t crc;
812 	size_t i;
813 
814 	crc = 0xffffffffU;	/* initial value */
815 
816 	for (i = 0; i < len; i++) {
817 		crc ^= buf[i];
818 		crc = (crc >> 4) ^ crctab[crc & 0xf];
819 		crc = (crc >> 4) ^ crctab[crc & 0xf];
820 	}
821 
822 	return (crc);
823 }
824 #endif
825 
826 uint32_t
827 ether_crc32_be(const uint8_t *buf, size_t len)
828 {
829 	uint32_t c, crc, carry;
830 	size_t i, j;
831 
832 	crc = 0xffffffffU;	/* initial value */
833 
834 	for (i = 0; i < len; i++) {
835 		c = buf[i];
836 		for (j = 0; j < 8; j++) {
837 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
838 			crc <<= 1;
839 			c >>= 1;
840 			if (carry)
841 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
842 		}
843 	}
844 
845 	return (crc);
846 }
847 
848 /*
849  * find the size of ethernet header, and call classifier
850  */
851 void
852 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
853 		   struct altq_pktattr *pktattr)
854 {
855 	struct ether_header *eh;
856 	uint16_t ether_type;
857 	int hlen, af, hdrsize;
858 	caddr_t hdr;
859 
860 	hlen = sizeof(struct ether_header);
861 	eh = mtod(m, struct ether_header *);
862 
863 	ether_type = ntohs(eh->ether_type);
864 	if (ether_type < ETHERMTU) {
865 		/* ick! LLC/SNAP */
866 		struct llc *llc = (struct llc *)(eh + 1);
867 		hlen += 8;
868 
869 		if (m->m_len < hlen ||
870 		    llc->llc_dsap != LLC_SNAP_LSAP ||
871 		    llc->llc_ssap != LLC_SNAP_LSAP ||
872 		    llc->llc_control != LLC_UI)
873 			goto bad;  /* not snap! */
874 
875 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
876 	}
877 
878 	if (ether_type == ETHERTYPE_IP) {
879 		af = AF_INET;
880 		hdrsize = 20;  /* sizeof(struct ip) */
881 #ifdef INET6
882 	} else if (ether_type == ETHERTYPE_IPV6) {
883 		af = AF_INET6;
884 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
885 #endif
886 	} else
887 		goto bad;
888 
889 	while (m->m_len <= hlen) {
890 		hlen -= m->m_len;
891 		m = m->m_next;
892 	}
893 	hdr = m->m_data + hlen;
894 	if (m->m_len < hlen + hdrsize) {
895 		/*
896 		 * ip header is not in a single mbuf.  this should not
897 		 * happen in the current code.
898 		 * (todo: use m_pulldown in the future)
899 		 */
900 		goto bad;
901 	}
902 	m->m_data += hlen;
903 	m->m_len -= hlen;
904 	ifq_classify(ifq, m, af, pktattr);
905 	m->m_data -= hlen;
906 	m->m_len += hlen;
907 
908 	return;
909 
910 bad:
911 	pktattr->pattr_class = NULL;
912 	pktattr->pattr_hdr = NULL;
913 	pktattr->pattr_af = AF_UNSPEC;
914 }
915 
916 static void
917 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
918 		     const struct ether_header *save_eh)
919 {
920 	struct mbuf *m = *m0;
921 
922 	ether_restore_hdr++;
923 
924 	/*
925 	 * Prepend the header, optimize for the common case of
926 	 * eh pointing into the mbuf.
927 	 */
928 	if ((const void *)(eh + 1) == (void *)m->m_data) {
929 		m->m_data -= ETHER_HDR_LEN;
930 		m->m_len += ETHER_HDR_LEN;
931 		m->m_pkthdr.len += ETHER_HDR_LEN;
932 	} else {
933 		ether_prepend_hdr++;
934 
935 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
936 		if (m != NULL) {
937 			bcopy(save_eh, mtod(m, struct ether_header *),
938 			      ETHER_HDR_LEN);
939 		}
940 	}
941 	*m0 = m;
942 }
943 
944 /*
945  * Upper layer processing for a received Ethernet packet.
946  */
947 void
948 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
949 {
950 	struct ether_header *eh;
951 	int isr, discard = 0;
952 	u_short ether_type;
953 	struct ip_fw *rule = NULL;
954 
955 	M_ASSERTPKTHDR(m);
956 	KASSERT(m->m_len >= ETHER_HDR_LEN,
957 		("ether header is not contiguous!"));
958 
959 	eh = mtod(m, struct ether_header *);
960 
961 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
962 		struct m_tag *mtag;
963 
964 		/* Extract info from dummynet tag */
965 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
966 		KKASSERT(mtag != NULL);
967 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
968 		KKASSERT(rule != NULL);
969 
970 		m_tag_delete(m, mtag);
971 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
972 
973 		/* packet is passing the second time */
974 		goto post_stats;
975 	}
976 
977 	/*
978 	 * We got a packet which was unicast to a different Ethernet
979 	 * address.  If the driver is working properly, then this
980 	 * situation can only happen when the interface is in
981 	 * promiscuous mode.  We defer the packet discarding until the
982 	 * vlan processing is done, so that vlan/bridge or vlan/netgraph
983 	 * could work.
984 	 */
985 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
986 	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
987 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
988 		if (ether_debug & 1) {
989 			kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
990 				"%02x:%02x:%02x:%02x:%02x:%02x "
991 				"%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
992 				eh->ether_dhost[0],
993 				eh->ether_dhost[1],
994 				eh->ether_dhost[2],
995 				eh->ether_dhost[3],
996 				eh->ether_dhost[4],
997 				eh->ether_dhost[5],
998 				eh->ether_shost[0],
999 				eh->ether_shost[1],
1000 				eh->ether_shost[2],
1001 				eh->ether_shost[3],
1002 				eh->ether_shost[4],
1003 				eh->ether_shost[5],
1004 				eh->ether_type,
1005 				((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1006 				((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1007 				((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1008 				((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1009 				((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1010 				((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1011 			);
1012 		}
1013 		if ((ether_debug & 2) == 0)
1014 			discard = 1;
1015 	}
1016 
1017 post_stats:
1018 	if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
1019 		struct ether_header save_eh = *eh;
1020 
1021 		/* XXX old crufty stuff, needs to be removed */
1022 		m_adj(m, sizeof(struct ether_header));
1023 
1024 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1025 			m_freem(m);
1026 			return;
1027 		}
1028 
1029 		ether_restore_header(&m, eh, &save_eh);
1030 		if (m == NULL)
1031 			return;
1032 		eh = mtod(m, struct ether_header *);
1033 	}
1034 
1035 	ether_type = ntohs(eh->ether_type);
1036 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1037 
1038 	if (m->m_flags & M_VLANTAG) {
1039 		void (*vlan_input_func)(struct mbuf *);
1040 
1041 		vlan_input_func = vlan_input_p;
1042 		if (vlan_input_func != NULL) {
1043 			vlan_input_func(m);
1044 		} else {
1045 			m->m_pkthdr.rcvif->if_noproto++;
1046 			m_freem(m);
1047 		}
1048 		return;
1049 	}
1050 
1051 	/*
1052 	 * If we have been asked to discard this packet
1053 	 * (e.g. not for us), drop it before entering
1054 	 * the upper layer.
1055 	 */
1056 	if (discard) {
1057 		m_freem(m);
1058 		return;
1059 	}
1060 
1061 	/*
1062 	 * Clear protocol specific flags,
1063 	 * before entering the upper layer.
1064 	 */
1065 	m->m_flags &= ~M_ETHER_FLAGS;
1066 
1067 	/* Strip ethernet header. */
1068 	m_adj(m, sizeof(struct ether_header));
1069 
1070 	switch (ether_type) {
1071 #ifdef INET
1072 	case ETHERTYPE_IP:
1073 		if ((m->m_flags & M_LENCHECKED) == 0) {
1074 			if (!ip_lengthcheck(&m, 0))
1075 				return;
1076 		}
1077 		if (ipflow_fastforward(m))
1078 			return;
1079 		isr = NETISR_IP;
1080 		break;
1081 
1082 	case ETHERTYPE_ARP:
1083 		if (ifp->if_flags & IFF_NOARP) {
1084 			/* Discard packet if ARP is disabled on interface */
1085 			m_freem(m);
1086 			return;
1087 		}
1088 		isr = NETISR_ARP;
1089 		break;
1090 #endif
1091 
1092 #ifdef INET6
1093 	case ETHERTYPE_IPV6:
1094 		isr = NETISR_IPV6;
1095 		break;
1096 #endif
1097 
1098 #ifdef IPX
1099 	case ETHERTYPE_IPX:
1100 		if (ef_inputp) {
1101 			/*
1102 			 * Hold BGL and recheck ef_inputp
1103 			 */
1104 			get_mplock();
1105 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1106 				rel_mplock();
1107 				return;
1108 			}
1109 			rel_mplock();
1110 		}
1111 		isr = NETISR_IPX;
1112 		break;
1113 #endif
1114 
1115 #ifdef MPLS
1116 	case ETHERTYPE_MPLS:
1117 	case ETHERTYPE_MPLS_MCAST:
1118 		/* Should have been set by ether_input_pkt(). */
1119 		KKASSERT(m->m_flags & M_MPLSLABELED);
1120 		isr = NETISR_MPLS;
1121 		break;
1122 #endif
1123 
1124 	default:
1125 		/*
1126 		 * The accurate msgport is not determined before
1127 		 * we reach here, so recharacterize packet.
1128 		 */
1129 		m->m_flags &= ~M_HASH;
1130 #ifdef IPX
1131 		if (ef_inputp) {
1132 			/*
1133 			 * Hold BGL and recheck ef_inputp
1134 			 */
1135 			get_mplock();
1136 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1137 				rel_mplock();
1138 				return;
1139 			}
1140 			rel_mplock();
1141 		}
1142 #endif
1143 		if (ng_ether_input_orphan_p != NULL) {
1144 			/*
1145 			 * Put back the ethernet header so netgraph has a
1146 			 * consistent view of inbound packets.
1147 			 */
1148 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1149 			if (m == NULL) {
1150 				/*
1151 				 * M_PREPEND frees the mbuf in case of failure.
1152 				 */
1153 				return;
1154 			}
1155 			/*
1156 			 * Hold BGL and recheck ng_ether_input_orphan_p
1157 			 */
1158 			get_mplock();
1159 			if (ng_ether_input_orphan_p != NULL) {
1160 				ng_ether_input_orphan_p(ifp, m);
1161 				rel_mplock();
1162 				return;
1163 			}
1164 			rel_mplock();
1165 		}
1166 		m_freem(m);
1167 		return;
1168 	}
1169 
1170 	if (m->m_flags & M_HASH) {
1171 		if (&curthread->td_msgport == cpu_portfn(m->m_pkthdr.hash)) {
1172 			netisr_handle(isr, m);
1173 			return;
1174 		} else {
1175 			/*
1176 			 * XXX Something is wrong,
1177 			 * we probably should panic here!
1178 			 */
1179 			m->m_flags &= ~M_HASH;
1180 			atomic_add_long(&ether_input_wronghash, 1);
1181 		}
1182 	}
1183 #ifdef RSS_DEBUG
1184 	atomic_add_long(&ether_input_requeue, 1);
1185 #endif
1186 	netisr_queue(isr, m);
1187 }
1188 
1189 /*
1190  * First we perform any link layer operations, then continue to the
1191  * upper layers with ether_demux_oncpu().
1192  */
1193 static void
1194 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1195 {
1196 #ifdef CARP
1197 	void *carp;
1198 #endif
1199 
1200 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1201 		/*
1202 		 * Receiving interface's flags are changed, when this
1203 		 * packet is waiting for processing; discard it.
1204 		 */
1205 		m_freem(m);
1206 		return;
1207 	}
1208 
1209 	/*
1210 	 * Tap the packet off here for a bridge.  bridge_input()
1211 	 * will return NULL if it has consumed the packet, otherwise
1212 	 * it gets processed as normal.  Note that bridge_input()
1213 	 * will always return the original packet if we need to
1214 	 * process it locally.
1215 	 */
1216 	if (ifp->if_bridge) {
1217 		KASSERT(bridge_input_p != NULL,
1218 			("%s: if_bridge not loaded!", __func__));
1219 
1220 		if(m->m_flags & M_ETHER_BRIDGED) {
1221 			m->m_flags &= ~M_ETHER_BRIDGED;
1222 		} else {
1223 			m = bridge_input_p(ifp, m);
1224 			if (m == NULL)
1225 				return;
1226 
1227 			KASSERT(ifp == m->m_pkthdr.rcvif,
1228 				("bridge_input_p changed rcvif"));
1229 		}
1230 	}
1231 
1232 #ifdef CARP
1233 	carp = ifp->if_carp;
1234 	if (carp) {
1235 		m = carp_input(carp, m);
1236 		if (m == NULL)
1237 			return;
1238 		KASSERT(ifp == m->m_pkthdr.rcvif,
1239 		    ("carp_input changed rcvif"));
1240 	}
1241 #endif
1242 
1243 	/* Handle ng_ether(4) processing, if any */
1244 	if (ng_ether_input_p != NULL) {
1245 		/*
1246 		 * Hold BGL and recheck ng_ether_input_p
1247 		 */
1248 		get_mplock();
1249 		if (ng_ether_input_p != NULL)
1250 			ng_ether_input_p(ifp, &m);
1251 		rel_mplock();
1252 
1253 		if (m == NULL)
1254 			return;
1255 	}
1256 
1257 	/* Continue with upper layer processing */
1258 	ether_demux_oncpu(ifp, m);
1259 }
1260 
1261 /*
1262  * Perform certain functions of ether_input_pkt():
1263  * - Test IFF_UP
1264  * - Update statistics
1265  * - Run bpf(4) tap if requested
1266  * Then pass the packet to ether_input_oncpu().
1267  *
1268  * This function should be used by pseudo interface (e.g. vlan(4)),
1269  * when it tries to claim that the packet is received by it.
1270  *
1271  * REINPUT_KEEPRCVIF
1272  * REINPUT_RUNBPF
1273  */
1274 void
1275 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1276 {
1277 	/* Discard packet if interface is not up */
1278 	if (!(ifp->if_flags & IFF_UP)) {
1279 		m_freem(m);
1280 		return;
1281 	}
1282 
1283 	/*
1284 	 * Change receiving interface.  The bridge will often pass a flag to
1285 	 * ask that this not be done so ARPs get applied to the correct
1286 	 * side.
1287 	 */
1288 	if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1289 	    m->m_pkthdr.rcvif == NULL) {
1290 		m->m_pkthdr.rcvif = ifp;
1291 	}
1292 
1293 	/* Update statistics */
1294 	ifp->if_ipackets++;
1295 	ifp->if_ibytes += m->m_pkthdr.len;
1296 	if (m->m_flags & (M_MCAST | M_BCAST))
1297 		ifp->if_imcasts++;
1298 
1299 	if (reinput_flags & REINPUT_RUNBPF)
1300 		BPF_MTAP(ifp, m);
1301 
1302 	ether_input_oncpu(ifp, m);
1303 }
1304 
1305 static __inline boolean_t
1306 ether_vlancheck(struct mbuf **m0)
1307 {
1308 	struct mbuf *m = *m0;
1309 	struct ether_header *eh;
1310 	uint16_t ether_type;
1311 
1312 	eh = mtod(m, struct ether_header *);
1313 	ether_type = ntohs(eh->ether_type);
1314 
1315 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1316 		/*
1317 		 * Extract vlan tag if hardware does not do it for us
1318 		 */
1319 		vlan_ether_decap(&m);
1320 		if (m == NULL)
1321 			goto failed;
1322 
1323 		eh = mtod(m, struct ether_header *);
1324 		ether_type = ntohs(eh->ether_type);
1325 	}
1326 
1327 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1328 		/*
1329 		 * To prevent possible dangerous recursion,
1330 		 * we don't do vlan-in-vlan
1331 		 */
1332 		m->m_pkthdr.rcvif->if_noproto++;
1333 		goto failed;
1334 	}
1335 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1336 
1337 	m->m_flags |= M_ETHER_VLANCHECKED;
1338 	*m0 = m;
1339 	return TRUE;
1340 failed:
1341 	if (m != NULL)
1342 		m_freem(m);
1343 	*m0 = NULL;
1344 	return FALSE;
1345 }
1346 
1347 static void
1348 ether_input_handler(netmsg_t nmsg)
1349 {
1350 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1351 	struct ether_header *eh;
1352 	struct ifnet *ifp;
1353 	struct mbuf *m;
1354 
1355 	m = nmp->nm_packet;
1356 	M_ASSERTPKTHDR(m);
1357 
1358 	if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1359 		if (!ether_vlancheck(&m)) {
1360 			KKASSERT(m == NULL);
1361 			return;
1362 		}
1363 	}
1364 	if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH)
1365 #ifdef RSS_DEBUG
1366 	    || ether_input_ckhash
1367 #endif
1368 	    ) {
1369 		int isr;
1370 
1371 		/*
1372 		 * Need to verify the hash supplied by the hardware
1373 		 * which could be wrong.
1374 		 */
1375 		m->m_flags &= ~(M_HASH | M_CKHASH);
1376 		isr = ether_characterize(&m);
1377 		if (m == NULL)
1378 			return;
1379 		KKASSERT(m->m_flags & M_HASH);
1380 
1381 		if (m->m_pkthdr.hash != mycpuid) {
1382 			/*
1383 			 * Wrong hardware supplied hash; redispatch
1384 			 */
1385 			ether_dispatch(isr, m);
1386 #ifdef RSS_DEBUG
1387 			atomic_add_long(&ether_input_wronghwhash, 1);
1388 #endif
1389 			return;
1390 		}
1391 	}
1392 	ifp = m->m_pkthdr.rcvif;
1393 
1394 	eh = mtod(m, struct ether_header *);
1395 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1396 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1397 			 ifp->if_addrlen) == 0)
1398 			m->m_flags |= M_BCAST;
1399 		else
1400 			m->m_flags |= M_MCAST;
1401 		ifp->if_imcasts++;
1402 	}
1403 
1404 	ether_input_oncpu(ifp, m);
1405 }
1406 
1407 /*
1408  * Send the packet to the target msgport
1409  *
1410  * At this point the packet had better be characterized (M_HASH set),
1411  * so we know which cpu to send it to.
1412  */
1413 static void
1414 ether_dispatch(int isr, struct mbuf *m)
1415 {
1416 	struct netmsg_packet *pmsg;
1417 
1418 	KKASSERT(m->m_flags & M_HASH);
1419 	pmsg = &m->m_hdr.mh_netmsg;
1420 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1421 		    0, ether_input_handler);
1422 	pmsg->nm_packet = m;
1423 	pmsg->base.lmsg.u.ms_result = isr;
1424 
1425 	logether(disp_beg, NULL);
1426 	lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg);
1427 	logether(disp_end, NULL);
1428 }
1429 
1430 /*
1431  * Process a received Ethernet packet.
1432  *
1433  * The ethernet header is assumed to be in the mbuf so the caller
1434  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1435  * bytes in the first mbuf.
1436  */
1437 void
1438 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi)
1439 {
1440 	int isr;
1441 
1442 	M_ASSERTPKTHDR(m);
1443 
1444 	/* Discard packet if interface is not up */
1445 	if (!(ifp->if_flags & IFF_UP)) {
1446 		m_freem(m);
1447 		return;
1448 	}
1449 
1450 	if (m->m_len < sizeof(struct ether_header)) {
1451 		/* XXX error in the caller. */
1452 		m_freem(m);
1453 		return;
1454 	}
1455 
1456 	m->m_pkthdr.rcvif = ifp;
1457 
1458 	logether(pkt_beg, ifp);
1459 
1460 	ETHER_BPF_MTAP(ifp, m);
1461 
1462 	ifp->if_ibytes += m->m_pkthdr.len;
1463 
1464 	if (ifp->if_flags & IFF_MONITOR) {
1465 		struct ether_header *eh;
1466 
1467 		eh = mtod(m, struct ether_header *);
1468 		if (ETHER_IS_MULTICAST(eh->ether_dhost))
1469 			ifp->if_imcasts++;
1470 
1471 		/*
1472 		 * Interface marked for monitoring; discard packet.
1473 		 */
1474 		m_freem(m);
1475 
1476 		logether(pkt_end, ifp);
1477 		return;
1478 	}
1479 
1480 	/*
1481 	 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1482 	 * we can dispatch it immediately without further inspection.
1483 	 */
1484 	if (pi != NULL && (m->m_flags & M_HASH)) {
1485 #ifdef RSS_DEBUG
1486 		atomic_add_long(&ether_pktinfo_try, 1);
1487 #endif
1488 		netisr_hashcheck(pi->pi_netisr, m, pi);
1489 		if (m->m_flags & M_HASH) {
1490 			ether_dispatch(pi->pi_netisr, m);
1491 #ifdef RSS_DEBUG
1492 			atomic_add_long(&ether_pktinfo_hit, 1);
1493 #endif
1494 			logether(pkt_end, ifp);
1495 			return;
1496 		}
1497 	}
1498 #ifdef RSS_DEBUG
1499 	else if (ifp->if_capenable & IFCAP_RSS) {
1500 		if (pi == NULL)
1501 			atomic_add_long(&ether_rss_nopi, 1);
1502 		else
1503 			atomic_add_long(&ether_rss_nohash, 1);
1504 	}
1505 #endif
1506 
1507 	/*
1508 	 * Packet hash will be recalculated by software, so clear
1509 	 * the M_HASH and M_CKHASH flag set by the driver; the hash
1510 	 * value calculated by the hardware may not be exactly what
1511 	 * we want.
1512 	 */
1513 	m->m_flags &= ~(M_HASH | M_CKHASH);
1514 
1515 	if (!ether_vlancheck(&m)) {
1516 		KKASSERT(m == NULL);
1517 		logether(pkt_end, ifp);
1518 		return;
1519 	}
1520 
1521 	isr = ether_characterize(&m);
1522 	if (m == NULL) {
1523 		logether(pkt_end, ifp);
1524 		return;
1525 	}
1526 
1527 	/*
1528 	 * Finally dispatch it
1529 	 */
1530 	ether_dispatch(isr, m);
1531 
1532 	logether(pkt_end, ifp);
1533 }
1534 
1535 static int
1536 ether_characterize(struct mbuf **m0)
1537 {
1538 	struct mbuf *m = *m0;
1539 	struct ether_header *eh;
1540 	uint16_t ether_type;
1541 	int isr;
1542 
1543 	eh = mtod(m, struct ether_header *);
1544 	ether_type = ntohs(eh->ether_type);
1545 
1546 	/*
1547 	 * Map ether type to netisr id.
1548 	 */
1549 	switch (ether_type) {
1550 #ifdef INET
1551 	case ETHERTYPE_IP:
1552 		isr = NETISR_IP;
1553 		break;
1554 
1555 	case ETHERTYPE_ARP:
1556 		isr = NETISR_ARP;
1557 		break;
1558 #endif
1559 
1560 #ifdef INET6
1561 	case ETHERTYPE_IPV6:
1562 		isr = NETISR_IPV6;
1563 		break;
1564 #endif
1565 
1566 #ifdef IPX
1567 	case ETHERTYPE_IPX:
1568 		isr = NETISR_IPX;
1569 		break;
1570 #endif
1571 
1572 #ifdef MPLS
1573 	case ETHERTYPE_MPLS:
1574 	case ETHERTYPE_MPLS_MCAST:
1575 		m->m_flags |= M_MPLSLABELED;
1576 		isr = NETISR_MPLS;
1577 		break;
1578 #endif
1579 
1580 	default:
1581 		/*
1582 		 * NETISR_MAX is an invalid value; it is chosen to let
1583 		 * netisr_characterize() know that we have no clear
1584 		 * idea where this packet should go.
1585 		 */
1586 		isr = NETISR_MAX;
1587 		break;
1588 	}
1589 
1590 	/*
1591 	 * Ask the isr to characterize the packet since we couldn't.
1592 	 * This is an attempt to optimally get us onto the correct protocol
1593 	 * thread.
1594 	 */
1595 	netisr_characterize(isr, &m, sizeof(struct ether_header));
1596 
1597 	*m0 = m;
1598 	return isr;
1599 }
1600 
1601 static void
1602 ether_demux_handler(netmsg_t nmsg)
1603 {
1604 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1605 	struct ifnet *ifp;
1606 	struct mbuf *m;
1607 
1608 	m = nmp->nm_packet;
1609 	M_ASSERTPKTHDR(m);
1610 	ifp = m->m_pkthdr.rcvif;
1611 
1612 	ether_demux_oncpu(ifp, m);
1613 }
1614 
1615 void
1616 ether_demux(struct mbuf *m)
1617 {
1618 	struct netmsg_packet *pmsg;
1619 	int isr;
1620 
1621 	isr = ether_characterize(&m);
1622 	if (m == NULL)
1623 		return;
1624 
1625 	KKASSERT(m->m_flags & M_HASH);
1626 	pmsg = &m->m_hdr.mh_netmsg;
1627 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1628 	    0, ether_demux_handler);
1629 	pmsg->nm_packet = m;
1630 	pmsg->base.lmsg.u.ms_result = isr;
1631 
1632 	lwkt_sendmsg(cpu_portfn(m->m_pkthdr.hash), &pmsg->base.lmsg);
1633 }
1634 
1635 boolean_t
1636 ether_tso_pullup(struct mbuf **mp, int *hoff0, struct ip **ip, int *iphlen,
1637     struct tcphdr **th, int *thoff)
1638 {
1639 	struct mbuf *m = *mp;
1640 	struct ether_header *eh;
1641 	uint16_t type;
1642 	int hoff;
1643 
1644 	KASSERT(M_WRITABLE(m), ("not writable"));
1645 
1646 	hoff = ETHER_HDR_LEN;
1647 	if (m->m_len < hoff) {
1648 		m = m_pullup(m, hoff);
1649 		if (m == NULL)
1650 			goto failed;
1651 	}
1652 	eh = mtod(m, struct ether_header *);
1653 	type = eh->ether_type;
1654 
1655 	if (type == htons(ETHERTYPE_VLAN)) {
1656 		struct ether_vlan_header *evh;
1657 
1658 		hoff += EVL_ENCAPLEN;
1659 		if (m->m_len < hoff) {
1660 			m = m_pullup(m, hoff);
1661 			if (m == NULL)
1662 				goto failed;
1663 		}
1664 		evh = mtod(m, struct ether_vlan_header *);
1665 		type = evh->evl_proto;
1666 	}
1667 	KASSERT(type == htons(ETHERTYPE_IP), ("not IP %d", ntohs(type)));
1668 
1669 	*mp = m;
1670 	*hoff0 = hoff;
1671 	return tcp_tso_pullup(mp, hoff, ip, iphlen, th, thoff);
1672 
1673 failed:
1674 	if (m != NULL)
1675 		m_freem(m);
1676 	*mp = NULL;
1677 	return FALSE;
1678 }
1679 
1680 MODULE_VERSION(ether, 1);
1681