xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 6b7bde44184a7a4502ef3ec6f061b4268b50ea82)
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  * $DragonFly: src/sys/net/if_ethersubr.c,v 1.85 2008/08/23 08:26:04 sephe Exp $
36  */
37 
38 #include "opt_atalk.h"
39 #include "opt_inet.h"
40 #include "opt_inet6.h"
41 #include "opt_ipx.h"
42 #include "opt_mpls.h"
43 #include "opt_netgraph.h"
44 #include "opt_carp.h"
45 #include "opt_ethernet.h"
46 
47 #include <sys/param.h>
48 #include <sys/systm.h>
49 #include <sys/globaldata.h>
50 #include <sys/kernel.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 #include <sys/thread2.h>
59 
60 #include <net/if.h>
61 #include <net/netisr.h>
62 #include <net/route.h>
63 #include <net/if_llc.h>
64 #include <net/if_dl.h>
65 #include <net/if_types.h>
66 #include <net/ifq_var.h>
67 #include <net/bpf.h>
68 #include <net/ethernet.h>
69 #include <net/vlan/if_vlan_ether.h>
70 #include <net/netmsg2.h>
71 
72 #if defined(INET) || defined(INET6)
73 #include <netinet/in.h>
74 #include <netinet/in_var.h>
75 #include <netinet/if_ether.h>
76 #include <net/ipfw/ip_fw.h>
77 #include <net/dummynet/ip_dummynet.h>
78 #endif
79 #ifdef INET6
80 #include <netinet6/nd6.h>
81 #endif
82 
83 #ifdef CARP
84 #include <netinet/ip_carp.h>
85 #endif
86 
87 #ifdef IPX
88 #include <netproto/ipx/ipx.h>
89 #include <netproto/ipx/ipx_if.h>
90 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
91 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
92 		  short *tp, int *hlen);
93 #endif
94 
95 #ifdef NS
96 #include <netns/ns.h>
97 #include <netns/ns_if.h>
98 ushort ns_nettype;
99 int ether_outputdebug = 0;
100 int ether_inputdebug = 0;
101 #endif
102 
103 #ifdef NETATALK
104 #include <netproto/atalk/at.h>
105 #include <netproto/atalk/at_var.h>
106 #include <netproto/atalk/at_extern.h>
107 
108 #define	llc_snap_org_code	llc_un.type_snap.org_code
109 #define	llc_snap_ether_type	llc_un.type_snap.ether_type
110 
111 extern u_char	at_org_code[3];
112 extern u_char	aarp_org_code[3];
113 #endif /* NETATALK */
114 
115 #ifdef MPLS
116 #include <netproto/mpls/mpls.h>
117 #endif
118 
119 /* netgraph node hooks for ng_ether(4) */
120 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
121 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp,
122 		struct mbuf *m, const struct ether_header *eh);
123 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
124 void	(*ng_ether_attach_p)(struct ifnet *ifp);
125 void	(*ng_ether_detach_p)(struct ifnet *ifp);
126 
127 void	(*vlan_input2_p)(struct mbuf *);
128 
129 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
130 			struct rtentry *);
131 static void ether_restore_header(struct mbuf **, const struct ether_header *,
132 				 const struct ether_header *);
133 
134 /*
135  * if_bridge support
136  */
137 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
138 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
139 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
140 
141 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
142 			      struct sockaddr *);
143 
144 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
145 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
146 };
147 
148 #define gotoerr(e) do { error = (e); goto bad; } while (0)
149 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
150 
151 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
152 				struct ip_fw **rule,
153 				const struct ether_header *eh);
154 
155 static int ether_ipfw;
156 static u_int ether_restore_hdr;
157 static u_int ether_prepend_hdr;
158 
159 SYSCTL_DECL(_net_link);
160 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
161 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
162 	   &ether_ipfw, 0, "Pass ether pkts through firewall");
163 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
164 	    &ether_restore_hdr, 0, "# of ether header restoration");
165 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
166 	    &ether_prepend_hdr, 0,
167 	    "# of ether header restoration which prepends mbuf");
168 
169 /*
170  * Ethernet output routine.
171  * Encapsulate a packet of type family for the local net.
172  * Use trailer local net encapsulation if enough data in first
173  * packet leaves a multiple of 512 bytes of data in remainder.
174  * Assumes that ifp is actually pointer to arpcom structure.
175  */
176 static int
177 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
178 	     struct rtentry *rt)
179 {
180 	struct ether_header *eh, *deh;
181 	u_char *edst;
182 	int loop_copy = 0;
183 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
184 	struct arpcom *ac = IFP2AC(ifp);
185 	int error;
186 
187 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
188 
189 	if (ifp->if_flags & IFF_MONITOR)
190 		gotoerr(ENETDOWN);
191 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
192 		gotoerr(ENETDOWN);
193 
194 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
195 	if (m == NULL)
196 		return (ENOBUFS);
197 	eh = mtod(m, struct ether_header *);
198 	edst = eh->ether_dhost;
199 
200 	/*
201 	 * Fill in the destination ethernet address and frame type.
202 	 */
203 	switch (dst->sa_family) {
204 #ifdef INET
205 	case AF_INET:
206 		if (!arpresolve(ifp, rt, m, dst, edst))
207 			return (0);	/* if not yet resolved */
208 #ifdef MPLS
209 		if (m->m_flags & M_MPLSLABELED)
210 			eh->ether_type = htons(ETHERTYPE_MPLS);
211 		else
212 #endif
213 			eh->ether_type = htons(ETHERTYPE_IP);
214 		break;
215 #endif
216 #ifdef INET6
217 	case AF_INET6:
218 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
219 			return (0);		/* Something bad happenned. */
220 		eh->ether_type = htons(ETHERTYPE_IPV6);
221 		break;
222 #endif
223 #ifdef IPX
224 	case AF_IPX:
225 		if (ef_outputp != NULL) {
226 			error = ef_outputp(ifp, &m, dst, &eh->ether_type,
227 					   &hlen);
228 			if (error)
229 				goto bad;
230 		} else {
231 			eh->ether_type = htons(ETHERTYPE_IPX);
232 			bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
233 			      edst, ETHER_ADDR_LEN);
234 		}
235 		break;
236 #endif
237 #ifdef NETATALK
238 	case AF_APPLETALK: {
239 		struct at_ifaddr *aa;
240 
241 		if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
242 			error = 0;	/* XXX */
243 			goto bad;
244 		}
245 		/*
246 		 * In the phase 2 case, need to prepend an mbuf for
247 		 * the llc header.  Since we must preserve the value
248 		 * of m, which is passed to us by value, we m_copy()
249 		 * the first mbuf, and use it for our llc header.
250 		 */
251 		if (aa->aa_flags & AFA_PHASE2) {
252 			struct llc llc;
253 
254 			M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
255 			eh = mtod(m, struct ether_header *);
256 			edst = eh->ether_dhost;
257 			llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
258 			llc.llc_control = LLC_UI;
259 			bcopy(at_org_code, llc.llc_snap_org_code,
260 			      sizeof at_org_code);
261 			llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
262 			bcopy(&llc,
263 			      mtod(m, caddr_t) + sizeof(struct ether_header),
264 			      sizeof(struct llc));
265 			eh->ether_type = htons(m->m_pkthdr.len);
266 			hlen = sizeof(struct llc) + ETHER_HDR_LEN;
267 		} else {
268 			eh->ether_type = htons(ETHERTYPE_AT);
269 		}
270 		if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst))
271 			return (0);
272 		break;
273 	  }
274 #endif
275 #ifdef NS
276 	case AF_NS:
277 		switch(ns_nettype) {
278 		default:
279 		case 0x8137:	/* Novell Ethernet_II Ethernet TYPE II */
280 			eh->ether_type = 0x8137;
281 			break;
282 		case 0x0:	/* Novell 802.3 */
283 			eh->ether_type = htons(m->m_pkthdr.len);
284 			break;
285 		case 0xe0e0:	/* Novell 802.2 and Token-Ring */
286 			M_PREPEND(m, 3, MB_DONTWAIT);
287 			eh = mtod(m, struct ether_header *);
288 			edst = eh->ether_dhost;
289 			eh->ether_type = htons(m->m_pkthdr.len);
290 			cp = mtod(m, u_char *) + sizeof(struct ether_header);
291 			*cp++ = 0xE0;
292 			*cp++ = 0xE0;
293 			*cp++ = 0x03;
294 			break;
295 		}
296 		bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
297 		      ETHER_ADDR_LEN);
298 		/*
299 		 * XXX if ns_thishost is the same as the node's ethernet
300 		 * address then just the default code will catch this anyhow.
301 		 * So I'm not sure if this next clause should be here at all?
302 		 * [JRE]
303 		 */
304 		if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
305 			m->m_pkthdr.rcvif = ifp;
306 			netisr_dispatch(NETISR_NS, m);
307 			return (error);
308 		}
309 		if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
310 			m->m_flags |= M_BCAST;
311 		break;
312 #endif
313 	case pseudo_AF_HDRCMPLT:
314 	case AF_UNSPEC:
315 		loop_copy = -1; /* if this is for us, don't do it */
316 		deh = (struct ether_header *)dst->sa_data;
317 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
318 		eh->ether_type = deh->ether_type;
319 		break;
320 
321 	default:
322 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
323 		gotoerr(EAFNOSUPPORT);
324 	}
325 
326 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
327 		memcpy(eh->ether_shost,
328 		       ((struct ether_header *)dst->sa_data)->ether_shost,
329 		       ETHER_ADDR_LEN);
330 	else
331 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
332 
333 	/*
334 	 * Bridges require special output handling.
335 	 */
336 	if (ifp->if_bridge) {
337 		KASSERT(bridge_output_p != NULL,
338 			("%s: if_bridge not loaded!", __func__));
339 		return bridge_output_p(ifp, m);
340 	}
341 
342 	/*
343 	 * If a simplex interface, and the packet is being sent to our
344 	 * Ethernet address or a broadcast address, loopback a copy.
345 	 * XXX To make a simplex device behave exactly like a duplex
346 	 * device, we should copy in the case of sending to our own
347 	 * ethernet address (thus letting the original actually appear
348 	 * on the wire). However, we don't do that here for security
349 	 * reasons and compatibility with the original behavior.
350 	 */
351 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
352 		int csum_flags = 0;
353 
354 		if (m->m_pkthdr.csum_flags & CSUM_IP)
355 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
356 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
357 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
358 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
359 			struct mbuf *n;
360 
361 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
362 				n->m_pkthdr.csum_flags |= csum_flags;
363 				if (csum_flags & CSUM_DATA_VALID)
364 					n->m_pkthdr.csum_data = 0xffff;
365 				if_simloop(ifp, n, dst->sa_family, hlen);
366 			} else
367 				ifp->if_iqdrops++;
368 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
369 				ETHER_ADDR_LEN) == 0) {
370 			m->m_pkthdr.csum_flags |= csum_flags;
371 			if (csum_flags & CSUM_DATA_VALID)
372 				m->m_pkthdr.csum_data = 0xffff;
373 			if_simloop(ifp, m, dst->sa_family, hlen);
374 			return (0);	/* XXX */
375 		}
376 	}
377 
378 #ifdef CARP
379 	if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL)))
380 		goto bad;
381 #endif
382 
383 
384 	/* Handle ng_ether(4) processing, if any */
385 	if (ng_ether_output_p != NULL) {
386 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0)
387 			goto bad;
388 		if (m == NULL)
389 			return (0);
390 	}
391 
392 	/* Continue with link-layer output */
393 	return ether_output_frame(ifp, m);
394 
395 bad:
396 	m_freem(m);
397 	return (error);
398 }
399 
400 /*
401  * Ethernet link layer output routine to send a raw frame to the device.
402  *
403  * This assumes that the 14 byte Ethernet header is present and contiguous
404  * in the first mbuf.
405  */
406 int
407 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
408 {
409 	struct ip_fw *rule = NULL;
410 	int error = 0;
411 	struct altq_pktattr pktattr;
412 
413 	ASSERT_NOT_SERIALIZED(ifp->if_serializer);
414 
415 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
416 		struct m_tag *mtag;
417 
418 		/* Extract info from dummynet tag */
419 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
420 		KKASSERT(mtag != NULL);
421 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
422 		KKASSERT(rule != NULL);
423 
424 		m_tag_delete(m, mtag);
425 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
426 	}
427 
428 	if (ifq_is_enabled(&ifp->if_snd))
429 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
430 	crit_enter();
431 	if (IPFW_LOADED && ether_ipfw != 0) {
432 		struct ether_header save_eh, *eh;
433 
434 		eh = mtod(m, struct ether_header *);
435 		save_eh = *eh;
436 		m_adj(m, ETHER_HDR_LEN);
437 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
438 			crit_exit();
439 			if (m != NULL) {
440 				m_freem(m);
441 				return ENOBUFS; /* pkt dropped */
442 			} else
443 				return 0;	/* consumed e.g. in a pipe */
444 		}
445 
446 		/* packet was ok, restore the ethernet header */
447 		ether_restore_header(&m, eh, &save_eh);
448 		if (m == NULL) {
449 			crit_exit();
450 			return ENOBUFS;
451 		}
452 	}
453 	crit_exit();
454 
455 	/*
456 	 * Queue message on interface, update output statistics if
457 	 * successful, and start output if interface not yet active.
458 	 */
459 	error = ifq_dispatch(ifp, m, &pktattr);
460 	return (error);
461 }
462 
463 /*
464  * ipfw processing for ethernet packets (in and out).
465  * The second parameter is NULL from ether_demux(), and ifp from
466  * ether_output_frame().
467  */
468 static boolean_t
469 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
470 	       const struct ether_header *eh)
471 {
472 	struct ether_header save_eh = *eh;	/* might be a ptr in m */
473 	struct ip_fw_args args;
474 	struct m_tag *mtag;
475 	int i;
476 
477 	if (*rule != NULL && fw_one_pass)
478 		return TRUE; /* dummynet packet, already partially processed */
479 
480 	/*
481 	 * I need some amount of data to be contiguous.
482 	 */
483 	i = min((*m0)->m_pkthdr.len, max_protohdr);
484 	if ((*m0)->m_len < i) {
485 		*m0 = m_pullup(*m0, i);
486 		if (*m0 == NULL)
487 			return FALSE;
488 	}
489 
490 	/*
491 	 * Clean up tags
492 	 */
493 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
494 		m_tag_delete(*m0, mtag);
495 	if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
496 		mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
497 		KKASSERT(mtag != NULL);
498 		m_tag_delete(*m0, mtag);
499 		(*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
500 	}
501 
502 	args.m = *m0;		/* the packet we are looking at		*/
503 	args.oif = dst;		/* destination, if any			*/
504 	args.rule = *rule;	/* matching rule to restart		*/
505 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
506 	i = ip_fw_chk_ptr(&args);
507 	*m0 = args.m;
508 	*rule = args.rule;
509 
510 	if ((i & IP_FW_PORT_DENY_FLAG) || *m0 == NULL)	/* drop */
511 		return FALSE;
512 
513 	if (i == 0)					/* a PASS rule.  */
514 		return TRUE;
515 
516 	if (i & IP_FW_PORT_DYNT_FLAG) {
517 		/*
518 		 * Pass the pkt to dummynet, which consumes it.
519 		 */
520 		struct mbuf *m;
521 
522 		m = *m0;	/* pass the original to dummynet */
523 		*m0 = NULL;	/* and nothing back to the caller */
524 
525 		ether_restore_header(&m, eh, &save_eh);
526 		if (m == NULL)
527 			return FALSE;
528 
529 		ip_fw_dn_io_ptr(m, (i & 0xffff),
530 			dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
531 		return FALSE;
532 	}
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 
540 static void
541 ether_input(struct ifnet *ifp, struct mbuf *m)
542 {
543 	ether_input_chain2(ifp, m, NULL);
544 }
545 
546 /*
547  * Perform common duties while attaching to interface list
548  */
549 void
550 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
551 {
552 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
553 			   serializer);
554 }
555 
556 void
557 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
558 		   lwkt_serialize_t serializer)
559 {
560 	struct sockaddr_dl *sdl;
561 
562 	ifp->if_type = IFT_ETHER;
563 	ifp->if_addrlen = ETHER_ADDR_LEN;
564 	ifp->if_hdrlen = ETHER_HDR_LEN;
565 	if_attach(ifp, serializer);
566 	ifp->if_mtu = ETHERMTU;
567 	if (ifp->if_baudrate == 0)
568 		ifp->if_baudrate = 10000000;
569 	ifp->if_output = ether_output;
570 	ifp->if_input = ether_input;
571 	ifp->if_resolvemulti = ether_resolvemulti;
572 	ifp->if_broadcastaddr = etherbroadcastaddr;
573 	sdl = IF_LLSOCKADDR(ifp);
574 	sdl->sdl_type = IFT_ETHER;
575 	sdl->sdl_alen = ifp->if_addrlen;
576 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
577 	/*
578 	 * XXX Keep the current drivers happy.
579 	 * XXX Remove once all drivers have been cleaned up
580 	 */
581 	if (lla != IFP2AC(ifp)->ac_enaddr)
582 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
583 	bpfattach(ifp, dlt, hdrlen);
584 	if (ng_ether_attach_p != NULL)
585 		(*ng_ether_attach_p)(ifp);
586 
587 	if_printf(ifp, "MAC address: %6D\n", lla, ":");
588 }
589 
590 /*
591  * Perform common duties while detaching an Ethernet interface
592  */
593 void
594 ether_ifdetach(struct ifnet *ifp)
595 {
596 	if_down(ifp);
597 
598 	if (ng_ether_detach_p != NULL)
599 		(*ng_ether_detach_p)(ifp);
600 	bpfdetach(ifp);
601 	if_detach(ifp);
602 }
603 
604 int
605 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
606 {
607 	struct ifaddr *ifa = (struct ifaddr *) data;
608 	struct ifreq *ifr = (struct ifreq *) data;
609 	int error = 0;
610 
611 #define IF_INIT(ifp) \
612 do { \
613 	if (((ifp)->if_flags & IFF_UP) == 0) { \
614 		(ifp)->if_flags |= IFF_UP; \
615 		(ifp)->if_init((ifp)->if_softc); \
616 	} \
617 } while (0)
618 
619 	ASSERT_SERIALIZED(ifp->if_serializer);
620 
621 	switch (command) {
622 	case SIOCSIFADDR:
623 		switch (ifa->ifa_addr->sa_family) {
624 #ifdef INET
625 		case AF_INET:
626 			IF_INIT(ifp);	/* before arpwhohas */
627 			arp_ifinit(ifp, ifa);
628 			break;
629 #endif
630 #ifdef IPX
631 		/*
632 		 * XXX - This code is probably wrong
633 		 */
634 		case AF_IPX:
635 			{
636 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
637 			struct arpcom *ac = IFP2AC(ifp);
638 
639 			if (ipx_nullhost(*ina))
640 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
641 			else
642 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
643 				      sizeof ac->ac_enaddr);
644 
645 			IF_INIT(ifp);	/* Set new address. */
646 			break;
647 			}
648 #endif
649 #ifdef NS
650 		/*
651 		 * XXX - This code is probably wrong
652 		 */
653 		case AF_NS:
654 		{
655 			struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
656 			struct arpcom *ac = IFP2AC(ifp);
657 
658 			if (ns_nullhost(*ina))
659 				ina->x_host = *(union ns_host *)(ac->ac_enaddr);
660 			else
661 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
662 				      sizeof ac->ac_enaddr);
663 
664 			/*
665 			 * Set new address
666 			 */
667 			IF_INIT(ifp);
668 			break;
669 		}
670 #endif
671 		default:
672 			IF_INIT(ifp);
673 			break;
674 		}
675 		break;
676 
677 	case SIOCGIFADDR:
678 		bcopy(IFP2AC(ifp)->ac_enaddr,
679 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
680 		      ETHER_ADDR_LEN);
681 		break;
682 
683 	case SIOCSIFMTU:
684 		/*
685 		 * Set the interface MTU.
686 		 */
687 		if (ifr->ifr_mtu > ETHERMTU) {
688 			error = EINVAL;
689 		} else {
690 			ifp->if_mtu = ifr->ifr_mtu;
691 		}
692 		break;
693 	default:
694 		error = EINVAL;
695 		break;
696 	}
697 	return (error);
698 
699 #undef IF_INIT
700 }
701 
702 int
703 ether_resolvemulti(
704 	struct ifnet *ifp,
705 	struct sockaddr **llsa,
706 	struct sockaddr *sa)
707 {
708 	struct sockaddr_dl *sdl;
709 	struct sockaddr_in *sin;
710 #ifdef INET6
711 	struct sockaddr_in6 *sin6;
712 #endif
713 	u_char *e_addr;
714 
715 	switch(sa->sa_family) {
716 	case AF_LINK:
717 		/*
718 		 * No mapping needed. Just check that it's a valid MC address.
719 		 */
720 		sdl = (struct sockaddr_dl *)sa;
721 		e_addr = LLADDR(sdl);
722 		if ((e_addr[0] & 1) != 1)
723 			return EADDRNOTAVAIL;
724 		*llsa = 0;
725 		return 0;
726 
727 #ifdef INET
728 	case AF_INET:
729 		sin = (struct sockaddr_in *)sa;
730 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
731 			return EADDRNOTAVAIL;
732 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
733 		       M_WAITOK | M_ZERO);
734 		sdl->sdl_len = sizeof *sdl;
735 		sdl->sdl_family = AF_LINK;
736 		sdl->sdl_index = ifp->if_index;
737 		sdl->sdl_type = IFT_ETHER;
738 		sdl->sdl_alen = ETHER_ADDR_LEN;
739 		e_addr = LLADDR(sdl);
740 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
741 		*llsa = (struct sockaddr *)sdl;
742 		return 0;
743 #endif
744 #ifdef INET6
745 	case AF_INET6:
746 		sin6 = (struct sockaddr_in6 *)sa;
747 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
748 			/*
749 			 * An IP6 address of 0 means listen to all
750 			 * of the Ethernet multicast address used for IP6.
751 			 * (This is used for multicast routers.)
752 			 */
753 			ifp->if_flags |= IFF_ALLMULTI;
754 			*llsa = 0;
755 			return 0;
756 		}
757 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
758 			return EADDRNOTAVAIL;
759 		MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
760 		       M_WAITOK | M_ZERO);
761 		sdl->sdl_len = sizeof *sdl;
762 		sdl->sdl_family = AF_LINK;
763 		sdl->sdl_index = ifp->if_index;
764 		sdl->sdl_type = IFT_ETHER;
765 		sdl->sdl_alen = ETHER_ADDR_LEN;
766 		e_addr = LLADDR(sdl);
767 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
768 		*llsa = (struct sockaddr *)sdl;
769 		return 0;
770 #endif
771 
772 	default:
773 		/*
774 		 * Well, the text isn't quite right, but it's the name
775 		 * that counts...
776 		 */
777 		return EAFNOSUPPORT;
778 	}
779 }
780 
781 #if 0
782 /*
783  * This is for reference.  We have a table-driven version
784  * of the little-endian crc32 generator, which is faster
785  * than the double-loop.
786  */
787 uint32_t
788 ether_crc32_le(const uint8_t *buf, size_t len)
789 {
790 	uint32_t c, crc, carry;
791 	size_t i, j;
792 
793 	crc = 0xffffffffU;	/* initial value */
794 
795 	for (i = 0; i < len; i++) {
796 		c = buf[i];
797 		for (j = 0; j < 8; j++) {
798 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
799 			crc >>= 1;
800 			c >>= 1;
801 			if (carry)
802 				crc = (crc ^ ETHER_CRC_POLY_LE);
803 		}
804 	}
805 
806 	return (crc);
807 }
808 #else
809 uint32_t
810 ether_crc32_le(const uint8_t *buf, size_t len)
811 {
812 	static const uint32_t crctab[] = {
813 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
814 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
815 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
816 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
817 	};
818 	uint32_t crc;
819 	size_t i;
820 
821 	crc = 0xffffffffU;	/* initial value */
822 
823 	for (i = 0; i < len; i++) {
824 		crc ^= buf[i];
825 		crc = (crc >> 4) ^ crctab[crc & 0xf];
826 		crc = (crc >> 4) ^ crctab[crc & 0xf];
827 	}
828 
829 	return (crc);
830 }
831 #endif
832 
833 uint32_t
834 ether_crc32_be(const uint8_t *buf, size_t len)
835 {
836 	uint32_t c, crc, carry;
837 	size_t i, j;
838 
839 	crc = 0xffffffffU;	/* initial value */
840 
841 	for (i = 0; i < len; i++) {
842 		c = buf[i];
843 		for (j = 0; j < 8; j++) {
844 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
845 			crc <<= 1;
846 			c >>= 1;
847 			if (carry)
848 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
849 		}
850 	}
851 
852 	return (crc);
853 }
854 
855 /*
856  * find the size of ethernet header, and call classifier
857  */
858 void
859 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
860 		   struct altq_pktattr *pktattr)
861 {
862 	struct ether_header *eh;
863 	uint16_t ether_type;
864 	int hlen, af, hdrsize;
865 	caddr_t hdr;
866 
867 	hlen = sizeof(struct ether_header);
868 	eh = mtod(m, struct ether_header *);
869 
870 	ether_type = ntohs(eh->ether_type);
871 	if (ether_type < ETHERMTU) {
872 		/* ick! LLC/SNAP */
873 		struct llc *llc = (struct llc *)(eh + 1);
874 		hlen += 8;
875 
876 		if (m->m_len < hlen ||
877 		    llc->llc_dsap != LLC_SNAP_LSAP ||
878 		    llc->llc_ssap != LLC_SNAP_LSAP ||
879 		    llc->llc_control != LLC_UI)
880 			goto bad;  /* not snap! */
881 
882 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
883 	}
884 
885 	if (ether_type == ETHERTYPE_IP) {
886 		af = AF_INET;
887 		hdrsize = 20;  /* sizeof(struct ip) */
888 #ifdef INET6
889 	} else if (ether_type == ETHERTYPE_IPV6) {
890 		af = AF_INET6;
891 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
892 #endif
893 	} else
894 		goto bad;
895 
896 	while (m->m_len <= hlen) {
897 		hlen -= m->m_len;
898 		m = m->m_next;
899 	}
900 	hdr = m->m_data + hlen;
901 	if (m->m_len < hlen + hdrsize) {
902 		/*
903 		 * ip header is not in a single mbuf.  this should not
904 		 * happen in the current code.
905 		 * (todo: use m_pulldown in the future)
906 		 */
907 		goto bad;
908 	}
909 	m->m_data += hlen;
910 	m->m_len -= hlen;
911 	ifq_classify(ifq, m, af, pktattr);
912 	m->m_data -= hlen;
913 	m->m_len += hlen;
914 
915 	return;
916 
917 bad:
918 	pktattr->pattr_class = NULL;
919 	pktattr->pattr_hdr = NULL;
920 	pktattr->pattr_af = AF_UNSPEC;
921 }
922 
923 static void
924 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
925 		     const struct ether_header *save_eh)
926 {
927 	struct mbuf *m = *m0;
928 
929 	ether_restore_hdr++;
930 
931 	/*
932 	 * Prepend the header, optimize for the common case of
933 	 * eh pointing into the mbuf.
934 	 */
935 	if ((const void *)(eh + 1) == (void *)m->m_data) {
936 		m->m_data -= ETHER_HDR_LEN;
937 		m->m_len += ETHER_HDR_LEN;
938 		m->m_pkthdr.len += ETHER_HDR_LEN;
939 	} else {
940 		ether_prepend_hdr++;
941 
942 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
943 		if (m != NULL) {
944 			bcopy(save_eh, mtod(m, struct ether_header *),
945 			      ETHER_HDR_LEN);
946 		}
947 	}
948 	*m0 = m;
949 }
950 
951 #ifdef ETHER_INPUT_CHAIN
952 
953 static void
954 ether_input_ipifunc(void *arg)
955 {
956 	struct mbuf *m, *next;
957 	lwkt_port_t port;
958 
959 	m = arg;
960 	do {
961 		next = m->m_nextpkt;
962 		m->m_nextpkt = NULL;
963 
964 		port = m->m_pkthdr.header;
965 		m->m_pkthdr.header = NULL;
966 
967 		lwkt_sendmsg(port,
968 		&m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
969 
970 		m = next;
971 	} while (m != NULL);
972 }
973 
974 void
975 ether_input_dispatch(struct mbuf_chain *chain)
976 {
977 #ifdef SMP
978 	int i;
979 
980 	for (i = 0; i < ncpus; ++i) {
981 		if (chain[i].mc_head != NULL) {
982 			lwkt_send_ipiq(globaldata_find(i),
983 			ether_input_ipifunc, chain[i].mc_head);
984 		}
985 	}
986 #else
987 	if (chain->mc_head != NULL)
988 		ether_input_ipifunc(chain->mc_head);
989 #endif
990 }
991 
992 void
993 ether_input_chain_init(struct mbuf_chain *chain)
994 {
995 #ifdef SMP
996 	int i;
997 
998 	for (i = 0; i < ncpus; ++i)
999 		chain[i].mc_head = chain[i].mc_tail = NULL;
1000 #else
1001 	chain->mc_head = chain->mc_tail = NULL;
1002 #endif
1003 }
1004 
1005 #endif	/* ETHER_INPUT_CHAIN */
1006 
1007 /*
1008  * Upper layer processing for a received Ethernet packet.
1009  */
1010 void
1011 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1012 {
1013 	struct ether_header *eh;
1014 	int isr, redispatch;
1015 	u_short ether_type;
1016 	struct ip_fw *rule = NULL;
1017 #ifdef NETATALK
1018 	struct llc *l;
1019 #endif
1020 
1021 	M_ASSERTPKTHDR(m);
1022 	KASSERT(m->m_len >= ETHER_HDR_LEN,
1023 		("ether header is no contiguous!\n"));
1024 
1025 	eh = mtod(m, struct ether_header *);
1026 
1027 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
1028 		struct m_tag *mtag;
1029 
1030 		/* Extract info from dummynet tag */
1031 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1032 		KKASSERT(mtag != NULL);
1033 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1034 		KKASSERT(rule != NULL);
1035 
1036 		m_tag_delete(m, mtag);
1037 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
1038 
1039 		/* packet is passing the second time */
1040 		goto post_stats;
1041 	}
1042 
1043 #ifdef CARP
1044 	/*
1045 	 * XXX: Okay, we need to call carp_forus() and - if it is for
1046 	 * us jump over code that does the normal check
1047 	 * "ac_enaddr == ether_dhost". The check sequence is a bit
1048 	 * different from OpenBSD, so we jump over as few code as
1049 	 * possible, to catch _all_ sanity checks. This needs
1050 	 * evaluation, to see if the carp ether_dhost values break any
1051 	 * of these checks!
1052 	 */
1053 	if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost))
1054 		goto post_stats;
1055 #endif
1056 
1057 	/*
1058 	 * Discard packet if upper layers shouldn't see it because
1059 	 * it was unicast to a different Ethernet address.  If the
1060 	 * driver is working properly, then this situation can only
1061 	 * happen when the interface is in promiscuous mode.
1062 	 */
1063 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1064 	    (eh->ether_dhost[0] & 1) == 0 &&
1065 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1066 		m_freem(m);
1067 		return;
1068 	}
1069 
1070 post_stats:
1071 	if (IPFW_LOADED && ether_ipfw != 0) {
1072 		struct ether_header save_eh = *eh;
1073 
1074 		/* XXX old crufty stuff, needs to be removed */
1075 		m_adj(m, sizeof(struct ether_header));
1076 
1077 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1078 			m_freem(m);
1079 			return;
1080 		}
1081 
1082 		ether_restore_header(&m, eh, &save_eh);
1083 		if (m == NULL)
1084 			return;
1085 		eh = mtod(m, struct ether_header *);
1086 	}
1087 
1088 	ether_type = ntohs(eh->ether_type);
1089 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1090 
1091 	if (m->m_flags & M_VLANTAG) {
1092 		if (vlan_input2_p != NULL) {
1093 			vlan_input2_p(m);
1094 		} else {
1095 			m->m_pkthdr.rcvif->if_noproto++;
1096 			m_freem(m);
1097 		}
1098 		return;
1099 	}
1100 
1101 	m_adj(m, sizeof(struct ether_header));
1102 	redispatch = 0;
1103 
1104 	switch (ether_type) {
1105 #ifdef INET
1106 	case ETHERTYPE_IP:
1107 		if (ipflow_fastforward(m))
1108 			return;
1109 		isr = NETISR_IP;
1110 		break;
1111 
1112 	case ETHERTYPE_ARP:
1113 		if (ifp->if_flags & IFF_NOARP) {
1114 			/* Discard packet if ARP is disabled on interface */
1115 			m_freem(m);
1116 			return;
1117 		}
1118 		isr = NETISR_ARP;
1119 		break;
1120 #endif
1121 
1122 #ifdef INET6
1123 	case ETHERTYPE_IPV6:
1124 		isr = NETISR_IPV6;
1125 		break;
1126 #endif
1127 
1128 #ifdef IPX
1129 	case ETHERTYPE_IPX:
1130 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1131 			return;
1132 		isr = NETISR_IPX;
1133 		break;
1134 #endif
1135 
1136 #ifdef NS
1137 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1138 		isr = NETISR_NS;
1139 		break;
1140 
1141 #endif
1142 
1143 #ifdef NETATALK
1144 	case ETHERTYPE_AT:
1145 		isr = NETISR_ATALK1;
1146 		break;
1147 	case ETHERTYPE_AARP:
1148 		isr = NETISR_AARP;
1149 		break;
1150 #endif
1151 
1152 #ifdef MPLS
1153 	case ETHERTYPE_MPLS:
1154 	case ETHERTYPE_MPLS_MCAST:
1155 		/* Should have been set by ether_input_chain2(). */
1156 		KKASSERT(m->m_flags & M_MPLSLABELED);
1157 		isr = NETISR_MPLS;
1158 		break;
1159 #endif
1160 
1161 	default:
1162 		/*
1163 		 * The accurate msgport is not determined before
1164 		 * we reach here, so redo the dispatching
1165 		 */
1166 		redispatch = 1;
1167 #ifdef IPX
1168 		if (ef_inputp && ef_inputp(ifp, eh, m) == 0)
1169 			return;
1170 #endif
1171 #ifdef NS
1172 		checksum = mtod(m, ushort *);
1173 		/* Novell 802.3 */
1174 		if ((ether_type <= ETHERMTU) &&
1175 		    ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1176 			if (*checksum == 0xE0E0) {
1177 				m->m_pkthdr.len -= 3;
1178 				m->m_len -= 3;
1179 				m->m_data += 3;
1180 			}
1181 			isr = NETISR_NS;
1182 			break;
1183 		}
1184 #endif
1185 #ifdef NETATALK
1186 		if (ether_type > ETHERMTU)
1187 			goto dropanyway;
1188 		l = mtod(m, struct llc *);
1189 		if (l->llc_dsap == LLC_SNAP_LSAP &&
1190 		    l->llc_ssap == LLC_SNAP_LSAP &&
1191 		    l->llc_control == LLC_UI) {
1192 			if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1193 				 sizeof at_org_code) == 0 &&
1194 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1195 				m_adj(m, sizeof(struct llc));
1196 				isr = NETISR_ATALK2;
1197 				break;
1198 			}
1199 			if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1200 				 sizeof aarp_org_code) == 0 &&
1201 			    ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1202 				m_adj(m, sizeof(struct llc));
1203 				isr = NETISR_AARP;
1204 				break;
1205 			}
1206 		}
1207 dropanyway:
1208 #endif
1209 		if (ng_ether_input_orphan_p != NULL)
1210 			ng_ether_input_orphan_p(ifp, m, eh);
1211 		else
1212 			m_freem(m);
1213 		return;
1214 	}
1215 
1216 	if (!redispatch)
1217 		netisr_run(isr, m);
1218 	else
1219 		netisr_dispatch(isr, m);
1220 }
1221 
1222 /*
1223  * First we perform any link layer operations, then continue to the
1224  * upper layers with ether_demux_oncpu().
1225  */
1226 void
1227 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1228 {
1229 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1230 		/*
1231 		 * Receiving interface's flags are changed, when this
1232 		 * packet is waiting for processing; discard it.
1233 		 */
1234 		m_freem(m);
1235 		return;
1236 	}
1237 
1238 	/*
1239 	 * Tap the packet off here for a bridge.  bridge_input()
1240 	 * will return NULL if it has consumed the packet, otherwise
1241 	 * it gets processed as normal.  Note that bridge_input()
1242 	 * will always return the original packet if we need to
1243 	 * process it locally.
1244 	 */
1245 	if (ifp->if_bridge) {
1246 		KASSERT(bridge_input_p != NULL,
1247 			("%s: if_bridge not loaded!", __func__));
1248 
1249 		if(m->m_flags & M_PROTO1) {
1250 			m->m_flags &= ~M_PROTO1;
1251 		} else {
1252 			/* clear M_PROMISC, in case the packets comes from a vlan */
1253 			/* m->m_flags &= ~M_PROMISC; */
1254 			m = bridge_input_p(ifp, m);
1255 			if (m == NULL)
1256 				return;
1257 
1258 			KASSERT(ifp == m->m_pkthdr.rcvif,
1259 				("bridge_input_p changed rcvif\n"));
1260 		}
1261 	}
1262 
1263 	/* Handle ng_ether(4) processing, if any */
1264 	if (ng_ether_input_p != NULL) {
1265 		ng_ether_input_p(ifp, &m);
1266 		if (m == NULL)
1267 			return;
1268 	}
1269 
1270 	/* Continue with upper layer processing */
1271 	ether_demux_oncpu(ifp, m);
1272 }
1273 
1274 static void
1275 ether_input_handler(struct netmsg *nmsg)
1276 {
1277 	struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1278 	struct ifnet *ifp;
1279 	struct mbuf *m;
1280 
1281 	m = nmp->nm_packet;
1282 	M_ASSERTPKTHDR(m);
1283 	ifp = m->m_pkthdr.rcvif;
1284 
1285 	ether_input_oncpu(ifp, m);
1286 }
1287 
1288 static __inline void
1289 ether_init_netpacket(int num, struct mbuf *m)
1290 {
1291 	struct netmsg_packet *pmsg;
1292 
1293 	pmsg = &m->m_hdr.mh_netmsg;
1294 	netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, 0,
1295 		    ether_input_handler);
1296 	pmsg->nm_packet = m;
1297 	pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1298 }
1299 
1300 static __inline struct lwkt_port *
1301 ether_mport(int num, struct mbuf **m)
1302 {
1303 	if (num == NETISR_MAX) {
1304 		/*
1305 		 * All packets whose target msgports can't be
1306 		 * determined here are dispatched to netisr0,
1307 		 * where further dispatching may happen.
1308 		 */
1309 		return cpu_portfn(0);
1310 	}
1311 	return netisr_find_port(num, m);
1312 }
1313 
1314 /*
1315  * Process a received Ethernet packet.
1316  *
1317  * The ethernet header is assumed to be in the mbuf so the caller
1318  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1319  * bytes in the first mbuf.
1320  *
1321  * We first try to find the target msgport for this ether frame, if
1322  * there is no target msgport for it, this ether frame is discarded,
1323  * else we do following processing according to whether 'chain' is
1324  * NULL or not:
1325  * - If 'chain' is NULL, this ether frame is sent to the target msgport
1326  *   immediately.  This situation happens when ether_input_chain2 is
1327  *   accessed through ifnet.if_input.
1328  * - If 'chain' is not NULL, this ether frame is queued to the 'chain'
1329  *   bucket indexed by the target msgport's cpuid and the target msgport
1330  *   is saved in mbuf's m_pkthdr.m_head.  Caller of ether_input_chain2
1331  *   must initialize 'chain' by calling ether_input_chain_init().
1332  *   ether_input_dispatch must be called later to send ether frames
1333  *   queued on 'chain' to their target msgport.
1334  */
1335 void
1336 ether_input_chain2(struct ifnet *ifp, struct mbuf *m, struct mbuf_chain *chain)
1337 {
1338 	struct ether_header *eh, *save_eh, save_eh0;
1339 	struct lwkt_port *port;
1340 	uint16_t ether_type;
1341 	int isr;
1342 
1343 	ASSERT_SERIALIZED(ifp->if_serializer);
1344 	M_ASSERTPKTHDR(m);
1345 
1346 	/* Discard packet if interface is not up */
1347 	if (!(ifp->if_flags & IFF_UP)) {
1348 		m_freem(m);
1349 		return;
1350 	}
1351 
1352 	if (m->m_len < sizeof(struct ether_header)) {
1353 		/* XXX error in the caller. */
1354 		m_freem(m);
1355 		return;
1356 	}
1357 	eh = mtod(m, struct ether_header *);
1358 
1359 	m->m_pkthdr.rcvif = ifp;
1360 
1361 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1362 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1363 			 ifp->if_addrlen) == 0)
1364 			m->m_flags |= M_BCAST;
1365 		else
1366 			m->m_flags |= M_MCAST;
1367 		ifp->if_imcasts++;
1368 	}
1369 
1370 	ETHER_BPF_MTAP(ifp, m);
1371 
1372 	ifp->if_ibytes += m->m_pkthdr.len;
1373 
1374 	if (ifp->if_flags & IFF_MONITOR) {
1375 		/*
1376 		 * Interface marked for monitoring; discard packet.
1377 		 */
1378 		m_freem(m);
1379 		return;
1380 	}
1381 
1382 	if (ntohs(eh->ether_type) == ETHERTYPE_VLAN &&
1383 	    (m->m_flags & M_VLANTAG) == 0) {
1384 		/*
1385 		 * Extract vlan tag if hardware does not do it for us
1386 		 */
1387 		vlan_ether_decap(&m);
1388 		if (m == NULL)
1389 			return;
1390 		eh = mtod(m, struct ether_header *);
1391 	}
1392 	ether_type = ntohs(eh->ether_type);
1393 
1394 	if ((m->m_flags & M_VLANTAG) && ether_type == ETHERTYPE_VLAN) {
1395 		/*
1396 		 * To prevent possible dangerous recursion,
1397 		 * we don't do vlan-in-vlan
1398 		 */
1399 		ifp->if_noproto++;
1400 		m_freem(m);
1401 		return;
1402 	}
1403 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1404 
1405 	/*
1406 	 * Map ether type to netisr id.
1407 	 */
1408 	switch (ether_type) {
1409 #ifdef INET
1410 	case ETHERTYPE_IP:
1411 		isr = NETISR_IP;
1412 		break;
1413 
1414 	case ETHERTYPE_ARP:
1415 		isr = NETISR_ARP;
1416 		break;
1417 #endif
1418 
1419 #ifdef INET6
1420 	case ETHERTYPE_IPV6:
1421 		isr = NETISR_IPV6;
1422 		break;
1423 #endif
1424 
1425 #ifdef IPX
1426 	case ETHERTYPE_IPX:
1427 		isr = NETISR_IPX;
1428 		break;
1429 #endif
1430 
1431 #ifdef NS
1432 	case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1433 		isr = NETISR_NS;
1434 		break;
1435 #endif
1436 
1437 #ifdef NETATALK
1438 	case ETHERTYPE_AT:
1439 		isr = NETISR_ATALK1;
1440 		break;
1441 	case ETHERTYPE_AARP:
1442 		isr = NETISR_AARP;
1443 		break;
1444 #endif
1445 
1446 #ifdef MPLS
1447 	case ETHERTYPE_MPLS:
1448 	case ETHERTYPE_MPLS_MCAST:
1449 		m->m_flags |= M_MPLSLABELED;
1450 		isr = NETISR_MPLS;
1451 		break;
1452 #endif
1453 
1454 	default:
1455 		/*
1456 		 * NETISR_MAX is an invalid value; it is chosen to let
1457 		 * ether_mport() know that we are not able to decide
1458 		 * this packet's msgport here.
1459 		 */
1460 		isr = NETISR_MAX;
1461 		break;
1462 	}
1463 
1464 	/*
1465 	 * If the packet is in contiguous memory, following
1466 	 * m_adj() could ensure that the hidden ether header
1467 	 * will not be destroyed, else we will have to save
1468 	 * the ether header for the later restoration.
1469 	 */
1470 	if (m->m_pkthdr.len != m->m_len) {
1471 		save_eh0 = *eh;
1472 		save_eh = &save_eh0;
1473 	} else {
1474 		save_eh = NULL;
1475 	}
1476 
1477 	/*
1478 	 * Temporarily remove ether header; ether_mport()
1479 	 * expects a packet without ether header.
1480 	 */
1481 	m_adj(m, sizeof(struct ether_header));
1482 
1483 	/*
1484 	 * Find the packet's target msgport.
1485 	 */
1486 	port = ether_mport(isr, &m);
1487 	if (port == NULL) {
1488 		KKASSERT(m == NULL);
1489 		return;
1490 	}
1491 
1492 	/*
1493 	 * Restore ether header.
1494 	 */
1495 	if (save_eh != NULL) {
1496 		ether_restore_header(&m, eh, save_eh);
1497 		if (m == NULL)
1498 			return;
1499 	} else {
1500 		m->m_data -= ETHER_HDR_LEN;
1501 		m->m_len += ETHER_HDR_LEN;
1502 		m->m_pkthdr.len += ETHER_HDR_LEN;
1503 	}
1504 
1505 	/*
1506 	 * Initialize mbuf's netmsg packet _after_ possible
1507 	 * ether header restoration, else the initialized
1508 	 * netmsg packet may be lost during ether header
1509 	 * restoration.
1510 	 */
1511 	ether_init_netpacket(isr, m);
1512 
1513 #ifdef ETHER_INPUT_CHAIN
1514 	if (chain != NULL) {
1515 		struct mbuf_chain *c;
1516 		int cpuid;
1517 
1518 		m->m_pkthdr.header = port; /* XXX */
1519 		cpuid = port->mpu_td->td_gd->gd_cpuid;
1520 
1521 		c = &chain[cpuid];
1522 		if (c->mc_head == NULL) {
1523 			c->mc_head = c->mc_tail = m;
1524 		} else {
1525 			c->mc_tail->m_nextpkt = m;
1526 			c->mc_tail = m;
1527 		}
1528 		m->m_nextpkt = NULL;
1529 	} else
1530 #endif	/* ETHER_INPUT_CHAIN */
1531 		lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1532 }
1533