xref: /dflybsd-src/sys/net/if_ethersubr.c (revision 7fd4e1a12ffdc583c9780b5a1cd21b3252cc19ae)
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 #include <net/netisr2.h>
75 
76 #if defined(INET) || defined(INET6)
77 #include <netinet/in.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/tcp_var.h>
80 #include <netinet/if_ether.h>
81 #include <netinet/ip_flow.h>
82 #include <net/ipfw/ip_fw.h>
83 #include <net/dummynet/ip_dummynet.h>
84 #endif
85 #ifdef INET6
86 #include <netinet6/nd6.h>
87 #endif
88 
89 #ifdef CARP
90 #include <netinet/ip_carp.h>
91 #endif
92 
93 #ifdef IPX
94 #include <netproto/ipx/ipx.h>
95 #include <netproto/ipx/ipx_if.h>
96 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
97 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
98 		  short *tp, int *hlen);
99 #endif
100 
101 #ifdef MPLS
102 #include <netproto/mpls/mpls.h>
103 #endif
104 
105 /* netgraph node hooks for ng_ether(4) */
106 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
107 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
108 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
109 void	(*ng_ether_attach_p)(struct ifnet *ifp);
110 void	(*ng_ether_detach_p)(struct ifnet *ifp);
111 
112 void	(*vlan_input_p)(struct mbuf *);
113 
114 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
115 			struct rtentry *);
116 static void ether_restore_header(struct mbuf **, const struct ether_header *,
117 				 const struct ether_header *);
118 static int ether_characterize(struct mbuf **);
119 static void ether_dispatch(int, struct mbuf *);
120 
121 /*
122  * if_bridge support
123  */
124 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
125 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
126 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
127 struct ifnet *(*bridge_interface_p)(void *if_bridge);
128 
129 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
130 			      struct sockaddr *);
131 
132 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
133 	0xff, 0xff, 0xff, 0xff, 0xff, 0xff
134 };
135 
136 #define gotoerr(e) do { error = (e); goto bad; } while (0)
137 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
138 
139 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
140 				struct ip_fw **rule,
141 				const struct ether_header *eh);
142 
143 static int ether_ipfw;
144 static u_long ether_restore_hdr;
145 static u_long ether_prepend_hdr;
146 static u_long ether_input_wronghash;
147 static int ether_debug;
148 
149 #ifdef RSS_DEBUG
150 static u_long ether_pktinfo_try;
151 static u_long ether_pktinfo_hit;
152 static u_long ether_rss_nopi;
153 static u_long ether_rss_nohash;
154 static u_long ether_input_requeue;
155 #endif
156 static u_long ether_input_wronghwhash;
157 static int ether_input_ckhash;
158 
159 #define ETHER_TSOLEN_DEFAULT	(4 * ETHERMTU)
160 
161 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
162 TUNABLE_INT("net.link.ether.tsolen", &ether_tsolen_default);
163 
164 SYSCTL_DECL(_net_link);
165 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
166 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
167     &ether_debug, 0, "Ether debug");
168 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
169     &ether_ipfw, 0, "Pass ether pkts through firewall");
170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
171     &ether_restore_hdr, 0, "# of ether header restoration");
172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
173     &ether_prepend_hdr, 0,
174     "# of ether header restoration which prepends mbuf");
175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
176     &ether_input_wronghash, 0, "# of input packets with wrong hash");
177 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
178     &ether_tsolen_default, 0, "Default max TSO length");
179 
180 #ifdef RSS_DEBUG
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
182     &ether_rss_nopi, 0, "# of packets do not have pktinfo");
183 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
184     &ether_rss_nohash, 0, "# of packets do not have hash");
185 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
186     &ether_pktinfo_try, 0,
187     "# of tries to find packets' msgport using pktinfo");
188 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
189     &ether_pktinfo_hit, 0,
190     "# of packets whose msgport are found using pktinfo");
191 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
192     &ether_input_requeue, 0, "# of input packets gets requeued");
193 #endif
194 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
195     &ether_input_wronghwhash, 0, "# of input packets with wrong hw hash");
196 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
197     &ether_input_ckhash, 0, "always check hash");
198 
199 #define ETHER_KTR_STR		"ifp=%p"
200 #define ETHER_KTR_ARGS	struct ifnet *ifp
201 #ifndef KTR_ETHERNET
202 #define KTR_ETHERNET		KTR_ALL
203 #endif
204 KTR_INFO_MASTER(ether);
205 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
206 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
207 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
208 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
209 #define logether(name, arg)	KTR_LOG(ether_ ## name, arg)
210 
211 /*
212  * Ethernet output routine.
213  * Encapsulate a packet of type family for the local net.
214  * Use trailer local net encapsulation if enough data in first
215  * packet leaves a multiple of 512 bytes of data in remainder.
216  * Assumes that ifp is actually pointer to arpcom structure.
217  */
218 static int
219 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
220 	     struct rtentry *rt)
221 {
222 	struct ether_header *eh, *deh;
223 	u_char *edst;
224 	int loop_copy = 0;
225 	int hlen = ETHER_HDR_LEN;	/* link layer header length */
226 	struct arpcom *ac = IFP2AC(ifp);
227 	int error;
228 
229 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
230 
231 	if (ifp->if_flags & IFF_MONITOR)
232 		gotoerr(ENETDOWN);
233 	if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
234 		gotoerr(ENETDOWN);
235 
236 	M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
237 	if (m == NULL)
238 		return (ENOBUFS);
239 	m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
240 	eh = mtod(m, struct ether_header *);
241 	edst = eh->ether_dhost;
242 
243 	/*
244 	 * Fill in the destination ethernet address and frame type.
245 	 */
246 	switch (dst->sa_family) {
247 #ifdef INET
248 	case AF_INET:
249 		if (!arpresolve(ifp, rt, m, dst, edst))
250 			return (0);	/* if not yet resolved */
251 #ifdef MPLS
252 		if (m->m_flags & M_MPLSLABELED)
253 			eh->ether_type = htons(ETHERTYPE_MPLS);
254 		else
255 #endif
256 			eh->ether_type = htons(ETHERTYPE_IP);
257 		break;
258 #endif
259 #ifdef INET6
260 	case AF_INET6:
261 		if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
262 			return (0);		/* Something bad happenned. */
263 		eh->ether_type = htons(ETHERTYPE_IPV6);
264 		break;
265 #endif
266 #ifdef IPX
267 	case AF_IPX:
268 		if (ef_outputp != NULL) {
269 			/*
270 			 * Hold BGL and recheck ef_outputp
271 			 */
272 			get_mplock();
273 			if (ef_outputp != NULL) {
274 				error = ef_outputp(ifp, &m, dst,
275 						   &eh->ether_type, &hlen);
276 				rel_mplock();
277 				if (error)
278 					goto bad;
279 				else
280 					break;
281 			}
282 			rel_mplock();
283 		}
284 		eh->ether_type = htons(ETHERTYPE_IPX);
285 		bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
286 		      edst, ETHER_ADDR_LEN);
287 		break;
288 #endif
289 	case pseudo_AF_HDRCMPLT:
290 	case AF_UNSPEC:
291 		loop_copy = -1; /* if this is for us, don't do it */
292 		deh = (struct ether_header *)dst->sa_data;
293 		memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
294 		eh->ether_type = deh->ether_type;
295 		break;
296 
297 	default:
298 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
299 		gotoerr(EAFNOSUPPORT);
300 	}
301 
302 	if (dst->sa_family == pseudo_AF_HDRCMPLT)	/* unlikely */
303 		memcpy(eh->ether_shost,
304 		       ((struct ether_header *)dst->sa_data)->ether_shost,
305 		       ETHER_ADDR_LEN);
306 	else
307 		memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
308 
309 	/*
310 	 * Bridges require special output handling.
311 	 */
312 	if (ifp->if_bridge) {
313 		KASSERT(bridge_output_p != NULL,
314 			("%s: if_bridge not loaded!", __func__));
315 		return bridge_output_p(ifp, m);
316 	}
317 
318 	/*
319 	 * If a simplex interface, and the packet is being sent to our
320 	 * Ethernet address or a broadcast address, loopback a copy.
321 	 * XXX To make a simplex device behave exactly like a duplex
322 	 * device, we should copy in the case of sending to our own
323 	 * ethernet address (thus letting the original actually appear
324 	 * on the wire). However, we don't do that here for security
325 	 * reasons and compatibility with the original behavior.
326 	 */
327 	if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
328 		int csum_flags = 0;
329 
330 		if (m->m_pkthdr.csum_flags & CSUM_IP)
331 			csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
332 		if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
333 			csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
334 		if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
335 			struct mbuf *n;
336 
337 			if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
338 				n->m_pkthdr.csum_flags |= csum_flags;
339 				if (csum_flags & CSUM_DATA_VALID)
340 					n->m_pkthdr.csum_data = 0xffff;
341 				if_simloop(ifp, n, dst->sa_family, hlen);
342 			} else
343 				IFNET_STAT_INC(ifp, iqdrops, 1);
344 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
345 				ETHER_ADDR_LEN) == 0) {
346 			m->m_pkthdr.csum_flags |= csum_flags;
347 			if (csum_flags & CSUM_DATA_VALID)
348 				m->m_pkthdr.csum_data = 0xffff;
349 			if_simloop(ifp, m, dst->sa_family, hlen);
350 			return (0);	/* XXX */
351 		}
352 	}
353 
354 #ifdef CARP
355 	if (ifp->if_type == IFT_CARP) {
356 		ifp = carp_parent(ifp);
357 		if (ifp == NULL)
358 			gotoerr(ENETUNREACH);
359 
360 		ac = IFP2AC(ifp);
361 
362 		/*
363 		 * Check precondition again
364 		 */
365 		ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
366 
367 		if (ifp->if_flags & IFF_MONITOR)
368 			gotoerr(ENETDOWN);
369 		if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
370 		    (IFF_UP | IFF_RUNNING))
371 			gotoerr(ENETDOWN);
372 	}
373 #endif
374 
375 	/* Handle ng_ether(4) processing, if any */
376 	if (ng_ether_output_p != NULL) {
377 		/*
378 		 * Hold BGL and recheck ng_ether_output_p
379 		 */
380 		get_mplock();
381 		if (ng_ether_output_p != NULL) {
382 			if ((error = ng_ether_output_p(ifp, &m)) != 0) {
383 				rel_mplock();
384 				goto bad;
385 			}
386 			if (m == NULL) {
387 				rel_mplock();
388 				return (0);
389 			}
390 		}
391 		rel_mplock();
392 	}
393 
394 	/* Continue with link-layer output */
395 	return ether_output_frame(ifp, m);
396 
397 bad:
398 	m_freem(m);
399 	return (error);
400 }
401 
402 /*
403  * Returns the bridge interface an ifp is associated
404  * with.
405  *
406  * Only call if ifp->if_bridge != NULL.
407  */
408 struct ifnet *
409 ether_bridge_interface(struct ifnet *ifp)
410 {
411 	if (bridge_interface_p)
412 		return(bridge_interface_p(ifp->if_bridge));
413 	return (ifp);
414 }
415 
416 /*
417  * Ethernet link layer output routine to send a raw frame to the device.
418  *
419  * This assumes that the 14 byte Ethernet header is present and contiguous
420  * in the first mbuf.
421  */
422 int
423 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
424 {
425 	struct ip_fw *rule = NULL;
426 	int error = 0;
427 	struct altq_pktattr pktattr;
428 
429 	ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
430 
431 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
432 		struct m_tag *mtag;
433 
434 		/* Extract info from dummynet tag */
435 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
436 		KKASSERT(mtag != NULL);
437 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
438 		KKASSERT(rule != NULL);
439 
440 		m_tag_delete(m, mtag);
441 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
442 	}
443 
444 	if (ifq_is_enabled(&ifp->if_snd))
445 		altq_etherclassify(&ifp->if_snd, m, &pktattr);
446 	crit_enter();
447 	if (IPFW_LOADED && ether_ipfw != 0) {
448 		struct ether_header save_eh, *eh;
449 
450 		eh = mtod(m, struct ether_header *);
451 		save_eh = *eh;
452 		m_adj(m, ETHER_HDR_LEN);
453 		if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
454 			crit_exit();
455 			if (m != NULL) {
456 				m_freem(m);
457 				return ENOBUFS; /* pkt dropped */
458 			} else
459 				return 0;	/* consumed e.g. in a pipe */
460 		}
461 
462 		/* packet was ok, restore the ethernet header */
463 		ether_restore_header(&m, eh, &save_eh);
464 		if (m == NULL) {
465 			crit_exit();
466 			return ENOBUFS;
467 		}
468 	}
469 	crit_exit();
470 
471 	/*
472 	 * Queue message on interface, update output statistics if
473 	 * successful, and start output if interface not yet active.
474 	 */
475 	error = ifq_dispatch(ifp, m, &pktattr);
476 	return (error);
477 }
478 
479 /*
480  * ipfw processing for ethernet packets (in and out).
481  * The second parameter is NULL from ether_demux(), and ifp from
482  * ether_output_frame().
483  */
484 static boolean_t
485 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
486 	       const struct ether_header *eh)
487 {
488 	struct ether_header save_eh = *eh;	/* might be a ptr in *m0 */
489 	struct ip_fw_args args;
490 	struct m_tag *mtag;
491 	struct mbuf *m;
492 	int i;
493 
494 	if (*rule != NULL && fw_one_pass)
495 		return TRUE; /* dummynet packet, already partially processed */
496 
497 	/*
498 	 * I need some amount of data to be contiguous.
499 	 */
500 	i = min((*m0)->m_pkthdr.len, max_protohdr);
501 	if ((*m0)->m_len < i) {
502 		*m0 = m_pullup(*m0, i);
503 		if (*m0 == NULL)
504 			return FALSE;
505 	}
506 
507 	/*
508 	 * Clean up tags
509 	 */
510 	if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
511 		m_tag_delete(*m0, mtag);
512 	if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
513 		mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
514 		KKASSERT(mtag != NULL);
515 		m_tag_delete(*m0, mtag);
516 		(*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
517 	}
518 
519 	args.m = *m0;		/* the packet we are looking at		*/
520 	args.oif = dst;		/* destination, if any			*/
521 	args.rule = *rule;	/* matching rule to restart		*/
522 	args.eh = &save_eh;	/* MAC header for bridged/MAC packets	*/
523 	i = ip_fw_chk_ptr(&args);
524 	*m0 = args.m;
525 	*rule = args.rule;
526 
527 	if (*m0 == NULL)
528 		return FALSE;
529 
530 	switch (i) {
531 	case IP_FW_PASS:
532 		return TRUE;
533 
534 	case IP_FW_DIVERT:
535 	case IP_FW_TEE:
536 	case IP_FW_DENY:
537 		/*
538 		 * XXX at some point add support for divert/forward actions.
539 		 * If none of the above matches, we have to drop the pkt.
540 		 */
541 		return FALSE;
542 
543 	case IP_FW_DUMMYNET:
544 		/*
545 		 * Pass the pkt to dummynet, which consumes it.
546 		 */
547 		m = *m0;	/* pass the original to dummynet */
548 		*m0 = NULL;	/* and nothing back to the caller */
549 
550 		ether_restore_header(&m, eh, &save_eh);
551 		if (m == NULL)
552 			return FALSE;
553 
554 		ip_fw_dn_io_ptr(m, args.cookie,
555 				dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
556 		ip_dn_queue(m);
557 		return FALSE;
558 
559 	default:
560 		panic("unknown ipfw return value: %d", i);
561 	}
562 }
563 
564 static void
565 ether_input(struct ifnet *ifp, struct mbuf *m)
566 {
567 	ether_input_pkt(ifp, m, NULL);
568 }
569 
570 /*
571  * Perform common duties while attaching to interface list
572  */
573 void
574 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
575 {
576 	ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
577 			   serializer);
578 }
579 
580 void
581 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
582 		   lwkt_serialize_t serializer)
583 {
584 	struct sockaddr_dl *sdl;
585 	char ethstr[ETHER_ADDRSTRLEN + 1];
586 
587 	ifp->if_type = IFT_ETHER;
588 	ifp->if_addrlen = ETHER_ADDR_LEN;
589 	ifp->if_hdrlen = ETHER_HDR_LEN;
590 	if_attach(ifp, serializer);
591 	ifp->if_mtu = ETHERMTU;
592 	if (ifp->if_tsolen <= 0) {
593 		if ((ether_tsolen_default / ETHERMTU) < 2) {
594 			kprintf("ether TSO maxlen %d -> %d\n",
595 			    ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
596 			ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
597 		}
598 		ifp->if_tsolen = ether_tsolen_default;
599 	}
600 	if (ifp->if_baudrate == 0)
601 		ifp->if_baudrate = 10000000;
602 	ifp->if_output = ether_output;
603 	ifp->if_input = ether_input;
604 	ifp->if_resolvemulti = ether_resolvemulti;
605 	ifp->if_broadcastaddr = etherbroadcastaddr;
606 	sdl = IF_LLSOCKADDR(ifp);
607 	sdl->sdl_type = IFT_ETHER;
608 	sdl->sdl_alen = ifp->if_addrlen;
609 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
610 	/*
611 	 * XXX Keep the current drivers happy.
612 	 * XXX Remove once all drivers have been cleaned up
613 	 */
614 	if (lla != IFP2AC(ifp)->ac_enaddr)
615 		bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
616 	bpfattach(ifp, dlt, hdrlen);
617 	if (ng_ether_attach_p != NULL)
618 		(*ng_ether_attach_p)(ifp);
619 
620 	if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
621 }
622 
623 /*
624  * Perform common duties while detaching an Ethernet interface
625  */
626 void
627 ether_ifdetach(struct ifnet *ifp)
628 {
629 	if_down(ifp);
630 
631 	if (ng_ether_detach_p != NULL)
632 		(*ng_ether_detach_p)(ifp);
633 	bpfdetach(ifp);
634 	if_detach(ifp);
635 }
636 
637 int
638 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
639 {
640 	struct ifaddr *ifa = (struct ifaddr *) data;
641 	struct ifreq *ifr = (struct ifreq *) data;
642 	int error = 0;
643 
644 #define IF_INIT(ifp) \
645 do { \
646 	if (((ifp)->if_flags & IFF_UP) == 0) { \
647 		(ifp)->if_flags |= IFF_UP; \
648 		(ifp)->if_init((ifp)->if_softc); \
649 	} \
650 } while (0)
651 
652 	ASSERT_IFNET_SERIALIZED_ALL(ifp);
653 
654 	switch (command) {
655 	case SIOCSIFADDR:
656 		switch (ifa->ifa_addr->sa_family) {
657 #ifdef INET
658 		case AF_INET:
659 			IF_INIT(ifp);	/* before arpwhohas */
660 			arp_ifinit(ifp, ifa);
661 			break;
662 #endif
663 #ifdef IPX
664 		/*
665 		 * XXX - This code is probably wrong
666 		 */
667 		case AF_IPX:
668 			{
669 			struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
670 			struct arpcom *ac = IFP2AC(ifp);
671 
672 			if (ipx_nullhost(*ina))
673 				ina->x_host = *(union ipx_host *) ac->ac_enaddr;
674 			else
675 				bcopy(ina->x_host.c_host, ac->ac_enaddr,
676 				      sizeof ac->ac_enaddr);
677 
678 			IF_INIT(ifp);	/* Set new address. */
679 			break;
680 			}
681 #endif
682 		default:
683 			IF_INIT(ifp);
684 			break;
685 		}
686 		break;
687 
688 	case SIOCGIFADDR:
689 		bcopy(IFP2AC(ifp)->ac_enaddr,
690 		      ((struct sockaddr *)ifr->ifr_data)->sa_data,
691 		      ETHER_ADDR_LEN);
692 		break;
693 
694 	case SIOCSIFMTU:
695 		/*
696 		 * Set the interface MTU.
697 		 */
698 		if (ifr->ifr_mtu > ETHERMTU) {
699 			error = EINVAL;
700 		} else {
701 			ifp->if_mtu = ifr->ifr_mtu;
702 		}
703 		break;
704 	default:
705 		error = EINVAL;
706 		break;
707 	}
708 	return (error);
709 
710 #undef IF_INIT
711 }
712 
713 int
714 ether_resolvemulti(
715 	struct ifnet *ifp,
716 	struct sockaddr **llsa,
717 	struct sockaddr *sa)
718 {
719 	struct sockaddr_dl *sdl;
720 #ifdef INET
721 	struct sockaddr_in *sin;
722 #endif
723 #ifdef INET6
724 	struct sockaddr_in6 *sin6;
725 #endif
726 	u_char *e_addr;
727 
728 	switch(sa->sa_family) {
729 	case AF_LINK:
730 		/*
731 		 * No mapping needed. Just check that it's a valid MC address.
732 		 */
733 		sdl = (struct sockaddr_dl *)sa;
734 		e_addr = LLADDR(sdl);
735 		if ((e_addr[0] & 1) != 1)
736 			return EADDRNOTAVAIL;
737 		*llsa = NULL;
738 		return 0;
739 
740 #ifdef INET
741 	case AF_INET:
742 		sin = (struct sockaddr_in *)sa;
743 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
744 			return EADDRNOTAVAIL;
745 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
746 		sdl->sdl_len = sizeof *sdl;
747 		sdl->sdl_family = AF_LINK;
748 		sdl->sdl_index = ifp->if_index;
749 		sdl->sdl_type = IFT_ETHER;
750 		sdl->sdl_alen = ETHER_ADDR_LEN;
751 		e_addr = LLADDR(sdl);
752 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
753 		*llsa = (struct sockaddr *)sdl;
754 		return 0;
755 #endif
756 #ifdef INET6
757 	case AF_INET6:
758 		sin6 = (struct sockaddr_in6 *)sa;
759 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
760 			/*
761 			 * An IP6 address of 0 means listen to all
762 			 * of the Ethernet multicast address used for IP6.
763 			 * (This is used for multicast routers.)
764 			 */
765 			ifp->if_flags |= IFF_ALLMULTI;
766 			*llsa = NULL;
767 			return 0;
768 		}
769 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
770 			return EADDRNOTAVAIL;
771 		sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
772 		sdl->sdl_len = sizeof *sdl;
773 		sdl->sdl_family = AF_LINK;
774 		sdl->sdl_index = ifp->if_index;
775 		sdl->sdl_type = IFT_ETHER;
776 		sdl->sdl_alen = ETHER_ADDR_LEN;
777 		e_addr = LLADDR(sdl);
778 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
779 		*llsa = (struct sockaddr *)sdl;
780 		return 0;
781 #endif
782 
783 	default:
784 		/*
785 		 * Well, the text isn't quite right, but it's the name
786 		 * that counts...
787 		 */
788 		return EAFNOSUPPORT;
789 	}
790 }
791 
792 #if 0
793 /*
794  * This is for reference.  We have a table-driven version
795  * of the little-endian crc32 generator, which is faster
796  * than the double-loop.
797  */
798 uint32_t
799 ether_crc32_le(const uint8_t *buf, size_t len)
800 {
801 	uint32_t c, crc, carry;
802 	size_t i, j;
803 
804 	crc = 0xffffffffU;	/* initial value */
805 
806 	for (i = 0; i < len; i++) {
807 		c = buf[i];
808 		for (j = 0; j < 8; j++) {
809 			carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
810 			crc >>= 1;
811 			c >>= 1;
812 			if (carry)
813 				crc = (crc ^ ETHER_CRC_POLY_LE);
814 		}
815 	}
816 
817 	return (crc);
818 }
819 #else
820 uint32_t
821 ether_crc32_le(const uint8_t *buf, size_t len)
822 {
823 	static const uint32_t crctab[] = {
824 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
825 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
826 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
827 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
828 	};
829 	uint32_t crc;
830 	size_t i;
831 
832 	crc = 0xffffffffU;	/* initial value */
833 
834 	for (i = 0; i < len; i++) {
835 		crc ^= buf[i];
836 		crc = (crc >> 4) ^ crctab[crc & 0xf];
837 		crc = (crc >> 4) ^ crctab[crc & 0xf];
838 	}
839 
840 	return (crc);
841 }
842 #endif
843 
844 uint32_t
845 ether_crc32_be(const uint8_t *buf, size_t len)
846 {
847 	uint32_t c, crc, carry;
848 	size_t i, j;
849 
850 	crc = 0xffffffffU;	/* initial value */
851 
852 	for (i = 0; i < len; i++) {
853 		c = buf[i];
854 		for (j = 0; j < 8; j++) {
855 			carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
856 			crc <<= 1;
857 			c >>= 1;
858 			if (carry)
859 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
860 		}
861 	}
862 
863 	return (crc);
864 }
865 
866 /*
867  * find the size of ethernet header, and call classifier
868  */
869 void
870 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
871 		   struct altq_pktattr *pktattr)
872 {
873 	struct ether_header *eh;
874 	uint16_t ether_type;
875 	int hlen, af, hdrsize;
876 
877 	hlen = sizeof(struct ether_header);
878 	eh = mtod(m, struct ether_header *);
879 
880 	ether_type = ntohs(eh->ether_type);
881 	if (ether_type < ETHERMTU) {
882 		/* ick! LLC/SNAP */
883 		struct llc *llc = (struct llc *)(eh + 1);
884 		hlen += 8;
885 
886 		if (m->m_len < hlen ||
887 		    llc->llc_dsap != LLC_SNAP_LSAP ||
888 		    llc->llc_ssap != LLC_SNAP_LSAP ||
889 		    llc->llc_control != LLC_UI)
890 			goto bad;  /* not snap! */
891 
892 		ether_type = ntohs(llc->llc_un.type_snap.ether_type);
893 	}
894 
895 	if (ether_type == ETHERTYPE_IP) {
896 		af = AF_INET;
897 		hdrsize = 20;  /* sizeof(struct ip) */
898 #ifdef INET6
899 	} else if (ether_type == ETHERTYPE_IPV6) {
900 		af = AF_INET6;
901 		hdrsize = 40;  /* sizeof(struct ip6_hdr) */
902 #endif
903 	} else
904 		goto bad;
905 
906 	while (m->m_len <= hlen) {
907 		hlen -= m->m_len;
908 		m = m->m_next;
909 	}
910 	if (m->m_len < hlen + hdrsize) {
911 		/*
912 		 * ip header is not in a single mbuf.  this should not
913 		 * happen in the current code.
914 		 * (todo: use m_pulldown in the future)
915 		 */
916 		goto bad;
917 	}
918 	m->m_data += hlen;
919 	m->m_len -= hlen;
920 	ifq_classify(ifq, m, af, pktattr);
921 	m->m_data -= hlen;
922 	m->m_len += hlen;
923 
924 	return;
925 
926 bad:
927 	pktattr->pattr_class = NULL;
928 	pktattr->pattr_hdr = NULL;
929 	pktattr->pattr_af = AF_UNSPEC;
930 }
931 
932 static void
933 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
934 		     const struct ether_header *save_eh)
935 {
936 	struct mbuf *m = *m0;
937 
938 	ether_restore_hdr++;
939 
940 	/*
941 	 * Prepend the header, optimize for the common case of
942 	 * eh pointing into the mbuf.
943 	 */
944 	if ((const void *)(eh + 1) == (void *)m->m_data) {
945 		m->m_data -= ETHER_HDR_LEN;
946 		m->m_len += ETHER_HDR_LEN;
947 		m->m_pkthdr.len += ETHER_HDR_LEN;
948 	} else {
949 		ether_prepend_hdr++;
950 
951 		M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
952 		if (m != NULL) {
953 			bcopy(save_eh, mtod(m, struct ether_header *),
954 			      ETHER_HDR_LEN);
955 		}
956 	}
957 	*m0 = m;
958 }
959 
960 /*
961  * Upper layer processing for a received Ethernet packet.
962  */
963 void
964 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
965 {
966 	struct ether_header *eh;
967 	int isr, discard = 0;
968 	u_short ether_type;
969 	struct ip_fw *rule = NULL;
970 
971 	M_ASSERTPKTHDR(m);
972 	KASSERT(m->m_len >= ETHER_HDR_LEN,
973 		("ether header is not contiguous!"));
974 
975 	eh = mtod(m, struct ether_header *);
976 
977 	if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
978 		struct m_tag *mtag;
979 
980 		/* Extract info from dummynet tag */
981 		mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
982 		KKASSERT(mtag != NULL);
983 		rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
984 		KKASSERT(rule != NULL);
985 
986 		m_tag_delete(m, mtag);
987 		m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
988 
989 		/* packet is passing the second time */
990 		goto post_stats;
991 	}
992 
993 	/*
994 	 * We got a packet which was unicast to a different Ethernet
995 	 * address.  If the driver is working properly, then this
996 	 * situation can only happen when the interface is in
997 	 * promiscuous mode.  We defer the packet discarding until the
998 	 * vlan processing is done, so that vlan/bridge or vlan/netgraph
999 	 * could work.
1000 	 */
1001 	if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1002 	    !ETHER_IS_MULTICAST(eh->ether_dhost) &&
1003 	    bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1004 		if (ether_debug & 1) {
1005 			kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
1006 				"%02x:%02x:%02x:%02x:%02x:%02x "
1007 				"%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
1008 				eh->ether_dhost[0],
1009 				eh->ether_dhost[1],
1010 				eh->ether_dhost[2],
1011 				eh->ether_dhost[3],
1012 				eh->ether_dhost[4],
1013 				eh->ether_dhost[5],
1014 				eh->ether_shost[0],
1015 				eh->ether_shost[1],
1016 				eh->ether_shost[2],
1017 				eh->ether_shost[3],
1018 				eh->ether_shost[4],
1019 				eh->ether_shost[5],
1020 				eh->ether_type,
1021 				((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1022 				((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1023 				((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1024 				((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1025 				((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1026 				((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1027 			);
1028 		}
1029 		if ((ether_debug & 2) == 0)
1030 			discard = 1;
1031 	}
1032 
1033 post_stats:
1034 	if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
1035 		struct ether_header save_eh = *eh;
1036 
1037 		/* XXX old crufty stuff, needs to be removed */
1038 		m_adj(m, sizeof(struct ether_header));
1039 
1040 		if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1041 			m_freem(m);
1042 			return;
1043 		}
1044 
1045 		ether_restore_header(&m, eh, &save_eh);
1046 		if (m == NULL)
1047 			return;
1048 		eh = mtod(m, struct ether_header *);
1049 	}
1050 
1051 	ether_type = ntohs(eh->ether_type);
1052 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1053 
1054 	if (m->m_flags & M_VLANTAG) {
1055 		void (*vlan_input_func)(struct mbuf *);
1056 
1057 		vlan_input_func = vlan_input_p;
1058 		if (vlan_input_func != NULL) {
1059 			vlan_input_func(m);
1060 		} else {
1061 			IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1062 			m_freem(m);
1063 		}
1064 		return;
1065 	}
1066 
1067 	/*
1068 	 * If we have been asked to discard this packet
1069 	 * (e.g. not for us), drop it before entering
1070 	 * the upper layer.
1071 	 */
1072 	if (discard) {
1073 		m_freem(m);
1074 		return;
1075 	}
1076 
1077 	/*
1078 	 * Clear protocol specific flags,
1079 	 * before entering the upper layer.
1080 	 */
1081 	m->m_flags &= ~M_ETHER_FLAGS;
1082 
1083 	/* Strip ethernet header. */
1084 	m_adj(m, sizeof(struct ether_header));
1085 
1086 	switch (ether_type) {
1087 #ifdef INET
1088 	case ETHERTYPE_IP:
1089 		if ((m->m_flags & M_LENCHECKED) == 0) {
1090 			if (!ip_lengthcheck(&m, 0))
1091 				return;
1092 		}
1093 		if (ipflow_fastforward(m))
1094 			return;
1095 		isr = NETISR_IP;
1096 		break;
1097 
1098 	case ETHERTYPE_ARP:
1099 		if (ifp->if_flags & IFF_NOARP) {
1100 			/* Discard packet if ARP is disabled on interface */
1101 			m_freem(m);
1102 			return;
1103 		}
1104 		isr = NETISR_ARP;
1105 		break;
1106 #endif
1107 
1108 #ifdef INET6
1109 	case ETHERTYPE_IPV6:
1110 		isr = NETISR_IPV6;
1111 		break;
1112 #endif
1113 
1114 #ifdef IPX
1115 	case ETHERTYPE_IPX:
1116 		if (ef_inputp) {
1117 			/*
1118 			 * Hold BGL and recheck ef_inputp
1119 			 */
1120 			get_mplock();
1121 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1122 				rel_mplock();
1123 				return;
1124 			}
1125 			rel_mplock();
1126 		}
1127 		isr = NETISR_IPX;
1128 		break;
1129 #endif
1130 
1131 #ifdef MPLS
1132 	case ETHERTYPE_MPLS:
1133 	case ETHERTYPE_MPLS_MCAST:
1134 		/* Should have been set by ether_input_pkt(). */
1135 		KKASSERT(m->m_flags & M_MPLSLABELED);
1136 		isr = NETISR_MPLS;
1137 		break;
1138 #endif
1139 
1140 	default:
1141 		/*
1142 		 * The accurate msgport is not determined before
1143 		 * we reach here, so recharacterize packet.
1144 		 */
1145 		m->m_flags &= ~M_HASH;
1146 #ifdef IPX
1147 		if (ef_inputp) {
1148 			/*
1149 			 * Hold BGL and recheck ef_inputp
1150 			 */
1151 			get_mplock();
1152 			if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1153 				rel_mplock();
1154 				return;
1155 			}
1156 			rel_mplock();
1157 		}
1158 #endif
1159 		if (ng_ether_input_orphan_p != NULL) {
1160 			/*
1161 			 * Put back the ethernet header so netgraph has a
1162 			 * consistent view of inbound packets.
1163 			 */
1164 			M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1165 			if (m == NULL) {
1166 				/*
1167 				 * M_PREPEND frees the mbuf in case of failure.
1168 				 */
1169 				return;
1170 			}
1171 			/*
1172 			 * Hold BGL and recheck ng_ether_input_orphan_p
1173 			 */
1174 			get_mplock();
1175 			if (ng_ether_input_orphan_p != NULL) {
1176 				ng_ether_input_orphan_p(ifp, m);
1177 				rel_mplock();
1178 				return;
1179 			}
1180 			rel_mplock();
1181 		}
1182 		m_freem(m);
1183 		return;
1184 	}
1185 
1186 	if (m->m_flags & M_HASH) {
1187 		if (&curthread->td_msgport ==
1188 		    netisr_hashport(m->m_pkthdr.hash)) {
1189 			netisr_handle(isr, m);
1190 			return;
1191 		} else {
1192 			/*
1193 			 * XXX Something is wrong,
1194 			 * we probably should panic here!
1195 			 */
1196 			m->m_flags &= ~M_HASH;
1197 			atomic_add_long(&ether_input_wronghash, 1);
1198 		}
1199 	}
1200 #ifdef RSS_DEBUG
1201 	atomic_add_long(&ether_input_requeue, 1);
1202 #endif
1203 	netisr_queue(isr, m);
1204 }
1205 
1206 /*
1207  * First we perform any link layer operations, then continue to the
1208  * upper layers with ether_demux_oncpu().
1209  */
1210 static void
1211 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1212 {
1213 #ifdef CARP
1214 	void *carp;
1215 #endif
1216 
1217 	if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1218 		/*
1219 		 * Receiving interface's flags are changed, when this
1220 		 * packet is waiting for processing; discard it.
1221 		 */
1222 		m_freem(m);
1223 		return;
1224 	}
1225 
1226 	/*
1227 	 * Tap the packet off here for a bridge.  bridge_input()
1228 	 * will return NULL if it has consumed the packet, otherwise
1229 	 * it gets processed as normal.  Note that bridge_input()
1230 	 * will always return the original packet if we need to
1231 	 * process it locally.
1232 	 */
1233 	if (ifp->if_bridge) {
1234 		KASSERT(bridge_input_p != NULL,
1235 			("%s: if_bridge not loaded!", __func__));
1236 
1237 		if(m->m_flags & M_ETHER_BRIDGED) {
1238 			m->m_flags &= ~M_ETHER_BRIDGED;
1239 		} else {
1240 			m = bridge_input_p(ifp, m);
1241 			if (m == NULL)
1242 				return;
1243 
1244 			KASSERT(ifp == m->m_pkthdr.rcvif,
1245 				("bridge_input_p changed rcvif"));
1246 		}
1247 	}
1248 
1249 #ifdef CARP
1250 	carp = ifp->if_carp;
1251 	if (carp) {
1252 		m = carp_input(carp, m);
1253 		if (m == NULL)
1254 			return;
1255 		KASSERT(ifp == m->m_pkthdr.rcvif,
1256 		    ("carp_input changed rcvif"));
1257 	}
1258 #endif
1259 
1260 	/* Handle ng_ether(4) processing, if any */
1261 	if (ng_ether_input_p != NULL) {
1262 		/*
1263 		 * Hold BGL and recheck ng_ether_input_p
1264 		 */
1265 		get_mplock();
1266 		if (ng_ether_input_p != NULL)
1267 			ng_ether_input_p(ifp, &m);
1268 		rel_mplock();
1269 
1270 		if (m == NULL)
1271 			return;
1272 	}
1273 
1274 	/* Continue with upper layer processing */
1275 	ether_demux_oncpu(ifp, m);
1276 }
1277 
1278 /*
1279  * Perform certain functions of ether_input_pkt():
1280  * - Test IFF_UP
1281  * - Update statistics
1282  * - Run bpf(4) tap if requested
1283  * Then pass the packet to ether_input_oncpu().
1284  *
1285  * This function should be used by pseudo interface (e.g. vlan(4)),
1286  * when it tries to claim that the packet is received by it.
1287  *
1288  * REINPUT_KEEPRCVIF
1289  * REINPUT_RUNBPF
1290  */
1291 void
1292 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1293 {
1294 	/* Discard packet if interface is not up */
1295 	if (!(ifp->if_flags & IFF_UP)) {
1296 		m_freem(m);
1297 		return;
1298 	}
1299 
1300 	/*
1301 	 * Change receiving interface.  The bridge will often pass a flag to
1302 	 * ask that this not be done so ARPs get applied to the correct
1303 	 * side.
1304 	 */
1305 	if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1306 	    m->m_pkthdr.rcvif == NULL) {
1307 		m->m_pkthdr.rcvif = ifp;
1308 	}
1309 
1310 	/* Update statistics */
1311 	IFNET_STAT_INC(ifp, ipackets, 1);
1312 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1313 	if (m->m_flags & (M_MCAST | M_BCAST))
1314 		IFNET_STAT_INC(ifp, imcasts, 1);
1315 
1316 	if (reinput_flags & REINPUT_RUNBPF)
1317 		BPF_MTAP(ifp, m);
1318 
1319 	ether_input_oncpu(ifp, m);
1320 }
1321 
1322 static __inline boolean_t
1323 ether_vlancheck(struct mbuf **m0)
1324 {
1325 	struct mbuf *m = *m0;
1326 	struct ether_header *eh;
1327 	uint16_t ether_type;
1328 
1329 	eh = mtod(m, struct ether_header *);
1330 	ether_type = ntohs(eh->ether_type);
1331 
1332 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1333 		/*
1334 		 * Extract vlan tag if hardware does not do it for us
1335 		 */
1336 		vlan_ether_decap(&m);
1337 		if (m == NULL)
1338 			goto failed;
1339 
1340 		eh = mtod(m, struct ether_header *);
1341 		ether_type = ntohs(eh->ether_type);
1342 	}
1343 
1344 	if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1345 		/*
1346 		 * To prevent possible dangerous recursion,
1347 		 * we don't do vlan-in-vlan
1348 		 */
1349 		IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1350 		goto failed;
1351 	}
1352 	KKASSERT(ether_type != ETHERTYPE_VLAN);
1353 
1354 	m->m_flags |= M_ETHER_VLANCHECKED;
1355 	*m0 = m;
1356 	return TRUE;
1357 failed:
1358 	if (m != NULL)
1359 		m_freem(m);
1360 	*m0 = NULL;
1361 	return FALSE;
1362 }
1363 
1364 static void
1365 ether_input_handler(netmsg_t nmsg)
1366 {
1367 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1368 	struct ether_header *eh;
1369 	struct ifnet *ifp;
1370 	struct mbuf *m;
1371 
1372 	m = nmp->nm_packet;
1373 	M_ASSERTPKTHDR(m);
1374 
1375 	if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1376 		if (!ether_vlancheck(&m)) {
1377 			KKASSERT(m == NULL);
1378 			return;
1379 		}
1380 	}
1381 	if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1382 	    __predict_false(ether_input_ckhash)) {
1383 		int isr;
1384 
1385 		/*
1386 		 * Need to verify the hash supplied by the hardware
1387 		 * which could be wrong.
1388 		 */
1389 		m->m_flags &= ~(M_HASH | M_CKHASH);
1390 		isr = ether_characterize(&m);
1391 		if (m == NULL)
1392 			return;
1393 		KKASSERT(m->m_flags & M_HASH);
1394 
1395 		if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1396 			/*
1397 			 * Wrong hardware supplied hash; redispatch
1398 			 */
1399 			ether_dispatch(isr, m);
1400 			if (__predict_false(ether_input_ckhash))
1401 				atomic_add_long(&ether_input_wronghwhash, 1);
1402 			return;
1403 		}
1404 	}
1405 	ifp = m->m_pkthdr.rcvif;
1406 
1407 	eh = mtod(m, struct ether_header *);
1408 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1409 		if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1410 			 ifp->if_addrlen) == 0)
1411 			m->m_flags |= M_BCAST;
1412 		else
1413 			m->m_flags |= M_MCAST;
1414 		IFNET_STAT_INC(ifp, imcasts, 1);
1415 	}
1416 
1417 	ether_input_oncpu(ifp, m);
1418 }
1419 
1420 /*
1421  * Send the packet to the target msgport
1422  *
1423  * At this point the packet had better be characterized (M_HASH set),
1424  * so we know which cpu to send it to.
1425  */
1426 static void
1427 ether_dispatch(int isr, struct mbuf *m)
1428 {
1429 	struct netmsg_packet *pmsg;
1430 
1431 	KKASSERT(m->m_flags & M_HASH);
1432 	pmsg = &m->m_hdr.mh_netmsg;
1433 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1434 		    0, ether_input_handler);
1435 	pmsg->nm_packet = m;
1436 	pmsg->base.lmsg.u.ms_result = isr;
1437 
1438 	logether(disp_beg, NULL);
1439 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1440 	logether(disp_end, NULL);
1441 }
1442 
1443 /*
1444  * Process a received Ethernet packet.
1445  *
1446  * The ethernet header is assumed to be in the mbuf so the caller
1447  * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1448  * bytes in the first mbuf.
1449  */
1450 void
1451 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi)
1452 {
1453 	int isr;
1454 
1455 	M_ASSERTPKTHDR(m);
1456 
1457 	/* Discard packet if interface is not up */
1458 	if (!(ifp->if_flags & IFF_UP)) {
1459 		m_freem(m);
1460 		return;
1461 	}
1462 
1463 	if (m->m_len < sizeof(struct ether_header)) {
1464 		/* XXX error in the caller. */
1465 		m_freem(m);
1466 		return;
1467 	}
1468 
1469 	m->m_pkthdr.rcvif = ifp;
1470 
1471 	logether(pkt_beg, ifp);
1472 
1473 	ETHER_BPF_MTAP(ifp, m);
1474 
1475 	IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1476 
1477 	if (ifp->if_flags & IFF_MONITOR) {
1478 		struct ether_header *eh;
1479 
1480 		eh = mtod(m, struct ether_header *);
1481 		if (ETHER_IS_MULTICAST(eh->ether_dhost))
1482 			IFNET_STAT_INC(ifp, imcasts, 1);
1483 
1484 		/*
1485 		 * Interface marked for monitoring; discard packet.
1486 		 */
1487 		m_freem(m);
1488 
1489 		logether(pkt_end, ifp);
1490 		return;
1491 	}
1492 
1493 	/*
1494 	 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1495 	 * we can dispatch it immediately without further inspection.
1496 	 */
1497 	if (pi != NULL && (m->m_flags & M_HASH)) {
1498 #ifdef RSS_DEBUG
1499 		atomic_add_long(&ether_pktinfo_try, 1);
1500 #endif
1501 		netisr_hashcheck(pi->pi_netisr, m, pi);
1502 		if (m->m_flags & M_HASH) {
1503 			ether_dispatch(pi->pi_netisr, m);
1504 #ifdef RSS_DEBUG
1505 			atomic_add_long(&ether_pktinfo_hit, 1);
1506 #endif
1507 			logether(pkt_end, ifp);
1508 			return;
1509 		}
1510 	}
1511 #ifdef RSS_DEBUG
1512 	else if (ifp->if_capenable & IFCAP_RSS) {
1513 		if (pi == NULL)
1514 			atomic_add_long(&ether_rss_nopi, 1);
1515 		else
1516 			atomic_add_long(&ether_rss_nohash, 1);
1517 	}
1518 #endif
1519 
1520 	/*
1521 	 * Packet hash will be recalculated by software, so clear
1522 	 * the M_HASH and M_CKHASH flag set by the driver; the hash
1523 	 * value calculated by the hardware may not be exactly what
1524 	 * we want.
1525 	 */
1526 	m->m_flags &= ~(M_HASH | M_CKHASH);
1527 
1528 	if (!ether_vlancheck(&m)) {
1529 		KKASSERT(m == NULL);
1530 		logether(pkt_end, ifp);
1531 		return;
1532 	}
1533 
1534 	isr = ether_characterize(&m);
1535 	if (m == NULL) {
1536 		logether(pkt_end, ifp);
1537 		return;
1538 	}
1539 
1540 	/*
1541 	 * Finally dispatch it
1542 	 */
1543 	ether_dispatch(isr, m);
1544 
1545 	logether(pkt_end, ifp);
1546 }
1547 
1548 static int
1549 ether_characterize(struct mbuf **m0)
1550 {
1551 	struct mbuf *m = *m0;
1552 	struct ether_header *eh;
1553 	uint16_t ether_type;
1554 	int isr;
1555 
1556 	eh = mtod(m, struct ether_header *);
1557 	ether_type = ntohs(eh->ether_type);
1558 
1559 	/*
1560 	 * Map ether type to netisr id.
1561 	 */
1562 	switch (ether_type) {
1563 #ifdef INET
1564 	case ETHERTYPE_IP:
1565 		isr = NETISR_IP;
1566 		break;
1567 
1568 	case ETHERTYPE_ARP:
1569 		isr = NETISR_ARP;
1570 		break;
1571 #endif
1572 
1573 #ifdef INET6
1574 	case ETHERTYPE_IPV6:
1575 		isr = NETISR_IPV6;
1576 		break;
1577 #endif
1578 
1579 #ifdef IPX
1580 	case ETHERTYPE_IPX:
1581 		isr = NETISR_IPX;
1582 		break;
1583 #endif
1584 
1585 #ifdef MPLS
1586 	case ETHERTYPE_MPLS:
1587 	case ETHERTYPE_MPLS_MCAST:
1588 		m->m_flags |= M_MPLSLABELED;
1589 		isr = NETISR_MPLS;
1590 		break;
1591 #endif
1592 
1593 	default:
1594 		/*
1595 		 * NETISR_MAX is an invalid value; it is chosen to let
1596 		 * netisr_characterize() know that we have no clear
1597 		 * idea where this packet should go.
1598 		 */
1599 		isr = NETISR_MAX;
1600 		break;
1601 	}
1602 
1603 	/*
1604 	 * Ask the isr to characterize the packet since we couldn't.
1605 	 * This is an attempt to optimally get us onto the correct protocol
1606 	 * thread.
1607 	 */
1608 	netisr_characterize(isr, &m, sizeof(struct ether_header));
1609 
1610 	*m0 = m;
1611 	return isr;
1612 }
1613 
1614 static void
1615 ether_demux_handler(netmsg_t nmsg)
1616 {
1617 	struct netmsg_packet *nmp = &nmsg->packet;	/* actual size */
1618 	struct ifnet *ifp;
1619 	struct mbuf *m;
1620 
1621 	m = nmp->nm_packet;
1622 	M_ASSERTPKTHDR(m);
1623 	ifp = m->m_pkthdr.rcvif;
1624 
1625 	ether_demux_oncpu(ifp, m);
1626 }
1627 
1628 void
1629 ether_demux(struct mbuf *m)
1630 {
1631 	struct netmsg_packet *pmsg;
1632 	int isr;
1633 
1634 	isr = ether_characterize(&m);
1635 	if (m == NULL)
1636 		return;
1637 
1638 	KKASSERT(m->m_flags & M_HASH);
1639 	pmsg = &m->m_hdr.mh_netmsg;
1640 	netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1641 	    0, ether_demux_handler);
1642 	pmsg->nm_packet = m;
1643 	pmsg->base.lmsg.u.ms_result = isr;
1644 
1645 	lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1646 }
1647 
1648 boolean_t
1649 ether_tso_pullup(struct mbuf **mp, int *hoff0, struct ip **ip, int *iphlen,
1650     struct tcphdr **th, int *thoff)
1651 {
1652 	struct mbuf *m = *mp;
1653 	struct ether_header *eh;
1654 	uint16_t type;
1655 	int hoff;
1656 
1657 	KASSERT(M_WRITABLE(m), ("not writable"));
1658 
1659 	hoff = ETHER_HDR_LEN;
1660 	if (m->m_len < hoff) {
1661 		m = m_pullup(m, hoff);
1662 		if (m == NULL)
1663 			goto failed;
1664 	}
1665 	eh = mtod(m, struct ether_header *);
1666 	type = eh->ether_type;
1667 
1668 	if (type == htons(ETHERTYPE_VLAN)) {
1669 		struct ether_vlan_header *evh;
1670 
1671 		hoff += EVL_ENCAPLEN;
1672 		if (m->m_len < hoff) {
1673 			m = m_pullup(m, hoff);
1674 			if (m == NULL)
1675 				goto failed;
1676 		}
1677 		evh = mtod(m, struct ether_vlan_header *);
1678 		type = evh->evl_proto;
1679 	}
1680 	KASSERT(type == htons(ETHERTYPE_IP), ("not IP %d", ntohs(type)));
1681 
1682 	*mp = m;
1683 	*hoff0 = hoff;
1684 	return tcp_tso_pullup(mp, hoff, ip, iphlen, th, thoff);
1685 
1686 failed:
1687 	if (m != NULL)
1688 		m_freem(m);
1689 	*mp = NULL;
1690 	return FALSE;
1691 }
1692 
1693 u_char *
1694 kether_aton(const char *macstr, u_char *addr)
1695 {
1696         unsigned int o0, o1, o2, o3, o4, o5;
1697         int n;
1698 
1699         if (macstr == NULL || addr == NULL)
1700                 return NULL;
1701 
1702         n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1703             &o3, &o4, &o5);
1704         if (n != 6)
1705                 return NULL;
1706 
1707         addr[0] = o0;
1708         addr[1] = o1;
1709         addr[2] = o2;
1710         addr[3] = o3;
1711         addr[4] = o4;
1712         addr[5] = o5;
1713 
1714         return addr;
1715 }
1716 
1717 char *
1718 kether_ntoa(const u_char *addr, char *buf)
1719 {
1720         int len = ETHER_ADDRSTRLEN + 1;
1721         int n;
1722 
1723         n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1724             addr[1], addr[2], addr[3], addr[4], addr[5]);
1725 
1726         if (n < 17)
1727                 return NULL;
1728 
1729         return buf;
1730 }
1731 
1732 MODULE_VERSION(ether, 1);
1733