xref: /freebsd/sys/net/if_ethersubr.c (revision bbb29a3c0f2c4565eff6fda70426807b6ed97f8b)
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  * 4. Neither the name of the University nor the names of its contributors
14  *    may be used to endorse or promote products derived from this software
15  *    without specific prior written permission.
16  *
17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27  * SUCH DAMAGE.
28  *
29  *	@(#)if_ethersubr.c	8.1 (Berkeley) 6/10/93
30  * $FreeBSD$
31  */
32 
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 #include "opt_netgraph.h"
36 #include "opt_mbuf_profiling.h"
37 #include "opt_rss.h"
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/module.h>
45 #include <sys/mbuf.h>
46 #include <sys/random.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/sysctl.h>
50 #include <sys/uuid.h>
51 
52 #include <net/if.h>
53 #include <net/if_var.h>
54 #include <net/if_arp.h>
55 #include <net/netisr.h>
56 #include <net/route.h>
57 #include <net/if_llc.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/bpf.h>
61 #include <net/ethernet.h>
62 #include <net/if_bridgevar.h>
63 #include <net/if_vlan_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/pfil.h>
66 #include <net/vnet.h>
67 
68 #include <netpfil/pf/pf_mtag.h>
69 
70 #if defined(INET) || defined(INET6)
71 #include <netinet/in.h>
72 #include <netinet/in_var.h>
73 #include <netinet/if_ether.h>
74 #include <netinet/in_rss.h>
75 #include <netinet/ip_carp.h>
76 #include <netinet/ip_var.h>
77 #endif
78 #ifdef INET6
79 #include <netinet6/nd6.h>
80 #endif
81 #include <security/mac/mac_framework.h>
82 
83 #ifdef CTASSERT
84 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
85 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
86 #endif
87 
88 VNET_DEFINE(struct pfil_head, link_pfil_hook);	/* Packet filter hooks */
89 
90 /* netgraph node hooks for ng_ether(4) */
91 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
92 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
93 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
94 void	(*ng_ether_attach_p)(struct ifnet *ifp);
95 void	(*ng_ether_detach_p)(struct ifnet *ifp);
96 
97 void	(*vlan_input_p)(struct ifnet *, struct mbuf *);
98 
99 /* if_bridge(4) support */
100 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
101 int	(*bridge_output_p)(struct ifnet *, struct mbuf *,
102 		struct sockaddr *, struct rtentry *);
103 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
104 
105 /* if_lagg(4) support */
106 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
107 
108 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
109 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
110 
111 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
112 		struct sockaddr *);
113 #ifdef VIMAGE
114 static	void ether_reassign(struct ifnet *, struct vnet *, char *);
115 #endif
116 
117 #define	ETHER_IS_BROADCAST(addr) \
118 	(bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
119 
120 #define senderr(e) do { error = (e); goto bad;} while (0)
121 
122 static void
123 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
124 {
125 	int csum_flags = 0;
126 
127 	if (src->m_pkthdr.csum_flags & CSUM_IP)
128 		csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
129 	if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
130 		csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
131 	if (src->m_pkthdr.csum_flags & CSUM_SCTP)
132 		csum_flags |= CSUM_SCTP_VALID;
133 	dst->m_pkthdr.csum_flags |= csum_flags;
134 	if (csum_flags & CSUM_DATA_VALID)
135 		dst->m_pkthdr.csum_data = 0xffff;
136 }
137 
138 /*
139  * Ethernet output routine.
140  * Encapsulate a packet of type family for the local net.
141  * Use trailer local net encapsulation if enough data in first
142  * packet leaves a multiple of 512 bytes of data in remainder.
143  */
144 int
145 ether_output(struct ifnet *ifp, struct mbuf *m,
146 	const struct sockaddr *dst, struct route *ro)
147 {
148 	short type;
149 	int error = 0, hdrcmplt = 0;
150 	u_char edst[ETHER_ADDR_LEN];
151 	struct llentry *lle = NULL;
152 	struct rtentry *rt0 = NULL;
153 	struct ether_header *eh;
154 	struct pf_mtag *t;
155 	int loop_copy = 1;
156 	int hlen;	/* link layer header length */
157 	int is_gw = 0;
158 	uint32_t pflags = 0;
159 
160 	if (ro != NULL) {
161 		if (!(m->m_flags & (M_BCAST | M_MCAST))) {
162 			lle = ro->ro_lle;
163 			if (lle != NULL)
164 				pflags = lle->la_flags;
165 		}
166 		rt0 = ro->ro_rt;
167 		if (rt0 != NULL && (rt0->rt_flags & RTF_GATEWAY) != 0)
168 			is_gw = 1;
169 	}
170 #ifdef MAC
171 	error = mac_ifnet_check_transmit(ifp, m);
172 	if (error)
173 		senderr(error);
174 #endif
175 
176 	M_PROFILE(m);
177 	if (ifp->if_flags & IFF_MONITOR)
178 		senderr(ENETDOWN);
179 	if (!((ifp->if_flags & IFF_UP) &&
180 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
181 		senderr(ENETDOWN);
182 
183 	hlen = ETHER_HDR_LEN;
184 	switch (dst->sa_family) {
185 #ifdef INET
186 	case AF_INET:
187 		if (lle != NULL && (pflags & LLE_VALID) != 0)
188 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
189 		else
190 			error = arpresolve(ifp, is_gw, m, dst, edst, &pflags);
191 		if (error)
192 			return (error == EWOULDBLOCK ? 0 : error);
193 		type = htons(ETHERTYPE_IP);
194 		break;
195 	case AF_ARP:
196 	{
197 		struct arphdr *ah;
198 		ah = mtod(m, struct arphdr *);
199 		ah->ar_hrd = htons(ARPHRD_ETHER);
200 
201 		loop_copy = 0; /* if this is for us, don't do it */
202 
203 		switch(ntohs(ah->ar_op)) {
204 		case ARPOP_REVREQUEST:
205 		case ARPOP_REVREPLY:
206 			type = htons(ETHERTYPE_REVARP);
207 			break;
208 		case ARPOP_REQUEST:
209 		case ARPOP_REPLY:
210 		default:
211 			type = htons(ETHERTYPE_ARP);
212 			break;
213 		}
214 
215 		if (m->m_flags & M_BCAST)
216 			bcopy(ifp->if_broadcastaddr, edst, ETHER_ADDR_LEN);
217 		else
218 			bcopy(ar_tha(ah), edst, ETHER_ADDR_LEN);
219 
220 	}
221 	break;
222 #endif
223 #ifdef INET6
224 	case AF_INET6:
225 		if (lle != NULL && (pflags & LLE_VALID))
226 			memcpy(edst, &lle->ll_addr.mac16, sizeof(edst));
227 		else
228 			error = nd6_storelladdr(ifp, m, dst, (u_char *)edst,
229 			    &pflags);
230 		if (error)
231 			return error;
232 		type = htons(ETHERTYPE_IPV6);
233 		break;
234 #endif
235 	case pseudo_AF_HDRCMPLT:
236 	    {
237 		const struct ether_header *eh;
238 
239 		hdrcmplt = 1;
240 		/* FALLTHROUGH */
241 
242 	case AF_UNSPEC:
243 		loop_copy = 0; /* if this is for us, don't do it */
244 		eh = (const struct ether_header *)dst->sa_data;
245 		(void)memcpy(edst, eh->ether_dhost, sizeof (edst));
246 		type = eh->ether_type;
247 		break;
248             }
249 	default:
250 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
251 		senderr(EAFNOSUPPORT);
252 	}
253 
254 	if ((pflags & LLE_IFADDR) != 0) {
255 		update_mbuf_csumflags(m, m);
256 		return (if_simloop(ifp, m, dst->sa_family, 0));
257 	}
258 
259 	/*
260 	 * Add local net header.  If no space in first mbuf,
261 	 * allocate another.
262 	 */
263 	M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
264 	if (m == NULL)
265 		senderr(ENOBUFS);
266 	eh = mtod(m, struct ether_header *);
267 	if (hdrcmplt == 0) {
268 		memcpy(&eh->ether_type, &type, sizeof(eh->ether_type));
269 		memcpy(eh->ether_dhost, edst, sizeof (edst));
270 		memcpy(eh->ether_shost, IF_LLADDR(ifp),sizeof(eh->ether_shost));
271 	}
272 
273 	/*
274 	 * If a simplex interface, and the packet is being sent to our
275 	 * Ethernet address or a broadcast address, loopback a copy.
276 	 * XXX To make a simplex device behave exactly like a duplex
277 	 * device, we should copy in the case of sending to our own
278 	 * ethernet address (thus letting the original actually appear
279 	 * on the wire). However, we don't do that here for security
280 	 * reasons and compatibility with the original behavior.
281 	 */
282 	if ((ifp->if_flags & IFF_SIMPLEX) && loop_copy &&
283 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
284 		if (m->m_flags & M_BCAST) {
285 			struct mbuf *n;
286 
287 			/*
288 			 * Because if_simloop() modifies the packet, we need a
289 			 * writable copy through m_dup() instead of a readonly
290 			 * one as m_copy[m] would give us. The alternative would
291 			 * be to modify if_simloop() to handle the readonly mbuf,
292 			 * but performancewise it is mostly equivalent (trading
293 			 * extra data copying vs. extra locking).
294 			 *
295 			 * XXX This is a local workaround.  A number of less
296 			 * often used kernel parts suffer from the same bug.
297 			 * See PR kern/105943 for a proposed general solution.
298 			 */
299 			if ((n = m_dup(m, M_NOWAIT)) != NULL) {
300 				update_mbuf_csumflags(m, n);
301 				(void)if_simloop(ifp, n, dst->sa_family, hlen);
302 			} else
303 				if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
304 		} else if (bcmp(eh->ether_dhost, eh->ether_shost,
305 				ETHER_ADDR_LEN) == 0) {
306 			update_mbuf_csumflags(m, m);
307 			(void) if_simloop(ifp, m, dst->sa_family, hlen);
308 			return (0);	/* XXX */
309 		}
310 	}
311 
312        /*
313 	* Bridges require special output handling.
314 	*/
315 	if (ifp->if_bridge) {
316 		BRIDGE_OUTPUT(ifp, m, error);
317 		return (error);
318 	}
319 
320 #if defined(INET) || defined(INET6)
321 	if (ifp->if_carp &&
322 	    (error = (*carp_output_p)(ifp, m, dst)))
323 		goto bad;
324 #endif
325 
326 	/* Handle ng_ether(4) processing, if any */
327 	if (ifp->if_l2com != NULL) {
328 		KASSERT(ng_ether_output_p != NULL,
329 		    ("ng_ether_output_p is NULL"));
330 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
331 bad:			if (m != NULL)
332 				m_freem(m);
333 			return (error);
334 		}
335 		if (m == NULL)
336 			return (0);
337 	}
338 
339 	/* Continue with link-layer output */
340 	return ether_output_frame(ifp, m);
341 }
342 
343 /*
344  * Ethernet link layer output routine to send a raw frame to the device.
345  *
346  * This assumes that the 14 byte Ethernet header is present and contiguous
347  * in the first mbuf (if BRIDGE'ing).
348  */
349 int
350 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
351 {
352 	int i;
353 
354 	if (PFIL_HOOKED(&V_link_pfil_hook)) {
355 		i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL);
356 
357 		if (i != 0)
358 			return (EACCES);
359 
360 		if (m == NULL)
361 			return (0);
362 	}
363 
364 	/*
365 	 * Queue message on interface, update output statistics if
366 	 * successful, and start output if interface not yet active.
367 	 */
368 	return ((ifp->if_transmit)(ifp, m));
369 }
370 
371 #if defined(INET) || defined(INET6)
372 #endif
373 
374 /*
375  * Process a received Ethernet packet; the packet is in the
376  * mbuf chain m with the ethernet header at the front.
377  */
378 static void
379 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
380 {
381 	struct ether_header *eh;
382 	u_short etype;
383 
384 	if ((ifp->if_flags & IFF_UP) == 0) {
385 		m_freem(m);
386 		return;
387 	}
388 #ifdef DIAGNOSTIC
389 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
390 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
391 		m_freem(m);
392 		return;
393 	}
394 #endif
395 	/*
396 	 * Do consistency checks to verify assumptions
397 	 * made by code past this point.
398 	 */
399 	if ((m->m_flags & M_PKTHDR) == 0) {
400 		if_printf(ifp, "discard frame w/o packet header\n");
401 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
402 		m_freem(m);
403 		return;
404 	}
405 	if (m->m_len < ETHER_HDR_LEN) {
406 		/* XXX maybe should pullup? */
407 		if_printf(ifp, "discard frame w/o leading ethernet "
408 				"header (len %u pkt len %u)\n",
409 				m->m_len, m->m_pkthdr.len);
410 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
411 		m_freem(m);
412 		return;
413 	}
414 	eh = mtod(m, struct ether_header *);
415 	etype = ntohs(eh->ether_type);
416 	if (m->m_pkthdr.rcvif == NULL) {
417 		if_printf(ifp, "discard frame w/o interface pointer\n");
418 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
419 		m_freem(m);
420 		return;
421 	}
422 #ifdef DIAGNOSTIC
423 	if (m->m_pkthdr.rcvif != ifp) {
424 		if_printf(ifp, "Warning, frame marked as received on %s\n",
425 			m->m_pkthdr.rcvif->if_xname);
426 	}
427 #endif
428 
429 	CURVNET_SET_QUIET(ifp->if_vnet);
430 
431 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
432 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
433 			m->m_flags |= M_BCAST;
434 		else
435 			m->m_flags |= M_MCAST;
436 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
437 	}
438 
439 #ifdef MAC
440 	/*
441 	 * Tag the mbuf with an appropriate MAC label before any other
442 	 * consumers can get to it.
443 	 */
444 	mac_ifnet_create_mbuf(ifp, m);
445 #endif
446 
447 	/*
448 	 * Give bpf a chance at the packet.
449 	 */
450 	ETHER_BPF_MTAP(ifp, m);
451 
452 	/*
453 	 * If the CRC is still on the packet, trim it off. We do this once
454 	 * and once only in case we are re-entered. Nothing else on the
455 	 * Ethernet receive path expects to see the FCS.
456 	 */
457 	if (m->m_flags & M_HASFCS) {
458 		m_adj(m, -ETHER_CRC_LEN);
459 		m->m_flags &= ~M_HASFCS;
460 	}
461 
462 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
463 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
464 
465 	/* Allow monitor mode to claim this frame, after stats are updated. */
466 	if (ifp->if_flags & IFF_MONITOR) {
467 		m_freem(m);
468 		CURVNET_RESTORE();
469 		return;
470 	}
471 
472 	/* Handle input from a lagg(4) port */
473 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
474 		KASSERT(lagg_input_p != NULL,
475 		    ("%s: if_lagg not loaded!", __func__));
476 		m = (*lagg_input_p)(ifp, m);
477 		if (m != NULL)
478 			ifp = m->m_pkthdr.rcvif;
479 		else {
480 			CURVNET_RESTORE();
481 			return;
482 		}
483 	}
484 
485 	/*
486 	 * If the hardware did not process an 802.1Q tag, do this now,
487 	 * to allow 802.1P priority frames to be passed to the main input
488 	 * path correctly.
489 	 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
490 	 */
491 	if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
492 		struct ether_vlan_header *evl;
493 
494 		if (m->m_len < sizeof(*evl) &&
495 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
496 #ifdef DIAGNOSTIC
497 			if_printf(ifp, "cannot pullup VLAN header\n");
498 #endif
499 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
500 			m_freem(m);
501 			CURVNET_RESTORE();
502 			return;
503 		}
504 
505 		evl = mtod(m, struct ether_vlan_header *);
506 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
507 		m->m_flags |= M_VLANTAG;
508 
509 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
510 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
511 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
512 		eh = mtod(m, struct ether_header *);
513 	}
514 
515 	M_SETFIB(m, ifp->if_fib);
516 
517 	/* Allow ng_ether(4) to claim this frame. */
518 	if (ifp->if_l2com != NULL) {
519 		KASSERT(ng_ether_input_p != NULL,
520 		    ("%s: ng_ether_input_p is NULL", __func__));
521 		m->m_flags &= ~M_PROMISC;
522 		(*ng_ether_input_p)(ifp, &m);
523 		if (m == NULL) {
524 			CURVNET_RESTORE();
525 			return;
526 		}
527 		eh = mtod(m, struct ether_header *);
528 	}
529 
530 	/*
531 	 * Allow if_bridge(4) to claim this frame.
532 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
533 	 * and the frame should be delivered locally.
534 	 */
535 	if (ifp->if_bridge != NULL) {
536 		m->m_flags &= ~M_PROMISC;
537 		BRIDGE_INPUT(ifp, m);
538 		if (m == NULL) {
539 			CURVNET_RESTORE();
540 			return;
541 		}
542 		eh = mtod(m, struct ether_header *);
543 	}
544 
545 #if defined(INET) || defined(INET6)
546 	/*
547 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
548 	 * mbuf flows up to Layer 3.
549 	 * FreeBSD's implementation of carp(4) uses the inprotosw
550 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
551 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
552 	 * is outside the scope of the M_PROMISC test below.
553 	 * TODO: Maintain a hash table of ethernet addresses other than
554 	 * ether_dhost which may be active on this ifp.
555 	 */
556 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
557 		m->m_flags &= ~M_PROMISC;
558 	} else
559 #endif
560 	{
561 		/*
562 		 * If the frame received was not for our MAC address, set the
563 		 * M_PROMISC flag on the mbuf chain. The frame may need to
564 		 * be seen by the rest of the Ethernet input path in case of
565 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
566 		 * seen by upper protocol layers.
567 		 */
568 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
569 		    bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
570 			m->m_flags |= M_PROMISC;
571 	}
572 
573 	random_harvest(&(m->m_data), 12, 2, RANDOM_NET_ETHER);
574 
575 	ether_demux(ifp, m);
576 	CURVNET_RESTORE();
577 }
578 
579 /*
580  * Ethernet input dispatch; by default, direct dispatch here regardless of
581  * global configuration.  However, if RSS is enabled, hook up RSS affinity
582  * so that when deferred or hybrid dispatch is enabled, we can redistribute
583  * load based on RSS.
584  *
585  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
586  * not it had already done work distribution via multi-queue.  Then we could
587  * direct dispatch in the event load balancing was already complete and
588  * handle the case of interfaces with different capabilities better.
589  *
590  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
591  * at multiple layers?
592  *
593  * XXXRW: For now, enable all this only if RSS is compiled in, although it
594  * works fine without RSS.  Need to characterise the performance overhead
595  * of the detour through the netisr code in the event the result is always
596  * direct dispatch.
597  */
598 static void
599 ether_nh_input(struct mbuf *m)
600 {
601 
602 	ether_input_internal(m->m_pkthdr.rcvif, m);
603 }
604 
605 static struct netisr_handler	ether_nh = {
606 	.nh_name = "ether",
607 	.nh_handler = ether_nh_input,
608 	.nh_proto = NETISR_ETHER,
609 #ifdef RSS
610 	.nh_policy = NETISR_POLICY_CPU,
611 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
612 	.nh_m2cpuid = rss_m2cpuid,
613 #else
614 	.nh_policy = NETISR_POLICY_SOURCE,
615 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
616 #endif
617 };
618 
619 static void
620 ether_init(__unused void *arg)
621 {
622 
623 	netisr_register(&ether_nh);
624 }
625 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
626 
627 static void
628 vnet_ether_init(__unused void *arg)
629 {
630 	int i;
631 
632 	/* Initialize packet filter hooks. */
633 	V_link_pfil_hook.ph_type = PFIL_TYPE_AF;
634 	V_link_pfil_hook.ph_af = AF_LINK;
635 	if ((i = pfil_head_register(&V_link_pfil_hook)) != 0)
636 		printf("%s: WARNING: unable to register pfil link hook, "
637 			"error %d\n", __func__, i);
638 }
639 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
640     vnet_ether_init, NULL);
641 
642 static void
643 vnet_ether_destroy(__unused void *arg)
644 {
645 	int i;
646 
647 	if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0)
648 		printf("%s: WARNING: unable to unregister pfil link hook, "
649 			"error %d\n", __func__, i);
650 }
651 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
652     vnet_ether_destroy, NULL);
653 
654 
655 
656 static void
657 ether_input(struct ifnet *ifp, struct mbuf *m)
658 {
659 
660 	struct mbuf *mn;
661 
662 	/*
663 	 * The drivers are allowed to pass in a chain of packets linked with
664 	 * m_nextpkt. We split them up into separate packets here and pass
665 	 * them up. This allows the drivers to amortize the receive lock.
666 	 */
667 	while (m) {
668 		mn = m->m_nextpkt;
669 		m->m_nextpkt = NULL;
670 
671 		/*
672 		 * We will rely on rcvif being set properly in the deferred context,
673 		 * so assert it is correct here.
674 		 */
675 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch", __func__));
676 		netisr_dispatch(NETISR_ETHER, m);
677 		m = mn;
678 	}
679 }
680 
681 /*
682  * Upper layer processing for a received Ethernet packet.
683  */
684 void
685 ether_demux(struct ifnet *ifp, struct mbuf *m)
686 {
687 	struct ether_header *eh;
688 	int i, isr;
689 	u_short ether_type;
690 
691 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
692 
693 	/* Do not grab PROMISC frames in case we are re-entered. */
694 	if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) {
695 		i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL);
696 
697 		if (i != 0 || m == NULL)
698 			return;
699 	}
700 
701 	eh = mtod(m, struct ether_header *);
702 	ether_type = ntohs(eh->ether_type);
703 
704 	/*
705 	 * If this frame has a VLAN tag other than 0, call vlan_input()
706 	 * if its module is loaded. Otherwise, drop.
707 	 */
708 	if ((m->m_flags & M_VLANTAG) &&
709 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
710 		if (ifp->if_vlantrunk == NULL) {
711 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
712 			m_freem(m);
713 			return;
714 		}
715 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
716 		    __func__));
717 		/* Clear before possibly re-entering ether_input(). */
718 		m->m_flags &= ~M_PROMISC;
719 		(*vlan_input_p)(ifp, m);
720 		return;
721 	}
722 
723 	/*
724 	 * Pass promiscuously received frames to the upper layer if the user
725 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
726 	 */
727 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
728 		m_freem(m);
729 		return;
730 	}
731 
732 	/*
733 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
734 	 * Strip off Ethernet header.
735 	 */
736 	m->m_flags &= ~M_VLANTAG;
737 	m_clrprotoflags(m);
738 	m_adj(m, ETHER_HDR_LEN);
739 
740 	/*
741 	 * Dispatch frame to upper layer.
742 	 */
743 	switch (ether_type) {
744 #ifdef INET
745 	case ETHERTYPE_IP:
746 		if ((m = ip_fastforward(m)) == NULL)
747 			return;
748 		isr = NETISR_IP;
749 		break;
750 
751 	case ETHERTYPE_ARP:
752 		if (ifp->if_flags & IFF_NOARP) {
753 			/* Discard packet if ARP is disabled on interface */
754 			m_freem(m);
755 			return;
756 		}
757 		isr = NETISR_ARP;
758 		break;
759 #endif
760 #ifdef INET6
761 	case ETHERTYPE_IPV6:
762 		isr = NETISR_IPV6;
763 		break;
764 #endif
765 	default:
766 		goto discard;
767 	}
768 	netisr_dispatch(isr, m);
769 	return;
770 
771 discard:
772 	/*
773 	 * Packet is to be discarded.  If netgraph is present,
774 	 * hand the packet to it for last chance processing;
775 	 * otherwise dispose of it.
776 	 */
777 	if (ifp->if_l2com != NULL) {
778 		KASSERT(ng_ether_input_orphan_p != NULL,
779 		    ("ng_ether_input_orphan_p is NULL"));
780 		/*
781 		 * Put back the ethernet header so netgraph has a
782 		 * consistent view of inbound packets.
783 		 */
784 		M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
785 		(*ng_ether_input_orphan_p)(ifp, m);
786 		return;
787 	}
788 	m_freem(m);
789 }
790 
791 /*
792  * Convert Ethernet address to printable (loggable) representation.
793  * This routine is for compatibility; it's better to just use
794  *
795  *	printf("%6D", <pointer to address>, ":");
796  *
797  * since there's no static buffer involved.
798  */
799 char *
800 ether_sprintf(const u_char *ap)
801 {
802 	static char etherbuf[18];
803 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
804 	return (etherbuf);
805 }
806 
807 /*
808  * Perform common duties while attaching to interface list
809  */
810 void
811 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
812 {
813 	int i;
814 	struct ifaddr *ifa;
815 	struct sockaddr_dl *sdl;
816 
817 	ifp->if_addrlen = ETHER_ADDR_LEN;
818 	ifp->if_hdrlen = ETHER_HDR_LEN;
819 	if_attach(ifp);
820 	ifp->if_mtu = ETHERMTU;
821 	ifp->if_output = ether_output;
822 	ifp->if_input = ether_input;
823 	ifp->if_resolvemulti = ether_resolvemulti;
824 #ifdef VIMAGE
825 	ifp->if_reassign = ether_reassign;
826 #endif
827 	if (ifp->if_baudrate == 0)
828 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
829 	ifp->if_broadcastaddr = etherbroadcastaddr;
830 
831 	ifa = ifp->if_addr;
832 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
833 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
834 	sdl->sdl_type = IFT_ETHER;
835 	sdl->sdl_alen = ifp->if_addrlen;
836 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
837 
838 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
839 	if (ng_ether_attach_p != NULL)
840 		(*ng_ether_attach_p)(ifp);
841 
842 	/* Announce Ethernet MAC address if non-zero. */
843 	for (i = 0; i < ifp->if_addrlen; i++)
844 		if (lla[i] != 0)
845 			break;
846 	if (i != ifp->if_addrlen)
847 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
848 
849 	uuid_ether_add(LLADDR(sdl));
850 }
851 
852 /*
853  * Perform common duties while detaching an Ethernet interface
854  */
855 void
856 ether_ifdetach(struct ifnet *ifp)
857 {
858 	struct sockaddr_dl *sdl;
859 
860 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
861 	uuid_ether_del(LLADDR(sdl));
862 
863 	if (ifp->if_l2com != NULL) {
864 		KASSERT(ng_ether_detach_p != NULL,
865 		    ("ng_ether_detach_p is NULL"));
866 		(*ng_ether_detach_p)(ifp);
867 	}
868 
869 	bpfdetach(ifp);
870 	if_detach(ifp);
871 }
872 
873 #ifdef VIMAGE
874 void
875 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
876 {
877 
878 	if (ifp->if_l2com != NULL) {
879 		KASSERT(ng_ether_detach_p != NULL,
880 		    ("ng_ether_detach_p is NULL"));
881 		(*ng_ether_detach_p)(ifp);
882 	}
883 
884 	if (ng_ether_attach_p != NULL) {
885 		CURVNET_SET_QUIET(new_vnet);
886 		(*ng_ether_attach_p)(ifp);
887 		CURVNET_RESTORE();
888 	}
889 }
890 #endif
891 
892 SYSCTL_DECL(_net_link);
893 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
894 
895 #if 0
896 /*
897  * This is for reference.  We have a table-driven version
898  * of the little-endian crc32 generator, which is faster
899  * than the double-loop.
900  */
901 uint32_t
902 ether_crc32_le(const uint8_t *buf, size_t len)
903 {
904 	size_t i;
905 	uint32_t crc;
906 	int bit;
907 	uint8_t data;
908 
909 	crc = 0xffffffff;	/* initial value */
910 
911 	for (i = 0; i < len; i++) {
912 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
913 			carry = (crc ^ data) & 1;
914 			crc >>= 1;
915 			if (carry)
916 				crc = (crc ^ ETHER_CRC_POLY_LE);
917 		}
918 	}
919 
920 	return (crc);
921 }
922 #else
923 uint32_t
924 ether_crc32_le(const uint8_t *buf, size_t len)
925 {
926 	static const uint32_t crctab[] = {
927 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
928 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
929 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
930 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
931 	};
932 	size_t i;
933 	uint32_t crc;
934 
935 	crc = 0xffffffff;	/* initial value */
936 
937 	for (i = 0; i < len; i++) {
938 		crc ^= buf[i];
939 		crc = (crc >> 4) ^ crctab[crc & 0xf];
940 		crc = (crc >> 4) ^ crctab[crc & 0xf];
941 	}
942 
943 	return (crc);
944 }
945 #endif
946 
947 uint32_t
948 ether_crc32_be(const uint8_t *buf, size_t len)
949 {
950 	size_t i;
951 	uint32_t crc, carry;
952 	int bit;
953 	uint8_t data;
954 
955 	crc = 0xffffffff;	/* initial value */
956 
957 	for (i = 0; i < len; i++) {
958 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
959 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
960 			crc <<= 1;
961 			if (carry)
962 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
963 		}
964 	}
965 
966 	return (crc);
967 }
968 
969 int
970 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
971 {
972 	struct ifaddr *ifa = (struct ifaddr *) data;
973 	struct ifreq *ifr = (struct ifreq *) data;
974 	int error = 0;
975 
976 	switch (command) {
977 	case SIOCSIFADDR:
978 		ifp->if_flags |= IFF_UP;
979 
980 		switch (ifa->ifa_addr->sa_family) {
981 #ifdef INET
982 		case AF_INET:
983 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
984 			arp_ifinit(ifp, ifa);
985 			break;
986 #endif
987 		default:
988 			ifp->if_init(ifp->if_softc);
989 			break;
990 		}
991 		break;
992 
993 	case SIOCGIFADDR:
994 		{
995 			struct sockaddr *sa;
996 
997 			sa = (struct sockaddr *) & ifr->ifr_data;
998 			bcopy(IF_LLADDR(ifp),
999 			      (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1000 		}
1001 		break;
1002 
1003 	case SIOCSIFMTU:
1004 		/*
1005 		 * Set the interface MTU.
1006 		 */
1007 		if (ifr->ifr_mtu > ETHERMTU) {
1008 			error = EINVAL;
1009 		} else {
1010 			ifp->if_mtu = ifr->ifr_mtu;
1011 		}
1012 		break;
1013 	default:
1014 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1015 		break;
1016 	}
1017 	return (error);
1018 }
1019 
1020 static int
1021 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1022 	struct sockaddr *sa)
1023 {
1024 	struct sockaddr_dl *sdl;
1025 #ifdef INET
1026 	struct sockaddr_in *sin;
1027 #endif
1028 #ifdef INET6
1029 	struct sockaddr_in6 *sin6;
1030 #endif
1031 	u_char *e_addr;
1032 
1033 	switch(sa->sa_family) {
1034 	case AF_LINK:
1035 		/*
1036 		 * No mapping needed. Just check that it's a valid MC address.
1037 		 */
1038 		sdl = (struct sockaddr_dl *)sa;
1039 		e_addr = LLADDR(sdl);
1040 		if (!ETHER_IS_MULTICAST(e_addr))
1041 			return EADDRNOTAVAIL;
1042 		*llsa = 0;
1043 		return 0;
1044 
1045 #ifdef INET
1046 	case AF_INET:
1047 		sin = (struct sockaddr_in *)sa;
1048 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1049 			return EADDRNOTAVAIL;
1050 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1051 		sdl->sdl_alen = ETHER_ADDR_LEN;
1052 		e_addr = LLADDR(sdl);
1053 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1054 		*llsa = (struct sockaddr *)sdl;
1055 		return 0;
1056 #endif
1057 #ifdef INET6
1058 	case AF_INET6:
1059 		sin6 = (struct sockaddr_in6 *)sa;
1060 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1061 			/*
1062 			 * An IP6 address of 0 means listen to all
1063 			 * of the Ethernet multicast address used for IP6.
1064 			 * (This is used for multicast routers.)
1065 			 */
1066 			ifp->if_flags |= IFF_ALLMULTI;
1067 			*llsa = 0;
1068 			return 0;
1069 		}
1070 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1071 			return EADDRNOTAVAIL;
1072 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1073 		sdl->sdl_alen = ETHER_ADDR_LEN;
1074 		e_addr = LLADDR(sdl);
1075 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1076 		*llsa = (struct sockaddr *)sdl;
1077 		return 0;
1078 #endif
1079 
1080 	default:
1081 		/*
1082 		 * Well, the text isn't quite right, but it's the name
1083 		 * that counts...
1084 		 */
1085 		return EAFNOSUPPORT;
1086 	}
1087 }
1088 
1089 static moduledata_t ether_mod = {
1090 	.name = "ether",
1091 };
1092 
1093 void
1094 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1095 {
1096 	struct ether_vlan_header vlan;
1097 	struct mbuf mv, mb;
1098 
1099 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1100 	    ("%s: vlan information not present", __func__));
1101 	KASSERT(m->m_len >= sizeof(struct ether_header),
1102 	    ("%s: mbuf not large enough for header", __func__));
1103 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1104 	vlan.evl_proto = vlan.evl_encap_proto;
1105 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1106 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1107 	m->m_len -= sizeof(struct ether_header);
1108 	m->m_data += sizeof(struct ether_header);
1109 	/*
1110 	 * If a data link has been supplied by the caller, then we will need to
1111 	 * re-create a stack allocated mbuf chain with the following structure:
1112 	 *
1113 	 * (1) mbuf #1 will contain the supplied data link
1114 	 * (2) mbuf #2 will contain the vlan header
1115 	 * (3) mbuf #3 will contain the original mbuf's packet data
1116 	 *
1117 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1118 	 */
1119 	if (data != NULL) {
1120 		mv.m_next = m;
1121 		mv.m_data = (caddr_t)&vlan;
1122 		mv.m_len = sizeof(vlan);
1123 		mb.m_next = &mv;
1124 		mb.m_data = data;
1125 		mb.m_len = dlen;
1126 		bpf_mtap(bp, &mb);
1127 	} else
1128 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1129 	m->m_len += sizeof(struct ether_header);
1130 	m->m_data -= sizeof(struct ether_header);
1131 }
1132 
1133 struct mbuf *
1134 ether_vlanencap(struct mbuf *m, uint16_t tag)
1135 {
1136 	struct ether_vlan_header *evl;
1137 
1138 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1139 	if (m == NULL)
1140 		return (NULL);
1141 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1142 
1143 	if (m->m_len < sizeof(*evl)) {
1144 		m = m_pullup(m, sizeof(*evl));
1145 		if (m == NULL)
1146 			return (NULL);
1147 	}
1148 
1149 	/*
1150 	 * Transform the Ethernet header into an Ethernet header
1151 	 * with 802.1Q encapsulation.
1152 	 */
1153 	evl = mtod(m, struct ether_vlan_header *);
1154 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1155 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1156 	evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1157 	evl->evl_tag = htons(tag);
1158 	return (m);
1159 }
1160 
1161 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1162 MODULE_VERSION(ether, 1);
1163