xref: /freebsd/sys/net/if_ethersubr.c (revision 87bea33a67cad31661a6fb9ea4c62a5fc266cd98)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1982, 1989, 1993
5  *	The Regents of the University of California.  All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the University nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 #include "opt_inet.h"
33 #include "opt_inet6.h"
34 #include "opt_netgraph.h"
35 #include "opt_mbuf_profiling.h"
36 #include "opt_rss.h"
37 
38 #include <sys/param.h>
39 #include <sys/systm.h>
40 #include <sys/devctl.h>
41 #include <sys/eventhandler.h>
42 #include <sys/jail.h>
43 #include <sys/kernel.h>
44 #include <sys/lock.h>
45 #include <sys/malloc.h>
46 #include <sys/mbuf.h>
47 #include <sys/module.h>
48 #include <sys/msan.h>
49 #include <sys/proc.h>
50 #include <sys/priv.h>
51 #include <sys/random.h>
52 #include <sys/socket.h>
53 #include <sys/sockio.h>
54 #include <sys/sysctl.h>
55 #include <sys/uuid.h>
56 #ifdef KDB
57 #include <sys/kdb.h>
58 #endif
59 
60 #include <net/ieee_oui.h>
61 #include <net/if.h>
62 #include <net/if_var.h>
63 #include <net/if_private.h>
64 #include <net/if_arp.h>
65 #include <net/netisr.h>
66 #include <net/route.h>
67 #include <net/if_llc.h>
68 #include <net/if_dl.h>
69 #include <net/if_types.h>
70 #include <net/bpf.h>
71 #include <net/ethernet.h>
72 #include <net/if_bridgevar.h>
73 #include <net/if_vlan_var.h>
74 #include <net/if_llatbl.h>
75 #include <net/pfil.h>
76 #include <net/rss_config.h>
77 #include <net/vnet.h>
78 
79 #include <netpfil/pf/pf_mtag.h>
80 
81 #if defined(INET) || defined(INET6)
82 #include <netinet/in.h>
83 #include <netinet/in_var.h>
84 #include <netinet/if_ether.h>
85 #include <netinet/ip_carp.h>
86 #include <netinet/ip_var.h>
87 #endif
88 #ifdef INET6
89 #include <netinet6/nd6.h>
90 #endif
91 #include <security/mac/mac_framework.h>
92 
93 #include <crypto/sha1.h>
94 
95 VNET_DEFINE(pfil_head_t, link_pfil_head);	/* Packet filter hooks */
96 
97 /* netgraph node hooks for ng_ether(4) */
98 void	(*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
99 void	(*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
100 int	(*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
101 
102 /* if_bridge(4) support */
103 void	(*bridge_dn_p)(struct mbuf *, struct ifnet *);
104 bool	(*bridge_same_p)(const void *, const void *);
105 void	*(*bridge_get_softc_p)(struct ifnet *);
106 bool	(*bridge_member_ifaddrs_p)(void);
107 
108 /* if_lagg(4) support */
109 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
110 
111 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
112 			{ 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
113 
114 static	int ether_resolvemulti(struct ifnet *, struct sockaddr **,
115 		struct sockaddr *);
116 static	int ether_requestencap(struct ifnet *, struct if_encap_req *);
117 
118 static inline bool ether_do_pcp(struct ifnet *, struct mbuf *);
119 
120 #define senderr(e) do { error = (e); goto bad;} while (0)
121 
122 static void
update_mbuf_csumflags(struct mbuf * src,struct mbuf * dst)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  * Handle link-layer encapsulation requests.
140  */
141 static int
ether_requestencap(struct ifnet * ifp,struct if_encap_req * req)142 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
143 {
144 	struct ether_header *eh;
145 	struct arphdr *ah;
146 	uint16_t etype;
147 	const u_char *lladdr;
148 
149 	if (req->rtype != IFENCAP_LL)
150 		return (EOPNOTSUPP);
151 
152 	if (req->bufsize < ETHER_HDR_LEN)
153 		return (ENOMEM);
154 
155 	eh = (struct ether_header *)req->buf;
156 	lladdr = req->lladdr;
157 	req->lladdr_off = 0;
158 
159 	switch (req->family) {
160 	case AF_INET:
161 		etype = htons(ETHERTYPE_IP);
162 		break;
163 	case AF_INET6:
164 		etype = htons(ETHERTYPE_IPV6);
165 		break;
166 	case AF_ARP:
167 		ah = (struct arphdr *)req->hdata;
168 		ah->ar_hrd = htons(ARPHRD_ETHER);
169 
170 		switch(ntohs(ah->ar_op)) {
171 		case ARPOP_REVREQUEST:
172 		case ARPOP_REVREPLY:
173 			etype = htons(ETHERTYPE_REVARP);
174 			break;
175 		case ARPOP_REQUEST:
176 		case ARPOP_REPLY:
177 		default:
178 			etype = htons(ETHERTYPE_ARP);
179 			break;
180 		}
181 
182 		if (req->flags & IFENCAP_FLAG_BROADCAST)
183 			lladdr = ifp->if_broadcastaddr;
184 		break;
185 	default:
186 		return (EAFNOSUPPORT);
187 	}
188 
189 	memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
190 	memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
191 	memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
192 	req->bufsize = sizeof(struct ether_header);
193 
194 	return (0);
195 }
196 
197 static int
ether_resolve_addr(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro,u_char * phdr,uint32_t * pflags,struct llentry ** plle)198 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
199 	const struct sockaddr *dst, struct route *ro, u_char *phdr,
200 	uint32_t *pflags, struct llentry **plle)
201 {
202 	uint32_t lleflags = 0;
203 	int error = 0;
204 #if defined(INET) || defined(INET6)
205 	struct ether_header *eh = (struct ether_header *)phdr;
206 	uint16_t etype;
207 #endif
208 
209 	if (plle)
210 		*plle = NULL;
211 
212 	switch (dst->sa_family) {
213 #ifdef INET
214 	case AF_INET:
215 		if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
216 			error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
217 			    plle);
218 		else {
219 			if (m->m_flags & M_BCAST)
220 				memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
221 				    ETHER_ADDR_LEN);
222 			else {
223 				const struct in_addr *a;
224 				a = &(((const struct sockaddr_in *)dst)->sin_addr);
225 				ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
226 			}
227 			etype = htons(ETHERTYPE_IP);
228 			memcpy(&eh->ether_type, &etype, sizeof(etype));
229 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
230 		}
231 		break;
232 #endif
233 #ifdef INET6
234 	case AF_INET6:
235 		if ((m->m_flags & M_MCAST) == 0) {
236 			int af = RO_GET_FAMILY(ro, dst);
237 			error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
238 			    &lleflags, plle);
239 		} else {
240 			const struct in6_addr *a6;
241 			a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
242 			ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
243 			etype = htons(ETHERTYPE_IPV6);
244 			memcpy(&eh->ether_type, &etype, sizeof(etype));
245 			memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
246 		}
247 		break;
248 #endif
249 	default:
250 		if_printf(ifp, "can't handle af%d\n", dst->sa_family);
251 		if (m != NULL)
252 			m_freem(m);
253 		return (EAFNOSUPPORT);
254 	}
255 
256 	if (error == EHOSTDOWN) {
257 		if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
258 			error = EHOSTUNREACH;
259 	}
260 
261 	if (error != 0)
262 		return (error);
263 
264 	*pflags = RT_MAY_LOOP;
265 	if (lleflags & LLE_IFADDR)
266 		*pflags |= RT_L2_ME;
267 
268 	return (0);
269 }
270 
271 /*
272  * Ethernet output routine.
273  * Encapsulate a packet of type family for the local net.
274  * Use trailer local net encapsulation if enough data in first
275  * packet leaves a multiple of 512 bytes of data in remainder.
276  */
277 int
ether_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro)278 ether_output(struct ifnet *ifp, struct mbuf *m,
279 	const struct sockaddr *dst, struct route *ro)
280 {
281 	int error = 0;
282 	char linkhdr[ETHER_HDR_LEN], *phdr;
283 	struct ether_header *eh;
284 	struct pf_mtag *t;
285 	bool loop_copy;
286 	int hlen;	/* link layer header length */
287 	uint32_t pflags;
288 	struct llentry *lle = NULL;
289 	int addref = 0;
290 
291 	phdr = NULL;
292 	pflags = 0;
293 	if (ro != NULL) {
294 		/* XXX BPF uses ro_prepend */
295 		if (ro->ro_prepend != NULL) {
296 			phdr = ro->ro_prepend;
297 			hlen = ro->ro_plen;
298 		} else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
299 			if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
300 				lle = ro->ro_lle;
301 				if (lle != NULL &&
302 				    (lle->la_flags & LLE_VALID) == 0) {
303 					LLE_FREE(lle);
304 					lle = NULL;	/* redundant */
305 					ro->ro_lle = NULL;
306 				}
307 				if (lle == NULL) {
308 					/* if we lookup, keep cache */
309 					addref = 1;
310 				} else
311 					/*
312 					 * Notify LLE code that
313 					 * the entry was used
314 					 * by datapath.
315 					 */
316 					llentry_provide_feedback(lle);
317 			}
318 			if (lle != NULL) {
319 				phdr = lle->r_linkdata;
320 				hlen = lle->r_hdrlen;
321 				pflags = lle->r_flags;
322 			}
323 		}
324 	}
325 
326 #ifdef MAC
327 	error = mac_ifnet_check_transmit(ifp, m);
328 	if (error)
329 		senderr(error);
330 #endif
331 
332 	M_PROFILE(m);
333 	if (ifp->if_flags & IFF_MONITOR)
334 		senderr(ENETDOWN);
335 	if (!((ifp->if_flags & IFF_UP) &&
336 	    (ifp->if_drv_flags & IFF_DRV_RUNNING)))
337 		senderr(ENETDOWN);
338 
339 	if (phdr == NULL) {
340 		/* No prepend data supplied. Try to calculate ourselves. */
341 		phdr = linkhdr;
342 		hlen = ETHER_HDR_LEN;
343 		error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
344 		    addref ? &lle : NULL);
345 		if (addref && lle != NULL)
346 			ro->ro_lle = lle;
347 		if (error != 0)
348 			return (error == EWOULDBLOCK ? 0 : error);
349 	}
350 
351 	if ((pflags & RT_L2_ME) != 0) {
352 		update_mbuf_csumflags(m, m);
353 		return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
354 	}
355 	loop_copy = (pflags & RT_MAY_LOOP) != 0;
356 
357 	/*
358 	 * Add local net header.  If no space in first mbuf,
359 	 * allocate another.
360 	 *
361 	 * Note that we do prepend regardless of RT_HAS_HEADER flag.
362 	 * This is done because BPF code shifts m_data pointer
363 	 * to the end of ethernet header prior to calling if_output().
364 	 */
365 	M_PREPEND(m, hlen, M_NOWAIT);
366 	if (m == NULL)
367 		senderr(ENOBUFS);
368 	if ((pflags & RT_HAS_HEADER) == 0) {
369 		eh = mtod(m, struct ether_header *);
370 		memcpy(eh, phdr, hlen);
371 	}
372 
373 	/*
374 	 * If a simplex interface, and the packet is being sent to our
375 	 * Ethernet address or a broadcast address, loopback a copy.
376 	 * XXX To make a simplex device behave exactly like a duplex
377 	 * device, we should copy in the case of sending to our own
378 	 * ethernet address (thus letting the original actually appear
379 	 * on the wire). However, we don't do that here for security
380 	 * reasons and compatibility with the original behavior.
381 	 */
382 	if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
383 	    ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
384 		struct mbuf *n;
385 
386 		/*
387 		 * Because if_simloop() modifies the packet, we need a
388 		 * writable copy through m_dup() instead of a readonly
389 		 * one as m_copy[m] would give us. The alternative would
390 		 * be to modify if_simloop() to handle the readonly mbuf,
391 		 * but performancewise it is mostly equivalent (trading
392 		 * extra data copying vs. extra locking).
393 		 *
394 		 * XXX This is a local workaround.  A number of less
395 		 * often used kernel parts suffer from the same bug.
396 		 * See PR kern/105943 for a proposed general solution.
397 		 */
398 		if ((n = m_dup(m, M_NOWAIT)) != NULL) {
399 			update_mbuf_csumflags(m, n);
400 			(void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
401 		} else
402 			if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
403 	}
404 
405        /*
406 	* Bridges require special output handling.
407 	*/
408 	if (ifp->if_bridge) {
409 		BRIDGE_OUTPUT(ifp, m, error);
410 		return (error);
411 	}
412 
413 #if defined(INET) || defined(INET6)
414 	if (ifp->if_carp &&
415 	    (error = (*carp_output_p)(ifp, m, dst)))
416 		goto bad;
417 #endif
418 
419 	/* Handle ng_ether(4) processing, if any */
420 	if (ifp->if_l2com != NULL) {
421 		KASSERT(ng_ether_output_p != NULL,
422 		    ("ng_ether_output_p is NULL"));
423 		if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
424 bad:			if (m != NULL)
425 				m_freem(m);
426 			return (error);
427 		}
428 		if (m == NULL)
429 			return (0);
430 	}
431 
432 	/* Continue with link-layer output */
433 	return ether_output_frame(ifp, m);
434 }
435 
436 static bool
ether_set_pcp(struct mbuf ** mp,struct ifnet * ifp,uint8_t pcp)437 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
438 {
439 	struct ether_8021q_tag qtag;
440 	struct ether_header *eh;
441 
442 	eh = mtod(*mp, struct ether_header *);
443 	if (eh->ether_type == htons(ETHERTYPE_VLAN) ||
444 	    eh->ether_type == htons(ETHERTYPE_QINQ)) {
445 		(*mp)->m_flags &= ~M_VLANTAG;
446 		return (true);
447 	}
448 
449 	qtag.vid = 0;
450 	qtag.pcp = pcp;
451 	qtag.proto = ETHERTYPE_VLAN;
452 	if (ether_8021q_frame(mp, ifp, ifp, &qtag))
453 		return (true);
454 	if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
455 	return (false);
456 }
457 
458 /*
459  * Ethernet link layer output routine to send a raw frame to the device.
460  *
461  * This assumes that the 14 byte Ethernet header is present and contiguous
462  * in the first mbuf (if BRIDGE'ing).
463  */
464 int
ether_output_frame(struct ifnet * ifp,struct mbuf * m)465 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
466 {
467 	if (ether_do_pcp(ifp, m) && !ether_set_pcp(&m, ifp, ifp->if_pcp))
468 		return (0);
469 
470 	if (PFIL_HOOKED_OUT(V_link_pfil_head))
471 		switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, NULL)) {
472 		case PFIL_DROPPED:
473 			return (EACCES);
474 		case PFIL_CONSUMED:
475 			return (0);
476 		}
477 
478 	/*
479 	 * Queue message on interface, update output statistics if successful,
480 	 * and start output if interface not yet active.
481 	 *
482 	 * If KMSAN is enabled, use it to verify that the data does not contain
483 	 * any uninitialized bytes.
484 	 */
485 	kmsan_check_mbuf(m, "ether_output");
486 	return ((ifp->if_transmit)(ifp, m));
487 }
488 
489 /*
490  * Process a received Ethernet packet; the packet is in the
491  * mbuf chain m with the ethernet header at the front.
492  */
493 static void
ether_input_internal(struct ifnet * ifp,struct mbuf * m)494 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
495 {
496 	struct ether_header *eh;
497 	u_short etype;
498 
499 	if ((ifp->if_flags & IFF_UP) == 0) {
500 		m_freem(m);
501 		return;
502 	}
503 #ifdef DIAGNOSTIC
504 	if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
505 		if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
506 		m_freem(m);
507 		return;
508 	}
509 #endif
510 	if (__predict_false(m->m_len < ETHER_HDR_LEN)) {
511 		/* Drivers should pullup and ensure the mbuf is valid */
512 		if_printf(ifp, "discard frame w/o leading ethernet "
513 				"header (len %d pkt len %d)\n",
514 				m->m_len, m->m_pkthdr.len);
515 		if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
516 		m_freem(m);
517 		return;
518 	}
519 	eh = mtod(m, struct ether_header *);
520 	etype = ntohs(eh->ether_type);
521 	random_harvest_queue_ether(m, sizeof(*m));
522 
523 	CURVNET_SET_QUIET(ifp->if_vnet);
524 
525 	if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
526 		if (ETHER_IS_BROADCAST(eh->ether_dhost))
527 			m->m_flags |= M_BCAST;
528 		else
529 			m->m_flags |= M_MCAST;
530 		if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
531 	}
532 
533 #ifdef MAC
534 	/*
535 	 * Tag the mbuf with an appropriate MAC label before any other
536 	 * consumers can get to it.
537 	 */
538 	mac_ifnet_create_mbuf(ifp, m);
539 #endif
540 
541 	/*
542 	 * Give bpf a chance at the packet.
543 	 */
544 	ETHER_BPF_MTAP(ifp, m);
545 
546 	if (!(ifp->if_capenable & IFCAP_HWSTATS))
547 		if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
548 
549 	/* Allow monitor mode to claim this frame, after stats are updated. */
550 	if (ifp->if_flags & IFF_MONITOR) {
551 		m_freem(m);
552 		CURVNET_RESTORE();
553 		return;
554 	}
555 
556 	/* Handle input from a lagg(4) port */
557 	if (ifp->if_type == IFT_IEEE8023ADLAG) {
558 		KASSERT(lagg_input_ethernet_p != NULL,
559 		    ("%s: if_lagg not loaded!", __func__));
560 		m = (*lagg_input_ethernet_p)(ifp, m);
561 		if (m != NULL)
562 			ifp = m->m_pkthdr.rcvif;
563 		else {
564 			CURVNET_RESTORE();
565 			return;
566 		}
567 	}
568 
569 	/*
570 	 * If the hardware did not process an 802.1Q tag, do this now,
571 	 * to allow 802.1P priority frames to be passed to the main input
572 	 * path correctly.
573 	 */
574 	if ((m->m_flags & M_VLANTAG) == 0 &&
575 	    ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
576 		struct ether_vlan_header *evl;
577 
578 		if (m->m_len < sizeof(*evl) &&
579 		    (m = m_pullup(m, sizeof(*evl))) == NULL) {
580 #ifdef DIAGNOSTIC
581 			if_printf(ifp, "cannot pullup VLAN header\n");
582 #endif
583 			if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
584 			CURVNET_RESTORE();
585 			return;
586 		}
587 
588 		evl = mtod(m, struct ether_vlan_header *);
589 		m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
590 		m->m_flags |= M_VLANTAG;
591 
592 		bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
593 		    ETHER_HDR_LEN - ETHER_TYPE_LEN);
594 		m_adj(m, ETHER_VLAN_ENCAP_LEN);
595 		eh = mtod(m, struct ether_header *);
596 	}
597 
598 	M_SETFIB(m, ifp->if_fib);
599 
600 	/* Allow ng_ether(4) to claim this frame. */
601 	if (ifp->if_l2com != NULL) {
602 		KASSERT(ng_ether_input_p != NULL,
603 		    ("%s: ng_ether_input_p is NULL", __func__));
604 		m->m_flags &= ~M_PROMISC;
605 		(*ng_ether_input_p)(ifp, &m);
606 		if (m == NULL) {
607 			CURVNET_RESTORE();
608 			return;
609 		}
610 		eh = mtod(m, struct ether_header *);
611 	}
612 
613 	/*
614 	 * Allow if_bridge(4) to claim this frame.
615 	 *
616 	 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
617 	 * and the frame should be delivered locally.
618 	 *
619 	 * If M_BRIDGE_INJECT is set, the packet was received directly by the
620 	 * bridge via netmap, so "ifp" is the bridge itself and the packet
621 	 * should be re-examined.
622 	 */
623 	if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) {
624 		m->m_flags &= ~M_PROMISC;
625 		BRIDGE_INPUT(ifp, m);
626 		if (m == NULL) {
627 			CURVNET_RESTORE();
628 			return;
629 		}
630 		eh = mtod(m, struct ether_header *);
631 	}
632 
633 #if defined(INET) || defined(INET6)
634 	/*
635 	 * Clear M_PROMISC on frame so that carp(4) will see it when the
636 	 * mbuf flows up to Layer 3.
637 	 * FreeBSD's implementation of carp(4) uses the inprotosw
638 	 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
639 	 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
640 	 * is outside the scope of the M_PROMISC test below.
641 	 * TODO: Maintain a hash table of ethernet addresses other than
642 	 * ether_dhost which may be active on this ifp.
643 	 */
644 	if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
645 		m->m_flags &= ~M_PROMISC;
646 	} else
647 #endif
648 	{
649 		/*
650 		 * If the frame received was not for our MAC address, set the
651 		 * M_PROMISC flag on the mbuf chain. The frame may need to
652 		 * be seen by the rest of the Ethernet input path in case of
653 		 * re-entry (e.g. bridge, vlan, netgraph) but should not be
654 		 * seen by upper protocol layers.
655 		 */
656 		if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
657 		    memcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
658 			m->m_flags |= M_PROMISC;
659 	}
660 
661 	ether_demux(ifp, m);
662 	CURVNET_RESTORE();
663 }
664 
665 /*
666  * Ethernet input dispatch; by default, direct dispatch here regardless of
667  * global configuration.  However, if RSS is enabled, hook up RSS affinity
668  * so that when deferred or hybrid dispatch is enabled, we can redistribute
669  * load based on RSS.
670  *
671  * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
672  * not it had already done work distribution via multi-queue.  Then we could
673  * direct dispatch in the event load balancing was already complete and
674  * handle the case of interfaces with different capabilities better.
675  *
676  * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
677  * at multiple layers?
678  *
679  * XXXRW: For now, enable all this only if RSS is compiled in, although it
680  * works fine without RSS.  Need to characterise the performance overhead
681  * of the detour through the netisr code in the event the result is always
682  * direct dispatch.
683  */
684 static void
ether_nh_input(struct mbuf * m)685 ether_nh_input(struct mbuf *m)
686 {
687 
688 	M_ASSERTPKTHDR(m);
689 	KASSERT(m->m_pkthdr.rcvif != NULL,
690 	    ("%s: NULL interface pointer", __func__));
691 	ether_input_internal(m->m_pkthdr.rcvif, m);
692 }
693 
694 static struct netisr_handler	ether_nh = {
695 	.nh_name = "ether",
696 	.nh_handler = ether_nh_input,
697 	.nh_proto = NETISR_ETHER,
698 #ifdef RSS
699 	.nh_policy = NETISR_POLICY_CPU,
700 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
701 	.nh_m2cpuid = rss_m2cpuid,
702 #else
703 	.nh_policy = NETISR_POLICY_SOURCE,
704 	.nh_dispatch = NETISR_DISPATCH_DIRECT,
705 #endif
706 };
707 
708 static void
ether_init(__unused void * arg)709 ether_init(__unused void *arg)
710 {
711 
712 	netisr_register(&ether_nh);
713 }
714 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
715 
716 static void
vnet_ether_init(const __unused void * arg)717 vnet_ether_init(const __unused void *arg)
718 {
719 	struct pfil_head_args args;
720 
721 	args.pa_version = PFIL_VERSION;
722 	args.pa_flags = PFIL_IN | PFIL_OUT;
723 	args.pa_type = PFIL_TYPE_ETHERNET;
724 	args.pa_headname = PFIL_ETHER_NAME;
725 	V_link_pfil_head = pfil_head_register(&args);
726 
727 #ifdef VIMAGE
728 	netisr_register_vnet(&ether_nh);
729 #endif
730 }
731 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
732     vnet_ether_init, NULL);
733 
734 #ifdef VIMAGE
735 static void
vnet_ether_pfil_destroy(const __unused void * arg)736 vnet_ether_pfil_destroy(const __unused void *arg)
737 {
738 
739 	pfil_head_unregister(V_link_pfil_head);
740 }
741 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
742     vnet_ether_pfil_destroy, NULL);
743 
744 static void
vnet_ether_destroy(__unused void * arg)745 vnet_ether_destroy(__unused void *arg)
746 {
747 
748 	netisr_unregister_vnet(&ether_nh);
749 }
750 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
751     vnet_ether_destroy, NULL);
752 #endif
753 
754 static void
ether_input(struct ifnet * ifp,struct mbuf * m)755 ether_input(struct ifnet *ifp, struct mbuf *m)
756 {
757 	struct epoch_tracker et;
758 	struct mbuf *mn;
759 	bool needs_epoch;
760 
761 	needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH);
762 #ifdef INVARIANTS
763 	/*
764 	 * This temporary code is here to prevent epoch unaware and unmarked
765 	 * drivers to panic the system.  Once all drivers are taken care of,
766 	 * the whole INVARIANTS block should go away.
767 	 */
768 	if (!needs_epoch && !in_epoch(net_epoch_preempt)) {
769 		static bool printedonce;
770 
771 		needs_epoch = true;
772 		if (!printedonce) {
773 			printedonce = true;
774 			if_printf(ifp, "called %s w/o net epoch! "
775 			    "PLEASE file a bug report.", __func__);
776 #ifdef KDB
777 			kdb_backtrace();
778 #endif
779 		}
780 	}
781 #endif
782 
783 	/*
784 	 * The drivers are allowed to pass in a chain of packets linked with
785 	 * m_nextpkt. We split them up into separate packets here and pass
786 	 * them up. This allows the drivers to amortize the receive lock.
787 	 */
788 	CURVNET_SET_QUIET(ifp->if_vnet);
789 	if (__predict_false(needs_epoch))
790 		NET_EPOCH_ENTER(et);
791 	while (m) {
792 		mn = m->m_nextpkt;
793 		m->m_nextpkt = NULL;
794 
795 		/*
796 		 * We will rely on rcvif being set properly in the deferred
797 		 * context, so assert it is correct here.
798 		 */
799 		MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
800 		KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
801 		    "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
802 		netisr_dispatch(NETISR_ETHER, m);
803 		m = mn;
804 	}
805 	if (__predict_false(needs_epoch))
806 		NET_EPOCH_EXIT(et);
807 	CURVNET_RESTORE();
808 }
809 
810 /*
811  * Upper layer processing for a received Ethernet packet.
812  */
813 void
ether_demux(struct ifnet * ifp,struct mbuf * m)814 ether_demux(struct ifnet *ifp, struct mbuf *m)
815 {
816 	struct ether_header *eh;
817 	int i, isr;
818 	u_short ether_type;
819 
820 	NET_EPOCH_ASSERT();
821 	KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
822 
823 	/* Do not grab PROMISC frames in case we are re-entered. */
824 	if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
825 		i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL);
826 		if (i != PFIL_PASS)
827 			return;
828 	}
829 
830 	eh = mtod(m, struct ether_header *);
831 	ether_type = ntohs(eh->ether_type);
832 
833 	/*
834 	 * If this frame has a VLAN tag other than 0, call vlan_input()
835 	 * if its module is loaded. Otherwise, drop.
836 	 */
837 	if ((m->m_flags & M_VLANTAG) &&
838 	    EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
839 		if (ifp->if_vlantrunk == NULL) {
840 			if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
841 			m_freem(m);
842 			return;
843 		}
844 		KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
845 		    __func__));
846 		/* Clear before possibly re-entering ether_input(). */
847 		m->m_flags &= ~M_PROMISC;
848 		(*vlan_input_p)(ifp, m);
849 		return;
850 	}
851 
852 	/*
853 	 * Pass promiscuously received frames to the upper layer if the user
854 	 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
855 	 */
856 	if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
857 		m_freem(m);
858 		return;
859 	}
860 
861 	/*
862 	 * Reset layer specific mbuf flags to avoid confusing upper layers.
863 	 */
864 	m->m_flags &= ~M_VLANTAG;
865 	m_clrprotoflags(m);
866 
867 	/*
868 	 * Dispatch frame to upper layer.
869 	 */
870 	switch (ether_type) {
871 #ifdef INET
872 	case ETHERTYPE_IP:
873 		isr = NETISR_IP;
874 		break;
875 
876 	case ETHERTYPE_ARP:
877 		if (ifp->if_flags & IFF_NOARP) {
878 			/* Discard packet if ARP is disabled on interface */
879 			m_freem(m);
880 			return;
881 		}
882 		isr = NETISR_ARP;
883 		break;
884 #endif
885 #ifdef INET6
886 	case ETHERTYPE_IPV6:
887 		isr = NETISR_IPV6;
888 		break;
889 #endif
890 	default:
891 		goto discard;
892 	}
893 
894 	/* Strip off Ethernet header. */
895 	m_adj(m, ETHER_HDR_LEN);
896 
897 	netisr_dispatch(isr, m);
898 	return;
899 
900 discard:
901 	/*
902 	 * Packet is to be discarded.  If netgraph is present,
903 	 * hand the packet to it for last chance processing;
904 	 * otherwise dispose of it.
905 	 */
906 	if (ifp->if_l2com != NULL) {
907 		KASSERT(ng_ether_input_orphan_p != NULL,
908 		    ("ng_ether_input_orphan_p is NULL"));
909 		(*ng_ether_input_orphan_p)(ifp, m);
910 		return;
911 	}
912 	m_freem(m);
913 }
914 
915 /*
916  * Convert Ethernet address to printable (loggable) representation.
917  * This routine is for compatibility; it's better to just use
918  *
919  *	printf("%6D", <pointer to address>, ":");
920  *
921  * since there's no static buffer involved.
922  */
923 char *
ether_sprintf(const u_char * ap)924 ether_sprintf(const u_char *ap)
925 {
926 	static char etherbuf[18];
927 	snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
928 	return (etherbuf);
929 }
930 
931 /*
932  * Perform common duties while attaching to interface list
933  */
934 void
ether_ifattach(struct ifnet * ifp,const u_int8_t * lla)935 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
936 {
937 	int i;
938 	struct ifaddr *ifa;
939 	struct sockaddr_dl *sdl;
940 
941 	ifp->if_addrlen = ETHER_ADDR_LEN;
942 	ifp->if_hdrlen = (ifp->if_capabilities & IFCAP_VLAN_MTU) != 0 ?
943 	    ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN : ETHER_HDR_LEN;
944 	ifp->if_mtu = ETHERMTU;
945 	if_attach(ifp);
946 	ifp->if_output = ether_output;
947 	ifp->if_input = ether_input;
948 	ifp->if_resolvemulti = ether_resolvemulti;
949 	ifp->if_requestencap = ether_requestencap;
950 	if (ifp->if_baudrate == 0)
951 		ifp->if_baudrate = IF_Mbps(10);		/* just a default */
952 	ifp->if_broadcastaddr = etherbroadcastaddr;
953 
954 	ifa = ifp->if_addr;
955 	KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
956 	sdl = (struct sockaddr_dl *)ifa->ifa_addr;
957 	sdl->sdl_type = IFT_ETHER;
958 	sdl->sdl_alen = ifp->if_addrlen;
959 	bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
960 
961 	if (ifp->if_hw_addr != NULL)
962 		bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
963 
964 	bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
965 
966 	/* Announce Ethernet MAC address if non-zero. */
967 	for (i = 0; i < ifp->if_addrlen; i++)
968 		if (lla[i] != 0)
969 			break;
970 	if (i != ifp->if_addrlen)
971 		if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
972 
973 	uuid_ether_add(LLADDR(sdl));
974 
975 	/* Add necessary bits are setup; announce it now. */
976 	EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
977 	if (IS_DEFAULT_VNET(curvnet))
978 		devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
979 }
980 
981 /*
982  * Perform common duties while detaching an Ethernet interface
983  */
984 void
ether_ifdetach(struct ifnet * ifp)985 ether_ifdetach(struct ifnet *ifp)
986 {
987 	struct sockaddr_dl *sdl;
988 
989 	sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
990 	uuid_ether_del(LLADDR(sdl));
991 
992 	bpfdetach(ifp);
993 	if_detach(ifp);
994 }
995 
996 SYSCTL_DECL(_net_link);
997 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
998     "Ethernet");
999 
1000 #if 0
1001 /*
1002  * This is for reference.  We have a table-driven version
1003  * of the little-endian crc32 generator, which is faster
1004  * than the double-loop.
1005  */
1006 uint32_t
1007 ether_crc32_le(const uint8_t *buf, size_t len)
1008 {
1009 	size_t i;
1010 	uint32_t crc;
1011 	int bit;
1012 	uint8_t data;
1013 
1014 	crc = 0xffffffff;	/* initial value */
1015 
1016 	for (i = 0; i < len; i++) {
1017 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1018 			carry = (crc ^ data) & 1;
1019 			crc >>= 1;
1020 			if (carry)
1021 				crc = (crc ^ ETHER_CRC_POLY_LE);
1022 		}
1023 	}
1024 
1025 	return (crc);
1026 }
1027 #else
1028 uint32_t
ether_crc32_le(const uint8_t * buf,size_t len)1029 ether_crc32_le(const uint8_t *buf, size_t len)
1030 {
1031 	static const uint32_t crctab[] = {
1032 		0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1033 		0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1034 		0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1035 		0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1036 	};
1037 	size_t i;
1038 	uint32_t crc;
1039 
1040 	crc = 0xffffffff;	/* initial value */
1041 
1042 	for (i = 0; i < len; i++) {
1043 		crc ^= buf[i];
1044 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1045 		crc = (crc >> 4) ^ crctab[crc & 0xf];
1046 	}
1047 
1048 	return (crc);
1049 }
1050 #endif
1051 
1052 uint32_t
ether_crc32_be(const uint8_t * buf,size_t len)1053 ether_crc32_be(const uint8_t *buf, size_t len)
1054 {
1055 	size_t i;
1056 	uint32_t crc, carry;
1057 	int bit;
1058 	uint8_t data;
1059 
1060 	crc = 0xffffffff;	/* initial value */
1061 
1062 	for (i = 0; i < len; i++) {
1063 		for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1064 			carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1065 			crc <<= 1;
1066 			if (carry)
1067 				crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1068 		}
1069 	}
1070 
1071 	return (crc);
1072 }
1073 
1074 int
ether_ioctl(struct ifnet * ifp,u_long command,caddr_t data)1075 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1076 {
1077 	struct ifaddr *ifa = (struct ifaddr *) data;
1078 	struct ifreq *ifr = (struct ifreq *) data;
1079 	int error = 0;
1080 
1081 	switch (command) {
1082 	case SIOCSIFADDR:
1083 		ifp->if_flags |= IFF_UP;
1084 
1085 		switch (ifa->ifa_addr->sa_family) {
1086 #ifdef INET
1087 		case AF_INET:
1088 			ifp->if_init(ifp->if_softc);	/* before arpwhohas */
1089 			arp_ifinit(ifp, ifa);
1090 			break;
1091 #endif
1092 		default:
1093 			ifp->if_init(ifp->if_softc);
1094 			break;
1095 		}
1096 		break;
1097 
1098 	case SIOCGIFADDR:
1099 		bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1100 		    ETHER_ADDR_LEN);
1101 		break;
1102 
1103 	case SIOCSIFMTU:
1104 		/*
1105 		 * Set the interface MTU.
1106 		 */
1107 		if (ifr->ifr_mtu > ETHERMTU) {
1108 			error = EINVAL;
1109 		} else {
1110 			ifp->if_mtu = ifr->ifr_mtu;
1111 		}
1112 		break;
1113 
1114 	case SIOCSLANPCP:
1115 		error = priv_check(curthread, PRIV_NET_SETLANPCP);
1116 		if (error != 0)
1117 			break;
1118 		if (ifr->ifr_lan_pcp > 7 &&
1119 		    ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1120 			error = EINVAL;
1121 		} else {
1122 			ifp->if_pcp = ifr->ifr_lan_pcp;
1123 			/* broadcast event about PCP change */
1124 			EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1125 		}
1126 		break;
1127 
1128 	case SIOCGLANPCP:
1129 		ifr->ifr_lan_pcp = ifp->if_pcp;
1130 		break;
1131 
1132 	default:
1133 		error = EINVAL;			/* XXX netbsd has ENOTTY??? */
1134 		break;
1135 	}
1136 	return (error);
1137 }
1138 
1139 static int
ether_resolvemulti(struct ifnet * ifp,struct sockaddr ** llsa,struct sockaddr * sa)1140 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1141 	struct sockaddr *sa)
1142 {
1143 	struct sockaddr_dl *sdl;
1144 #ifdef INET
1145 	struct sockaddr_in *sin;
1146 #endif
1147 #ifdef INET6
1148 	struct sockaddr_in6 *sin6;
1149 #endif
1150 	u_char *e_addr;
1151 
1152 	switch(sa->sa_family) {
1153 	case AF_LINK:
1154 		/*
1155 		 * No mapping needed. Just check that it's a valid MC address.
1156 		 */
1157 		sdl = (struct sockaddr_dl *)sa;
1158 		e_addr = LLADDR(sdl);
1159 		if (!ETHER_IS_MULTICAST(e_addr))
1160 			return EADDRNOTAVAIL;
1161 		*llsa = NULL;
1162 		return 0;
1163 
1164 #ifdef INET
1165 	case AF_INET:
1166 		sin = (struct sockaddr_in *)sa;
1167 		if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1168 			return EADDRNOTAVAIL;
1169 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1170 		sdl->sdl_alen = ETHER_ADDR_LEN;
1171 		e_addr = LLADDR(sdl);
1172 		ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1173 		*llsa = (struct sockaddr *)sdl;
1174 		return 0;
1175 #endif
1176 #ifdef INET6
1177 	case AF_INET6:
1178 		sin6 = (struct sockaddr_in6 *)sa;
1179 		if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1180 			/*
1181 			 * An IP6 address of 0 means listen to all
1182 			 * of the Ethernet multicast address used for IP6.
1183 			 * (This is used for multicast routers.)
1184 			 */
1185 			ifp->if_flags |= IFF_ALLMULTI;
1186 			*llsa = NULL;
1187 			return 0;
1188 		}
1189 		if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1190 			return EADDRNOTAVAIL;
1191 		sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1192 		sdl->sdl_alen = ETHER_ADDR_LEN;
1193 		e_addr = LLADDR(sdl);
1194 		ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1195 		*llsa = (struct sockaddr *)sdl;
1196 		return 0;
1197 #endif
1198 
1199 	default:
1200 		/*
1201 		 * Well, the text isn't quite right, but it's the name
1202 		 * that counts...
1203 		 */
1204 		return EAFNOSUPPORT;
1205 	}
1206 }
1207 
1208 static moduledata_t ether_mod = {
1209 	.name = "ether",
1210 };
1211 
1212 void
ether_vlan_mtap(struct bpf_if * bp,struct mbuf * m,void * data,u_int dlen)1213 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1214 {
1215 	struct ether_vlan_header vlan;
1216 	struct mbuf mv, mb;
1217 
1218 	KASSERT((m->m_flags & M_VLANTAG) != 0,
1219 	    ("%s: vlan information not present", __func__));
1220 	KASSERT(m->m_len >= sizeof(struct ether_header),
1221 	    ("%s: mbuf not large enough for header", __func__));
1222 	bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1223 	vlan.evl_proto = vlan.evl_encap_proto;
1224 	vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1225 	vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1226 	m->m_len -= sizeof(struct ether_header);
1227 	m->m_data += sizeof(struct ether_header);
1228 	/*
1229 	 * If a data link has been supplied by the caller, then we will need to
1230 	 * re-create a stack allocated mbuf chain with the following structure:
1231 	 *
1232 	 * (1) mbuf #1 will contain the supplied data link
1233 	 * (2) mbuf #2 will contain the vlan header
1234 	 * (3) mbuf #3 will contain the original mbuf's packet data
1235 	 *
1236 	 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1237 	 */
1238 	if (data != NULL) {
1239 		mv.m_next = m;
1240 		mv.m_data = (caddr_t)&vlan;
1241 		mv.m_len = sizeof(vlan);
1242 		mb.m_next = &mv;
1243 		mb.m_data = data;
1244 		mb.m_len = dlen;
1245 		bpf_mtap(bp, &mb);
1246 	} else
1247 		bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1248 	m->m_len += sizeof(struct ether_header);
1249 	m->m_data -= sizeof(struct ether_header);
1250 }
1251 
1252 struct mbuf *
ether_vlanencap_proto(struct mbuf * m,uint16_t tag,uint16_t proto)1253 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1254 {
1255 	struct ether_vlan_header *evl;
1256 
1257 	M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1258 	if (m == NULL)
1259 		return (NULL);
1260 	/* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1261 
1262 	if (m->m_len < sizeof(*evl)) {
1263 		m = m_pullup(m, sizeof(*evl));
1264 		if (m == NULL)
1265 			return (NULL);
1266 	}
1267 
1268 	/*
1269 	 * Transform the Ethernet header into an Ethernet header
1270 	 * with 802.1Q encapsulation.
1271 	 */
1272 	evl = mtod(m, struct ether_vlan_header *);
1273 	bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1274 	    (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1275 	evl->evl_encap_proto = htons(proto);
1276 	evl->evl_tag = htons(tag);
1277 	return (m);
1278 }
1279 
1280 void
ether_bpf_mtap_if(struct ifnet * ifp,struct mbuf * m)1281 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m)
1282 {
1283 	if (bpf_peers_present(ifp->if_bpf)) {
1284 		M_ASSERTVALID(m);
1285 		if ((m->m_flags & M_VLANTAG) != 0)
1286 			ether_vlan_mtap(ifp->if_bpf, m, NULL, 0);
1287 		else
1288 			bpf_mtap(ifp->if_bpf, m);
1289 	}
1290 }
1291 
1292 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1293     "IEEE 802.1Q VLAN");
1294 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1295     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1296     "for consistency");
1297 
1298 VNET_DEFINE_STATIC(int, soft_pad);
1299 #define	V_soft_pad	VNET(soft_pad)
1300 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1301     &VNET_NAME(soft_pad), 0,
1302     "pad short frames before tagging");
1303 
1304 /*
1305  * For now, make preserving PCP via an mbuf tag optional, as it increases
1306  * per-packet memory allocations and frees.  In the future, it would be
1307  * preferable to reuse ether_vtag for this, or similar.
1308  */
1309 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1310 #define	V_vlan_mtag_pcp	VNET(vlan_mtag_pcp)
1311 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1312     &VNET_NAME(vlan_mtag_pcp), 0,
1313     "Retain VLAN PCP information as packets are passed up the stack");
1314 
1315 static inline bool
ether_do_pcp(struct ifnet * ifp,struct mbuf * m)1316 ether_do_pcp(struct ifnet *ifp, struct mbuf *m)
1317 {
1318 	if (ifp->if_type == IFT_L2VLAN)
1319 		return (false);
1320 	if (ifp->if_pcp != IFNET_PCP_NONE || (m->m_flags & M_VLANTAG) != 0)
1321 		return (true);
1322 	if (V_vlan_mtag_pcp &&
1323 	    m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_OUT, NULL) != NULL)
1324 		return (true);
1325 	return (false);
1326 }
1327 
1328 bool
ether_8021q_frame(struct mbuf ** mp,struct ifnet * ife,struct ifnet * p,const struct ether_8021q_tag * qtag)1329 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1330     const struct ether_8021q_tag *qtag)
1331 {
1332 	struct m_tag *mtag;
1333 	int n;
1334 	uint16_t tag;
1335 	uint8_t pcp = qtag->pcp;
1336 	static const char pad[8];	/* just zeros */
1337 
1338 	/*
1339 	 * Pad the frame to the minimum size allowed if told to.
1340 	 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1341 	 * paragraph C.4.4.3.b.  It can help to work around buggy
1342 	 * bridges that violate paragraph C.4.4.3.a from the same
1343 	 * document, i.e., fail to pad short frames after untagging.
1344 	 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1345 	 * untagging it will produce a 62-byte frame, which is a runt
1346 	 * and requires padding.  There are VLAN-enabled network
1347 	 * devices that just discard such runts instead or mishandle
1348 	 * them somehow.
1349 	 */
1350 	if (V_soft_pad && p->if_type == IFT_ETHER) {
1351 		for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1352 		     n > 0; n -= sizeof(pad)) {
1353 			if (!m_append(*mp, min(n, sizeof(pad)), pad))
1354 				break;
1355 		}
1356 		if (n > 0) {
1357 			m_freem(*mp);
1358 			*mp = NULL;
1359 			if_printf(ife, "cannot pad short frame");
1360 			return (false);
1361 		}
1362 	}
1363 
1364 	/*
1365 	 * If PCP is set in mbuf, use it
1366 	 */
1367 	if ((*mp)->m_flags & M_VLANTAG) {
1368 		pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1369 	}
1370 
1371 	/*
1372 	 * If underlying interface can do VLAN tag insertion itself,
1373 	 * just pass the packet along. However, we need some way to
1374 	 * tell the interface where the packet came from so that it
1375 	 * knows how to find the VLAN tag to use, so we attach a
1376 	 * packet tag that holds it.
1377 	 */
1378 	if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1379 	    MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1380 		tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1381 	else
1382 		tag = EVL_MAKETAG(qtag->vid, pcp, 0);
1383 	if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1384 	    (qtag->proto == ETHERTYPE_VLAN)) {
1385 		(*mp)->m_pkthdr.ether_vtag = tag;
1386 		(*mp)->m_flags |= M_VLANTAG;
1387 	} else {
1388 		*mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1389 		if (*mp == NULL) {
1390 			if_printf(ife, "unable to prepend 802.1Q header");
1391 			return (false);
1392 		}
1393 		(*mp)->m_flags &= ~M_VLANTAG;
1394 	}
1395 	return (true);
1396 }
1397 
1398 /*
1399  * Allocate an address from the FreeBSD Foundation OUI.  This uses a
1400  * cryptographic hash function on the containing jail's name, UUID and the
1401  * interface name to attempt to provide a unique but stable address.
1402  * Pseudo-interfaces which require a MAC address should use this function to
1403  * allocate non-locally-administered addresses.
1404  */
1405 void
ether_gen_addr_byname(const char * nameunit,struct ether_addr * hwaddr)1406 ether_gen_addr_byname(const char *nameunit, struct ether_addr *hwaddr)
1407 {
1408 	SHA1_CTX ctx;
1409 	char *buf;
1410 	char uuid[HOSTUUIDLEN + 1];
1411 	uint64_t addr;
1412 	int i, sz;
1413 	unsigned char digest[SHA1_RESULTLEN];
1414 	char jailname[MAXHOSTNAMELEN];
1415 
1416 	getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1417 	if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1418 		/* Fall back to a random mac address. */
1419 		goto rando;
1420 	}
1421 
1422 	/* If each (vnet) jail would also have a unique hostuuid this would not
1423 	 * be necessary. */
1424 	getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1425 	sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, nameunit,
1426 	    jailname);
1427 	if (sz < 0) {
1428 		/* Fall back to a random mac address. */
1429 		goto rando;
1430 	}
1431 
1432 	SHA1Init(&ctx);
1433 	SHA1Update(&ctx, buf, sz);
1434 	SHA1Final(digest, &ctx);
1435 	free(buf, M_TEMP);
1436 
1437 	addr = (digest[0] << 8) | digest[1] | OUI_FREEBSD_GENERATED_LOW;
1438 	for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1439 		hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1440 		    0xFF;
1441 	}
1442 
1443 	return;
1444 rando:
1445 	arc4rand(hwaddr, sizeof(*hwaddr), 0);
1446 	/* Unicast */
1447 	hwaddr->octet[0] &= 0xFE;
1448 	/* Locally administered. */
1449 	hwaddr->octet[0] |= 0x02;
1450 }
1451 
1452 void
ether_gen_addr(struct ifnet * ifp,struct ether_addr * hwaddr)1453 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1454 {
1455 	ether_gen_addr_byname(if_name(ifp), hwaddr);
1456 }
1457 
1458 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1459 MODULE_VERSION(ether, 1);
1460