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 #ifdef EXPERIMENTAL
479 #if defined(INET6) && defined(INET)
480 /* draft-ietf-6man-ipv6only-flag */
481 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
482 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
483 struct ether_header *eh;
484
485 eh = mtod(m, struct ether_header *);
486 switch (ntohs(eh->ether_type)) {
487 case ETHERTYPE_IP:
488 case ETHERTYPE_ARP:
489 case ETHERTYPE_REVARP:
490 m_freem(m);
491 return (EAFNOSUPPORT);
492 /* NOTREACHED */
493 break;
494 };
495 }
496 #endif
497 #endif
498
499 /*
500 * Queue message on interface, update output statistics if successful,
501 * and start output if interface not yet active.
502 *
503 * If KMSAN is enabled, use it to verify that the data does not contain
504 * any uninitialized bytes.
505 */
506 kmsan_check_mbuf(m, "ether_output");
507 return ((ifp->if_transmit)(ifp, m));
508 }
509
510 /*
511 * Process a received Ethernet packet; the packet is in the
512 * mbuf chain m with the ethernet header at the front.
513 */
514 static void
ether_input_internal(struct ifnet * ifp,struct mbuf * m)515 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
516 {
517 struct ether_header *eh;
518 u_short etype;
519
520 if ((ifp->if_flags & IFF_UP) == 0) {
521 m_freem(m);
522 return;
523 }
524 #ifdef DIAGNOSTIC
525 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
526 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
527 m_freem(m);
528 return;
529 }
530 #endif
531 if (__predict_false(m->m_len < ETHER_HDR_LEN)) {
532 /* Drivers should pullup and ensure the mbuf is valid */
533 if_printf(ifp, "discard frame w/o leading ethernet "
534 "header (len %d pkt len %d)\n",
535 m->m_len, m->m_pkthdr.len);
536 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
537 m_freem(m);
538 return;
539 }
540 eh = mtod(m, struct ether_header *);
541 etype = ntohs(eh->ether_type);
542 random_harvest_queue_ether(m, sizeof(*m));
543
544 #ifdef EXPERIMENTAL
545 #if defined(INET6) && defined(INET)
546 /* draft-ietf-6man-ipv6only-flag */
547 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
548 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
549 switch (etype) {
550 case ETHERTYPE_IP:
551 case ETHERTYPE_ARP:
552 case ETHERTYPE_REVARP:
553 m_freem(m);
554 return;
555 /* NOTREACHED */
556 break;
557 };
558 }
559 #endif
560 #endif
561
562 CURVNET_SET_QUIET(ifp->if_vnet);
563
564 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
565 if (ETHER_IS_BROADCAST(eh->ether_dhost))
566 m->m_flags |= M_BCAST;
567 else
568 m->m_flags |= M_MCAST;
569 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
570 }
571
572 #ifdef MAC
573 /*
574 * Tag the mbuf with an appropriate MAC label before any other
575 * consumers can get to it.
576 */
577 mac_ifnet_create_mbuf(ifp, m);
578 #endif
579
580 /*
581 * Give bpf a chance at the packet.
582 */
583 ETHER_BPF_MTAP(ifp, m);
584
585 if (!(ifp->if_capenable & IFCAP_HWSTATS))
586 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
587
588 /* Allow monitor mode to claim this frame, after stats are updated. */
589 if (ifp->if_flags & IFF_MONITOR) {
590 m_freem(m);
591 CURVNET_RESTORE();
592 return;
593 }
594
595 /* Handle input from a lagg(4) port */
596 if (ifp->if_type == IFT_IEEE8023ADLAG) {
597 KASSERT(lagg_input_ethernet_p != NULL,
598 ("%s: if_lagg not loaded!", __func__));
599 m = (*lagg_input_ethernet_p)(ifp, m);
600 if (m != NULL)
601 ifp = m->m_pkthdr.rcvif;
602 else {
603 CURVNET_RESTORE();
604 return;
605 }
606 }
607
608 /*
609 * If the hardware did not process an 802.1Q tag, do this now,
610 * to allow 802.1P priority frames to be passed to the main input
611 * path correctly.
612 */
613 if ((m->m_flags & M_VLANTAG) == 0 &&
614 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
615 struct ether_vlan_header *evl;
616
617 if (m->m_len < sizeof(*evl) &&
618 (m = m_pullup(m, sizeof(*evl))) == NULL) {
619 #ifdef DIAGNOSTIC
620 if_printf(ifp, "cannot pullup VLAN header\n");
621 #endif
622 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
623 CURVNET_RESTORE();
624 return;
625 }
626
627 evl = mtod(m, struct ether_vlan_header *);
628 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
629 m->m_flags |= M_VLANTAG;
630
631 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
632 ETHER_HDR_LEN - ETHER_TYPE_LEN);
633 m_adj(m, ETHER_VLAN_ENCAP_LEN);
634 eh = mtod(m, struct ether_header *);
635 }
636
637 M_SETFIB(m, ifp->if_fib);
638
639 /* Allow ng_ether(4) to claim this frame. */
640 if (ifp->if_l2com != NULL) {
641 KASSERT(ng_ether_input_p != NULL,
642 ("%s: ng_ether_input_p is NULL", __func__));
643 m->m_flags &= ~M_PROMISC;
644 (*ng_ether_input_p)(ifp, &m);
645 if (m == NULL) {
646 CURVNET_RESTORE();
647 return;
648 }
649 eh = mtod(m, struct ether_header *);
650 }
651
652 /*
653 * Allow if_bridge(4) to claim this frame.
654 *
655 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
656 * and the frame should be delivered locally.
657 *
658 * If M_BRIDGE_INJECT is set, the packet was received directly by the
659 * bridge via netmap, so "ifp" is the bridge itself and the packet
660 * should be re-examined.
661 */
662 if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) {
663 m->m_flags &= ~M_PROMISC;
664 BRIDGE_INPUT(ifp, m);
665 if (m == NULL) {
666 CURVNET_RESTORE();
667 return;
668 }
669 eh = mtod(m, struct ether_header *);
670 }
671
672 #if defined(INET) || defined(INET6)
673 /*
674 * Clear M_PROMISC on frame so that carp(4) will see it when the
675 * mbuf flows up to Layer 3.
676 * FreeBSD's implementation of carp(4) uses the inprotosw
677 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
678 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
679 * is outside the scope of the M_PROMISC test below.
680 * TODO: Maintain a hash table of ethernet addresses other than
681 * ether_dhost which may be active on this ifp.
682 */
683 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
684 m->m_flags &= ~M_PROMISC;
685 } else
686 #endif
687 {
688 /*
689 * If the frame received was not for our MAC address, set the
690 * M_PROMISC flag on the mbuf chain. The frame may need to
691 * be seen by the rest of the Ethernet input path in case of
692 * re-entry (e.g. bridge, vlan, netgraph) but should not be
693 * seen by upper protocol layers.
694 */
695 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
696 memcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
697 m->m_flags |= M_PROMISC;
698 }
699
700 ether_demux(ifp, m);
701 CURVNET_RESTORE();
702 }
703
704 /*
705 * Ethernet input dispatch; by default, direct dispatch here regardless of
706 * global configuration. However, if RSS is enabled, hook up RSS affinity
707 * so that when deferred or hybrid dispatch is enabled, we can redistribute
708 * load based on RSS.
709 *
710 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
711 * not it had already done work distribution via multi-queue. Then we could
712 * direct dispatch in the event load balancing was already complete and
713 * handle the case of interfaces with different capabilities better.
714 *
715 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
716 * at multiple layers?
717 *
718 * XXXRW: For now, enable all this only if RSS is compiled in, although it
719 * works fine without RSS. Need to characterise the performance overhead
720 * of the detour through the netisr code in the event the result is always
721 * direct dispatch.
722 */
723 static void
ether_nh_input(struct mbuf * m)724 ether_nh_input(struct mbuf *m)
725 {
726
727 M_ASSERTPKTHDR(m);
728 KASSERT(m->m_pkthdr.rcvif != NULL,
729 ("%s: NULL interface pointer", __func__));
730 ether_input_internal(m->m_pkthdr.rcvif, m);
731 }
732
733 static struct netisr_handler ether_nh = {
734 .nh_name = "ether",
735 .nh_handler = ether_nh_input,
736 .nh_proto = NETISR_ETHER,
737 #ifdef RSS
738 .nh_policy = NETISR_POLICY_CPU,
739 .nh_dispatch = NETISR_DISPATCH_DIRECT,
740 .nh_m2cpuid = rss_m2cpuid,
741 #else
742 .nh_policy = NETISR_POLICY_SOURCE,
743 .nh_dispatch = NETISR_DISPATCH_DIRECT,
744 #endif
745 };
746
747 static void
ether_init(__unused void * arg)748 ether_init(__unused void *arg)
749 {
750
751 netisr_register(ðer_nh);
752 }
753 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
754
755 static void
vnet_ether_init(const __unused void * arg)756 vnet_ether_init(const __unused void *arg)
757 {
758 struct pfil_head_args args;
759
760 args.pa_version = PFIL_VERSION;
761 args.pa_flags = PFIL_IN | PFIL_OUT;
762 args.pa_type = PFIL_TYPE_ETHERNET;
763 args.pa_headname = PFIL_ETHER_NAME;
764 V_link_pfil_head = pfil_head_register(&args);
765
766 #ifdef VIMAGE
767 netisr_register_vnet(ðer_nh);
768 #endif
769 }
770 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
771 vnet_ether_init, NULL);
772
773 #ifdef VIMAGE
774 static void
vnet_ether_pfil_destroy(const __unused void * arg)775 vnet_ether_pfil_destroy(const __unused void *arg)
776 {
777
778 pfil_head_unregister(V_link_pfil_head);
779 }
780 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
781 vnet_ether_pfil_destroy, NULL);
782
783 static void
vnet_ether_destroy(__unused void * arg)784 vnet_ether_destroy(__unused void *arg)
785 {
786
787 netisr_unregister_vnet(ðer_nh);
788 }
789 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
790 vnet_ether_destroy, NULL);
791 #endif
792
793 static void
ether_input(struct ifnet * ifp,struct mbuf * m)794 ether_input(struct ifnet *ifp, struct mbuf *m)
795 {
796 struct epoch_tracker et;
797 struct mbuf *mn;
798 bool needs_epoch;
799
800 needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH);
801 #ifdef INVARIANTS
802 /*
803 * This temporary code is here to prevent epoch unaware and unmarked
804 * drivers to panic the system. Once all drivers are taken care of,
805 * the whole INVARIANTS block should go away.
806 */
807 if (!needs_epoch && !in_epoch(net_epoch_preempt)) {
808 static bool printedonce;
809
810 needs_epoch = true;
811 if (!printedonce) {
812 printedonce = true;
813 if_printf(ifp, "called %s w/o net epoch! "
814 "PLEASE file a bug report.", __func__);
815 #ifdef KDB
816 kdb_backtrace();
817 #endif
818 }
819 }
820 #endif
821
822 /*
823 * The drivers are allowed to pass in a chain of packets linked with
824 * m_nextpkt. We split them up into separate packets here and pass
825 * them up. This allows the drivers to amortize the receive lock.
826 */
827 CURVNET_SET_QUIET(ifp->if_vnet);
828 if (__predict_false(needs_epoch))
829 NET_EPOCH_ENTER(et);
830 while (m) {
831 mn = m->m_nextpkt;
832 m->m_nextpkt = NULL;
833
834 /*
835 * We will rely on rcvif being set properly in the deferred
836 * context, so assert it is correct here.
837 */
838 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
839 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
840 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
841 netisr_dispatch(NETISR_ETHER, m);
842 m = mn;
843 }
844 if (__predict_false(needs_epoch))
845 NET_EPOCH_EXIT(et);
846 CURVNET_RESTORE();
847 }
848
849 /*
850 * Upper layer processing for a received Ethernet packet.
851 */
852 void
ether_demux(struct ifnet * ifp,struct mbuf * m)853 ether_demux(struct ifnet *ifp, struct mbuf *m)
854 {
855 struct ether_header *eh;
856 int i, isr;
857 u_short ether_type;
858
859 NET_EPOCH_ASSERT();
860 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
861
862 /* Do not grab PROMISC frames in case we are re-entered. */
863 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
864 i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL);
865 if (i != PFIL_PASS)
866 return;
867 }
868
869 eh = mtod(m, struct ether_header *);
870 ether_type = ntohs(eh->ether_type);
871
872 /*
873 * If this frame has a VLAN tag other than 0, call vlan_input()
874 * if its module is loaded. Otherwise, drop.
875 */
876 if ((m->m_flags & M_VLANTAG) &&
877 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
878 if (ifp->if_vlantrunk == NULL) {
879 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
880 m_freem(m);
881 return;
882 }
883 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
884 __func__));
885 /* Clear before possibly re-entering ether_input(). */
886 m->m_flags &= ~M_PROMISC;
887 (*vlan_input_p)(ifp, m);
888 return;
889 }
890
891 /*
892 * Pass promiscuously received frames to the upper layer if the user
893 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
894 */
895 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
896 m_freem(m);
897 return;
898 }
899
900 /*
901 * Reset layer specific mbuf flags to avoid confusing upper layers.
902 */
903 m->m_flags &= ~M_VLANTAG;
904 m_clrprotoflags(m);
905
906 /*
907 * Dispatch frame to upper layer.
908 */
909 switch (ether_type) {
910 #ifdef INET
911 case ETHERTYPE_IP:
912 isr = NETISR_IP;
913 break;
914
915 case ETHERTYPE_ARP:
916 if (ifp->if_flags & IFF_NOARP) {
917 /* Discard packet if ARP is disabled on interface */
918 m_freem(m);
919 return;
920 }
921 isr = NETISR_ARP;
922 break;
923 #endif
924 #ifdef INET6
925 case ETHERTYPE_IPV6:
926 isr = NETISR_IPV6;
927 break;
928 #endif
929 default:
930 goto discard;
931 }
932
933 /* Strip off Ethernet header. */
934 m_adj(m, ETHER_HDR_LEN);
935
936 netisr_dispatch(isr, m);
937 return;
938
939 discard:
940 /*
941 * Packet is to be discarded. If netgraph is present,
942 * hand the packet to it for last chance processing;
943 * otherwise dispose of it.
944 */
945 if (ifp->if_l2com != NULL) {
946 KASSERT(ng_ether_input_orphan_p != NULL,
947 ("ng_ether_input_orphan_p is NULL"));
948 (*ng_ether_input_orphan_p)(ifp, m);
949 return;
950 }
951 m_freem(m);
952 }
953
954 /*
955 * Convert Ethernet address to printable (loggable) representation.
956 * This routine is for compatibility; it's better to just use
957 *
958 * printf("%6D", <pointer to address>, ":");
959 *
960 * since there's no static buffer involved.
961 */
962 char *
ether_sprintf(const u_char * ap)963 ether_sprintf(const u_char *ap)
964 {
965 static char etherbuf[18];
966 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
967 return (etherbuf);
968 }
969
970 /*
971 * Perform common duties while attaching to interface list
972 */
973 void
ether_ifattach(struct ifnet * ifp,const u_int8_t * lla)974 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
975 {
976 int i;
977 struct ifaddr *ifa;
978 struct sockaddr_dl *sdl;
979
980 ifp->if_addrlen = ETHER_ADDR_LEN;
981 ifp->if_hdrlen = (ifp->if_capabilities & IFCAP_VLAN_MTU) != 0 ?
982 ETHER_HDR_LEN + ETHER_VLAN_ENCAP_LEN : ETHER_HDR_LEN;
983 ifp->if_mtu = ETHERMTU;
984 if_attach(ifp);
985 ifp->if_output = ether_output;
986 ifp->if_input = ether_input;
987 ifp->if_resolvemulti = ether_resolvemulti;
988 ifp->if_requestencap = ether_requestencap;
989 if (ifp->if_baudrate == 0)
990 ifp->if_baudrate = IF_Mbps(10); /* just a default */
991 ifp->if_broadcastaddr = etherbroadcastaddr;
992
993 ifa = ifp->if_addr;
994 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
995 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
996 sdl->sdl_type = IFT_ETHER;
997 sdl->sdl_alen = ifp->if_addrlen;
998 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
999
1000 if (ifp->if_hw_addr != NULL)
1001 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
1002
1003 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1004
1005 /* Announce Ethernet MAC address if non-zero. */
1006 for (i = 0; i < ifp->if_addrlen; i++)
1007 if (lla[i] != 0)
1008 break;
1009 if (i != ifp->if_addrlen)
1010 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1011
1012 uuid_ether_add(LLADDR(sdl));
1013
1014 /* Add necessary bits are setup; announce it now. */
1015 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1016 if (IS_DEFAULT_VNET(curvnet))
1017 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1018 }
1019
1020 /*
1021 * Perform common duties while detaching an Ethernet interface
1022 */
1023 void
ether_ifdetach(struct ifnet * ifp)1024 ether_ifdetach(struct ifnet *ifp)
1025 {
1026 struct sockaddr_dl *sdl;
1027
1028 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1029 uuid_ether_del(LLADDR(sdl));
1030
1031 bpfdetach(ifp);
1032 if_detach(ifp);
1033 }
1034
1035 SYSCTL_DECL(_net_link);
1036 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1037 "Ethernet");
1038
1039 #if 0
1040 /*
1041 * This is for reference. We have a table-driven version
1042 * of the little-endian crc32 generator, which is faster
1043 * than the double-loop.
1044 */
1045 uint32_t
1046 ether_crc32_le(const uint8_t *buf, size_t len)
1047 {
1048 size_t i;
1049 uint32_t crc;
1050 int bit;
1051 uint8_t data;
1052
1053 crc = 0xffffffff; /* initial value */
1054
1055 for (i = 0; i < len; i++) {
1056 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1057 carry = (crc ^ data) & 1;
1058 crc >>= 1;
1059 if (carry)
1060 crc = (crc ^ ETHER_CRC_POLY_LE);
1061 }
1062 }
1063
1064 return (crc);
1065 }
1066 #else
1067 uint32_t
ether_crc32_le(const uint8_t * buf,size_t len)1068 ether_crc32_le(const uint8_t *buf, size_t len)
1069 {
1070 static const uint32_t crctab[] = {
1071 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1072 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1073 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1074 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1075 };
1076 size_t i;
1077 uint32_t crc;
1078
1079 crc = 0xffffffff; /* initial value */
1080
1081 for (i = 0; i < len; i++) {
1082 crc ^= buf[i];
1083 crc = (crc >> 4) ^ crctab[crc & 0xf];
1084 crc = (crc >> 4) ^ crctab[crc & 0xf];
1085 }
1086
1087 return (crc);
1088 }
1089 #endif
1090
1091 uint32_t
ether_crc32_be(const uint8_t * buf,size_t len)1092 ether_crc32_be(const uint8_t *buf, size_t len)
1093 {
1094 size_t i;
1095 uint32_t crc, carry;
1096 int bit;
1097 uint8_t data;
1098
1099 crc = 0xffffffff; /* initial value */
1100
1101 for (i = 0; i < len; i++) {
1102 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1103 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1104 crc <<= 1;
1105 if (carry)
1106 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1107 }
1108 }
1109
1110 return (crc);
1111 }
1112
1113 int
ether_ioctl(struct ifnet * ifp,u_long command,caddr_t data)1114 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1115 {
1116 struct ifaddr *ifa = (struct ifaddr *) data;
1117 struct ifreq *ifr = (struct ifreq *) data;
1118 int error = 0;
1119
1120 switch (command) {
1121 case SIOCSIFADDR:
1122 ifp->if_flags |= IFF_UP;
1123
1124 switch (ifa->ifa_addr->sa_family) {
1125 #ifdef INET
1126 case AF_INET:
1127 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1128 arp_ifinit(ifp, ifa);
1129 break;
1130 #endif
1131 default:
1132 ifp->if_init(ifp->if_softc);
1133 break;
1134 }
1135 break;
1136
1137 case SIOCGIFADDR:
1138 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1139 ETHER_ADDR_LEN);
1140 break;
1141
1142 case SIOCSIFMTU:
1143 /*
1144 * Set the interface MTU.
1145 */
1146 if (ifr->ifr_mtu > ETHERMTU) {
1147 error = EINVAL;
1148 } else {
1149 ifp->if_mtu = ifr->ifr_mtu;
1150 }
1151 break;
1152
1153 case SIOCSLANPCP:
1154 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1155 if (error != 0)
1156 break;
1157 if (ifr->ifr_lan_pcp > 7 &&
1158 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1159 error = EINVAL;
1160 } else {
1161 ifp->if_pcp = ifr->ifr_lan_pcp;
1162 /* broadcast event about PCP change */
1163 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1164 }
1165 break;
1166
1167 case SIOCGLANPCP:
1168 ifr->ifr_lan_pcp = ifp->if_pcp;
1169 break;
1170
1171 default:
1172 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1173 break;
1174 }
1175 return (error);
1176 }
1177
1178 static int
ether_resolvemulti(struct ifnet * ifp,struct sockaddr ** llsa,struct sockaddr * sa)1179 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1180 struct sockaddr *sa)
1181 {
1182 struct sockaddr_dl *sdl;
1183 #ifdef INET
1184 struct sockaddr_in *sin;
1185 #endif
1186 #ifdef INET6
1187 struct sockaddr_in6 *sin6;
1188 #endif
1189 u_char *e_addr;
1190
1191 switch(sa->sa_family) {
1192 case AF_LINK:
1193 /*
1194 * No mapping needed. Just check that it's a valid MC address.
1195 */
1196 sdl = (struct sockaddr_dl *)sa;
1197 e_addr = LLADDR(sdl);
1198 if (!ETHER_IS_MULTICAST(e_addr))
1199 return EADDRNOTAVAIL;
1200 *llsa = NULL;
1201 return 0;
1202
1203 #ifdef INET
1204 case AF_INET:
1205 sin = (struct sockaddr_in *)sa;
1206 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1207 return EADDRNOTAVAIL;
1208 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1209 sdl->sdl_alen = ETHER_ADDR_LEN;
1210 e_addr = LLADDR(sdl);
1211 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1212 *llsa = (struct sockaddr *)sdl;
1213 return 0;
1214 #endif
1215 #ifdef INET6
1216 case AF_INET6:
1217 sin6 = (struct sockaddr_in6 *)sa;
1218 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1219 /*
1220 * An IP6 address of 0 means listen to all
1221 * of the Ethernet multicast address used for IP6.
1222 * (This is used for multicast routers.)
1223 */
1224 ifp->if_flags |= IFF_ALLMULTI;
1225 *llsa = NULL;
1226 return 0;
1227 }
1228 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1229 return EADDRNOTAVAIL;
1230 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1231 sdl->sdl_alen = ETHER_ADDR_LEN;
1232 e_addr = LLADDR(sdl);
1233 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1234 *llsa = (struct sockaddr *)sdl;
1235 return 0;
1236 #endif
1237
1238 default:
1239 /*
1240 * Well, the text isn't quite right, but it's the name
1241 * that counts...
1242 */
1243 return EAFNOSUPPORT;
1244 }
1245 }
1246
1247 static moduledata_t ether_mod = {
1248 .name = "ether",
1249 };
1250
1251 void
ether_vlan_mtap(struct bpf_if * bp,struct mbuf * m,void * data,u_int dlen)1252 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1253 {
1254 struct ether_vlan_header vlan;
1255 struct mbuf mv, mb;
1256
1257 KASSERT((m->m_flags & M_VLANTAG) != 0,
1258 ("%s: vlan information not present", __func__));
1259 KASSERT(m->m_len >= sizeof(struct ether_header),
1260 ("%s: mbuf not large enough for header", __func__));
1261 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1262 vlan.evl_proto = vlan.evl_encap_proto;
1263 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1264 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1265 m->m_len -= sizeof(struct ether_header);
1266 m->m_data += sizeof(struct ether_header);
1267 /*
1268 * If a data link has been supplied by the caller, then we will need to
1269 * re-create a stack allocated mbuf chain with the following structure:
1270 *
1271 * (1) mbuf #1 will contain the supplied data link
1272 * (2) mbuf #2 will contain the vlan header
1273 * (3) mbuf #3 will contain the original mbuf's packet data
1274 *
1275 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1276 */
1277 if (data != NULL) {
1278 mv.m_next = m;
1279 mv.m_data = (caddr_t)&vlan;
1280 mv.m_len = sizeof(vlan);
1281 mb.m_next = &mv;
1282 mb.m_data = data;
1283 mb.m_len = dlen;
1284 bpf_mtap(bp, &mb);
1285 } else
1286 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1287 m->m_len += sizeof(struct ether_header);
1288 m->m_data -= sizeof(struct ether_header);
1289 }
1290
1291 struct mbuf *
ether_vlanencap_proto(struct mbuf * m,uint16_t tag,uint16_t proto)1292 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1293 {
1294 struct ether_vlan_header *evl;
1295
1296 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1297 if (m == NULL)
1298 return (NULL);
1299 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1300
1301 if (m->m_len < sizeof(*evl)) {
1302 m = m_pullup(m, sizeof(*evl));
1303 if (m == NULL)
1304 return (NULL);
1305 }
1306
1307 /*
1308 * Transform the Ethernet header into an Ethernet header
1309 * with 802.1Q encapsulation.
1310 */
1311 evl = mtod(m, struct ether_vlan_header *);
1312 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1313 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1314 evl->evl_encap_proto = htons(proto);
1315 evl->evl_tag = htons(tag);
1316 return (m);
1317 }
1318
1319 void
ether_bpf_mtap_if(struct ifnet * ifp,struct mbuf * m)1320 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m)
1321 {
1322 if (bpf_peers_present(ifp->if_bpf)) {
1323 M_ASSERTVALID(m);
1324 if ((m->m_flags & M_VLANTAG) != 0)
1325 ether_vlan_mtap(ifp->if_bpf, m, NULL, 0);
1326 else
1327 bpf_mtap(ifp->if_bpf, m);
1328 }
1329 }
1330
1331 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1332 "IEEE 802.1Q VLAN");
1333 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1334 CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1335 "for consistency");
1336
1337 VNET_DEFINE_STATIC(int, soft_pad);
1338 #define V_soft_pad VNET(soft_pad)
1339 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1340 &VNET_NAME(soft_pad), 0,
1341 "pad short frames before tagging");
1342
1343 /*
1344 * For now, make preserving PCP via an mbuf tag optional, as it increases
1345 * per-packet memory allocations and frees. In the future, it would be
1346 * preferable to reuse ether_vtag for this, or similar.
1347 */
1348 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1349 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp)
1350 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1351 &VNET_NAME(vlan_mtag_pcp), 0,
1352 "Retain VLAN PCP information as packets are passed up the stack");
1353
1354 static inline bool
ether_do_pcp(struct ifnet * ifp,struct mbuf * m)1355 ether_do_pcp(struct ifnet *ifp, struct mbuf *m)
1356 {
1357 if (ifp->if_type == IFT_L2VLAN)
1358 return (false);
1359 if (ifp->if_pcp != IFNET_PCP_NONE || (m->m_flags & M_VLANTAG) != 0)
1360 return (true);
1361 if (V_vlan_mtag_pcp &&
1362 m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_OUT, NULL) != NULL)
1363 return (true);
1364 return (false);
1365 }
1366
1367 bool
ether_8021q_frame(struct mbuf ** mp,struct ifnet * ife,struct ifnet * p,const struct ether_8021q_tag * qtag)1368 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1369 const struct ether_8021q_tag *qtag)
1370 {
1371 struct m_tag *mtag;
1372 int n;
1373 uint16_t tag;
1374 uint8_t pcp = qtag->pcp;
1375 static const char pad[8]; /* just zeros */
1376
1377 /*
1378 * Pad the frame to the minimum size allowed if told to.
1379 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1380 * paragraph C.4.4.3.b. It can help to work around buggy
1381 * bridges that violate paragraph C.4.4.3.a from the same
1382 * document, i.e., fail to pad short frames after untagging.
1383 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1384 * untagging it will produce a 62-byte frame, which is a runt
1385 * and requires padding. There are VLAN-enabled network
1386 * devices that just discard such runts instead or mishandle
1387 * them somehow.
1388 */
1389 if (V_soft_pad && p->if_type == IFT_ETHER) {
1390 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1391 n > 0; n -= sizeof(pad)) {
1392 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1393 break;
1394 }
1395 if (n > 0) {
1396 m_freem(*mp);
1397 *mp = NULL;
1398 if_printf(ife, "cannot pad short frame");
1399 return (false);
1400 }
1401 }
1402
1403 /*
1404 * If PCP is set in mbuf, use it
1405 */
1406 if ((*mp)->m_flags & M_VLANTAG) {
1407 pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1408 }
1409
1410 /*
1411 * If underlying interface can do VLAN tag insertion itself,
1412 * just pass the packet along. However, we need some way to
1413 * tell the interface where the packet came from so that it
1414 * knows how to find the VLAN tag to use, so we attach a
1415 * packet tag that holds it.
1416 */
1417 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1418 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1419 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1420 else
1421 tag = EVL_MAKETAG(qtag->vid, pcp, 0);
1422 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1423 (qtag->proto == ETHERTYPE_VLAN)) {
1424 (*mp)->m_pkthdr.ether_vtag = tag;
1425 (*mp)->m_flags |= M_VLANTAG;
1426 } else {
1427 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1428 if (*mp == NULL) {
1429 if_printf(ife, "unable to prepend 802.1Q header");
1430 return (false);
1431 }
1432 (*mp)->m_flags &= ~M_VLANTAG;
1433 }
1434 return (true);
1435 }
1436
1437 /*
1438 * Allocate an address from the FreeBSD Foundation OUI. This uses a
1439 * cryptographic hash function on the containing jail's name, UUID and the
1440 * interface name to attempt to provide a unique but stable address.
1441 * Pseudo-interfaces which require a MAC address should use this function to
1442 * allocate non-locally-administered addresses.
1443 */
1444 void
ether_gen_addr_byname(const char * nameunit,struct ether_addr * hwaddr)1445 ether_gen_addr_byname(const char *nameunit, struct ether_addr *hwaddr)
1446 {
1447 SHA1_CTX ctx;
1448 char *buf;
1449 char uuid[HOSTUUIDLEN + 1];
1450 uint64_t addr;
1451 int i, sz;
1452 unsigned char digest[SHA1_RESULTLEN];
1453 char jailname[MAXHOSTNAMELEN];
1454
1455 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1456 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1457 /* Fall back to a random mac address. */
1458 goto rando;
1459 }
1460
1461 /* If each (vnet) jail would also have a unique hostuuid this would not
1462 * be necessary. */
1463 getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1464 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, nameunit,
1465 jailname);
1466 if (sz < 0) {
1467 /* Fall back to a random mac address. */
1468 goto rando;
1469 }
1470
1471 SHA1Init(&ctx);
1472 SHA1Update(&ctx, buf, sz);
1473 SHA1Final(digest, &ctx);
1474 free(buf, M_TEMP);
1475
1476 addr = (digest[0] << 8) | digest[1] | OUI_FREEBSD_GENERATED_LOW;
1477 for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1478 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1479 0xFF;
1480 }
1481
1482 return;
1483 rando:
1484 arc4rand(hwaddr, sizeof(*hwaddr), 0);
1485 /* Unicast */
1486 hwaddr->octet[0] &= 0xFE;
1487 /* Locally administered. */
1488 hwaddr->octet[0] |= 0x02;
1489 }
1490
1491 void
ether_gen_addr(struct ifnet * ifp,struct ether_addr * hwaddr)1492 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1493 {
1494 ether_gen_addr_byname(if_name(ifp), hwaddr);
1495 }
1496
1497 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1498 MODULE_VERSION(ether, 1);
1499