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 #ifdef CTASSERT
96 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
97 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
98 #endif
99
100 VNET_DEFINE(pfil_head_t, link_pfil_head); /* Packet filter hooks */
101
102 /* netgraph node hooks for ng_ether(4) */
103 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
104 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
105 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
106 void (*ng_ether_attach_p)(struct ifnet *ifp);
107 void (*ng_ether_detach_p)(struct ifnet *ifp);
108
109 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
110
111 /* if_bridge(4) support */
112 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
113 bool (*bridge_same_p)(const void *, const void *);
114 void *(*bridge_get_softc_p)(struct ifnet *);
115
116 /* if_lagg(4) support */
117 struct mbuf *(*lagg_input_ethernet_p)(struct ifnet *, struct mbuf *);
118
119 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
120 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
121
122 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
123 struct sockaddr *);
124 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
125
126 static inline bool ether_do_pcp(struct ifnet *, struct mbuf *);
127
128 #define senderr(e) do { error = (e); goto bad;} while (0)
129
130 static void
update_mbuf_csumflags(struct mbuf * src,struct mbuf * dst)131 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
132 {
133 int csum_flags = 0;
134
135 if (src->m_pkthdr.csum_flags & CSUM_IP)
136 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
137 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
138 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
139 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
140 csum_flags |= CSUM_SCTP_VALID;
141 dst->m_pkthdr.csum_flags |= csum_flags;
142 if (csum_flags & CSUM_DATA_VALID)
143 dst->m_pkthdr.csum_data = 0xffff;
144 }
145
146 /*
147 * Handle link-layer encapsulation requests.
148 */
149 static int
ether_requestencap(struct ifnet * ifp,struct if_encap_req * req)150 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
151 {
152 struct ether_header *eh;
153 struct arphdr *ah;
154 uint16_t etype;
155 const u_char *lladdr;
156
157 if (req->rtype != IFENCAP_LL)
158 return (EOPNOTSUPP);
159
160 if (req->bufsize < ETHER_HDR_LEN)
161 return (ENOMEM);
162
163 eh = (struct ether_header *)req->buf;
164 lladdr = req->lladdr;
165 req->lladdr_off = 0;
166
167 switch (req->family) {
168 case AF_INET:
169 etype = htons(ETHERTYPE_IP);
170 break;
171 case AF_INET6:
172 etype = htons(ETHERTYPE_IPV6);
173 break;
174 case AF_ARP:
175 ah = (struct arphdr *)req->hdata;
176 ah->ar_hrd = htons(ARPHRD_ETHER);
177
178 switch(ntohs(ah->ar_op)) {
179 case ARPOP_REVREQUEST:
180 case ARPOP_REVREPLY:
181 etype = htons(ETHERTYPE_REVARP);
182 break;
183 case ARPOP_REQUEST:
184 case ARPOP_REPLY:
185 default:
186 etype = htons(ETHERTYPE_ARP);
187 break;
188 }
189
190 if (req->flags & IFENCAP_FLAG_BROADCAST)
191 lladdr = ifp->if_broadcastaddr;
192 break;
193 default:
194 return (EAFNOSUPPORT);
195 }
196
197 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
198 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
199 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
200 req->bufsize = sizeof(struct ether_header);
201
202 return (0);
203 }
204
205 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)206 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
207 const struct sockaddr *dst, struct route *ro, u_char *phdr,
208 uint32_t *pflags, struct llentry **plle)
209 {
210 uint32_t lleflags = 0;
211 int error = 0;
212 #if defined(INET) || defined(INET6)
213 struct ether_header *eh = (struct ether_header *)phdr;
214 uint16_t etype;
215 #endif
216
217 if (plle)
218 *plle = NULL;
219
220 switch (dst->sa_family) {
221 #ifdef INET
222 case AF_INET:
223 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
224 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
225 plle);
226 else {
227 if (m->m_flags & M_BCAST)
228 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
229 ETHER_ADDR_LEN);
230 else {
231 const struct in_addr *a;
232 a = &(((const struct sockaddr_in *)dst)->sin_addr);
233 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
234 }
235 etype = htons(ETHERTYPE_IP);
236 memcpy(&eh->ether_type, &etype, sizeof(etype));
237 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
238 }
239 break;
240 #endif
241 #ifdef INET6
242 case AF_INET6:
243 if ((m->m_flags & M_MCAST) == 0) {
244 int af = RO_GET_FAMILY(ro, dst);
245 error = nd6_resolve(ifp, LLE_SF(af, 0), m, dst, phdr,
246 &lleflags, plle);
247 } else {
248 const struct in6_addr *a6;
249 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
250 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
251 etype = htons(ETHERTYPE_IPV6);
252 memcpy(&eh->ether_type, &etype, sizeof(etype));
253 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
254 }
255 break;
256 #endif
257 default:
258 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
259 if (m != NULL)
260 m_freem(m);
261 return (EAFNOSUPPORT);
262 }
263
264 if (error == EHOSTDOWN) {
265 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
266 error = EHOSTUNREACH;
267 }
268
269 if (error != 0)
270 return (error);
271
272 *pflags = RT_MAY_LOOP;
273 if (lleflags & LLE_IFADDR)
274 *pflags |= RT_L2_ME;
275
276 return (0);
277 }
278
279 /*
280 * Ethernet output routine.
281 * Encapsulate a packet of type family for the local net.
282 * Use trailer local net encapsulation if enough data in first
283 * packet leaves a multiple of 512 bytes of data in remainder.
284 */
285 int
ether_output(struct ifnet * ifp,struct mbuf * m,const struct sockaddr * dst,struct route * ro)286 ether_output(struct ifnet *ifp, struct mbuf *m,
287 const struct sockaddr *dst, struct route *ro)
288 {
289 int error = 0;
290 char linkhdr[ETHER_HDR_LEN], *phdr;
291 struct ether_header *eh;
292 struct pf_mtag *t;
293 bool loop_copy;
294 int hlen; /* link layer header length */
295 uint32_t pflags;
296 struct llentry *lle = NULL;
297 int addref = 0;
298
299 phdr = NULL;
300 pflags = 0;
301 if (ro != NULL) {
302 /* XXX BPF uses ro_prepend */
303 if (ro->ro_prepend != NULL) {
304 phdr = ro->ro_prepend;
305 hlen = ro->ro_plen;
306 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
307 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
308 lle = ro->ro_lle;
309 if (lle != NULL &&
310 (lle->la_flags & LLE_VALID) == 0) {
311 LLE_FREE(lle);
312 lle = NULL; /* redundant */
313 ro->ro_lle = NULL;
314 }
315 if (lle == NULL) {
316 /* if we lookup, keep cache */
317 addref = 1;
318 } else
319 /*
320 * Notify LLE code that
321 * the entry was used
322 * by datapath.
323 */
324 llentry_provide_feedback(lle);
325 }
326 if (lle != NULL) {
327 phdr = lle->r_linkdata;
328 hlen = lle->r_hdrlen;
329 pflags = lle->r_flags;
330 }
331 }
332 }
333
334 #ifdef MAC
335 error = mac_ifnet_check_transmit(ifp, m);
336 if (error)
337 senderr(error);
338 #endif
339
340 M_PROFILE(m);
341 if (ifp->if_flags & IFF_MONITOR)
342 senderr(ENETDOWN);
343 if (!((ifp->if_flags & IFF_UP) &&
344 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
345 senderr(ENETDOWN);
346
347 if (phdr == NULL) {
348 /* No prepend data supplied. Try to calculate ourselves. */
349 phdr = linkhdr;
350 hlen = ETHER_HDR_LEN;
351 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
352 addref ? &lle : NULL);
353 if (addref && lle != NULL)
354 ro->ro_lle = lle;
355 if (error != 0)
356 return (error == EWOULDBLOCK ? 0 : error);
357 }
358
359 if ((pflags & RT_L2_ME) != 0) {
360 update_mbuf_csumflags(m, m);
361 return (if_simloop(ifp, m, RO_GET_FAMILY(ro, dst), 0));
362 }
363 loop_copy = (pflags & RT_MAY_LOOP) != 0;
364
365 /*
366 * Add local net header. If no space in first mbuf,
367 * allocate another.
368 *
369 * Note that we do prepend regardless of RT_HAS_HEADER flag.
370 * This is done because BPF code shifts m_data pointer
371 * to the end of ethernet header prior to calling if_output().
372 */
373 M_PREPEND(m, hlen, M_NOWAIT);
374 if (m == NULL)
375 senderr(ENOBUFS);
376 if ((pflags & RT_HAS_HEADER) == 0) {
377 eh = mtod(m, struct ether_header *);
378 memcpy(eh, phdr, hlen);
379 }
380
381 /*
382 * If a simplex interface, and the packet is being sent to our
383 * Ethernet address or a broadcast address, loopback a copy.
384 * XXX To make a simplex device behave exactly like a duplex
385 * device, we should copy in the case of sending to our own
386 * ethernet address (thus letting the original actually appear
387 * on the wire). However, we don't do that here for security
388 * reasons and compatibility with the original behavior.
389 */
390 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
391 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
392 struct mbuf *n;
393
394 /*
395 * Because if_simloop() modifies the packet, we need a
396 * writable copy through m_dup() instead of a readonly
397 * one as m_copy[m] would give us. The alternative would
398 * be to modify if_simloop() to handle the readonly mbuf,
399 * but performancewise it is mostly equivalent (trading
400 * extra data copying vs. extra locking).
401 *
402 * XXX This is a local workaround. A number of less
403 * often used kernel parts suffer from the same bug.
404 * See PR kern/105943 for a proposed general solution.
405 */
406 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
407 update_mbuf_csumflags(m, n);
408 (void)if_simloop(ifp, n, RO_GET_FAMILY(ro, dst), hlen);
409 } else
410 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
411 }
412
413 /*
414 * Bridges require special output handling.
415 */
416 if (ifp->if_bridge) {
417 BRIDGE_OUTPUT(ifp, m, error);
418 return (error);
419 }
420
421 #if defined(INET) || defined(INET6)
422 if (ifp->if_carp &&
423 (error = (*carp_output_p)(ifp, m, dst)))
424 goto bad;
425 #endif
426
427 /* Handle ng_ether(4) processing, if any */
428 if (ifp->if_l2com != NULL) {
429 KASSERT(ng_ether_output_p != NULL,
430 ("ng_ether_output_p is NULL"));
431 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
432 bad: if (m != NULL)
433 m_freem(m);
434 return (error);
435 }
436 if (m == NULL)
437 return (0);
438 }
439
440 /* Continue with link-layer output */
441 return ether_output_frame(ifp, m);
442 }
443
444 static bool
ether_set_pcp(struct mbuf ** mp,struct ifnet * ifp,uint8_t pcp)445 ether_set_pcp(struct mbuf **mp, struct ifnet *ifp, uint8_t pcp)
446 {
447 struct ether_8021q_tag qtag;
448 struct ether_header *eh;
449
450 eh = mtod(*mp, struct ether_header *);
451 if (eh->ether_type == htons(ETHERTYPE_VLAN) ||
452 eh->ether_type == htons(ETHERTYPE_QINQ)) {
453 (*mp)->m_flags &= ~M_VLANTAG;
454 return (true);
455 }
456
457 qtag.vid = 0;
458 qtag.pcp = pcp;
459 qtag.proto = ETHERTYPE_VLAN;
460 if (ether_8021q_frame(mp, ifp, ifp, &qtag))
461 return (true);
462 if_inc_counter(ifp, IFCOUNTER_OERRORS, 1);
463 return (false);
464 }
465
466 /*
467 * Ethernet link layer output routine to send a raw frame to the device.
468 *
469 * This assumes that the 14 byte Ethernet header is present and contiguous
470 * in the first mbuf (if BRIDGE'ing).
471 */
472 int
ether_output_frame(struct ifnet * ifp,struct mbuf * m)473 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
474 {
475 if (ether_do_pcp(ifp, m) && !ether_set_pcp(&m, ifp, ifp->if_pcp))
476 return (0);
477
478 if (PFIL_HOOKED_OUT(V_link_pfil_head))
479 switch (pfil_mbuf_out(V_link_pfil_head, &m, ifp, NULL)) {
480 case PFIL_DROPPED:
481 return (EACCES);
482 case PFIL_CONSUMED:
483 return (0);
484 }
485
486 #ifdef EXPERIMENTAL
487 #if defined(INET6) && defined(INET)
488 /* draft-ietf-6man-ipv6only-flag */
489 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
490 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
491 struct ether_header *eh;
492
493 eh = mtod(m, struct ether_header *);
494 switch (ntohs(eh->ether_type)) {
495 case ETHERTYPE_IP:
496 case ETHERTYPE_ARP:
497 case ETHERTYPE_REVARP:
498 m_freem(m);
499 return (EAFNOSUPPORT);
500 /* NOTREACHED */
501 break;
502 };
503 }
504 #endif
505 #endif
506
507 /*
508 * Queue message on interface, update output statistics if successful,
509 * and start output if interface not yet active.
510 *
511 * If KMSAN is enabled, use it to verify that the data does not contain
512 * any uninitialized bytes.
513 */
514 kmsan_check_mbuf(m, "ether_output");
515 return ((ifp->if_transmit)(ifp, m));
516 }
517
518 /*
519 * Process a received Ethernet packet; the packet is in the
520 * mbuf chain m with the ethernet header at the front.
521 */
522 static void
ether_input_internal(struct ifnet * ifp,struct mbuf * m)523 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
524 {
525 struct ether_header *eh;
526 u_short etype;
527
528 if ((ifp->if_flags & IFF_UP) == 0) {
529 m_freem(m);
530 return;
531 }
532 #ifdef DIAGNOSTIC
533 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
534 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
535 m_freem(m);
536 return;
537 }
538 #endif
539 if (__predict_false(m->m_len < ETHER_HDR_LEN)) {
540 /* Drivers should pullup and ensure the mbuf is valid */
541 if_printf(ifp, "discard frame w/o leading ethernet "
542 "header (len %d pkt len %d)\n",
543 m->m_len, m->m_pkthdr.len);
544 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
545 m_freem(m);
546 return;
547 }
548 eh = mtod(m, struct ether_header *);
549 etype = ntohs(eh->ether_type);
550 random_harvest_queue_ether(m, sizeof(*m));
551
552 #ifdef EXPERIMENTAL
553 #if defined(INET6) && defined(INET)
554 /* draft-ietf-6man-ipv6only-flag */
555 /* Catch ETHERTYPE_IP, and ETHERTYPE_[REV]ARP if we are v6-only. */
556 if ((ND_IFINFO(ifp)->flags & ND6_IFF_IPV6_ONLY_MASK) != 0) {
557 switch (etype) {
558 case ETHERTYPE_IP:
559 case ETHERTYPE_ARP:
560 case ETHERTYPE_REVARP:
561 m_freem(m);
562 return;
563 /* NOTREACHED */
564 break;
565 };
566 }
567 #endif
568 #endif
569
570 CURVNET_SET_QUIET(ifp->if_vnet);
571
572 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
573 if (ETHER_IS_BROADCAST(eh->ether_dhost))
574 m->m_flags |= M_BCAST;
575 else
576 m->m_flags |= M_MCAST;
577 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
578 }
579
580 #ifdef MAC
581 /*
582 * Tag the mbuf with an appropriate MAC label before any other
583 * consumers can get to it.
584 */
585 mac_ifnet_create_mbuf(ifp, m);
586 #endif
587
588 /*
589 * Give bpf a chance at the packet.
590 */
591 ETHER_BPF_MTAP(ifp, m);
592
593 if (!(ifp->if_capenable & IFCAP_HWSTATS))
594 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
595
596 /* Allow monitor mode to claim this frame, after stats are updated. */
597 if (ifp->if_flags & IFF_MONITOR) {
598 m_freem(m);
599 CURVNET_RESTORE();
600 return;
601 }
602
603 /* Handle input from a lagg(4) port */
604 if (ifp->if_type == IFT_IEEE8023ADLAG) {
605 KASSERT(lagg_input_ethernet_p != NULL,
606 ("%s: if_lagg not loaded!", __func__));
607 m = (*lagg_input_ethernet_p)(ifp, m);
608 if (m != NULL)
609 ifp = m->m_pkthdr.rcvif;
610 else {
611 CURVNET_RESTORE();
612 return;
613 }
614 }
615
616 /*
617 * If the hardware did not process an 802.1Q tag, do this now,
618 * to allow 802.1P priority frames to be passed to the main input
619 * path correctly.
620 */
621 if ((m->m_flags & M_VLANTAG) == 0 &&
622 ((etype == ETHERTYPE_VLAN) || (etype == ETHERTYPE_QINQ))) {
623 struct ether_vlan_header *evl;
624
625 if (m->m_len < sizeof(*evl) &&
626 (m = m_pullup(m, sizeof(*evl))) == NULL) {
627 #ifdef DIAGNOSTIC
628 if_printf(ifp, "cannot pullup VLAN header\n");
629 #endif
630 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
631 CURVNET_RESTORE();
632 return;
633 }
634
635 evl = mtod(m, struct ether_vlan_header *);
636 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
637 m->m_flags |= M_VLANTAG;
638
639 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
640 ETHER_HDR_LEN - ETHER_TYPE_LEN);
641 m_adj(m, ETHER_VLAN_ENCAP_LEN);
642 eh = mtod(m, struct ether_header *);
643 }
644
645 M_SETFIB(m, ifp->if_fib);
646
647 /* Allow ng_ether(4) to claim this frame. */
648 if (ifp->if_l2com != NULL) {
649 KASSERT(ng_ether_input_p != NULL,
650 ("%s: ng_ether_input_p is NULL", __func__));
651 m->m_flags &= ~M_PROMISC;
652 (*ng_ether_input_p)(ifp, &m);
653 if (m == NULL) {
654 CURVNET_RESTORE();
655 return;
656 }
657 eh = mtod(m, struct ether_header *);
658 }
659
660 /*
661 * Allow if_bridge(4) to claim this frame.
662 *
663 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
664 * and the frame should be delivered locally.
665 *
666 * If M_BRIDGE_INJECT is set, the packet was received directly by the
667 * bridge via netmap, so "ifp" is the bridge itself and the packet
668 * should be re-examined.
669 */
670 if (ifp->if_bridge != NULL || (m->m_flags & M_BRIDGE_INJECT) != 0) {
671 m->m_flags &= ~M_PROMISC;
672 BRIDGE_INPUT(ifp, m);
673 if (m == NULL) {
674 CURVNET_RESTORE();
675 return;
676 }
677 eh = mtod(m, struct ether_header *);
678 }
679
680 #if defined(INET) || defined(INET6)
681 /*
682 * Clear M_PROMISC on frame so that carp(4) will see it when the
683 * mbuf flows up to Layer 3.
684 * FreeBSD's implementation of carp(4) uses the inprotosw
685 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
686 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
687 * is outside the scope of the M_PROMISC test below.
688 * TODO: Maintain a hash table of ethernet addresses other than
689 * ether_dhost which may be active on this ifp.
690 */
691 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
692 m->m_flags &= ~M_PROMISC;
693 } else
694 #endif
695 {
696 /*
697 * If the frame received was not for our MAC address, set the
698 * M_PROMISC flag on the mbuf chain. The frame may need to
699 * be seen by the rest of the Ethernet input path in case of
700 * re-entry (e.g. bridge, vlan, netgraph) but should not be
701 * seen by upper protocol layers.
702 */
703 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
704 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
705 m->m_flags |= M_PROMISC;
706 }
707
708 ether_demux(ifp, m);
709 CURVNET_RESTORE();
710 }
711
712 /*
713 * Ethernet input dispatch; by default, direct dispatch here regardless of
714 * global configuration. However, if RSS is enabled, hook up RSS affinity
715 * so that when deferred or hybrid dispatch is enabled, we can redistribute
716 * load based on RSS.
717 *
718 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
719 * not it had already done work distribution via multi-queue. Then we could
720 * direct dispatch in the event load balancing was already complete and
721 * handle the case of interfaces with different capabilities better.
722 *
723 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
724 * at multiple layers?
725 *
726 * XXXRW: For now, enable all this only if RSS is compiled in, although it
727 * works fine without RSS. Need to characterise the performance overhead
728 * of the detour through the netisr code in the event the result is always
729 * direct dispatch.
730 */
731 static void
ether_nh_input(struct mbuf * m)732 ether_nh_input(struct mbuf *m)
733 {
734
735 M_ASSERTPKTHDR(m);
736 KASSERT(m->m_pkthdr.rcvif != NULL,
737 ("%s: NULL interface pointer", __func__));
738 ether_input_internal(m->m_pkthdr.rcvif, m);
739 }
740
741 static struct netisr_handler ether_nh = {
742 .nh_name = "ether",
743 .nh_handler = ether_nh_input,
744 .nh_proto = NETISR_ETHER,
745 #ifdef RSS
746 .nh_policy = NETISR_POLICY_CPU,
747 .nh_dispatch = NETISR_DISPATCH_DIRECT,
748 .nh_m2cpuid = rss_m2cpuid,
749 #else
750 .nh_policy = NETISR_POLICY_SOURCE,
751 .nh_dispatch = NETISR_DISPATCH_DIRECT,
752 #endif
753 };
754
755 static void
ether_init(__unused void * arg)756 ether_init(__unused void *arg)
757 {
758
759 netisr_register(ðer_nh);
760 }
761 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
762
763 static void
vnet_ether_init(const __unused void * arg)764 vnet_ether_init(const __unused void *arg)
765 {
766 struct pfil_head_args args;
767
768 args.pa_version = PFIL_VERSION;
769 args.pa_flags = PFIL_IN | PFIL_OUT;
770 args.pa_type = PFIL_TYPE_ETHERNET;
771 args.pa_headname = PFIL_ETHER_NAME;
772 V_link_pfil_head = pfil_head_register(&args);
773
774 #ifdef VIMAGE
775 netisr_register_vnet(ðer_nh);
776 #endif
777 }
778 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
779 vnet_ether_init, NULL);
780
781 #ifdef VIMAGE
782 static void
vnet_ether_pfil_destroy(const __unused void * arg)783 vnet_ether_pfil_destroy(const __unused void *arg)
784 {
785
786 pfil_head_unregister(V_link_pfil_head);
787 }
788 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
789 vnet_ether_pfil_destroy, NULL);
790
791 static void
vnet_ether_destroy(__unused void * arg)792 vnet_ether_destroy(__unused void *arg)
793 {
794
795 netisr_unregister_vnet(ðer_nh);
796 }
797 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
798 vnet_ether_destroy, NULL);
799 #endif
800
801 static void
ether_input(struct ifnet * ifp,struct mbuf * m)802 ether_input(struct ifnet *ifp, struct mbuf *m)
803 {
804 struct epoch_tracker et;
805 struct mbuf *mn;
806 bool needs_epoch;
807
808 needs_epoch = (ifp->if_flags & IFF_NEEDSEPOCH);
809 #ifdef INVARIANTS
810 /*
811 * This temporary code is here to prevent epoch unaware and unmarked
812 * drivers to panic the system. Once all drivers are taken care of,
813 * the whole INVARIANTS block should go away.
814 */
815 if (!needs_epoch && !in_epoch(net_epoch_preempt)) {
816 static bool printedonce;
817
818 needs_epoch = true;
819 if (!printedonce) {
820 printedonce = true;
821 if_printf(ifp, "called %s w/o net epoch! "
822 "PLEASE file a bug report.", __func__);
823 #ifdef KDB
824 kdb_backtrace();
825 #endif
826 }
827 }
828 #endif
829
830 /*
831 * The drivers are allowed to pass in a chain of packets linked with
832 * m_nextpkt. We split them up into separate packets here and pass
833 * them up. This allows the drivers to amortize the receive lock.
834 */
835 CURVNET_SET_QUIET(ifp->if_vnet);
836 if (__predict_false(needs_epoch))
837 NET_EPOCH_ENTER(et);
838 while (m) {
839 mn = m->m_nextpkt;
840 m->m_nextpkt = NULL;
841
842 /*
843 * We will rely on rcvif being set properly in the deferred
844 * context, so assert it is correct here.
845 */
846 MPASS((m->m_pkthdr.csum_flags & CSUM_SND_TAG) == 0);
847 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
848 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
849 netisr_dispatch(NETISR_ETHER, m);
850 m = mn;
851 }
852 if (__predict_false(needs_epoch))
853 NET_EPOCH_EXIT(et);
854 CURVNET_RESTORE();
855 }
856
857 /*
858 * Upper layer processing for a received Ethernet packet.
859 */
860 void
ether_demux(struct ifnet * ifp,struct mbuf * m)861 ether_demux(struct ifnet *ifp, struct mbuf *m)
862 {
863 struct ether_header *eh;
864 int i, isr;
865 u_short ether_type;
866
867 NET_EPOCH_ASSERT();
868 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
869
870 /* Do not grab PROMISC frames in case we are re-entered. */
871 if (PFIL_HOOKED_IN(V_link_pfil_head) && !(m->m_flags & M_PROMISC)) {
872 i = pfil_mbuf_in(V_link_pfil_head, &m, ifp, NULL);
873 if (i != PFIL_PASS)
874 return;
875 }
876
877 eh = mtod(m, struct ether_header *);
878 ether_type = ntohs(eh->ether_type);
879
880 /*
881 * If this frame has a VLAN tag other than 0, call vlan_input()
882 * if its module is loaded. Otherwise, drop.
883 */
884 if ((m->m_flags & M_VLANTAG) &&
885 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
886 if (ifp->if_vlantrunk == NULL) {
887 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
888 m_freem(m);
889 return;
890 }
891 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
892 __func__));
893 /* Clear before possibly re-entering ether_input(). */
894 m->m_flags &= ~M_PROMISC;
895 (*vlan_input_p)(ifp, m);
896 return;
897 }
898
899 /*
900 * Pass promiscuously received frames to the upper layer if the user
901 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
902 */
903 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
904 m_freem(m);
905 return;
906 }
907
908 /*
909 * Reset layer specific mbuf flags to avoid confusing upper layers.
910 */
911 m->m_flags &= ~M_VLANTAG;
912 m_clrprotoflags(m);
913
914 /*
915 * Dispatch frame to upper layer.
916 */
917 switch (ether_type) {
918 #ifdef INET
919 case ETHERTYPE_IP:
920 isr = NETISR_IP;
921 break;
922
923 case ETHERTYPE_ARP:
924 if (ifp->if_flags & IFF_NOARP) {
925 /* Discard packet if ARP is disabled on interface */
926 m_freem(m);
927 return;
928 }
929 isr = NETISR_ARP;
930 break;
931 #endif
932 #ifdef INET6
933 case ETHERTYPE_IPV6:
934 isr = NETISR_IPV6;
935 break;
936 #endif
937 default:
938 goto discard;
939 }
940
941 /* Strip off Ethernet header. */
942 m_adj(m, ETHER_HDR_LEN);
943
944 netisr_dispatch(isr, m);
945 return;
946
947 discard:
948 /*
949 * Packet is to be discarded. If netgraph is present,
950 * hand the packet to it for last chance processing;
951 * otherwise dispose of it.
952 */
953 if (ifp->if_l2com != NULL) {
954 KASSERT(ng_ether_input_orphan_p != NULL,
955 ("ng_ether_input_orphan_p is NULL"));
956 (*ng_ether_input_orphan_p)(ifp, m);
957 return;
958 }
959 m_freem(m);
960 }
961
962 /*
963 * Convert Ethernet address to printable (loggable) representation.
964 * This routine is for compatibility; it's better to just use
965 *
966 * printf("%6D", <pointer to address>, ":");
967 *
968 * since there's no static buffer involved.
969 */
970 char *
ether_sprintf(const u_char * ap)971 ether_sprintf(const u_char *ap)
972 {
973 static char etherbuf[18];
974 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
975 return (etherbuf);
976 }
977
978 /*
979 * Perform common duties while attaching to interface list
980 */
981 void
ether_ifattach(struct ifnet * ifp,const u_int8_t * lla)982 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
983 {
984 int i;
985 struct ifaddr *ifa;
986 struct sockaddr_dl *sdl;
987
988 ifp->if_addrlen = ETHER_ADDR_LEN;
989 ifp->if_hdrlen = ETHER_HDR_LEN;
990 ifp->if_mtu = ETHERMTU;
991 if_attach(ifp);
992 ifp->if_output = ether_output;
993 ifp->if_input = ether_input;
994 ifp->if_resolvemulti = ether_resolvemulti;
995 ifp->if_requestencap = ether_requestencap;
996 #ifdef VIMAGE
997 ifp->if_reassign = ether_reassign;
998 #endif
999 if (ifp->if_baudrate == 0)
1000 ifp->if_baudrate = IF_Mbps(10); /* just a default */
1001 ifp->if_broadcastaddr = etherbroadcastaddr;
1002
1003 ifa = ifp->if_addr;
1004 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
1005 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
1006 sdl->sdl_type = IFT_ETHER;
1007 sdl->sdl_alen = ifp->if_addrlen;
1008 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
1009
1010 if (ifp->if_hw_addr != NULL)
1011 bcopy(lla, ifp->if_hw_addr, ifp->if_addrlen);
1012
1013 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
1014 if (ng_ether_attach_p != NULL)
1015 (*ng_ether_attach_p)(ifp);
1016
1017 /* Announce Ethernet MAC address if non-zero. */
1018 for (i = 0; i < ifp->if_addrlen; i++)
1019 if (lla[i] != 0)
1020 break;
1021 if (i != ifp->if_addrlen)
1022 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
1023
1024 uuid_ether_add(LLADDR(sdl));
1025
1026 /* Add necessary bits are setup; announce it now. */
1027 EVENTHANDLER_INVOKE(ether_ifattach_event, ifp);
1028 if (IS_DEFAULT_VNET(curvnet))
1029 devctl_notify("ETHERNET", ifp->if_xname, "IFATTACH", NULL);
1030 }
1031
1032 /*
1033 * Perform common duties while detaching an Ethernet interface
1034 */
1035 void
ether_ifdetach(struct ifnet * ifp)1036 ether_ifdetach(struct ifnet *ifp)
1037 {
1038 struct sockaddr_dl *sdl;
1039
1040 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
1041 uuid_ether_del(LLADDR(sdl));
1042
1043 if (ifp->if_l2com != NULL) {
1044 KASSERT(ng_ether_detach_p != NULL,
1045 ("ng_ether_detach_p is NULL"));
1046 (*ng_ether_detach_p)(ifp);
1047 }
1048
1049 bpfdetach(ifp);
1050 if_detach(ifp);
1051 }
1052
1053 #ifdef VIMAGE
1054 void
ether_reassign(struct ifnet * ifp,struct vnet * new_vnet,char * unused __unused)1055 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
1056 {
1057
1058 if (ifp->if_l2com != NULL) {
1059 KASSERT(ng_ether_detach_p != NULL,
1060 ("ng_ether_detach_p is NULL"));
1061 (*ng_ether_detach_p)(ifp);
1062 }
1063
1064 if (ng_ether_attach_p != NULL) {
1065 CURVNET_SET_QUIET(new_vnet);
1066 (*ng_ether_attach_p)(ifp);
1067 CURVNET_RESTORE();
1068 }
1069 }
1070 #endif
1071
1072 SYSCTL_DECL(_net_link);
1073 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1074 "Ethernet");
1075
1076 #if 0
1077 /*
1078 * This is for reference. We have a table-driven version
1079 * of the little-endian crc32 generator, which is faster
1080 * than the double-loop.
1081 */
1082 uint32_t
1083 ether_crc32_le(const uint8_t *buf, size_t len)
1084 {
1085 size_t i;
1086 uint32_t crc;
1087 int bit;
1088 uint8_t data;
1089
1090 crc = 0xffffffff; /* initial value */
1091
1092 for (i = 0; i < len; i++) {
1093 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1094 carry = (crc ^ data) & 1;
1095 crc >>= 1;
1096 if (carry)
1097 crc = (crc ^ ETHER_CRC_POLY_LE);
1098 }
1099 }
1100
1101 return (crc);
1102 }
1103 #else
1104 uint32_t
ether_crc32_le(const uint8_t * buf,size_t len)1105 ether_crc32_le(const uint8_t *buf, size_t len)
1106 {
1107 static const uint32_t crctab[] = {
1108 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1109 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1110 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1111 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1112 };
1113 size_t i;
1114 uint32_t crc;
1115
1116 crc = 0xffffffff; /* initial value */
1117
1118 for (i = 0; i < len; i++) {
1119 crc ^= buf[i];
1120 crc = (crc >> 4) ^ crctab[crc & 0xf];
1121 crc = (crc >> 4) ^ crctab[crc & 0xf];
1122 }
1123
1124 return (crc);
1125 }
1126 #endif
1127
1128 uint32_t
ether_crc32_be(const uint8_t * buf,size_t len)1129 ether_crc32_be(const uint8_t *buf, size_t len)
1130 {
1131 size_t i;
1132 uint32_t crc, carry;
1133 int bit;
1134 uint8_t data;
1135
1136 crc = 0xffffffff; /* initial value */
1137
1138 for (i = 0; i < len; i++) {
1139 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1140 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1141 crc <<= 1;
1142 if (carry)
1143 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1144 }
1145 }
1146
1147 return (crc);
1148 }
1149
1150 int
ether_ioctl(struct ifnet * ifp,u_long command,caddr_t data)1151 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1152 {
1153 struct ifaddr *ifa = (struct ifaddr *) data;
1154 struct ifreq *ifr = (struct ifreq *) data;
1155 int error = 0;
1156
1157 switch (command) {
1158 case SIOCSIFADDR:
1159 ifp->if_flags |= IFF_UP;
1160
1161 switch (ifa->ifa_addr->sa_family) {
1162 #ifdef INET
1163 case AF_INET:
1164 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1165 arp_ifinit(ifp, ifa);
1166 break;
1167 #endif
1168 default:
1169 ifp->if_init(ifp->if_softc);
1170 break;
1171 }
1172 break;
1173
1174 case SIOCGIFADDR:
1175 bcopy(IF_LLADDR(ifp), &ifr->ifr_addr.sa_data[0],
1176 ETHER_ADDR_LEN);
1177 break;
1178
1179 case SIOCSIFMTU:
1180 /*
1181 * Set the interface MTU.
1182 */
1183 if (ifr->ifr_mtu > ETHERMTU) {
1184 error = EINVAL;
1185 } else {
1186 ifp->if_mtu = ifr->ifr_mtu;
1187 }
1188 break;
1189
1190 case SIOCSLANPCP:
1191 error = priv_check(curthread, PRIV_NET_SETLANPCP);
1192 if (error != 0)
1193 break;
1194 if (ifr->ifr_lan_pcp > 7 &&
1195 ifr->ifr_lan_pcp != IFNET_PCP_NONE) {
1196 error = EINVAL;
1197 } else {
1198 ifp->if_pcp = ifr->ifr_lan_pcp;
1199 /* broadcast event about PCP change */
1200 EVENTHANDLER_INVOKE(ifnet_event, ifp, IFNET_EVENT_PCP);
1201 }
1202 break;
1203
1204 case SIOCGLANPCP:
1205 ifr->ifr_lan_pcp = ifp->if_pcp;
1206 break;
1207
1208 default:
1209 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1210 break;
1211 }
1212 return (error);
1213 }
1214
1215 static int
ether_resolvemulti(struct ifnet * ifp,struct sockaddr ** llsa,struct sockaddr * sa)1216 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1217 struct sockaddr *sa)
1218 {
1219 struct sockaddr_dl *sdl;
1220 #ifdef INET
1221 struct sockaddr_in *sin;
1222 #endif
1223 #ifdef INET6
1224 struct sockaddr_in6 *sin6;
1225 #endif
1226 u_char *e_addr;
1227
1228 switch(sa->sa_family) {
1229 case AF_LINK:
1230 /*
1231 * No mapping needed. Just check that it's a valid MC address.
1232 */
1233 sdl = (struct sockaddr_dl *)sa;
1234 e_addr = LLADDR(sdl);
1235 if (!ETHER_IS_MULTICAST(e_addr))
1236 return EADDRNOTAVAIL;
1237 *llsa = NULL;
1238 return 0;
1239
1240 #ifdef INET
1241 case AF_INET:
1242 sin = (struct sockaddr_in *)sa;
1243 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1244 return EADDRNOTAVAIL;
1245 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1246 sdl->sdl_alen = ETHER_ADDR_LEN;
1247 e_addr = LLADDR(sdl);
1248 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1249 *llsa = (struct sockaddr *)sdl;
1250 return 0;
1251 #endif
1252 #ifdef INET6
1253 case AF_INET6:
1254 sin6 = (struct sockaddr_in6 *)sa;
1255 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1256 /*
1257 * An IP6 address of 0 means listen to all
1258 * of the Ethernet multicast address used for IP6.
1259 * (This is used for multicast routers.)
1260 */
1261 ifp->if_flags |= IFF_ALLMULTI;
1262 *llsa = NULL;
1263 return 0;
1264 }
1265 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1266 return EADDRNOTAVAIL;
1267 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1268 sdl->sdl_alen = ETHER_ADDR_LEN;
1269 e_addr = LLADDR(sdl);
1270 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1271 *llsa = (struct sockaddr *)sdl;
1272 return 0;
1273 #endif
1274
1275 default:
1276 /*
1277 * Well, the text isn't quite right, but it's the name
1278 * that counts...
1279 */
1280 return EAFNOSUPPORT;
1281 }
1282 }
1283
1284 static moduledata_t ether_mod = {
1285 .name = "ether",
1286 };
1287
1288 void
ether_vlan_mtap(struct bpf_if * bp,struct mbuf * m,void * data,u_int dlen)1289 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1290 {
1291 struct ether_vlan_header vlan;
1292 struct mbuf mv, mb;
1293
1294 KASSERT((m->m_flags & M_VLANTAG) != 0,
1295 ("%s: vlan information not present", __func__));
1296 KASSERT(m->m_len >= sizeof(struct ether_header),
1297 ("%s: mbuf not large enough for header", __func__));
1298 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1299 vlan.evl_proto = vlan.evl_encap_proto;
1300 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1301 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1302 m->m_len -= sizeof(struct ether_header);
1303 m->m_data += sizeof(struct ether_header);
1304 /*
1305 * If a data link has been supplied by the caller, then we will need to
1306 * re-create a stack allocated mbuf chain with the following structure:
1307 *
1308 * (1) mbuf #1 will contain the supplied data link
1309 * (2) mbuf #2 will contain the vlan header
1310 * (3) mbuf #3 will contain the original mbuf's packet data
1311 *
1312 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1313 */
1314 if (data != NULL) {
1315 mv.m_next = m;
1316 mv.m_data = (caddr_t)&vlan;
1317 mv.m_len = sizeof(vlan);
1318 mb.m_next = &mv;
1319 mb.m_data = data;
1320 mb.m_len = dlen;
1321 bpf_mtap(bp, &mb);
1322 } else
1323 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1324 m->m_len += sizeof(struct ether_header);
1325 m->m_data -= sizeof(struct ether_header);
1326 }
1327
1328 struct mbuf *
ether_vlanencap_proto(struct mbuf * m,uint16_t tag,uint16_t proto)1329 ether_vlanencap_proto(struct mbuf *m, uint16_t tag, uint16_t proto)
1330 {
1331 struct ether_vlan_header *evl;
1332
1333 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1334 if (m == NULL)
1335 return (NULL);
1336 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1337
1338 if (m->m_len < sizeof(*evl)) {
1339 m = m_pullup(m, sizeof(*evl));
1340 if (m == NULL)
1341 return (NULL);
1342 }
1343
1344 /*
1345 * Transform the Ethernet header into an Ethernet header
1346 * with 802.1Q encapsulation.
1347 */
1348 evl = mtod(m, struct ether_vlan_header *);
1349 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1350 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1351 evl->evl_encap_proto = htons(proto);
1352 evl->evl_tag = htons(tag);
1353 return (m);
1354 }
1355
1356 void
ether_bpf_mtap_if(struct ifnet * ifp,struct mbuf * m)1357 ether_bpf_mtap_if(struct ifnet *ifp, struct mbuf *m)
1358 {
1359 if (bpf_peers_present(ifp->if_bpf)) {
1360 M_ASSERTVALID(m);
1361 if ((m->m_flags & M_VLANTAG) != 0)
1362 ether_vlan_mtap(ifp->if_bpf, m, NULL, 0);
1363 else
1364 bpf_mtap(ifp->if_bpf, m);
1365 }
1366 }
1367
1368 static SYSCTL_NODE(_net_link, IFT_L2VLAN, vlan, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1369 "IEEE 802.1Q VLAN");
1370 static SYSCTL_NODE(_net_link_vlan, PF_LINK, link,
1371 CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
1372 "for consistency");
1373
1374 VNET_DEFINE_STATIC(int, soft_pad);
1375 #define V_soft_pad VNET(soft_pad)
1376 SYSCTL_INT(_net_link_vlan, OID_AUTO, soft_pad, CTLFLAG_RW | CTLFLAG_VNET,
1377 &VNET_NAME(soft_pad), 0,
1378 "pad short frames before tagging");
1379
1380 /*
1381 * For now, make preserving PCP via an mbuf tag optional, as it increases
1382 * per-packet memory allocations and frees. In the future, it would be
1383 * preferable to reuse ether_vtag for this, or similar.
1384 */
1385 VNET_DEFINE(int, vlan_mtag_pcp) = 0;
1386 #define V_vlan_mtag_pcp VNET(vlan_mtag_pcp)
1387 SYSCTL_INT(_net_link_vlan, OID_AUTO, mtag_pcp, CTLFLAG_RW | CTLFLAG_VNET,
1388 &VNET_NAME(vlan_mtag_pcp), 0,
1389 "Retain VLAN PCP information as packets are passed up the stack");
1390
1391 static inline bool
ether_do_pcp(struct ifnet * ifp,struct mbuf * m)1392 ether_do_pcp(struct ifnet *ifp, struct mbuf *m)
1393 {
1394 if (ifp->if_type == IFT_L2VLAN)
1395 return (false);
1396 if (ifp->if_pcp != IFNET_PCP_NONE || (m->m_flags & M_VLANTAG) != 0)
1397 return (true);
1398 if (V_vlan_mtag_pcp &&
1399 m_tag_locate(m, MTAG_8021Q, MTAG_8021Q_PCP_OUT, NULL) != NULL)
1400 return (true);
1401 return (false);
1402 }
1403
1404 bool
ether_8021q_frame(struct mbuf ** mp,struct ifnet * ife,struct ifnet * p,const struct ether_8021q_tag * qtag)1405 ether_8021q_frame(struct mbuf **mp, struct ifnet *ife, struct ifnet *p,
1406 const struct ether_8021q_tag *qtag)
1407 {
1408 struct m_tag *mtag;
1409 int n;
1410 uint16_t tag;
1411 uint8_t pcp = qtag->pcp;
1412 static const char pad[8]; /* just zeros */
1413
1414 /*
1415 * Pad the frame to the minimum size allowed if told to.
1416 * This option is in accord with IEEE Std 802.1Q, 2003 Ed.,
1417 * paragraph C.4.4.3.b. It can help to work around buggy
1418 * bridges that violate paragraph C.4.4.3.a from the same
1419 * document, i.e., fail to pad short frames after untagging.
1420 * E.g., a tagged frame 66 bytes long (incl. FCS) is OK, but
1421 * untagging it will produce a 62-byte frame, which is a runt
1422 * and requires padding. There are VLAN-enabled network
1423 * devices that just discard such runts instead or mishandle
1424 * them somehow.
1425 */
1426 if (V_soft_pad && p->if_type == IFT_ETHER) {
1427 for (n = ETHERMIN + ETHER_HDR_LEN - (*mp)->m_pkthdr.len;
1428 n > 0; n -= sizeof(pad)) {
1429 if (!m_append(*mp, min(n, sizeof(pad)), pad))
1430 break;
1431 }
1432 if (n > 0) {
1433 m_freem(*mp);
1434 *mp = NULL;
1435 if_printf(ife, "cannot pad short frame");
1436 return (false);
1437 }
1438 }
1439
1440 /*
1441 * If PCP is set in mbuf, use it
1442 */
1443 if ((*mp)->m_flags & M_VLANTAG) {
1444 pcp = EVL_PRIOFTAG((*mp)->m_pkthdr.ether_vtag);
1445 }
1446
1447 /*
1448 * If underlying interface can do VLAN tag insertion itself,
1449 * just pass the packet along. However, we need some way to
1450 * tell the interface where the packet came from so that it
1451 * knows how to find the VLAN tag to use, so we attach a
1452 * packet tag that holds it.
1453 */
1454 if (V_vlan_mtag_pcp && (mtag = m_tag_locate(*mp, MTAG_8021Q,
1455 MTAG_8021Q_PCP_OUT, NULL)) != NULL)
1456 tag = EVL_MAKETAG(qtag->vid, *(uint8_t *)(mtag + 1), 0);
1457 else
1458 tag = EVL_MAKETAG(qtag->vid, pcp, 0);
1459 if ((p->if_capenable & IFCAP_VLAN_HWTAGGING) &&
1460 (qtag->proto == ETHERTYPE_VLAN)) {
1461 (*mp)->m_pkthdr.ether_vtag = tag;
1462 (*mp)->m_flags |= M_VLANTAG;
1463 } else {
1464 *mp = ether_vlanencap_proto(*mp, tag, qtag->proto);
1465 if (*mp == NULL) {
1466 if_printf(ife, "unable to prepend 802.1Q header");
1467 return (false);
1468 }
1469 (*mp)->m_flags &= ~M_VLANTAG;
1470 }
1471 return (true);
1472 }
1473
1474 /*
1475 * Allocate an address from the FreeBSD Foundation OUI. This uses a
1476 * cryptographic hash function on the containing jail's name, UUID and the
1477 * interface name to attempt to provide a unique but stable address.
1478 * Pseudo-interfaces which require a MAC address should use this function to
1479 * allocate non-locally-administered addresses.
1480 */
1481 void
ether_gen_addr_byname(const char * nameunit,struct ether_addr * hwaddr)1482 ether_gen_addr_byname(const char *nameunit, struct ether_addr *hwaddr)
1483 {
1484 SHA1_CTX ctx;
1485 char *buf;
1486 char uuid[HOSTUUIDLEN + 1];
1487 uint64_t addr;
1488 int i, sz;
1489 char digest[SHA1_RESULTLEN];
1490 char jailname[MAXHOSTNAMELEN];
1491
1492 getcredhostuuid(curthread->td_ucred, uuid, sizeof(uuid));
1493 if (strncmp(uuid, DEFAULT_HOSTUUID, sizeof(uuid)) == 0) {
1494 /* Fall back to a random mac address. */
1495 goto rando;
1496 }
1497
1498 /* If each (vnet) jail would also have a unique hostuuid this would not
1499 * be necessary. */
1500 getjailname(curthread->td_ucred, jailname, sizeof(jailname));
1501 sz = asprintf(&buf, M_TEMP, "%s-%s-%s", uuid, nameunit,
1502 jailname);
1503 if (sz < 0) {
1504 /* Fall back to a random mac address. */
1505 goto rando;
1506 }
1507
1508 SHA1Init(&ctx);
1509 SHA1Update(&ctx, buf, sz);
1510 SHA1Final(digest, &ctx);
1511 free(buf, M_TEMP);
1512
1513 addr = ((digest[0] << 16) | (digest[1] << 8) | digest[2]) &
1514 OUI_FREEBSD_GENERATED_MASK;
1515 addr = OUI_FREEBSD(addr);
1516 for (i = 0; i < ETHER_ADDR_LEN; ++i) {
1517 hwaddr->octet[i] = addr >> ((ETHER_ADDR_LEN - i - 1) * 8) &
1518 0xFF;
1519 }
1520
1521 return;
1522 rando:
1523 arc4rand(hwaddr, sizeof(*hwaddr), 0);
1524 /* Unicast */
1525 hwaddr->octet[0] &= 0xFE;
1526 /* Locally administered. */
1527 hwaddr->octet[0] |= 0x02;
1528 }
1529
1530 void
ether_gen_addr(struct ifnet * ifp,struct ether_addr * hwaddr)1531 ether_gen_addr(struct ifnet *ifp, struct ether_addr *hwaddr)
1532 {
1533 ether_gen_addr_byname(if_name(ifp), hwaddr);
1534 }
1535
1536 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1537 MODULE_VERSION(ether, 1);
1538