1 /*-
2 * Copyright (c) 1982, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
16 *
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
27 * SUCH DAMAGE.
28 *
29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
30 * $FreeBSD: releng/11.0/sys/net/if_ethersubr.c 301498 2016-06-06 10:13:48Z bz $
31 */
32
33 #include "opt_inet.h"
34 #include "opt_inet6.h"
35 #include "opt_netgraph.h"
36 #include "opt_mbuf_profiling.h"
37 #include "opt_rss.h"
38
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/kernel.h>
42 #include <sys/lock.h>
43 #include <sys/malloc.h>
44 #include <sys/module.h>
45 #include <sys/mbuf.h>
46 #include <sys/random.h>
47 #include <sys/socket.h>
48 #include <sys/sockio.h>
49 #include <sys/sysctl.h>
50 #include <sys/uuid.h>
51
52 #include <net/if.h>
53 #include <net/if_var.h>
54 #include <net/if_arp.h>
55 #include <net/netisr.h>
56 #include <net/route.h>
57 #include <net/if_llc.h>
58 #include <net/if_dl.h>
59 #include <net/if_types.h>
60 #include <net/bpf.h>
61 #include <net/ethernet.h>
62 #include <net/if_bridgevar.h>
63 #include <net/if_vlan_var.h>
64 #include <net/if_llatbl.h>
65 #include <net/pfil.h>
66 #include <net/rss_config.h>
67 #include <net/vnet.h>
68
69 #include <netpfil/pf/pf_mtag.h>
70
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74 #include <netinet/if_ether.h>
75 #include <netinet/ip_carp.h>
76 #include <netinet/ip_var.h>
77 #endif
78 #ifdef INET6
79 #include <netinet6/nd6.h>
80 #endif
81 #include <security/mac/mac_framework.h>
82
83 #ifdef CTASSERT
84 CTASSERT(sizeof (struct ether_header) == ETHER_ADDR_LEN * 2 + 2);
85 CTASSERT(sizeof (struct ether_addr) == ETHER_ADDR_LEN);
86 #endif
87
88 VNET_DEFINE(struct pfil_head, link_pfil_hook); /* Packet filter hooks */
89
90 /* netgraph node hooks for ng_ether(4) */
91 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
92 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
93 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
94 void (*ng_ether_attach_p)(struct ifnet *ifp);
95 void (*ng_ether_detach_p)(struct ifnet *ifp);
96
97 void (*vlan_input_p)(struct ifnet *, struct mbuf *);
98
99 /* if_bridge(4) support */
100 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
101 int (*bridge_output_p)(struct ifnet *, struct mbuf *,
102 struct sockaddr *, struct rtentry *);
103 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
104
105 /* if_lagg(4) support */
106 struct mbuf *(*lagg_input_p)(struct ifnet *, struct mbuf *);
107
108 static const u_char etherbroadcastaddr[ETHER_ADDR_LEN] =
109 { 0xff, 0xff, 0xff, 0xff, 0xff, 0xff };
110
111 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
112 struct sockaddr *);
113 #ifdef VIMAGE
114 static void ether_reassign(struct ifnet *, struct vnet *, char *);
115 #endif
116 static int ether_requestencap(struct ifnet *, struct if_encap_req *);
117
118 #define ETHER_IS_BROADCAST(addr) \
119 (bcmp(etherbroadcastaddr, (addr), ETHER_ADDR_LEN) == 0)
120
121 #define senderr(e) do { error = (e); goto bad;} while (0)
122
123 static void
124 update_mbuf_csumflags(struct mbuf *src, struct mbuf *dst)
125 {
126 int csum_flags = 0;
127
128 if (src->m_pkthdr.csum_flags & CSUM_IP)
129 csum_flags |= (CSUM_IP_CHECKED|CSUM_IP_VALID);
130 if (src->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
131 csum_flags |= (CSUM_DATA_VALID|CSUM_PSEUDO_HDR);
132 if (src->m_pkthdr.csum_flags & CSUM_SCTP)
133 csum_flags |= CSUM_SCTP_VALID;
134 dst->m_pkthdr.csum_flags |= csum_flags;
135 if (csum_flags & CSUM_DATA_VALID)
136 dst->m_pkthdr.csum_data = 0xffff;
137 }
138
139 /*
140 * Handle link-layer encapsulation requests.
141 */
142 static int
143 ether_requestencap(struct ifnet *ifp, struct if_encap_req *req)
144 {
145 struct ether_header *eh;
146 struct arphdr *ah;
147 uint16_t etype;
148 const u_char *lladdr;
149
150 if (req->rtype != IFENCAP_LL)
151 return (EOPNOTSUPP);
152
153 if (req->bufsize < ETHER_HDR_LEN)
154 return (ENOMEM);
155
156 eh = (struct ether_header *)req->buf;
157 lladdr = req->lladdr;
158 req->lladdr_off = 0;
159
160 switch (req->family) {
161 case AF_INET:
162 etype = htons(ETHERTYPE_IP);
163 break;
164 case AF_INET6:
165 etype = htons(ETHERTYPE_IPV6);
166 break;
167 case AF_ARP:
168 ah = (struct arphdr *)req->hdata;
169 ah->ar_hrd = htons(ARPHRD_ETHER);
170
171 switch(ntohs(ah->ar_op)) {
172 case ARPOP_REVREQUEST:
173 case ARPOP_REVREPLY:
174 etype = htons(ETHERTYPE_REVARP);
175 break;
176 case ARPOP_REQUEST:
177 case ARPOP_REPLY:
178 default:
179 etype = htons(ETHERTYPE_ARP);
180 break;
181 }
182
183 if (req->flags & IFENCAP_FLAG_BROADCAST)
184 lladdr = ifp->if_broadcastaddr;
185 break;
186 default:
187 return (EAFNOSUPPORT);
188 }
189
190 memcpy(&eh->ether_type, &etype, sizeof(eh->ether_type));
191 memcpy(eh->ether_dhost, lladdr, ETHER_ADDR_LEN);
192 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
193 req->bufsize = sizeof(struct ether_header);
194
195 return (0);
196 }
197
198
199 static int
200 ether_resolve_addr(struct ifnet *ifp, struct mbuf *m,
201 const struct sockaddr *dst, struct route *ro, u_char *phdr,
202 uint32_t *pflags, struct llentry **plle)
203 {
204 struct ether_header *eh;
205 uint32_t lleflags = 0;
206 int error = 0;
207 #if defined(INET) || defined(INET6)
208 uint16_t etype;
209 #endif
210
211 if (plle)
212 *plle = NULL;
213 eh = (struct ether_header *)phdr;
214
215 switch (dst->sa_family) {
216 #ifdef INET
217 case AF_INET:
218 if ((m->m_flags & (M_BCAST | M_MCAST)) == 0)
219 error = arpresolve(ifp, 0, m, dst, phdr, &lleflags,
220 plle);
221 else {
222 if (m->m_flags & M_BCAST)
223 memcpy(eh->ether_dhost, ifp->if_broadcastaddr,
224 ETHER_ADDR_LEN);
225 else {
226 const struct in_addr *a;
227 a = &(((const struct sockaddr_in *)dst)->sin_addr);
228 ETHER_MAP_IP_MULTICAST(a, eh->ether_dhost);
229 }
230 etype = htons(ETHERTYPE_IP);
231 memcpy(&eh->ether_type, &etype, sizeof(etype));
232 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
233 }
234 break;
235 #endif
236 #ifdef INET6
237 case AF_INET6:
238 if ((m->m_flags & M_MCAST) == 0)
239 error = nd6_resolve(ifp, 0, m, dst, phdr, &lleflags,
240 plle);
241 else {
242 const struct in6_addr *a6;
243 a6 = &(((const struct sockaddr_in6 *)dst)->sin6_addr);
244 ETHER_MAP_IPV6_MULTICAST(a6, eh->ether_dhost);
245 etype = htons(ETHERTYPE_IPV6);
246 memcpy(&eh->ether_type, &etype, sizeof(etype));
247 memcpy(eh->ether_shost, IF_LLADDR(ifp), ETHER_ADDR_LEN);
248 }
249 break;
250 #endif
251 default:
252 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
253 if (m != NULL)
254 m_freem(m);
255 return (EAFNOSUPPORT);
256 }
257
258 if (error == EHOSTDOWN) {
259 if (ro != NULL && (ro->ro_flags & RT_HAS_GW) != 0)
260 error = EHOSTUNREACH;
261 }
262
263 if (error != 0)
264 return (error);
265
266 *pflags = RT_MAY_LOOP;
267 if (lleflags & LLE_IFADDR)
268 *pflags |= RT_L2_ME;
269
270 return (0);
271 }
272
273 /*
274 * Ethernet output routine.
275 * Encapsulate a packet of type family for the local net.
276 * Use trailer local net encapsulation if enough data in first
277 * packet leaves a multiple of 512 bytes of data in remainder.
278 */
279 int
280 ether_output(struct ifnet *ifp, struct mbuf *m,
281 const struct sockaddr *dst, struct route *ro)
282 {
283 int error = 0;
284 char linkhdr[ETHER_HDR_LEN], *phdr;
285 struct ether_header *eh;
286 struct pf_mtag *t;
287 int loop_copy = 1;
288 int hlen; /* link layer header length */
289 uint32_t pflags;
290 struct llentry *lle = NULL;
291 struct rtentry *rt0 = NULL;
292 int addref = 0;
293
294 phdr = NULL;
295 pflags = 0;
296 if (ro != NULL) {
297 /* XXX BPF uses ro_prepend */
298 if (ro->ro_prepend != NULL) {
299 phdr = ro->ro_prepend;
300 hlen = ro->ro_plen;
301 } else if (!(m->m_flags & (M_BCAST | M_MCAST))) {
302 if ((ro->ro_flags & RT_LLE_CACHE) != 0) {
303 lle = ro->ro_lle;
304 if (lle != NULL &&
305 (lle->la_flags & LLE_VALID) == 0) {
306 LLE_FREE(lle);
307 lle = NULL; /* redundant */
308 ro->ro_lle = NULL;
309 }
310 if (lle == NULL) {
311 /* if we lookup, keep cache */
312 addref = 1;
313 }
314 }
315 if (lle != NULL) {
316 phdr = lle->r_linkdata;
317 hlen = lle->r_hdrlen;
318 pflags = lle->r_flags;
319 }
320 }
321 rt0 = ro->ro_rt;
322 }
323
324 #ifdef MAC
325 error = mac_ifnet_check_transmit(ifp, m);
326 if (error)
327 senderr(error);
328 #endif
329
330 M_PROFILE(m);
331 if (ifp->if_flags & IFF_MONITOR)
332 senderr(ENETDOWN);
333 if (!((ifp->if_flags & IFF_UP) &&
334 (ifp->if_drv_flags & IFF_DRV_RUNNING)))
335 senderr(ENETDOWN);
336
337 if (phdr == NULL) {
338 /* No prepend data supplied. Try to calculate ourselves. */
339 phdr = linkhdr;
340 hlen = ETHER_HDR_LEN;
341 error = ether_resolve_addr(ifp, m, dst, ro, phdr, &pflags,
342 addref ? &lle : NULL);
343 if (addref && lle != NULL)
344 ro->ro_lle = lle;
345 if (error != 0)
346 return (error == EWOULDBLOCK ? 0 : error);
347 }
348
349 if ((pflags & RT_L2_ME) != 0) {
350 update_mbuf_csumflags(m, m);
351 return (if_simloop(ifp, m, dst->sa_family, 0));
352 }
353 loop_copy = pflags & RT_MAY_LOOP;
354
355 /*
356 * Add local net header. If no space in first mbuf,
357 * allocate another.
358 *
359 * Note that we do prepend regardless of RT_HAS_HEADER flag.
360 * This is done because BPF code shifts m_data pointer
361 * to the end of ethernet header prior to calling if_output().
362 */
363 M_PREPEND(m, hlen, M_NOWAIT);
364 if (m == NULL)
365 senderr(ENOBUFS);
366 if ((pflags & RT_HAS_HEADER) == 0) {
367 eh = mtod(m, struct ether_header *);
368 memcpy(eh, phdr, hlen);
369 }
370
371 /*
372 * If a simplex interface, and the packet is being sent to our
373 * Ethernet address or a broadcast address, loopback a copy.
374 * XXX To make a simplex device behave exactly like a duplex
375 * device, we should copy in the case of sending to our own
376 * ethernet address (thus letting the original actually appear
377 * on the wire). However, we don't do that here for security
378 * reasons and compatibility with the original behavior.
379 */
380 if ((m->m_flags & M_BCAST) && loop_copy && (ifp->if_flags & IFF_SIMPLEX) &&
381 ((t = pf_find_mtag(m)) == NULL || !t->routed)) {
382 struct mbuf *n;
383
384 /*
385 * Because if_simloop() modifies the packet, we need a
386 * writable copy through m_dup() instead of a readonly
387 * one as m_copy[m] would give us. The alternative would
388 * be to modify if_simloop() to handle the readonly mbuf,
389 * but performancewise it is mostly equivalent (trading
390 * extra data copying vs. extra locking).
391 *
392 * XXX This is a local workaround. A number of less
393 * often used kernel parts suffer from the same bug.
394 * See PR kern/105943 for a proposed general solution.
395 */
396 if ((n = m_dup(m, M_NOWAIT)) != NULL) {
397 update_mbuf_csumflags(m, n);
398 (void)if_simloop(ifp, n, dst->sa_family, hlen);
399 } else
400 if_inc_counter(ifp, IFCOUNTER_IQDROPS, 1);
401 }
402
403 /*
404 * Bridges require special output handling.
405 */
406 if (ifp->if_bridge) {
407 BRIDGE_OUTPUT(ifp, m, error);
408 return (error);
409 }
410
411 #if defined(INET) || defined(INET6)
412 if (ifp->if_carp &&
413 (error = (*carp_output_p)(ifp, m, dst)))
414 goto bad;
415 #endif
416
417 /* Handle ng_ether(4) processing, if any */
418 if (ifp->if_l2com != NULL) {
419 KASSERT(ng_ether_output_p != NULL,
420 ("ng_ether_output_p is NULL"));
421 if ((error = (*ng_ether_output_p)(ifp, &m)) != 0) {
422 bad: if (m != NULL)
423 m_freem(m);
424 return (error);
425 }
426 if (m == NULL)
427 return (0);
428 }
429
430 /* Continue with link-layer output */
431 return ether_output_frame(ifp, m);
432 }
433
434 /*
435 * Ethernet link layer output routine to send a raw frame to the device.
436 *
437 * This assumes that the 14 byte Ethernet header is present and contiguous
438 * in the first mbuf (if BRIDGE'ing).
439 */
440 int
441 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
442 {
443 int i;
444
445 if (PFIL_HOOKED(&V_link_pfil_hook)) {
446 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_OUT, NULL);
447
448 if (i != 0)
449 return (EACCES);
450
451 if (m == NULL)
452 return (0);
453 }
454
455 /*
456 * Queue message on interface, update output statistics if
457 * successful, and start output if interface not yet active.
458 */
459 return ((ifp->if_transmit)(ifp, m));
460 }
461
462 /*
463 * Process a received Ethernet packet; the packet is in the
464 * mbuf chain m with the ethernet header at the front.
465 */
466 static void
467 ether_input_internal(struct ifnet *ifp, struct mbuf *m)
468 {
469 struct ether_header *eh;
470 u_short etype;
471
472 if ((ifp->if_flags & IFF_UP) == 0) {
473 m_freem(m);
474 return;
475 }
476 #ifdef DIAGNOSTIC
477 if ((ifp->if_drv_flags & IFF_DRV_RUNNING) == 0) {
478 if_printf(ifp, "discard frame at !IFF_DRV_RUNNING\n");
479 m_freem(m);
480 return;
481 }
482 #endif
483 if (m->m_len < ETHER_HDR_LEN) {
484 /* XXX maybe should pullup? */
485 if_printf(ifp, "discard frame w/o leading ethernet "
486 "header (len %u pkt len %u)\n",
487 m->m_len, m->m_pkthdr.len);
488 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
489 m_freem(m);
490 return;
491 }
492 eh = mtod(m, struct ether_header *);
493 etype = ntohs(eh->ether_type);
494 random_harvest_queue(m, sizeof(*m), 2, RANDOM_NET_ETHER);
495
496 CURVNET_SET_QUIET(ifp->if_vnet);
497
498 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
499 if (ETHER_IS_BROADCAST(eh->ether_dhost))
500 m->m_flags |= M_BCAST;
501 else
502 m->m_flags |= M_MCAST;
503 if_inc_counter(ifp, IFCOUNTER_IMCASTS, 1);
504 }
505
506 #ifdef MAC
507 /*
508 * Tag the mbuf with an appropriate MAC label before any other
509 * consumers can get to it.
510 */
511 mac_ifnet_create_mbuf(ifp, m);
512 #endif
513
514 /*
515 * Give bpf a chance at the packet.
516 */
517 ETHER_BPF_MTAP(ifp, m);
518
519 /*
520 * If the CRC is still on the packet, trim it off. We do this once
521 * and once only in case we are re-entered. Nothing else on the
522 * Ethernet receive path expects to see the FCS.
523 */
524 if (m->m_flags & M_HASFCS) {
525 m_adj(m, -ETHER_CRC_LEN);
526 m->m_flags &= ~M_HASFCS;
527 }
528
529 if (!(ifp->if_capenable & IFCAP_HWSTATS))
530 if_inc_counter(ifp, IFCOUNTER_IBYTES, m->m_pkthdr.len);
531
532 /* Allow monitor mode to claim this frame, after stats are updated. */
533 if (ifp->if_flags & IFF_MONITOR) {
534 m_freem(m);
535 CURVNET_RESTORE();
536 return;
537 }
538
539 /* Handle input from a lagg(4) port */
540 if (ifp->if_type == IFT_IEEE8023ADLAG) {
541 KASSERT(lagg_input_p != NULL,
542 ("%s: if_lagg not loaded!", __func__));
543 m = (*lagg_input_p)(ifp, m);
544 if (m != NULL)
545 ifp = m->m_pkthdr.rcvif;
546 else {
547 CURVNET_RESTORE();
548 return;
549 }
550 }
551
552 /*
553 * If the hardware did not process an 802.1Q tag, do this now,
554 * to allow 802.1P priority frames to be passed to the main input
555 * path correctly.
556 * TODO: Deal with Q-in-Q frames, but not arbitrary nesting levels.
557 */
558 if ((m->m_flags & M_VLANTAG) == 0 && etype == ETHERTYPE_VLAN) {
559 struct ether_vlan_header *evl;
560
561 if (m->m_len < sizeof(*evl) &&
562 (m = m_pullup(m, sizeof(*evl))) == NULL) {
563 #ifdef DIAGNOSTIC
564 if_printf(ifp, "cannot pullup VLAN header\n");
565 #endif
566 if_inc_counter(ifp, IFCOUNTER_IERRORS, 1);
567 CURVNET_RESTORE();
568 return;
569 }
570
571 evl = mtod(m, struct ether_vlan_header *);
572 m->m_pkthdr.ether_vtag = ntohs(evl->evl_tag);
573 m->m_flags |= M_VLANTAG;
574
575 bcopy((char *)evl, (char *)evl + ETHER_VLAN_ENCAP_LEN,
576 ETHER_HDR_LEN - ETHER_TYPE_LEN);
577 m_adj(m, ETHER_VLAN_ENCAP_LEN);
578 eh = mtod(m, struct ether_header *);
579 }
580
581 M_SETFIB(m, ifp->if_fib);
582
583 /* Allow ng_ether(4) to claim this frame. */
584 if (ifp->if_l2com != NULL) {
585 KASSERT(ng_ether_input_p != NULL,
586 ("%s: ng_ether_input_p is NULL", __func__));
587 m->m_flags &= ~M_PROMISC;
588 (*ng_ether_input_p)(ifp, &m);
589 if (m == NULL) {
590 CURVNET_RESTORE();
591 return;
592 }
593 eh = mtod(m, struct ether_header *);
594 }
595
596 /*
597 * Allow if_bridge(4) to claim this frame.
598 * The BRIDGE_INPUT() macro will update ifp if the bridge changed it
599 * and the frame should be delivered locally.
600 */
601 if (ifp->if_bridge != NULL) {
602 m->m_flags &= ~M_PROMISC;
603 BRIDGE_INPUT(ifp, m);
604 if (m == NULL) {
605 CURVNET_RESTORE();
606 return;
607 }
608 eh = mtod(m, struct ether_header *);
609 }
610
611 #if defined(INET) || defined(INET6)
612 /*
613 * Clear M_PROMISC on frame so that carp(4) will see it when the
614 * mbuf flows up to Layer 3.
615 * FreeBSD's implementation of carp(4) uses the inprotosw
616 * to dispatch IPPROTO_CARP. carp(4) also allocates its own
617 * Ethernet addresses of the form 00:00:5e:00:01:xx, which
618 * is outside the scope of the M_PROMISC test below.
619 * TODO: Maintain a hash table of ethernet addresses other than
620 * ether_dhost which may be active on this ifp.
621 */
622 if (ifp->if_carp && (*carp_forus_p)(ifp, eh->ether_dhost)) {
623 m->m_flags &= ~M_PROMISC;
624 } else
625 #endif
626 {
627 /*
628 * If the frame received was not for our MAC address, set the
629 * M_PROMISC flag on the mbuf chain. The frame may need to
630 * be seen by the rest of the Ethernet input path in case of
631 * re-entry (e.g. bridge, vlan, netgraph) but should not be
632 * seen by upper protocol layers.
633 */
634 if (!ETHER_IS_MULTICAST(eh->ether_dhost) &&
635 bcmp(IF_LLADDR(ifp), eh->ether_dhost, ETHER_ADDR_LEN) != 0)
636 m->m_flags |= M_PROMISC;
637 }
638
639 ether_demux(ifp, m);
640 CURVNET_RESTORE();
641 }
642
643 /*
644 * Ethernet input dispatch; by default, direct dispatch here regardless of
645 * global configuration. However, if RSS is enabled, hook up RSS affinity
646 * so that when deferred or hybrid dispatch is enabled, we can redistribute
647 * load based on RSS.
648 *
649 * XXXRW: Would be nice if the ifnet passed up a flag indicating whether or
650 * not it had already done work distribution via multi-queue. Then we could
651 * direct dispatch in the event load balancing was already complete and
652 * handle the case of interfaces with different capabilities better.
653 *
654 * XXXRW: Sort of want an M_DISTRIBUTED flag to avoid multiple distributions
655 * at multiple layers?
656 *
657 * XXXRW: For now, enable all this only if RSS is compiled in, although it
658 * works fine without RSS. Need to characterise the performance overhead
659 * of the detour through the netisr code in the event the result is always
660 * direct dispatch.
661 */
662 static void
663 ether_nh_input(struct mbuf *m)
664 {
665
666 M_ASSERTPKTHDR(m);
667 KASSERT(m->m_pkthdr.rcvif != NULL,
668 ("%s: NULL interface pointer", __func__));
669 ether_input_internal(m->m_pkthdr.rcvif, m);
670 }
671
672 static struct netisr_handler ether_nh = {
673 .nh_name = "ether",
674 .nh_handler = ether_nh_input,
675 .nh_proto = NETISR_ETHER,
676 #ifdef RSS
677 .nh_policy = NETISR_POLICY_CPU,
678 .nh_dispatch = NETISR_DISPATCH_DIRECT,
679 .nh_m2cpuid = rss_m2cpuid,
680 #else
681 .nh_policy = NETISR_POLICY_SOURCE,
682 .nh_dispatch = NETISR_DISPATCH_DIRECT,
683 #endif
684 };
685
686 static void
687 ether_init(__unused void *arg)
688 {
689
690 netisr_register(ðer_nh);
691 }
692 SYSINIT(ether, SI_SUB_INIT_IF, SI_ORDER_ANY, ether_init, NULL);
693
694 static void
695 vnet_ether_init(__unused void *arg)
696 {
697 int i;
698
699 /* Initialize packet filter hooks. */
700 V_link_pfil_hook.ph_type = PFIL_TYPE_AF;
701 V_link_pfil_hook.ph_af = AF_LINK;
702 if ((i = pfil_head_register(&V_link_pfil_hook)) != 0)
703 printf("%s: WARNING: unable to register pfil link hook, "
704 "error %d\n", __func__, i);
705 #ifdef VIMAGE
706 netisr_register_vnet(ðer_nh);
707 #endif
708 }
709 VNET_SYSINIT(vnet_ether_init, SI_SUB_PROTO_IF, SI_ORDER_ANY,
710 vnet_ether_init, NULL);
711
712 #ifdef VIMAGE
713 static void
714 vnet_ether_pfil_destroy(__unused void *arg)
715 {
716 int i;
717
718 if ((i = pfil_head_unregister(&V_link_pfil_hook)) != 0)
719 printf("%s: WARNING: unable to unregister pfil link hook, "
720 "error %d\n", __func__, i);
721 }
722 VNET_SYSUNINIT(vnet_ether_pfil_uninit, SI_SUB_PROTO_PFIL, SI_ORDER_ANY,
723 vnet_ether_pfil_destroy, NULL);
724
725 static void
726 vnet_ether_destroy(__unused void *arg)
727 {
728
729 netisr_unregister_vnet(ðer_nh);
730 }
731 VNET_SYSUNINIT(vnet_ether_uninit, SI_SUB_PROTO_IF, SI_ORDER_ANY,
732 vnet_ether_destroy, NULL);
733 #endif
734
735
736
737 static void
738 ether_input(struct ifnet *ifp, struct mbuf *m)
739 {
740
741 struct mbuf *mn;
742
743 /*
744 * The drivers are allowed to pass in a chain of packets linked with
745 * m_nextpkt. We split them up into separate packets here and pass
746 * them up. This allows the drivers to amortize the receive lock.
747 */
748 while (m) {
749 mn = m->m_nextpkt;
750 m->m_nextpkt = NULL;
751
752 /*
753 * We will rely on rcvif being set properly in the deferred context,
754 * so assert it is correct here.
755 */
756 KASSERT(m->m_pkthdr.rcvif == ifp, ("%s: ifnet mismatch m %p "
757 "rcvif %p ifp %p", __func__, m, m->m_pkthdr.rcvif, ifp));
758 CURVNET_SET_QUIET(ifp->if_vnet);
759 netisr_dispatch(NETISR_ETHER, m);
760 CURVNET_RESTORE();
761 m = mn;
762 }
763 }
764
765 /*
766 * Upper layer processing for a received Ethernet packet.
767 */
768 void
769 ether_demux(struct ifnet *ifp, struct mbuf *m)
770 {
771 struct ether_header *eh;
772 int i, isr;
773 u_short ether_type;
774
775 KASSERT(ifp != NULL, ("%s: NULL interface pointer", __func__));
776
777 /* Do not grab PROMISC frames in case we are re-entered. */
778 if (PFIL_HOOKED(&V_link_pfil_hook) && !(m->m_flags & M_PROMISC)) {
779 i = pfil_run_hooks(&V_link_pfil_hook, &m, ifp, PFIL_IN, NULL);
780
781 if (i != 0 || m == NULL)
782 return;
783 }
784
785 eh = mtod(m, struct ether_header *);
786 ether_type = ntohs(eh->ether_type);
787
788 /*
789 * If this frame has a VLAN tag other than 0, call vlan_input()
790 * if its module is loaded. Otherwise, drop.
791 */
792 if ((m->m_flags & M_VLANTAG) &&
793 EVL_VLANOFTAG(m->m_pkthdr.ether_vtag) != 0) {
794 if (ifp->if_vlantrunk == NULL) {
795 if_inc_counter(ifp, IFCOUNTER_NOPROTO, 1);
796 m_freem(m);
797 return;
798 }
799 KASSERT(vlan_input_p != NULL,("%s: VLAN not loaded!",
800 __func__));
801 /* Clear before possibly re-entering ether_input(). */
802 m->m_flags &= ~M_PROMISC;
803 (*vlan_input_p)(ifp, m);
804 return;
805 }
806
807 /*
808 * Pass promiscuously received frames to the upper layer if the user
809 * requested this by setting IFF_PPROMISC. Otherwise, drop them.
810 */
811 if ((ifp->if_flags & IFF_PPROMISC) == 0 && (m->m_flags & M_PROMISC)) {
812 m_freem(m);
813 return;
814 }
815
816 /*
817 * Reset layer specific mbuf flags to avoid confusing upper layers.
818 * Strip off Ethernet header.
819 */
820 m->m_flags &= ~M_VLANTAG;
821 m_clrprotoflags(m);
822 m_adj(m, ETHER_HDR_LEN);
823
824 /*
825 * Dispatch frame to upper layer.
826 */
827 switch (ether_type) {
828 #ifdef INET
829 case ETHERTYPE_IP:
830 isr = NETISR_IP;
831 break;
832
833 case ETHERTYPE_ARP:
834 if (ifp->if_flags & IFF_NOARP) {
835 /* Discard packet if ARP is disabled on interface */
836 m_freem(m);
837 return;
838 }
839 isr = NETISR_ARP;
840 break;
841 #endif
842 #ifdef INET6
843 case ETHERTYPE_IPV6:
844 isr = NETISR_IPV6;
845 break;
846 #endif
847 default:
848 goto discard;
849 }
850 netisr_dispatch(isr, m);
851 return;
852
853 discard:
854 /*
855 * Packet is to be discarded. If netgraph is present,
856 * hand the packet to it for last chance processing;
857 * otherwise dispose of it.
858 */
859 if (ifp->if_l2com != NULL) {
860 KASSERT(ng_ether_input_orphan_p != NULL,
861 ("ng_ether_input_orphan_p is NULL"));
862 /*
863 * Put back the ethernet header so netgraph has a
864 * consistent view of inbound packets.
865 */
866 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
867 (*ng_ether_input_orphan_p)(ifp, m);
868 return;
869 }
870 m_freem(m);
871 }
872
873 /*
874 * Convert Ethernet address to printable (loggable) representation.
875 * This routine is for compatibility; it's better to just use
876 *
877 * printf("%6D", <pointer to address>, ":");
878 *
879 * since there's no static buffer involved.
880 */
881 char *
882 ether_sprintf(const u_char *ap)
883 {
884 static char etherbuf[18];
885 snprintf(etherbuf, sizeof (etherbuf), "%6D", ap, ":");
886 return (etherbuf);
887 }
888
889 /*
890 * Perform common duties while attaching to interface list
891 */
892 void
893 ether_ifattach(struct ifnet *ifp, const u_int8_t *lla)
894 {
895 int i;
896 struct ifaddr *ifa;
897 struct sockaddr_dl *sdl;
898
899 ifp->if_addrlen = ETHER_ADDR_LEN;
900 ifp->if_hdrlen = ETHER_HDR_LEN;
901 if_attach(ifp);
902 ifp->if_mtu = ETHERMTU;
903 ifp->if_output = ether_output;
904 ifp->if_input = ether_input;
905 ifp->if_resolvemulti = ether_resolvemulti;
906 ifp->if_requestencap = ether_requestencap;
907 #ifdef VIMAGE
908 ifp->if_reassign = ether_reassign;
909 #endif
910 if (ifp->if_baudrate == 0)
911 ifp->if_baudrate = IF_Mbps(10); /* just a default */
912 ifp->if_broadcastaddr = etherbroadcastaddr;
913
914 ifa = ifp->if_addr;
915 KASSERT(ifa != NULL, ("%s: no lladdr!\n", __func__));
916 sdl = (struct sockaddr_dl *)ifa->ifa_addr;
917 sdl->sdl_type = IFT_ETHER;
918 sdl->sdl_alen = ifp->if_addrlen;
919 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
920
921 bpfattach(ifp, DLT_EN10MB, ETHER_HDR_LEN);
922 if (ng_ether_attach_p != NULL)
923 (*ng_ether_attach_p)(ifp);
924
925 /* Announce Ethernet MAC address if non-zero. */
926 for (i = 0; i < ifp->if_addrlen; i++)
927 if (lla[i] != 0)
928 break;
929 if (i != ifp->if_addrlen)
930 if_printf(ifp, "Ethernet address: %6D\n", lla, ":");
931
932 uuid_ether_add(LLADDR(sdl));
933 }
934
935 /*
936 * Perform common duties while detaching an Ethernet interface
937 */
938 void
939 ether_ifdetach(struct ifnet *ifp)
940 {
941 struct sockaddr_dl *sdl;
942
943 sdl = (struct sockaddr_dl *)(ifp->if_addr->ifa_addr);
944 uuid_ether_del(LLADDR(sdl));
945
946 if (ifp->if_l2com != NULL) {
947 KASSERT(ng_ether_detach_p != NULL,
948 ("ng_ether_detach_p is NULL"));
949 (*ng_ether_detach_p)(ifp);
950 }
951
952 bpfdetach(ifp);
953 if_detach(ifp);
954 }
955
956 #ifdef VIMAGE
957 void
958 ether_reassign(struct ifnet *ifp, struct vnet *new_vnet, char *unused __unused)
959 {
960
961 if (ifp->if_l2com != NULL) {
962 KASSERT(ng_ether_detach_p != NULL,
963 ("ng_ether_detach_p is NULL"));
964 (*ng_ether_detach_p)(ifp);
965 }
966
967 if (ng_ether_attach_p != NULL) {
968 CURVNET_SET_QUIET(new_vnet);
969 (*ng_ether_attach_p)(ifp);
970 CURVNET_RESTORE();
971 }
972 }
973 #endif
974
975 SYSCTL_DECL(_net_link);
976 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
977
978 #if 0
979 /*
980 * This is for reference. We have a table-driven version
981 * of the little-endian crc32 generator, which is faster
982 * than the double-loop.
983 */
984 uint32_t
985 ether_crc32_le(const uint8_t *buf, size_t len)
986 {
987 size_t i;
988 uint32_t crc;
989 int bit;
990 uint8_t data;
991
992 crc = 0xffffffff; /* initial value */
993
994 for (i = 0; i < len; i++) {
995 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
996 carry = (crc ^ data) & 1;
997 crc >>= 1;
998 if (carry)
999 crc = (crc ^ ETHER_CRC_POLY_LE);
1000 }
1001 }
1002
1003 return (crc);
1004 }
1005 #else
1006 uint32_t
1007 ether_crc32_le(const uint8_t *buf, size_t len)
1008 {
1009 static const uint32_t crctab[] = {
1010 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
1011 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
1012 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
1013 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
1014 };
1015 size_t i;
1016 uint32_t crc;
1017
1018 crc = 0xffffffff; /* initial value */
1019
1020 for (i = 0; i < len; i++) {
1021 crc ^= buf[i];
1022 crc = (crc >> 4) ^ crctab[crc & 0xf];
1023 crc = (crc >> 4) ^ crctab[crc & 0xf];
1024 }
1025
1026 return (crc);
1027 }
1028 #endif
1029
1030 uint32_t
1031 ether_crc32_be(const uint8_t *buf, size_t len)
1032 {
1033 size_t i;
1034 uint32_t crc, carry;
1035 int bit;
1036 uint8_t data;
1037
1038 crc = 0xffffffff; /* initial value */
1039
1040 for (i = 0; i < len; i++) {
1041 for (data = *buf++, bit = 0; bit < 8; bit++, data >>= 1) {
1042 carry = ((crc & 0x80000000) ? 1 : 0) ^ (data & 0x01);
1043 crc <<= 1;
1044 if (carry)
1045 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
1046 }
1047 }
1048
1049 return (crc);
1050 }
1051
1052 int
1053 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
1054 {
1055 struct ifaddr *ifa = (struct ifaddr *) data;
1056 struct ifreq *ifr = (struct ifreq *) data;
1057 int error = 0;
1058
1059 switch (command) {
1060 case SIOCSIFADDR:
1061 ifp->if_flags |= IFF_UP;
1062
1063 switch (ifa->ifa_addr->sa_family) {
1064 #ifdef INET
1065 case AF_INET:
1066 ifp->if_init(ifp->if_softc); /* before arpwhohas */
1067 arp_ifinit(ifp, ifa);
1068 break;
1069 #endif
1070 default:
1071 ifp->if_init(ifp->if_softc);
1072 break;
1073 }
1074 break;
1075
1076 case SIOCGIFADDR:
1077 {
1078 struct sockaddr *sa;
1079
1080 sa = (struct sockaddr *) & ifr->ifr_data;
1081 bcopy(IF_LLADDR(ifp),
1082 (caddr_t) sa->sa_data, ETHER_ADDR_LEN);
1083 }
1084 break;
1085
1086 case SIOCSIFMTU:
1087 /*
1088 * Set the interface MTU.
1089 */
1090 if (ifr->ifr_mtu > ETHERMTU) {
1091 error = EINVAL;
1092 } else {
1093 ifp->if_mtu = ifr->ifr_mtu;
1094 }
1095 break;
1096 default:
1097 error = EINVAL; /* XXX netbsd has ENOTTY??? */
1098 break;
1099 }
1100 return (error);
1101 }
1102
1103 static int
1104 ether_resolvemulti(struct ifnet *ifp, struct sockaddr **llsa,
1105 struct sockaddr *sa)
1106 {
1107 struct sockaddr_dl *sdl;
1108 #ifdef INET
1109 struct sockaddr_in *sin;
1110 #endif
1111 #ifdef INET6
1112 struct sockaddr_in6 *sin6;
1113 #endif
1114 u_char *e_addr;
1115
1116 switch(sa->sa_family) {
1117 case AF_LINK:
1118 /*
1119 * No mapping needed. Just check that it's a valid MC address.
1120 */
1121 sdl = (struct sockaddr_dl *)sa;
1122 e_addr = LLADDR(sdl);
1123 if (!ETHER_IS_MULTICAST(e_addr))
1124 return EADDRNOTAVAIL;
1125 *llsa = NULL;
1126 return 0;
1127
1128 #ifdef INET
1129 case AF_INET:
1130 sin = (struct sockaddr_in *)sa;
1131 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
1132 return EADDRNOTAVAIL;
1133 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1134 sdl->sdl_alen = ETHER_ADDR_LEN;
1135 e_addr = LLADDR(sdl);
1136 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
1137 *llsa = (struct sockaddr *)sdl;
1138 return 0;
1139 #endif
1140 #ifdef INET6
1141 case AF_INET6:
1142 sin6 = (struct sockaddr_in6 *)sa;
1143 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
1144 /*
1145 * An IP6 address of 0 means listen to all
1146 * of the Ethernet multicast address used for IP6.
1147 * (This is used for multicast routers.)
1148 */
1149 ifp->if_flags |= IFF_ALLMULTI;
1150 *llsa = NULL;
1151 return 0;
1152 }
1153 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
1154 return EADDRNOTAVAIL;
1155 sdl = link_init_sdl(ifp, *llsa, IFT_ETHER);
1156 sdl->sdl_alen = ETHER_ADDR_LEN;
1157 e_addr = LLADDR(sdl);
1158 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
1159 *llsa = (struct sockaddr *)sdl;
1160 return 0;
1161 #endif
1162
1163 default:
1164 /*
1165 * Well, the text isn't quite right, but it's the name
1166 * that counts...
1167 */
1168 return EAFNOSUPPORT;
1169 }
1170 }
1171
1172 static moduledata_t ether_mod = {
1173 .name = "ether",
1174 };
1175
1176 void
1177 ether_vlan_mtap(struct bpf_if *bp, struct mbuf *m, void *data, u_int dlen)
1178 {
1179 struct ether_vlan_header vlan;
1180 struct mbuf mv, mb;
1181
1182 KASSERT((m->m_flags & M_VLANTAG) != 0,
1183 ("%s: vlan information not present", __func__));
1184 KASSERT(m->m_len >= sizeof(struct ether_header),
1185 ("%s: mbuf not large enough for header", __func__));
1186 bcopy(mtod(m, char *), &vlan, sizeof(struct ether_header));
1187 vlan.evl_proto = vlan.evl_encap_proto;
1188 vlan.evl_encap_proto = htons(ETHERTYPE_VLAN);
1189 vlan.evl_tag = htons(m->m_pkthdr.ether_vtag);
1190 m->m_len -= sizeof(struct ether_header);
1191 m->m_data += sizeof(struct ether_header);
1192 /*
1193 * If a data link has been supplied by the caller, then we will need to
1194 * re-create a stack allocated mbuf chain with the following structure:
1195 *
1196 * (1) mbuf #1 will contain the supplied data link
1197 * (2) mbuf #2 will contain the vlan header
1198 * (3) mbuf #3 will contain the original mbuf's packet data
1199 *
1200 * Otherwise, submit the packet and vlan header via bpf_mtap2().
1201 */
1202 if (data != NULL) {
1203 mv.m_next = m;
1204 mv.m_data = (caddr_t)&vlan;
1205 mv.m_len = sizeof(vlan);
1206 mb.m_next = &mv;
1207 mb.m_data = data;
1208 mb.m_len = dlen;
1209 bpf_mtap(bp, &mb);
1210 } else
1211 bpf_mtap2(bp, &vlan, sizeof(vlan), m);
1212 m->m_len += sizeof(struct ether_header);
1213 m->m_data -= sizeof(struct ether_header);
1214 }
1215
1216 struct mbuf *
1217 ether_vlanencap(struct mbuf *m, uint16_t tag)
1218 {
1219 struct ether_vlan_header *evl;
1220
1221 M_PREPEND(m, ETHER_VLAN_ENCAP_LEN, M_NOWAIT);
1222 if (m == NULL)
1223 return (NULL);
1224 /* M_PREPEND takes care of m_len, m_pkthdr.len for us */
1225
1226 if (m->m_len < sizeof(*evl)) {
1227 m = m_pullup(m, sizeof(*evl));
1228 if (m == NULL)
1229 return (NULL);
1230 }
1231
1232 /*
1233 * Transform the Ethernet header into an Ethernet header
1234 * with 802.1Q encapsulation.
1235 */
1236 evl = mtod(m, struct ether_vlan_header *);
1237 bcopy((char *)evl + ETHER_VLAN_ENCAP_LEN,
1238 (char *)evl, ETHER_HDR_LEN - ETHER_TYPE_LEN);
1239 evl->evl_encap_proto = htons(ETHERTYPE_VLAN);
1240 evl->evl_tag = htons(tag);
1241 return (m);
1242 }
1243
1244 DECLARE_MODULE(ether, ether_mod, SI_SUB_INIT_IF, SI_ORDER_ANY);
1245 MODULE_VERSION(ether, 1);
Cache object: 6fa796bb5d0e613a8fc657a67b78a88f
|