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