2 * Copyright (c) 1982, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
34 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
35 * $DragonFly: src/sys/net/if_ethersubr.c,v 1.96 2008/11/22 04:00:53 sephe Exp $
38 #include "opt_atalk.h"
40 #include "opt_inet6.h"
43 #include "opt_netgraph.h"
46 #include <sys/param.h>
47 #include <sys/systm.h>
48 #include <sys/globaldata.h>
49 #include <sys/kernel.h>
52 #include <sys/malloc.h>
54 #include <sys/msgport.h>
55 #include <sys/socket.h>
56 #include <sys/sockio.h>
57 #include <sys/sysctl.h>
58 #include <sys/thread.h>
59 #include <sys/thread2.h>
62 #include <net/netisr.h>
63 #include <net/route.h>
64 #include <net/if_llc.h>
65 #include <net/if_dl.h>
66 #include <net/if_types.h>
67 #include <net/ifq_var.h>
69 #include <net/ethernet.h>
70 #include <net/vlan/if_vlan_ether.h>
71 #include <net/netmsg2.h>
73 #if defined(INET) || defined(INET6)
74 #include <netinet/in.h>
75 #include <netinet/ip_var.h>
76 #include <netinet/if_ether.h>
77 #include <netinet/ip_flow.h>
78 #include <net/ipfw/ip_fw.h>
79 #include <net/dummynet/ip_dummynet.h>
82 #include <netinet6/nd6.h>
86 #include <netinet/ip_carp.h>
90 #include <netproto/ipx/ipx.h>
91 #include <netproto/ipx/ipx_if.h>
92 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
93 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
94 short *tp, int *hlen);
99 #include <netns/ns_if.h>
101 int ether_outputdebug = 0;
102 int ether_inputdebug = 0;
106 #include <netproto/atalk/at.h>
107 #include <netproto/atalk/at_var.h>
108 #include <netproto/atalk/at_extern.h>
110 #define llc_snap_org_code llc_un.type_snap.org_code
111 #define llc_snap_ether_type llc_un.type_snap.ether_type
113 extern u_char at_org_code[3];
114 extern u_char aarp_org_code[3];
115 #endif /* NETATALK */
118 #include <netproto/mpls/mpls.h>
121 /* netgraph node hooks for ng_ether(4) */
122 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
123 void (*ng_ether_input_orphan_p)(struct ifnet *ifp,
124 struct mbuf *m, const struct ether_header *eh);
125 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
126 void (*ng_ether_attach_p)(struct ifnet *ifp);
127 void (*ng_ether_detach_p)(struct ifnet *ifp);
129 void (*vlan_input_p)(struct mbuf *);
131 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
133 static void ether_restore_header(struct mbuf **, const struct ether_header *,
134 const struct ether_header *);
139 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
140 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
141 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
143 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
146 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
147 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
150 #define gotoerr(e) do { error = (e); goto bad; } while (0)
151 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
153 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
155 const struct ether_header *eh);
157 static int ether_ipfw;
158 static u_int ether_restore_hdr;
159 static u_int ether_prepend_hdr;
161 SYSCTL_DECL(_net_link);
162 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
163 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
164 ðer_ipfw, 0, "Pass ether pkts through firewall");
165 SYSCTL_UINT(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
166 ðer_restore_hdr, 0, "# of ether header restoration");
167 SYSCTL_UINT(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
168 ðer_prepend_hdr, 0,
169 "# of ether header restoration which prepends mbuf");
171 #define ETHER_KTR_STR "ifp=%p"
172 #define ETHER_KTR_ARG_SIZE (sizeof(void *))
174 #define KTR_ETHERNET KTR_ALL
176 KTR_INFO_MASTER(ether);
177 KTR_INFO(KTR_ETHERNET, ether, chain_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
178 KTR_INFO(KTR_ETHERNET, ether, chain_end, 1, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
179 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
180 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARG_SIZE);
181 #define logether(name, arg) KTR_LOG(ether_ ## name, arg)
184 * Ethernet output routine.
185 * Encapsulate a packet of type family for the local net.
186 * Use trailer local net encapsulation if enough data in first
187 * packet leaves a multiple of 512 bytes of data in remainder.
188 * Assumes that ifp is actually pointer to arpcom structure.
191 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
194 struct ether_header *eh, *deh;
197 int hlen = ETHER_HDR_LEN; /* link layer header length */
198 struct arpcom *ac = IFP2AC(ifp);
201 ASSERT_NOT_SERIALIZED(ifp->if_serializer);
203 if (ifp->if_flags & IFF_MONITOR)
205 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
208 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
211 eh = mtod(m, struct ether_header *);
212 edst = eh->ether_dhost;
215 * Fill in the destination ethernet address and frame type.
217 switch (dst->sa_family) {
220 if (!arpresolve(ifp, rt, m, dst, edst))
221 return (0); /* if not yet resolved */
223 if (m->m_flags & M_MPLSLABELED)
224 eh->ether_type = htons(ETHERTYPE_MPLS);
227 eh->ether_type = htons(ETHERTYPE_IP);
232 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
233 return (0); /* Something bad happenned. */
234 eh->ether_type = htons(ETHERTYPE_IPV6);
239 if (ef_outputp != NULL) {
241 * Hold BGL and recheck ef_outputp
244 if (ef_outputp != NULL) {
245 error = ef_outputp(ifp, &m, dst,
246 &eh->ether_type, &hlen);
255 eh->ether_type = htons(ETHERTYPE_IPX);
256 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
257 edst, ETHER_ADDR_LEN);
262 struct at_ifaddr *aa;
269 if ((aa = at_ifawithnet((struct sockaddr_at *)dst)) == NULL) {
275 * In the phase 2 case, need to prepend an mbuf for
276 * the llc header. Since we must preserve the value
277 * of m, which is passed to us by value, we m_copy()
278 * the first mbuf, and use it for our llc header.
280 if (aa->aa_flags & AFA_PHASE2) {
283 M_PREPEND(m, sizeof(struct llc), MB_DONTWAIT);
284 eh = mtod(m, struct ether_header *);
285 edst = eh->ether_dhost;
286 llc.llc_dsap = llc.llc_ssap = LLC_SNAP_LSAP;
287 llc.llc_control = LLC_UI;
288 bcopy(at_org_code, llc.llc_snap_org_code,
290 llc.llc_snap_ether_type = htons(ETHERTYPE_AT);
292 mtod(m, caddr_t) + sizeof(struct ether_header),
294 eh->ether_type = htons(m->m_pkthdr.len);
295 hlen = sizeof(struct llc) + ETHER_HDR_LEN;
297 eh->ether_type = htons(ETHERTYPE_AT);
299 if (!aarpresolve(ac, m, (struct sockaddr_at *)dst, edst)) {
312 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
313 eh->ether_type = 0x8137;
315 case 0x0: /* Novell 802.3 */
316 eh->ether_type = htons(m->m_pkthdr.len);
318 case 0xe0e0: /* Novell 802.2 and Token-Ring */
319 M_PREPEND(m, 3, MB_DONTWAIT);
320 eh = mtod(m, struct ether_header *);
321 edst = eh->ether_dhost;
322 eh->ether_type = htons(m->m_pkthdr.len);
323 cp = mtod(m, u_char *) + sizeof(struct ether_header);
329 bcopy(&(((struct sockaddr_ns *)dst)->sns_addr.x_host), edst,
332 * XXX if ns_thishost is the same as the node's ethernet
333 * address then just the default code will catch this anyhow.
334 * So I'm not sure if this next clause should be here at all?
337 if (bcmp(edst, &ns_thishost, ETHER_ADDR_LEN) == 0) {
338 m->m_pkthdr.rcvif = ifp;
339 netisr_dispatch(NETISR_NS, m);
342 if (bcmp(edst, &ns_broadhost, ETHER_ADDR_LEN) == 0)
343 m->m_flags |= M_BCAST;
346 case pseudo_AF_HDRCMPLT:
348 loop_copy = -1; /* if this is for us, don't do it */
349 deh = (struct ether_header *)dst->sa_data;
350 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
351 eh->ether_type = deh->ether_type;
355 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
356 gotoerr(EAFNOSUPPORT);
359 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
360 memcpy(eh->ether_shost,
361 ((struct ether_header *)dst->sa_data)->ether_shost,
364 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
367 * Bridges require special output handling.
369 if (ifp->if_bridge) {
370 KASSERT(bridge_output_p != NULL,
371 ("%s: if_bridge not loaded!", __func__));
372 return bridge_output_p(ifp, m);
376 * If a simplex interface, and the packet is being sent to our
377 * Ethernet address or a broadcast address, loopback a copy.
378 * XXX To make a simplex device behave exactly like a duplex
379 * device, we should copy in the case of sending to our own
380 * ethernet address (thus letting the original actually appear
381 * on the wire). However, we don't do that here for security
382 * reasons and compatibility with the original behavior.
384 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
387 if (m->m_pkthdr.csum_flags & CSUM_IP)
388 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
389 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
390 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
391 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
394 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
395 n->m_pkthdr.csum_flags |= csum_flags;
396 if (csum_flags & CSUM_DATA_VALID)
397 n->m_pkthdr.csum_data = 0xffff;
398 if_simloop(ifp, n, dst->sa_family, hlen);
401 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
402 ETHER_ADDR_LEN) == 0) {
403 m->m_pkthdr.csum_flags |= csum_flags;
404 if (csum_flags & CSUM_DATA_VALID)
405 m->m_pkthdr.csum_data = 0xffff;
406 if_simloop(ifp, m, dst->sa_family, hlen);
407 return (0); /* XXX */
414 * Hold BGL and recheck ifp->if_carp
417 if (ifp->if_carp && (error = carp_output(ifp, m, dst, NULL))) {
426 /* Handle ng_ether(4) processing, if any */
427 if (ng_ether_output_p != NULL) {
429 * Hold BGL and recheck ng_ether_output_p
432 if (ng_ether_output_p != NULL) {
433 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
445 /* Continue with link-layer output */
446 return ether_output_frame(ifp, m);
454 * Ethernet link layer output routine to send a raw frame to the device.
456 * This assumes that the 14 byte Ethernet header is present and contiguous
460 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
462 struct ip_fw *rule = NULL;
464 struct altq_pktattr pktattr;
466 ASSERT_NOT_SERIALIZED(ifp->if_serializer);
468 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
471 /* Extract info from dummynet tag */
472 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
473 KKASSERT(mtag != NULL);
474 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
475 KKASSERT(rule != NULL);
477 m_tag_delete(m, mtag);
478 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
481 if (ifq_is_enabled(&ifp->if_snd))
482 altq_etherclassify(&ifp->if_snd, m, &pktattr);
484 if (IPFW_LOADED && ether_ipfw != 0) {
485 struct ether_header save_eh, *eh;
487 eh = mtod(m, struct ether_header *);
489 m_adj(m, ETHER_HDR_LEN);
490 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
494 return ENOBUFS; /* pkt dropped */
496 return 0; /* consumed e.g. in a pipe */
499 /* packet was ok, restore the ethernet header */
500 ether_restore_header(&m, eh, &save_eh);
509 * Queue message on interface, update output statistics if
510 * successful, and start output if interface not yet active.
512 error = ifq_dispatch(ifp, m, &pktattr);
517 * ipfw processing for ethernet packets (in and out).
518 * The second parameter is NULL from ether_demux(), and ifp from
519 * ether_output_frame().
522 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
523 const struct ether_header *eh)
525 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
526 struct ip_fw_args args;
531 if (*rule != NULL && fw_one_pass)
532 return TRUE; /* dummynet packet, already partially processed */
535 * I need some amount of data to be contiguous.
537 i = min((*m0)->m_pkthdr.len, max_protohdr);
538 if ((*m0)->m_len < i) {
539 *m0 = m_pullup(*m0, i);
547 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
548 m_tag_delete(*m0, mtag);
549 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
550 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
551 KKASSERT(mtag != NULL);
552 m_tag_delete(*m0, mtag);
553 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
556 args.m = *m0; /* the packet we are looking at */
557 args.oif = dst; /* destination, if any */
558 args.rule = *rule; /* matching rule to restart */
559 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
560 i = ip_fw_chk_ptr(&args);
575 * XXX at some point add support for divert/forward actions.
576 * If none of the above matches, we have to drop the pkt.
582 * Pass the pkt to dummynet, which consumes it.
584 m = *m0; /* pass the original to dummynet */
585 *m0 = NULL; /* and nothing back to the caller */
587 ether_restore_header(&m, eh, &save_eh);
591 ip_fw_dn_io_ptr(m, args.cookie,
592 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
597 panic("unknown ipfw return value: %d\n", i);
602 ether_input(struct ifnet *ifp, struct mbuf *m)
604 ether_input_chain(ifp, m, NULL, NULL);
608 * Perform common duties while attaching to interface list
611 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
613 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
618 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
619 lwkt_serialize_t serializer)
621 struct sockaddr_dl *sdl;
623 ifp->if_type = IFT_ETHER;
624 ifp->if_addrlen = ETHER_ADDR_LEN;
625 ifp->if_hdrlen = ETHER_HDR_LEN;
626 if_attach(ifp, serializer);
627 ifp->if_mtu = ETHERMTU;
628 if (ifp->if_baudrate == 0)
629 ifp->if_baudrate = 10000000;
630 ifp->if_output = ether_output;
631 ifp->if_input = ether_input;
632 ifp->if_resolvemulti = ether_resolvemulti;
633 ifp->if_broadcastaddr = etherbroadcastaddr;
634 sdl = IF_LLSOCKADDR(ifp);
635 sdl->sdl_type = IFT_ETHER;
636 sdl->sdl_alen = ifp->if_addrlen;
637 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
639 * XXX Keep the current drivers happy.
640 * XXX Remove once all drivers have been cleaned up
642 if (lla != IFP2AC(ifp)->ac_enaddr)
643 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
644 bpfattach(ifp, dlt, hdrlen);
645 if (ng_ether_attach_p != NULL)
646 (*ng_ether_attach_p)(ifp);
648 if_printf(ifp, "MAC address: %6D\n", lla, ":");
652 * Perform common duties while detaching an Ethernet interface
655 ether_ifdetach(struct ifnet *ifp)
659 if (ng_ether_detach_p != NULL)
660 (*ng_ether_detach_p)(ifp);
666 ether_ioctl(struct ifnet *ifp, int command, caddr_t data)
668 struct ifaddr *ifa = (struct ifaddr *) data;
669 struct ifreq *ifr = (struct ifreq *) data;
672 #define IF_INIT(ifp) \
674 if (((ifp)->if_flags & IFF_UP) == 0) { \
675 (ifp)->if_flags |= IFF_UP; \
676 (ifp)->if_init((ifp)->if_softc); \
680 ASSERT_SERIALIZED(ifp->if_serializer);
684 switch (ifa->ifa_addr->sa_family) {
687 IF_INIT(ifp); /* before arpwhohas */
688 arp_ifinit(ifp, ifa);
693 * XXX - This code is probably wrong
697 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
698 struct arpcom *ac = IFP2AC(ifp);
700 if (ipx_nullhost(*ina))
701 ina->x_host = *(union ipx_host *) ac->ac_enaddr;
703 bcopy(ina->x_host.c_host, ac->ac_enaddr,
704 sizeof ac->ac_enaddr);
706 IF_INIT(ifp); /* Set new address. */
712 * XXX - This code is probably wrong
716 struct ns_addr *ina = &(IA_SNS(ifa)->sns_addr);
717 struct arpcom *ac = IFP2AC(ifp);
719 if (ns_nullhost(*ina))
720 ina->x_host = *(union ns_host *)(ac->ac_enaddr);
722 bcopy(ina->x_host.c_host, ac->ac_enaddr,
723 sizeof ac->ac_enaddr);
739 bcopy(IFP2AC(ifp)->ac_enaddr,
740 ((struct sockaddr *)ifr->ifr_data)->sa_data,
746 * Set the interface MTU.
748 if (ifr->ifr_mtu > ETHERMTU) {
751 ifp->if_mtu = ifr->ifr_mtu;
766 struct sockaddr **llsa,
769 struct sockaddr_dl *sdl;
770 struct sockaddr_in *sin;
772 struct sockaddr_in6 *sin6;
776 switch(sa->sa_family) {
779 * No mapping needed. Just check that it's a valid MC address.
781 sdl = (struct sockaddr_dl *)sa;
782 e_addr = LLADDR(sdl);
783 if ((e_addr[0] & 1) != 1)
784 return EADDRNOTAVAIL;
790 sin = (struct sockaddr_in *)sa;
791 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
792 return EADDRNOTAVAIL;
793 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
795 sdl->sdl_len = sizeof *sdl;
796 sdl->sdl_family = AF_LINK;
797 sdl->sdl_index = ifp->if_index;
798 sdl->sdl_type = IFT_ETHER;
799 sdl->sdl_alen = ETHER_ADDR_LEN;
800 e_addr = LLADDR(sdl);
801 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
802 *llsa = (struct sockaddr *)sdl;
807 sin6 = (struct sockaddr_in6 *)sa;
808 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
810 * An IP6 address of 0 means listen to all
811 * of the Ethernet multicast address used for IP6.
812 * (This is used for multicast routers.)
814 ifp->if_flags |= IFF_ALLMULTI;
818 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
819 return EADDRNOTAVAIL;
820 MALLOC(sdl, struct sockaddr_dl *, sizeof *sdl, M_IFMADDR,
822 sdl->sdl_len = sizeof *sdl;
823 sdl->sdl_family = AF_LINK;
824 sdl->sdl_index = ifp->if_index;
825 sdl->sdl_type = IFT_ETHER;
826 sdl->sdl_alen = ETHER_ADDR_LEN;
827 e_addr = LLADDR(sdl);
828 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
829 *llsa = (struct sockaddr *)sdl;
835 * Well, the text isn't quite right, but it's the name
844 * This is for reference. We have a table-driven version
845 * of the little-endian crc32 generator, which is faster
846 * than the double-loop.
849 ether_crc32_le(const uint8_t *buf, size_t len)
851 uint32_t c, crc, carry;
854 crc = 0xffffffffU; /* initial value */
856 for (i = 0; i < len; i++) {
858 for (j = 0; j < 8; j++) {
859 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
863 crc = (crc ^ ETHER_CRC_POLY_LE);
871 ether_crc32_le(const uint8_t *buf, size_t len)
873 static const uint32_t crctab[] = {
874 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
875 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
876 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
877 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
882 crc = 0xffffffffU; /* initial value */
884 for (i = 0; i < len; i++) {
886 crc = (crc >> 4) ^ crctab[crc & 0xf];
887 crc = (crc >> 4) ^ crctab[crc & 0xf];
895 ether_crc32_be(const uint8_t *buf, size_t len)
897 uint32_t c, crc, carry;
900 crc = 0xffffffffU; /* initial value */
902 for (i = 0; i < len; i++) {
904 for (j = 0; j < 8; j++) {
905 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
909 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
917 * find the size of ethernet header, and call classifier
920 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
921 struct altq_pktattr *pktattr)
923 struct ether_header *eh;
925 int hlen, af, hdrsize;
928 hlen = sizeof(struct ether_header);
929 eh = mtod(m, struct ether_header *);
931 ether_type = ntohs(eh->ether_type);
932 if (ether_type < ETHERMTU) {
934 struct llc *llc = (struct llc *)(eh + 1);
937 if (m->m_len < hlen ||
938 llc->llc_dsap != LLC_SNAP_LSAP ||
939 llc->llc_ssap != LLC_SNAP_LSAP ||
940 llc->llc_control != LLC_UI)
941 goto bad; /* not snap! */
943 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
946 if (ether_type == ETHERTYPE_IP) {
948 hdrsize = 20; /* sizeof(struct ip) */
950 } else if (ether_type == ETHERTYPE_IPV6) {
952 hdrsize = 40; /* sizeof(struct ip6_hdr) */
957 while (m->m_len <= hlen) {
961 hdr = m->m_data + hlen;
962 if (m->m_len < hlen + hdrsize) {
964 * ip header is not in a single mbuf. this should not
965 * happen in the current code.
966 * (todo: use m_pulldown in the future)
972 ifq_classify(ifq, m, af, pktattr);
979 pktattr->pattr_class = NULL;
980 pktattr->pattr_hdr = NULL;
981 pktattr->pattr_af = AF_UNSPEC;
985 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
986 const struct ether_header *save_eh)
988 struct mbuf *m = *m0;
993 * Prepend the header, optimize for the common case of
994 * eh pointing into the mbuf.
996 if ((const void *)(eh + 1) == (void *)m->m_data) {
997 m->m_data -= ETHER_HDR_LEN;
998 m->m_len += ETHER_HDR_LEN;
999 m->m_pkthdr.len += ETHER_HDR_LEN;
1001 ether_prepend_hdr++;
1003 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1005 bcopy(save_eh, mtod(m, struct ether_header *),
1013 ether_input_ipifunc(void *arg)
1015 struct mbuf *m, *next;
1020 next = m->m_nextpkt;
1021 m->m_nextpkt = NULL;
1023 port = m->m_pkthdr.header;
1024 m->m_pkthdr.header = NULL;
1027 &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1030 } while (m != NULL);
1034 ether_input_dispatch(struct mbuf_chain *chain)
1039 logether(disp_beg, NULL);
1040 for (i = 0; i < ncpus; ++i) {
1041 if (chain[i].mc_head != NULL) {
1042 lwkt_send_ipiq(globaldata_find(i),
1043 ether_input_ipifunc, chain[i].mc_head);
1047 logether(disp_beg, NULL);
1048 if (chain->mc_head != NULL)
1049 ether_input_ipifunc(chain->mc_head);
1051 logether(disp_end, NULL);
1055 ether_input_chain_init(struct mbuf_chain *chain)
1060 for (i = 0; i < ncpus; ++i)
1061 chain[i].mc_head = chain[i].mc_tail = NULL;
1063 chain->mc_head = chain->mc_tail = NULL;
1068 * Upper layer processing for a received Ethernet packet.
1071 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
1073 struct ether_header *eh;
1074 int isr, redispatch;
1076 struct ip_fw *rule = NULL;
1082 KASSERT(m->m_len >= ETHER_HDR_LEN,
1083 ("ether header is no contiguous!\n"));
1085 eh = mtod(m, struct ether_header *);
1087 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
1090 /* Extract info from dummynet tag */
1091 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
1092 KKASSERT(mtag != NULL);
1093 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
1094 KKASSERT(rule != NULL);
1096 m_tag_delete(m, mtag);
1097 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
1099 /* packet is passing the second time */
1105 * XXX: Okay, we need to call carp_forus() and - if it is for
1106 * us jump over code that does the normal check
1107 * "ac_enaddr == ether_dhost". The check sequence is a bit
1108 * different from OpenBSD, so we jump over as few code as
1109 * possible, to catch _all_ sanity checks. This needs
1110 * evaluation, to see if the carp ether_dhost values break any
1115 * Hold BGL and recheck ifp->if_carp
1118 if (ifp->if_carp && carp_forus(ifp->if_carp, eh->ether_dhost)) {
1127 * Discard packet if upper layers shouldn't see it because
1128 * it was unicast to a different Ethernet address. If the
1129 * driver is working properly, then this situation can only
1130 * happen when the interface is in promiscuous mode.
1132 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1133 (eh->ether_dhost[0] & 1) == 0 &&
1134 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1140 if (IPFW_LOADED && ether_ipfw != 0) {
1141 struct ether_header save_eh = *eh;
1143 /* XXX old crufty stuff, needs to be removed */
1144 m_adj(m, sizeof(struct ether_header));
1146 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1151 ether_restore_header(&m, eh, &save_eh);
1154 eh = mtod(m, struct ether_header *);
1157 ether_type = ntohs(eh->ether_type);
1158 KKASSERT(ether_type != ETHERTYPE_VLAN);
1160 if (m->m_flags & M_VLANTAG) {
1161 void (*vlan_input_func)(struct mbuf *);
1163 vlan_input_func = vlan_input_p;
1164 if (vlan_input_func != NULL) {
1167 m->m_pkthdr.rcvif->if_noproto++;
1174 * Clear protocol specific flags,
1175 * before entering the upper layer.
1177 m->m_flags &= ~M_ETHER_FLAGS;
1179 /* Strip ethernet header. */
1180 m_adj(m, sizeof(struct ether_header));
1183 * By default, we don't need to do the redispatch; for the
1184 * most common packet types, e.g. IPv4, ether_input_chain()
1185 * has already picked up the correct target network msgport.
1189 switch (ether_type) {
1192 if ((m->m_flags & M_LENCHECKED) == 0) {
1193 if (!ip_lengthcheck(&m))
1196 if (ipflow_fastforward(m))
1202 if (ifp->if_flags & IFF_NOARP) {
1203 /* Discard packet if ARP is disabled on interface */
1212 case ETHERTYPE_IPV6:
1221 * Hold BGL and recheck ef_inputp
1224 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1235 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1243 isr = NETISR_ATALK1;
1245 case ETHERTYPE_AARP:
1251 case ETHERTYPE_MPLS:
1252 case ETHERTYPE_MPLS_MCAST:
1253 /* Should have been set by ether_input_chain(). */
1254 KKASSERT(m->m_flags & M_MPLSLABELED);
1261 * The accurate msgport is not determined before
1262 * we reach here, so redo the dispatching
1268 * Hold BGL and recheck ef_inputp
1271 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1279 checksum = mtod(m, ushort *);
1281 if ((ether_type <= ETHERMTU) &&
1282 ((*checksum == 0xffff) || (*checksum == 0xE0E0))) {
1283 if (*checksum == 0xE0E0) {
1284 m->m_pkthdr.len -= 3;
1293 if (ether_type > ETHERMTU)
1295 l = mtod(m, struct llc *);
1296 if (l->llc_dsap == LLC_SNAP_LSAP &&
1297 l->llc_ssap == LLC_SNAP_LSAP &&
1298 l->llc_control == LLC_UI) {
1299 if (bcmp(&(l->llc_snap_org_code)[0], at_org_code,
1300 sizeof at_org_code) == 0 &&
1301 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AT) {
1302 m_adj(m, sizeof(struct llc));
1303 isr = NETISR_ATALK2;
1306 if (bcmp(&(l->llc_snap_org_code)[0], aarp_org_code,
1307 sizeof aarp_org_code) == 0 &&
1308 ntohs(l->llc_snap_ether_type) == ETHERTYPE_AARP) {
1309 m_adj(m, sizeof(struct llc));
1316 if (ng_ether_input_orphan_p != NULL) {
1318 * Hold BGL and recheck ng_ether_input_orphan_p
1321 if (ng_ether_input_orphan_p != NULL) {
1322 ng_ether_input_orphan_p(ifp, m, eh);
1335 netisr_dispatch(isr, m);
1339 * First we perform any link layer operations, then continue to the
1340 * upper layers with ether_demux_oncpu().
1343 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1345 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1347 * Receiving interface's flags are changed, when this
1348 * packet is waiting for processing; discard it.
1355 * Tap the packet off here for a bridge. bridge_input()
1356 * will return NULL if it has consumed the packet, otherwise
1357 * it gets processed as normal. Note that bridge_input()
1358 * will always return the original packet if we need to
1359 * process it locally.
1361 if (ifp->if_bridge) {
1362 KASSERT(bridge_input_p != NULL,
1363 ("%s: if_bridge not loaded!", __func__));
1365 if(m->m_flags & M_ETHER_BRIDGED) {
1366 m->m_flags &= ~M_ETHER_BRIDGED;
1368 /* clear M_PROMISC, in case the packets comes from a vlan */
1369 /* m->m_flags &= ~M_PROMISC; */
1370 m = bridge_input_p(ifp, m);
1374 KASSERT(ifp == m->m_pkthdr.rcvif,
1375 ("bridge_input_p changed rcvif\n"));
1379 /* Handle ng_ether(4) processing, if any */
1380 if (ng_ether_input_p != NULL) {
1382 * Hold BGL and recheck ng_ether_input_p
1385 if (ng_ether_input_p != NULL)
1386 ng_ether_input_p(ifp, &m);
1393 /* Continue with upper layer processing */
1394 ether_demux_oncpu(ifp, m);
1398 * Perform certain functions of ether_input_chain():
1400 * - Update statistics
1401 * - Run bpf(4) tap if requested
1402 * Then pass the packet to ether_input_oncpu().
1404 * This function should be used by pseudo interface (e.g. vlan(4)),
1405 * when it tries to claim that the packet is received by it.
1408 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int run_bpf)
1410 /* Discard packet if interface is not up */
1411 if (!(ifp->if_flags & IFF_UP)) {
1416 /* Change receiving interface */
1417 m->m_pkthdr.rcvif = ifp;
1419 /* Update statistics */
1421 ifp->if_ibytes += m->m_pkthdr.len;
1422 if (m->m_flags & (M_MCAST | M_BCAST))
1428 ether_input_oncpu(ifp, m);
1431 static __inline boolean_t
1432 ether_vlancheck(struct mbuf **m0)
1434 struct mbuf *m = *m0;
1435 struct ether_header *eh;
1436 uint16_t ether_type;
1438 eh = mtod(m, struct ether_header *);
1439 ether_type = ntohs(eh->ether_type);
1441 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1443 * Extract vlan tag if hardware does not do it for us
1445 vlan_ether_decap(&m);
1449 eh = mtod(m, struct ether_header *);
1450 ether_type = ntohs(eh->ether_type);
1453 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1455 * To prevent possible dangerous recursion,
1456 * we don't do vlan-in-vlan
1458 m->m_pkthdr.rcvif->if_noproto++;
1461 KKASSERT(ether_type != ETHERTYPE_VLAN);
1463 m->m_flags |= M_ETHER_VLANCHECKED;
1474 ether_input_handler(struct netmsg *nmsg)
1476 struct netmsg_packet *nmp = (struct netmsg_packet *)nmsg;
1477 struct ether_header *eh;
1483 ifp = m->m_pkthdr.rcvif;
1485 eh = mtod(m, struct ether_header *);
1486 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1487 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1488 ifp->if_addrlen) == 0)
1489 m->m_flags |= M_BCAST;
1491 m->m_flags |= M_MCAST;
1495 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1496 if (!ether_vlancheck(&m)) {
1497 KKASSERT(m == NULL);
1502 ether_input_oncpu(ifp, m);
1505 static __inline void
1506 ether_init_netpacket(int num, struct mbuf *m)
1508 struct netmsg_packet *pmsg;
1510 pmsg = &m->m_hdr.mh_netmsg;
1511 netmsg_init(&pmsg->nm_netmsg, &netisr_apanic_rport, MSGF_MPSAFE,
1512 ether_input_handler);
1513 pmsg->nm_packet = m;
1514 pmsg->nm_netmsg.nm_lmsg.u.ms_result = num;
1517 static __inline struct lwkt_port *
1518 ether_mport(int num, struct mbuf **m)
1520 if (num == NETISR_MAX) {
1522 * All packets whose target msgports can't be
1523 * determined here are dispatched to netisr0,
1524 * where further dispatching may happen.
1526 return cpu_portfn(0);
1528 return netisr_find_port(num, m);
1532 * Send the packet to the target msgport or
1533 * queue it into 'chain'.
1536 ether_dispatch(int isr, struct lwkt_port *port, struct mbuf *m,
1537 struct mbuf_chain *chain)
1539 ether_init_netpacket(isr, m);
1541 if (chain != NULL) {
1542 struct mbuf_chain *c;
1545 m->m_pkthdr.header = port; /* XXX */
1546 cpuid = port->mpu_td->td_gd->gd_cpuid;
1549 if (c->mc_head == NULL) {
1550 c->mc_head = c->mc_tail = m;
1552 c->mc_tail->m_nextpkt = m;
1555 m->m_nextpkt = NULL;
1557 lwkt_sendmsg(port, &m->m_hdr.mh_netmsg.nm_netmsg.nm_lmsg);
1562 * Process a received Ethernet packet.
1564 * The ethernet header is assumed to be in the mbuf so the caller
1565 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1566 * bytes in the first mbuf.
1568 * We first try to find the target msgport for this ether frame, if
1569 * there is no target msgport for it, this ether frame is discarded,
1570 * else we do following processing according to whether 'chain' is
1572 * - If 'chain' is NULL, this ether frame is sent to the target msgport
1573 * immediately. This situation happens when ether_input_chain is
1574 * accessed through ifnet.if_input.
1575 * - If 'chain' is not NULL, this ether frame is queued to the 'chain'
1576 * bucket indexed by the target msgport's cpuid and the target msgport
1577 * is saved in mbuf's m_pkthdr.m_head. Caller of ether_input_chain
1578 * must initialize 'chain' by calling ether_input_chain_init().
1579 * ether_input_dispatch must be called later to send ether frames
1580 * queued on 'chain' to their target msgport.
1583 ether_input_chain(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi,
1584 struct mbuf_chain *chain)
1586 struct ether_header *eh, *save_eh, save_eh0;
1587 struct lwkt_port *port;
1588 uint16_t ether_type;
1591 ASSERT_SERIALIZED(ifp->if_serializer);
1594 /* Discard packet if interface is not up */
1595 if (!(ifp->if_flags & IFF_UP)) {
1600 if (m->m_len < sizeof(struct ether_header)) {
1601 /* XXX error in the caller. */
1606 m->m_pkthdr.rcvif = ifp;
1608 logether(chain_beg, ifp);
1610 ETHER_BPF_MTAP(ifp, m);
1612 ifp->if_ibytes += m->m_pkthdr.len;
1614 if (ifp->if_flags & IFF_MONITOR) {
1615 eh = mtod(m, struct ether_header *);
1616 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1620 * Interface marked for monitoring; discard packet.
1624 logether(chain_end, ifp);
1628 if (pi != NULL && (m->m_flags & M_HASH)) {
1629 /* Try finding the port using the packet info */
1630 port = netisr_find_pktinfo_port(pi, m);
1632 ether_dispatch(pi->pi_netisr, port, m, chain);
1634 logether(chain_end, ifp);
1639 * The packet info does not contain enough
1640 * information, we will have to check the
1646 * Packet hash will be recalculated by software,
1647 * so clear the M_HASH flag set by the driver;
1648 * the hash value calculated by the hardware may
1649 * not be exactly what we want.
1651 m->m_flags &= ~M_HASH;
1653 if (!ether_vlancheck(&m)) {
1654 KKASSERT(m == NULL);
1655 logether(chain_end, ifp);
1658 eh = mtod(m, struct ether_header *);
1659 ether_type = ntohs(eh->ether_type);
1662 * Map ether type to netisr id.
1664 switch (ether_type) {
1676 case ETHERTYPE_IPV6:
1688 case 0x8137: /* Novell Ethernet_II Ethernet TYPE II */
1695 isr = NETISR_ATALK1;
1697 case ETHERTYPE_AARP:
1703 case ETHERTYPE_MPLS:
1704 case ETHERTYPE_MPLS_MCAST:
1705 m->m_flags |= M_MPLSLABELED;
1712 * NETISR_MAX is an invalid value; it is chosen to let
1713 * ether_mport() know that we are not able to decide
1714 * this packet's msgport here.
1721 * If the packet is in contiguous memory, following
1722 * m_adj() could ensure that the hidden ether header
1723 * will not be destroyed, else we will have to save
1724 * the ether header for the later restoration.
1726 if (m->m_pkthdr.len != m->m_len) {
1728 save_eh = &save_eh0;
1734 * Temporarily remove ether header; ether_mport()
1735 * expects a packet without ether header.
1737 m_adj(m, sizeof(struct ether_header));
1740 * Find the packet's target msgport.
1742 port = ether_mport(isr, &m);
1744 KKASSERT(m == NULL);
1745 logether(chain_end, ifp);
1750 * Restore ether header.
1752 if (save_eh != NULL) {
1753 ether_restore_header(&m, eh, save_eh);
1755 logether(chain_end, ifp);
1759 m->m_data -= ETHER_HDR_LEN;
1760 m->m_len += ETHER_HDR_LEN;
1761 m->m_pkthdr.len += ETHER_HDR_LEN;
1764 ether_dispatch(isr, port, m, chain);
1766 logether(chain_end, ifp);