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 $
38 #include "opt_inet6.h"
41 #include "opt_netgraph.h"
45 #include <sys/param.h>
46 #include <sys/systm.h>
47 #include <sys/globaldata.h>
48 #include <sys/kernel.h>
51 #include <sys/malloc.h>
53 #include <sys/msgport.h>
54 #include <sys/socket.h>
55 #include <sys/sockio.h>
56 #include <sys/sysctl.h>
57 #include <sys/thread.h>
59 #include <sys/thread2.h>
60 #include <sys/mplock2.h>
63 #include <net/netisr.h>
64 #include <net/route.h>
65 #include <net/if_llc.h>
66 #include <net/if_dl.h>
67 #include <net/if_types.h>
68 #include <net/ifq_var.h>
70 #include <net/ethernet.h>
71 #include <net/vlan/if_vlan_ether.h>
72 #include <net/vlan/if_vlan_var.h>
73 #include <net/netmsg2.h>
74 #include <net/netisr2.h>
76 #if defined(INET) || defined(INET6)
77 #include <netinet/in.h>
78 #include <netinet/ip_var.h>
79 #include <netinet/tcp_var.h>
80 #include <netinet/if_ether.h>
81 #include <netinet/ip_flow.h>
82 #include <net/ipfw/ip_fw.h>
83 #include <net/dummynet/ip_dummynet.h>
86 #include <netinet6/nd6.h>
90 #include <netinet/ip_carp.h>
94 #include <netproto/ipx/ipx.h>
95 #include <netproto/ipx/ipx_if.h>
96 int (*ef_inputp)(struct ifnet*, const struct ether_header *eh, struct mbuf *m);
97 int (*ef_outputp)(struct ifnet *ifp, struct mbuf **mp, struct sockaddr *dst,
98 short *tp, int *hlen);
102 #include <netproto/mpls/mpls.h>
105 /* netgraph node hooks for ng_ether(4) */
106 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
107 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
108 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
109 void (*ng_ether_attach_p)(struct ifnet *ifp);
110 void (*ng_ether_detach_p)(struct ifnet *ifp);
112 void (*vlan_input_p)(struct mbuf *);
114 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
116 static void ether_restore_header(struct mbuf **, const struct ether_header *,
117 const struct ether_header *);
118 static int ether_characterize(struct mbuf **);
119 static void ether_dispatch(int, struct mbuf *);
124 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
125 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
126 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
127 struct ifnet *(*bridge_interface_p)(void *if_bridge);
129 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
132 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
133 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
136 #define gotoerr(e) do { error = (e); goto bad; } while (0)
137 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
139 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
141 const struct ether_header *eh);
143 static int ether_ipfw;
144 static u_long ether_restore_hdr;
145 static u_long ether_prepend_hdr;
146 static u_long ether_input_wronghash;
147 static int ether_debug;
150 static u_long ether_pktinfo_try;
151 static u_long ether_pktinfo_hit;
152 static u_long ether_rss_nopi;
153 static u_long ether_rss_nohash;
154 static u_long ether_input_requeue;
156 static u_long ether_input_wronghwhash;
157 static int ether_input_ckhash;
159 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU)
161 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
162 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default);
164 SYSCTL_DECL(_net_link);
165 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
166 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
167 ðer_debug, 0, "Ether debug");
168 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
169 ðer_ipfw, 0, "Pass ether pkts through firewall");
170 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
171 ðer_restore_hdr, 0, "# of ether header restoration");
172 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
173 ðer_prepend_hdr, 0,
174 "# of ether header restoration which prepends mbuf");
175 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
176 ðer_input_wronghash, 0, "# of input packets with wrong hash");
177 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
178 ðer_tsolen_default, 0, "Default max TSO length");
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
182 ðer_rss_nopi, 0, "# of packets do not have pktinfo");
183 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
184 ðer_rss_nohash, 0, "# of packets do not have hash");
185 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
186 ðer_pktinfo_try, 0,
187 "# of tries to find packets' msgport using pktinfo");
188 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
189 ðer_pktinfo_hit, 0,
190 "# of packets whose msgport are found using pktinfo");
191 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
192 ðer_input_requeue, 0, "# of input packets gets requeued");
194 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
195 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash");
196 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
197 ðer_input_ckhash, 0, "always check hash");
199 #define ETHER_KTR_STR "ifp=%p"
200 #define ETHER_KTR_ARGS struct ifnet *ifp
202 #define KTR_ETHERNET KTR_ALL
204 KTR_INFO_MASTER(ether);
205 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
206 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
207 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
208 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
209 #define logether(name, arg) KTR_LOG(ether_ ## name, arg)
212 * Ethernet output routine.
213 * Encapsulate a packet of type family for the local net.
214 * Use trailer local net encapsulation if enough data in first
215 * packet leaves a multiple of 512 bytes of data in remainder.
216 * Assumes that ifp is actually pointer to arpcom structure.
219 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
222 struct ether_header *eh, *deh;
225 int hlen = ETHER_HDR_LEN; /* link layer header length */
226 struct arpcom *ac = IFP2AC(ifp);
229 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
231 if (ifp->if_flags & IFF_MONITOR)
233 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
236 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
239 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
240 eh = mtod(m, struct ether_header *);
241 edst = eh->ether_dhost;
244 * Fill in the destination ethernet address and frame type.
246 switch (dst->sa_family) {
249 if (!arpresolve(ifp, rt, m, dst, edst))
250 return (0); /* if not yet resolved */
252 if (m->m_flags & M_MPLSLABELED)
253 eh->ether_type = htons(ETHERTYPE_MPLS);
256 eh->ether_type = htons(ETHERTYPE_IP);
261 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
262 return (0); /* Something bad happenned. */
263 eh->ether_type = htons(ETHERTYPE_IPV6);
268 if (ef_outputp != NULL) {
270 * Hold BGL and recheck ef_outputp
273 if (ef_outputp != NULL) {
274 error = ef_outputp(ifp, &m, dst,
275 &eh->ether_type, &hlen);
284 eh->ether_type = htons(ETHERTYPE_IPX);
285 bcopy(&(((struct sockaddr_ipx *)dst)->sipx_addr.x_host),
286 edst, ETHER_ADDR_LEN);
289 case pseudo_AF_HDRCMPLT:
291 loop_copy = -1; /* if this is for us, don't do it */
292 deh = (struct ether_header *)dst->sa_data;
293 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
294 eh->ether_type = deh->ether_type;
298 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
299 gotoerr(EAFNOSUPPORT);
302 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
303 memcpy(eh->ether_shost,
304 ((struct ether_header *)dst->sa_data)->ether_shost,
307 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
310 * Bridges require special output handling.
312 if (ifp->if_bridge) {
313 KASSERT(bridge_output_p != NULL,
314 ("%s: if_bridge not loaded!", __func__));
315 return bridge_output_p(ifp, m);
319 * If a simplex interface, and the packet is being sent to our
320 * Ethernet address or a broadcast address, loopback a copy.
321 * XXX To make a simplex device behave exactly like a duplex
322 * device, we should copy in the case of sending to our own
323 * ethernet address (thus letting the original actually appear
324 * on the wire). However, we don't do that here for security
325 * reasons and compatibility with the original behavior.
327 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
330 if (m->m_pkthdr.csum_flags & CSUM_IP)
331 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
332 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
333 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
334 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
337 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
338 n->m_pkthdr.csum_flags |= csum_flags;
339 if (csum_flags & CSUM_DATA_VALID)
340 n->m_pkthdr.csum_data = 0xffff;
341 if_simloop(ifp, n, dst->sa_family, hlen);
343 IFNET_STAT_INC(ifp, iqdrops, 1);
344 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
345 ETHER_ADDR_LEN) == 0) {
346 m->m_pkthdr.csum_flags |= csum_flags;
347 if (csum_flags & CSUM_DATA_VALID)
348 m->m_pkthdr.csum_data = 0xffff;
349 if_simloop(ifp, m, dst->sa_family, hlen);
350 return (0); /* XXX */
355 if (ifp->if_type == IFT_CARP) {
356 ifp = carp_parent(ifp);
358 gotoerr(ENETUNREACH);
363 * Check precondition again
365 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
367 if (ifp->if_flags & IFF_MONITOR)
369 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
370 (IFF_UP | IFF_RUNNING))
375 /* Handle ng_ether(4) processing, if any */
376 if (ng_ether_output_p != NULL) {
378 * Hold BGL and recheck ng_ether_output_p
381 if (ng_ether_output_p != NULL) {
382 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
394 /* Continue with link-layer output */
395 return ether_output_frame(ifp, m);
403 * Returns the bridge interface an ifp is associated
406 * Only call if ifp->if_bridge != NULL.
409 ether_bridge_interface(struct ifnet *ifp)
411 if (bridge_interface_p)
412 return(bridge_interface_p(ifp->if_bridge));
417 * Ethernet link layer output routine to send a raw frame to the device.
419 * This assumes that the 14 byte Ethernet header is present and contiguous
423 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
425 struct ip_fw *rule = NULL;
427 struct altq_pktattr pktattr;
429 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
431 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
434 /* Extract info from dummynet tag */
435 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
436 KKASSERT(mtag != NULL);
437 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
438 KKASSERT(rule != NULL);
440 m_tag_delete(m, mtag);
441 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
444 if (ifq_is_enabled(&ifp->if_snd))
445 altq_etherclassify(&ifp->if_snd, m, &pktattr);
447 if (IPFW_LOADED && ether_ipfw != 0) {
448 struct ether_header save_eh, *eh;
450 eh = mtod(m, struct ether_header *);
452 m_adj(m, ETHER_HDR_LEN);
453 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
457 return ENOBUFS; /* pkt dropped */
459 return 0; /* consumed e.g. in a pipe */
462 /* packet was ok, restore the ethernet header */
463 ether_restore_header(&m, eh, &save_eh);
472 * Queue message on interface, update output statistics if
473 * successful, and start output if interface not yet active.
475 error = ifq_dispatch(ifp, m, &pktattr);
480 * ipfw processing for ethernet packets (in and out).
481 * The second parameter is NULL from ether_demux(), and ifp from
482 * ether_output_frame().
485 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
486 const struct ether_header *eh)
488 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
489 struct ip_fw_args args;
494 if (*rule != NULL && fw_one_pass)
495 return TRUE; /* dummynet packet, already partially processed */
498 * I need some amount of data to be contiguous.
500 i = min((*m0)->m_pkthdr.len, max_protohdr);
501 if ((*m0)->m_len < i) {
502 *m0 = m_pullup(*m0, i);
510 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
511 m_tag_delete(*m0, mtag);
512 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
513 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
514 KKASSERT(mtag != NULL);
515 m_tag_delete(*m0, mtag);
516 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
519 args.m = *m0; /* the packet we are looking at */
520 args.oif = dst; /* destination, if any */
521 args.rule = *rule; /* matching rule to restart */
522 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
523 i = ip_fw_chk_ptr(&args);
538 * XXX at some point add support for divert/forward actions.
539 * If none of the above matches, we have to drop the pkt.
545 * Pass the pkt to dummynet, which consumes it.
547 m = *m0; /* pass the original to dummynet */
548 *m0 = NULL; /* and nothing back to the caller */
550 ether_restore_header(&m, eh, &save_eh);
554 ip_fw_dn_io_ptr(m, args.cookie,
555 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
560 panic("unknown ipfw return value: %d", i);
565 ether_input(struct ifnet *ifp, struct mbuf *m)
567 ether_input_pkt(ifp, m, NULL);
571 * Perform common duties while attaching to interface list
574 ether_ifattach(struct ifnet *ifp, uint8_t *lla, lwkt_serialize_t serializer)
576 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
581 ether_ifattach_bpf(struct ifnet *ifp, uint8_t *lla, u_int dlt, u_int hdrlen,
582 lwkt_serialize_t serializer)
584 struct sockaddr_dl *sdl;
585 char ethstr[ETHER_ADDRSTRLEN + 1];
587 ifp->if_type = IFT_ETHER;
588 ifp->if_addrlen = ETHER_ADDR_LEN;
589 ifp->if_hdrlen = ETHER_HDR_LEN;
590 if_attach(ifp, serializer);
591 ifp->if_mtu = ETHERMTU;
592 if (ifp->if_tsolen <= 0) {
593 if ((ether_tsolen_default / ETHERMTU) < 2) {
594 kprintf("ether TSO maxlen %d -> %d\n",
595 ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
596 ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
598 ifp->if_tsolen = ether_tsolen_default;
600 if (ifp->if_baudrate == 0)
601 ifp->if_baudrate = 10000000;
602 ifp->if_output = ether_output;
603 ifp->if_input = ether_input;
604 ifp->if_resolvemulti = ether_resolvemulti;
605 ifp->if_broadcastaddr = etherbroadcastaddr;
606 sdl = IF_LLSOCKADDR(ifp);
607 sdl->sdl_type = IFT_ETHER;
608 sdl->sdl_alen = ifp->if_addrlen;
609 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
611 * XXX Keep the current drivers happy.
612 * XXX Remove once all drivers have been cleaned up
614 if (lla != IFP2AC(ifp)->ac_enaddr)
615 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
616 bpfattach(ifp, dlt, hdrlen);
617 if (ng_ether_attach_p != NULL)
618 (*ng_ether_attach_p)(ifp);
620 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
624 * Perform common duties while detaching an Ethernet interface
627 ether_ifdetach(struct ifnet *ifp)
631 if (ng_ether_detach_p != NULL)
632 (*ng_ether_detach_p)(ifp);
638 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
640 struct ifaddr *ifa = (struct ifaddr *) data;
641 struct ifreq *ifr = (struct ifreq *) data;
644 #define IF_INIT(ifp) \
646 if (((ifp)->if_flags & IFF_UP) == 0) { \
647 (ifp)->if_flags |= IFF_UP; \
648 (ifp)->if_init((ifp)->if_softc); \
652 ASSERT_IFNET_SERIALIZED_ALL(ifp);
656 switch (ifa->ifa_addr->sa_family) {
659 IF_INIT(ifp); /* before arpwhohas */
660 arp_ifinit(ifp, ifa);
665 * XXX - This code is probably wrong
669 struct ipx_addr *ina = &IA_SIPX(ifa)->sipx_addr;
670 struct arpcom *ac = IFP2AC(ifp);
672 if (ipx_nullhost(*ina))
673 ina->x_host = *(union ipx_host *) ac->ac_enaddr;
675 bcopy(ina->x_host.c_host, ac->ac_enaddr,
676 sizeof ac->ac_enaddr);
678 IF_INIT(ifp); /* Set new address. */
689 bcopy(IFP2AC(ifp)->ac_enaddr,
690 ((struct sockaddr *)ifr->ifr_data)->sa_data,
696 * Set the interface MTU.
698 if (ifr->ifr_mtu > ETHERMTU) {
701 ifp->if_mtu = ifr->ifr_mtu;
716 struct sockaddr **llsa,
719 struct sockaddr_dl *sdl;
721 struct sockaddr_in *sin;
724 struct sockaddr_in6 *sin6;
728 switch(sa->sa_family) {
731 * No mapping needed. Just check that it's a valid MC address.
733 sdl = (struct sockaddr_dl *)sa;
734 e_addr = LLADDR(sdl);
735 if ((e_addr[0] & 1) != 1)
736 return EADDRNOTAVAIL;
742 sin = (struct sockaddr_in *)sa;
743 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
744 return EADDRNOTAVAIL;
745 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
746 sdl->sdl_len = sizeof *sdl;
747 sdl->sdl_family = AF_LINK;
748 sdl->sdl_index = ifp->if_index;
749 sdl->sdl_type = IFT_ETHER;
750 sdl->sdl_alen = ETHER_ADDR_LEN;
751 e_addr = LLADDR(sdl);
752 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
753 *llsa = (struct sockaddr *)sdl;
758 sin6 = (struct sockaddr_in6 *)sa;
759 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
761 * An IP6 address of 0 means listen to all
762 * of the Ethernet multicast address used for IP6.
763 * (This is used for multicast routers.)
765 ifp->if_flags |= IFF_ALLMULTI;
769 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
770 return EADDRNOTAVAIL;
771 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
772 sdl->sdl_len = sizeof *sdl;
773 sdl->sdl_family = AF_LINK;
774 sdl->sdl_index = ifp->if_index;
775 sdl->sdl_type = IFT_ETHER;
776 sdl->sdl_alen = ETHER_ADDR_LEN;
777 e_addr = LLADDR(sdl);
778 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
779 *llsa = (struct sockaddr *)sdl;
785 * Well, the text isn't quite right, but it's the name
794 * This is for reference. We have a table-driven version
795 * of the little-endian crc32 generator, which is faster
796 * than the double-loop.
799 ether_crc32_le(const uint8_t *buf, size_t len)
801 uint32_t c, crc, carry;
804 crc = 0xffffffffU; /* initial value */
806 for (i = 0; i < len; i++) {
808 for (j = 0; j < 8; j++) {
809 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
813 crc = (crc ^ ETHER_CRC_POLY_LE);
821 ether_crc32_le(const uint8_t *buf, size_t len)
823 static const uint32_t crctab[] = {
824 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
825 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
826 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
827 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
832 crc = 0xffffffffU; /* initial value */
834 for (i = 0; i < len; i++) {
836 crc = (crc >> 4) ^ crctab[crc & 0xf];
837 crc = (crc >> 4) ^ crctab[crc & 0xf];
845 ether_crc32_be(const uint8_t *buf, size_t len)
847 uint32_t c, crc, carry;
850 crc = 0xffffffffU; /* initial value */
852 for (i = 0; i < len; i++) {
854 for (j = 0; j < 8; j++) {
855 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
859 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
867 * find the size of ethernet header, and call classifier
870 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
871 struct altq_pktattr *pktattr)
873 struct ether_header *eh;
875 int hlen, af, hdrsize;
877 hlen = sizeof(struct ether_header);
878 eh = mtod(m, struct ether_header *);
880 ether_type = ntohs(eh->ether_type);
881 if (ether_type < ETHERMTU) {
883 struct llc *llc = (struct llc *)(eh + 1);
886 if (m->m_len < hlen ||
887 llc->llc_dsap != LLC_SNAP_LSAP ||
888 llc->llc_ssap != LLC_SNAP_LSAP ||
889 llc->llc_control != LLC_UI)
890 goto bad; /* not snap! */
892 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
895 if (ether_type == ETHERTYPE_IP) {
897 hdrsize = 20; /* sizeof(struct ip) */
899 } else if (ether_type == ETHERTYPE_IPV6) {
901 hdrsize = 40; /* sizeof(struct ip6_hdr) */
906 while (m->m_len <= hlen) {
910 if (m->m_len < hlen + hdrsize) {
912 * ip header is not in a single mbuf. this should not
913 * happen in the current code.
914 * (todo: use m_pulldown in the future)
920 ifq_classify(ifq, m, af, pktattr);
927 pktattr->pattr_class = NULL;
928 pktattr->pattr_hdr = NULL;
929 pktattr->pattr_af = AF_UNSPEC;
933 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
934 const struct ether_header *save_eh)
936 struct mbuf *m = *m0;
941 * Prepend the header, optimize for the common case of
942 * eh pointing into the mbuf.
944 if ((const void *)(eh + 1) == (void *)m->m_data) {
945 m->m_data -= ETHER_HDR_LEN;
946 m->m_len += ETHER_HDR_LEN;
947 m->m_pkthdr.len += ETHER_HDR_LEN;
951 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
953 bcopy(save_eh, mtod(m, struct ether_header *),
961 * Upper layer processing for a received Ethernet packet.
964 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
966 struct ether_header *eh;
967 int isr, discard = 0;
969 struct ip_fw *rule = NULL;
972 KASSERT(m->m_len >= ETHER_HDR_LEN,
973 ("ether header is not contiguous!"));
975 eh = mtod(m, struct ether_header *);
977 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
980 /* Extract info from dummynet tag */
981 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
982 KKASSERT(mtag != NULL);
983 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
984 KKASSERT(rule != NULL);
986 m_tag_delete(m, mtag);
987 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
989 /* packet is passing the second time */
994 * We got a packet which was unicast to a different Ethernet
995 * address. If the driver is working properly, then this
996 * situation can only happen when the interface is in
997 * promiscuous mode. We defer the packet discarding until the
998 * vlan processing is done, so that vlan/bridge or vlan/netgraph
1001 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
1002 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
1003 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
1004 if (ether_debug & 1) {
1005 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
1006 "%02x:%02x:%02x:%02x:%02x:%02x "
1007 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
1021 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1022 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1023 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1024 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1025 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1026 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1029 if ((ether_debug & 2) == 0)
1034 if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
1035 struct ether_header save_eh = *eh;
1037 /* XXX old crufty stuff, needs to be removed */
1038 m_adj(m, sizeof(struct ether_header));
1040 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1045 ether_restore_header(&m, eh, &save_eh);
1048 eh = mtod(m, struct ether_header *);
1051 ether_type = ntohs(eh->ether_type);
1052 KKASSERT(ether_type != ETHERTYPE_VLAN);
1054 if (m->m_flags & M_VLANTAG) {
1055 void (*vlan_input_func)(struct mbuf *);
1057 vlan_input_func = vlan_input_p;
1058 if (vlan_input_func != NULL) {
1061 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1068 * If we have been asked to discard this packet
1069 * (e.g. not for us), drop it before entering
1078 * Clear protocol specific flags,
1079 * before entering the upper layer.
1081 m->m_flags &= ~M_ETHER_FLAGS;
1083 /* Strip ethernet header. */
1084 m_adj(m, sizeof(struct ether_header));
1086 switch (ether_type) {
1089 if ((m->m_flags & M_LENCHECKED) == 0) {
1090 if (!ip_lengthcheck(&m, 0))
1093 if (ipflow_fastforward(m))
1099 if (ifp->if_flags & IFF_NOARP) {
1100 /* Discard packet if ARP is disabled on interface */
1109 case ETHERTYPE_IPV6:
1118 * Hold BGL and recheck ef_inputp
1121 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1132 case ETHERTYPE_MPLS:
1133 case ETHERTYPE_MPLS_MCAST:
1134 /* Should have been set by ether_input_pkt(). */
1135 KKASSERT(m->m_flags & M_MPLSLABELED);
1142 * The accurate msgport is not determined before
1143 * we reach here, so recharacterize packet.
1145 m->m_flags &= ~M_HASH;
1149 * Hold BGL and recheck ef_inputp
1152 if (ef_inputp && ef_inputp(ifp, eh, m) == 0) {
1159 if (ng_ether_input_orphan_p != NULL) {
1161 * Put back the ethernet header so netgraph has a
1162 * consistent view of inbound packets.
1164 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1167 * M_PREPEND frees the mbuf in case of failure.
1172 * Hold BGL and recheck ng_ether_input_orphan_p
1175 if (ng_ether_input_orphan_p != NULL) {
1176 ng_ether_input_orphan_p(ifp, m);
1186 if (m->m_flags & M_HASH) {
1187 if (&curthread->td_msgport ==
1188 netisr_hashport(m->m_pkthdr.hash)) {
1189 netisr_handle(isr, m);
1193 * XXX Something is wrong,
1194 * we probably should panic here!
1196 m->m_flags &= ~M_HASH;
1197 atomic_add_long(ðer_input_wronghash, 1);
1201 atomic_add_long(ðer_input_requeue, 1);
1203 netisr_queue(isr, m);
1207 * First we perform any link layer operations, then continue to the
1208 * upper layers with ether_demux_oncpu().
1211 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1217 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1219 * Receiving interface's flags are changed, when this
1220 * packet is waiting for processing; discard it.
1227 * Tap the packet off here for a bridge. bridge_input()
1228 * will return NULL if it has consumed the packet, otherwise
1229 * it gets processed as normal. Note that bridge_input()
1230 * will always return the original packet if we need to
1231 * process it locally.
1233 if (ifp->if_bridge) {
1234 KASSERT(bridge_input_p != NULL,
1235 ("%s: if_bridge not loaded!", __func__));
1237 if(m->m_flags & M_ETHER_BRIDGED) {
1238 m->m_flags &= ~M_ETHER_BRIDGED;
1240 m = bridge_input_p(ifp, m);
1244 KASSERT(ifp == m->m_pkthdr.rcvif,
1245 ("bridge_input_p changed rcvif"));
1250 carp = ifp->if_carp;
1252 m = carp_input(carp, m);
1255 KASSERT(ifp == m->m_pkthdr.rcvif,
1256 ("carp_input changed rcvif"));
1260 /* Handle ng_ether(4) processing, if any */
1261 if (ng_ether_input_p != NULL) {
1263 * Hold BGL and recheck ng_ether_input_p
1266 if (ng_ether_input_p != NULL)
1267 ng_ether_input_p(ifp, &m);
1274 /* Continue with upper layer processing */
1275 ether_demux_oncpu(ifp, m);
1279 * Perform certain functions of ether_input_pkt():
1281 * - Update statistics
1282 * - Run bpf(4) tap if requested
1283 * Then pass the packet to ether_input_oncpu().
1285 * This function should be used by pseudo interface (e.g. vlan(4)),
1286 * when it tries to claim that the packet is received by it.
1292 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1294 /* Discard packet if interface is not up */
1295 if (!(ifp->if_flags & IFF_UP)) {
1301 * Change receiving interface. The bridge will often pass a flag to
1302 * ask that this not be done so ARPs get applied to the correct
1305 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1306 m->m_pkthdr.rcvif == NULL) {
1307 m->m_pkthdr.rcvif = ifp;
1310 /* Update statistics */
1311 IFNET_STAT_INC(ifp, ipackets, 1);
1312 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1313 if (m->m_flags & (M_MCAST | M_BCAST))
1314 IFNET_STAT_INC(ifp, imcasts, 1);
1316 if (reinput_flags & REINPUT_RUNBPF)
1319 ether_input_oncpu(ifp, m);
1322 static __inline boolean_t
1323 ether_vlancheck(struct mbuf **m0)
1325 struct mbuf *m = *m0;
1326 struct ether_header *eh;
1327 uint16_t ether_type;
1329 eh = mtod(m, struct ether_header *);
1330 ether_type = ntohs(eh->ether_type);
1332 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1334 * Extract vlan tag if hardware does not do it for us
1336 vlan_ether_decap(&m);
1340 eh = mtod(m, struct ether_header *);
1341 ether_type = ntohs(eh->ether_type);
1344 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1346 * To prevent possible dangerous recursion,
1347 * we don't do vlan-in-vlan
1349 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1352 KKASSERT(ether_type != ETHERTYPE_VLAN);
1354 m->m_flags |= M_ETHER_VLANCHECKED;
1365 ether_input_handler(netmsg_t nmsg)
1367 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1368 struct ether_header *eh;
1375 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1376 if (!ether_vlancheck(&m)) {
1377 KKASSERT(m == NULL);
1381 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1382 __predict_false(ether_input_ckhash)) {
1386 * Need to verify the hash supplied by the hardware
1387 * which could be wrong.
1389 m->m_flags &= ~(M_HASH | M_CKHASH);
1390 isr = ether_characterize(&m);
1393 KKASSERT(m->m_flags & M_HASH);
1395 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1397 * Wrong hardware supplied hash; redispatch
1399 ether_dispatch(isr, m);
1400 if (__predict_false(ether_input_ckhash))
1401 atomic_add_long(ðer_input_wronghwhash, 1);
1405 ifp = m->m_pkthdr.rcvif;
1407 eh = mtod(m, struct ether_header *);
1408 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1409 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1410 ifp->if_addrlen) == 0)
1411 m->m_flags |= M_BCAST;
1413 m->m_flags |= M_MCAST;
1414 IFNET_STAT_INC(ifp, imcasts, 1);
1417 ether_input_oncpu(ifp, m);
1421 * Send the packet to the target msgport
1423 * At this point the packet had better be characterized (M_HASH set),
1424 * so we know which cpu to send it to.
1427 ether_dispatch(int isr, struct mbuf *m)
1429 struct netmsg_packet *pmsg;
1431 KKASSERT(m->m_flags & M_HASH);
1432 pmsg = &m->m_hdr.mh_netmsg;
1433 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1434 0, ether_input_handler);
1435 pmsg->nm_packet = m;
1436 pmsg->base.lmsg.u.ms_result = isr;
1438 logether(disp_beg, NULL);
1439 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1440 logether(disp_end, NULL);
1444 * Process a received Ethernet packet.
1446 * The ethernet header is assumed to be in the mbuf so the caller
1447 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1448 * bytes in the first mbuf.
1451 ether_input_pkt(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi)
1457 /* Discard packet if interface is not up */
1458 if (!(ifp->if_flags & IFF_UP)) {
1463 if (m->m_len < sizeof(struct ether_header)) {
1464 /* XXX error in the caller. */
1469 m->m_pkthdr.rcvif = ifp;
1471 logether(pkt_beg, ifp);
1473 ETHER_BPF_MTAP(ifp, m);
1475 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1477 if (ifp->if_flags & IFF_MONITOR) {
1478 struct ether_header *eh;
1480 eh = mtod(m, struct ether_header *);
1481 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1482 IFNET_STAT_INC(ifp, imcasts, 1);
1485 * Interface marked for monitoring; discard packet.
1489 logether(pkt_end, ifp);
1494 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1495 * we can dispatch it immediately without further inspection.
1497 if (pi != NULL && (m->m_flags & M_HASH)) {
1499 atomic_add_long(ðer_pktinfo_try, 1);
1501 netisr_hashcheck(pi->pi_netisr, m, pi);
1502 if (m->m_flags & M_HASH) {
1503 ether_dispatch(pi->pi_netisr, m);
1505 atomic_add_long(ðer_pktinfo_hit, 1);
1507 logether(pkt_end, ifp);
1512 else if (ifp->if_capenable & IFCAP_RSS) {
1514 atomic_add_long(ðer_rss_nopi, 1);
1516 atomic_add_long(ðer_rss_nohash, 1);
1521 * Packet hash will be recalculated by software, so clear
1522 * the M_HASH and M_CKHASH flag set by the driver; the hash
1523 * value calculated by the hardware may not be exactly what
1526 m->m_flags &= ~(M_HASH | M_CKHASH);
1528 if (!ether_vlancheck(&m)) {
1529 KKASSERT(m == NULL);
1530 logether(pkt_end, ifp);
1534 isr = ether_characterize(&m);
1536 logether(pkt_end, ifp);
1541 * Finally dispatch it
1543 ether_dispatch(isr, m);
1545 logether(pkt_end, ifp);
1549 ether_characterize(struct mbuf **m0)
1551 struct mbuf *m = *m0;
1552 struct ether_header *eh;
1553 uint16_t ether_type;
1556 eh = mtod(m, struct ether_header *);
1557 ether_type = ntohs(eh->ether_type);
1560 * Map ether type to netisr id.
1562 switch (ether_type) {
1574 case ETHERTYPE_IPV6:
1586 case ETHERTYPE_MPLS:
1587 case ETHERTYPE_MPLS_MCAST:
1588 m->m_flags |= M_MPLSLABELED;
1595 * NETISR_MAX is an invalid value; it is chosen to let
1596 * netisr_characterize() know that we have no clear
1597 * idea where this packet should go.
1604 * Ask the isr to characterize the packet since we couldn't.
1605 * This is an attempt to optimally get us onto the correct protocol
1608 netisr_characterize(isr, &m, sizeof(struct ether_header));
1615 ether_demux_handler(netmsg_t nmsg)
1617 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1623 ifp = m->m_pkthdr.rcvif;
1625 ether_demux_oncpu(ifp, m);
1629 ether_demux(struct mbuf *m)
1631 struct netmsg_packet *pmsg;
1634 isr = ether_characterize(&m);
1638 KKASSERT(m->m_flags & M_HASH);
1639 pmsg = &m->m_hdr.mh_netmsg;
1640 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1641 0, ether_demux_handler);
1642 pmsg->nm_packet = m;
1643 pmsg->base.lmsg.u.ms_result = isr;
1645 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1649 ether_tso_pullup(struct mbuf **mp, int *hoff0, struct ip **ip, int *iphlen,
1650 struct tcphdr **th, int *thoff)
1652 struct mbuf *m = *mp;
1653 struct ether_header *eh;
1657 KASSERT(M_WRITABLE(m), ("not writable"));
1659 hoff = ETHER_HDR_LEN;
1660 if (m->m_len < hoff) {
1661 m = m_pullup(m, hoff);
1665 eh = mtod(m, struct ether_header *);
1666 type = eh->ether_type;
1668 if (type == htons(ETHERTYPE_VLAN)) {
1669 struct ether_vlan_header *evh;
1671 hoff += EVL_ENCAPLEN;
1672 if (m->m_len < hoff) {
1673 m = m_pullup(m, hoff);
1677 evh = mtod(m, struct ether_vlan_header *);
1678 type = evh->evl_proto;
1680 KASSERT(type == htons(ETHERTYPE_IP), ("not IP %d", ntohs(type)));
1684 return tcp_tso_pullup(mp, hoff, ip, iphlen, th, thoff);
1694 kether_aton(const char *macstr, u_char *addr)
1696 unsigned int o0, o1, o2, o3, o4, o5;
1699 if (macstr == NULL || addr == NULL)
1702 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1718 kether_ntoa(const u_char *addr, char *buf)
1720 int len = ETHER_ADDRSTRLEN + 1;
1723 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1724 addr[1], addr[2], addr[3], addr[4], addr[5]);
1732 MODULE_VERSION(ether, 1);