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. 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.
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
29 * @(#)if_ethersubr.c 8.1 (Berkeley) 6/10/93
30 * $FreeBSD: src/sys/net/if_ethersubr.c,v 1.70.2.33 2003/04/28 15:45:53 archie Exp $
34 #include "opt_inet6.h"
36 #include "opt_netgraph.h"
40 #include <sys/param.h>
41 #include <sys/systm.h>
42 #include <sys/globaldata.h>
43 #include <sys/kernel.h>
46 #include <sys/malloc.h>
48 #include <sys/msgport.h>
49 #include <sys/socket.h>
50 #include <sys/sockio.h>
51 #include <sys/sysctl.h>
52 #include <sys/thread.h>
54 #include <sys/thread2.h>
55 #include <sys/mplock2.h>
58 #include <net/netisr.h>
59 #include <net/route.h>
60 #include <net/if_llc.h>
61 #include <net/if_dl.h>
62 #include <net/if_types.h>
63 #include <net/ifq_var.h>
65 #include <net/ethernet.h>
66 #include <net/vlan/if_vlan_ether.h>
67 #include <net/vlan/if_vlan_var.h>
68 #include <net/netmsg2.h>
69 #include <net/netisr2.h>
71 #if defined(INET) || defined(INET6)
72 #include <netinet/in.h>
73 #include <netinet/ip_var.h>
74 #include <netinet/tcp_var.h>
75 #include <netinet/if_ether.h>
76 #include <netinet/ip_flow.h>
77 #include <net/ipfw/ip_fw.h>
78 #include <net/dummynet/ip_dummynet.h>
81 #include <netinet6/nd6.h>
85 #include <netinet/ip_carp.h>
89 #include <netproto/mpls/mpls.h>
92 /* netgraph node hooks for ng_ether(4) */
93 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
94 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
95 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
96 void (*ng_ether_attach_p)(struct ifnet *ifp);
97 void (*ng_ether_detach_p)(struct ifnet *ifp);
99 void (*vlan_input_p)(struct mbuf *);
101 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
103 static void ether_restore_header(struct mbuf **, const struct ether_header *,
104 const struct ether_header *);
105 static int ether_characterize(struct mbuf **);
106 static void ether_dispatch(int, struct mbuf *, int);
111 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
112 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
113 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
114 struct ifnet *(*bridge_interface_p)(void *if_bridge);
116 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
122 void (*lagg_input_p)(struct ifnet *, struct mbuf *);
123 int (*lagg_output_p)(struct ifnet *, struct mbuf *);
125 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
126 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
129 #define gotoerr(e) do { error = (e); goto bad; } while (0)
130 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
132 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
134 const struct ether_header *eh);
136 static int ether_ipfw;
137 static u_long ether_restore_hdr;
138 static u_long ether_prepend_hdr;
139 static u_long ether_input_wronghash;
140 static int ether_debug;
143 static u_long ether_pktinfo_try;
144 static u_long ether_pktinfo_hit;
145 static u_long ether_rss_nopi;
146 static u_long ether_rss_nohash;
147 static u_long ether_input_requeue;
149 static u_long ether_input_wronghwhash;
150 static int ether_input_ckhash;
152 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU)
154 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
155 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default);
157 SYSCTL_DECL(_net_link);
158 SYSCTL_NODE(_net_link, IFT_ETHER, ether, CTLFLAG_RW, 0, "Ethernet");
159 SYSCTL_INT(_net_link_ether, OID_AUTO, debug, CTLFLAG_RW,
160 ðer_debug, 0, "Ether debug");
161 SYSCTL_INT(_net_link_ether, OID_AUTO, ipfw, CTLFLAG_RW,
162 ðer_ipfw, 0, "Pass ether pkts through firewall");
163 SYSCTL_ULONG(_net_link_ether, OID_AUTO, restore_hdr, CTLFLAG_RW,
164 ðer_restore_hdr, 0, "# of ether header restoration");
165 SYSCTL_ULONG(_net_link_ether, OID_AUTO, prepend_hdr, CTLFLAG_RW,
166 ðer_prepend_hdr, 0,
167 "# of ether header restoration which prepends mbuf");
168 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghash, CTLFLAG_RW,
169 ðer_input_wronghash, 0, "# of input packets with wrong hash");
170 SYSCTL_INT(_net_link_ether, OID_AUTO, tsolen, CTLFLAG_RW,
171 ðer_tsolen_default, 0, "Default max TSO length");
174 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nopi, CTLFLAG_RW,
175 ðer_rss_nopi, 0, "# of packets do not have pktinfo");
176 SYSCTL_ULONG(_net_link_ether, OID_AUTO, rss_nohash, CTLFLAG_RW,
177 ðer_rss_nohash, 0, "# of packets do not have hash");
178 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_try, CTLFLAG_RW,
179 ðer_pktinfo_try, 0,
180 "# of tries to find packets' msgport using pktinfo");
181 SYSCTL_ULONG(_net_link_ether, OID_AUTO, pktinfo_hit, CTLFLAG_RW,
182 ðer_pktinfo_hit, 0,
183 "# of packets whose msgport are found using pktinfo");
184 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_requeue, CTLFLAG_RW,
185 ðer_input_requeue, 0, "# of input packets gets requeued");
187 SYSCTL_ULONG(_net_link_ether, OID_AUTO, input_wronghwhash, CTLFLAG_RW,
188 ðer_input_wronghwhash, 0, "# of input packets with wrong hw hash");
189 SYSCTL_INT(_net_link_ether, OID_AUTO, always_ckhash, CTLFLAG_RW,
190 ðer_input_ckhash, 0, "always check hash");
192 #define ETHER_KTR_STR "ifp=%p"
193 #define ETHER_KTR_ARGS struct ifnet *ifp
195 #define KTR_ETHERNET KTR_ALL
197 KTR_INFO_MASTER(ether);
198 KTR_INFO(KTR_ETHERNET, ether, pkt_beg, 0, ETHER_KTR_STR, ETHER_KTR_ARGS);
199 KTR_INFO(KTR_ETHERNET, ether, pkt_end, 1, ETHER_KTR_STR, ETHER_KTR_ARGS);
200 KTR_INFO(KTR_ETHERNET, ether, disp_beg, 2, ETHER_KTR_STR, ETHER_KTR_ARGS);
201 KTR_INFO(KTR_ETHERNET, ether, disp_end, 3, ETHER_KTR_STR, ETHER_KTR_ARGS);
202 #define logether(name, arg) KTR_LOG(ether_ ## name, arg)
205 * Ethernet output routine.
206 * Encapsulate a packet of type family for the local net.
207 * Use trailer local net encapsulation if enough data in first
208 * packet leaves a multiple of 512 bytes of data in remainder.
209 * Assumes that ifp is actually pointer to arpcom structure.
212 ether_output(struct ifnet *ifp, struct mbuf *m, struct sockaddr *dst,
215 struct ether_header *eh, *deh;
218 int hlen = ETHER_HDR_LEN; /* link layer header length */
219 struct arpcom *ac = IFP2AC(ifp);
222 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
224 if (ifp->if_flags & IFF_MONITOR)
226 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) != (IFF_UP | IFF_RUNNING))
229 M_PREPEND(m, sizeof(struct ether_header), MB_DONTWAIT);
232 m->m_pkthdr.csum_lhlen = sizeof(struct ether_header);
233 eh = mtod(m, struct ether_header *);
234 edst = eh->ether_dhost;
237 * Fill in the destination ethernet address and frame type.
239 switch (dst->sa_family) {
242 if (!arpresolve(ifp, rt, m, dst, edst))
243 return (0); /* if not yet resolved */
245 if (m->m_flags & M_MPLSLABELED)
246 eh->ether_type = htons(ETHERTYPE_MPLS);
249 eh->ether_type = htons(ETHERTYPE_IP);
254 if (!nd6_storelladdr(&ac->ac_if, rt, m, dst, edst))
255 return (0); /* Something bad happenned. */
256 eh->ether_type = htons(ETHERTYPE_IPV6);
259 case pseudo_AF_HDRCMPLT:
261 loop_copy = -1; /* if this is for us, don't do it */
262 deh = (struct ether_header *)dst->sa_data;
263 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
264 eh->ether_type = deh->ether_type;
268 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
269 gotoerr(EAFNOSUPPORT);
272 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
273 memcpy(eh->ether_shost,
274 ((struct ether_header *)dst->sa_data)->ether_shost,
277 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
280 * Bridges require special output handling.
282 if (ifp->if_bridge) {
283 KASSERT(bridge_output_p != NULL,
284 ("%s: if_bridge not loaded!", __func__));
285 return bridge_output_p(ifp, m);
289 KASSERT(lagg_output_p != NULL,
290 ("%s: if_lagg not loaded!", __func__));
291 return lagg_output_p(ifp, m);
296 * If a simplex interface, and the packet is being sent to our
297 * Ethernet address or a broadcast address, loopback a copy.
298 * XXX To make a simplex device behave exactly like a duplex
299 * device, we should copy in the case of sending to our own
300 * ethernet address (thus letting the original actually appear
301 * on the wire). However, we don't do that here for security
302 * reasons and compatibility with the original behavior.
304 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
307 if (m->m_pkthdr.csum_flags & CSUM_IP)
308 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
309 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
310 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
311 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
314 if ((n = m_copypacket(m, MB_DONTWAIT)) != NULL) {
315 n->m_pkthdr.csum_flags |= csum_flags;
316 if (csum_flags & CSUM_DATA_VALID)
317 n->m_pkthdr.csum_data = 0xffff;
318 if_simloop(ifp, n, dst->sa_family, hlen);
320 IFNET_STAT_INC(ifp, iqdrops, 1);
321 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
322 ETHER_ADDR_LEN) == 0) {
323 m->m_pkthdr.csum_flags |= csum_flags;
324 if (csum_flags & CSUM_DATA_VALID)
325 m->m_pkthdr.csum_data = 0xffff;
326 if_simloop(ifp, m, dst->sa_family, hlen);
327 return (0); /* XXX */
332 if (ifp->if_type == IFT_CARP) {
333 ifp = carp_parent(ifp);
335 gotoerr(ENETUNREACH);
340 * Check precondition again
342 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
344 if (ifp->if_flags & IFF_MONITOR)
346 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
347 (IFF_UP | IFF_RUNNING))
352 /* Handle ng_ether(4) processing, if any */
353 if (ng_ether_output_p != NULL) {
355 * Hold BGL and recheck ng_ether_output_p
358 if (ng_ether_output_p != NULL) {
359 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
371 /* Continue with link-layer output */
372 return ether_output_frame(ifp, m);
380 * Returns the bridge interface an ifp is associated
383 * Only call if ifp->if_bridge != NULL.
386 ether_bridge_interface(struct ifnet *ifp)
388 if (bridge_interface_p)
389 return(bridge_interface_p(ifp->if_bridge));
394 * Ethernet link layer output routine to send a raw frame to the device.
396 * This assumes that the 14 byte Ethernet header is present and contiguous
400 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
402 struct ip_fw *rule = NULL;
404 struct altq_pktattr pktattr;
406 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
408 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
411 /* Extract info from dummynet tag */
412 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
413 KKASSERT(mtag != NULL);
414 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
415 KKASSERT(rule != NULL);
417 m_tag_delete(m, mtag);
418 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
421 if (ifq_is_enabled(&ifp->if_snd))
422 altq_etherclassify(&ifp->if_snd, m, &pktattr);
424 if (IPFW_LOADED && ether_ipfw != 0) {
425 struct ether_header save_eh, *eh;
427 eh = mtod(m, struct ether_header *);
429 m_adj(m, ETHER_HDR_LEN);
430 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
434 return ENOBUFS; /* pkt dropped */
436 return 0; /* consumed e.g. in a pipe */
439 /* packet was ok, restore the ethernet header */
440 ether_restore_header(&m, eh, &save_eh);
449 * Queue message on interface, update output statistics if
450 * successful, and start output if interface not yet active.
452 error = ifq_dispatch(ifp, m, &pktattr);
457 * ipfw processing for ethernet packets (in and out).
458 * The second parameter is NULL from ether_demux(), and ifp from
459 * ether_output_frame().
462 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
463 const struct ether_header *eh)
465 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
466 struct ip_fw_args args;
471 if (*rule != NULL && fw_one_pass)
472 return TRUE; /* dummynet packet, already partially processed */
475 * I need some amount of data to be contiguous.
477 i = min((*m0)->m_pkthdr.len, max_protohdr);
478 if ((*m0)->m_len < i) {
479 *m0 = m_pullup(*m0, i);
487 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
488 m_tag_delete(*m0, mtag);
489 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
490 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
491 KKASSERT(mtag != NULL);
492 m_tag_delete(*m0, mtag);
493 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
496 args.m = *m0; /* the packet we are looking at */
497 args.oif = dst; /* destination, if any */
498 args.rule = *rule; /* matching rule to restart */
499 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
500 i = ip_fw_chk_ptr(&args);
515 * XXX at some point add support for divert/forward actions.
516 * If none of the above matches, we have to drop the pkt.
522 * Pass the pkt to dummynet, which consumes it.
524 m = *m0; /* pass the original to dummynet */
525 *m0 = NULL; /* and nothing back to the caller */
527 ether_restore_header(&m, eh, &save_eh);
531 ip_fw_dn_io_ptr(m, args.cookie,
532 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
537 panic("unknown ipfw return value: %d", i);
542 * Perform common duties while attaching to interface list
545 ether_ifattach(struct ifnet *ifp, const uint8_t *lla,
546 lwkt_serialize_t serializer)
548 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
553 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla,
554 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer)
556 struct sockaddr_dl *sdl;
557 char ethstr[ETHER_ADDRSTRLEN + 1];
561 ifp->if_type = IFT_ETHER;
562 ifp->if_addrlen = ETHER_ADDR_LEN;
563 ifp->if_hdrlen = ETHER_HDR_LEN;
564 if_attach(ifp, serializer);
566 for (i = 0; i < ifq->altq_subq_cnt; ++i) {
567 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i);
569 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen *
570 (ETHER_MAX_LEN - ETHER_CRC_LEN);
572 ifp->if_mtu = ETHERMTU;
573 if (ifp->if_tsolen <= 0) {
574 if ((ether_tsolen_default / ETHERMTU) < 2) {
575 kprintf("ether TSO maxlen %d -> %d\n",
576 ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
577 ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
579 ifp->if_tsolen = ether_tsolen_default;
581 if (ifp->if_baudrate == 0)
582 ifp->if_baudrate = 10000000;
583 ifp->if_output = ether_output;
584 ifp->if_input = ether_input;
585 ifp->if_resolvemulti = ether_resolvemulti;
586 ifp->if_broadcastaddr = etherbroadcastaddr;
587 sdl = IF_LLSOCKADDR(ifp);
588 sdl->sdl_type = IFT_ETHER;
589 sdl->sdl_alen = ifp->if_addrlen;
590 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
592 * XXX Keep the current drivers happy.
593 * XXX Remove once all drivers have been cleaned up
595 if (lla != IFP2AC(ifp)->ac_enaddr)
596 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
597 bpfattach(ifp, dlt, hdrlen);
598 if (ng_ether_attach_p != NULL)
599 (*ng_ether_attach_p)(ifp);
601 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
605 * Perform common duties while detaching an Ethernet interface
608 ether_ifdetach(struct ifnet *ifp)
612 if (ng_ether_detach_p != NULL)
613 (*ng_ether_detach_p)(ifp);
619 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
621 struct ifaddr *ifa = (struct ifaddr *) data;
622 struct ifreq *ifr = (struct ifreq *) data;
625 #define IF_INIT(ifp) \
627 if (((ifp)->if_flags & IFF_UP) == 0) { \
628 (ifp)->if_flags |= IFF_UP; \
629 (ifp)->if_init((ifp)->if_softc); \
633 ASSERT_IFNET_SERIALIZED_ALL(ifp);
637 switch (ifa->ifa_addr->sa_family) {
640 IF_INIT(ifp); /* before arpwhohas */
641 arp_ifinit(ifp, ifa);
651 bcopy(IFP2AC(ifp)->ac_enaddr,
652 ((struct sockaddr *)ifr->ifr_data)->sa_data,
658 * Set the interface MTU.
660 if (ifr->ifr_mtu > ETHERMTU) {
663 ifp->if_mtu = ifr->ifr_mtu;
678 struct sockaddr **llsa,
681 struct sockaddr_dl *sdl;
683 struct sockaddr_in *sin;
686 struct sockaddr_in6 *sin6;
690 switch(sa->sa_family) {
693 * No mapping needed. Just check that it's a valid MC address.
695 sdl = (struct sockaddr_dl *)sa;
696 e_addr = LLADDR(sdl);
697 if ((e_addr[0] & 1) != 1)
698 return EADDRNOTAVAIL;
704 sin = (struct sockaddr_in *)sa;
705 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_addr)))
706 return EADDRNOTAVAIL;
707 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
708 sdl->sdl_len = sizeof *sdl;
709 sdl->sdl_family = AF_LINK;
710 sdl->sdl_index = ifp->if_index;
711 sdl->sdl_type = IFT_ETHER;
712 sdl->sdl_alen = ETHER_ADDR_LEN;
713 e_addr = LLADDR(sdl);
714 ETHER_MAP_IP_MULTICAST(&sin->sin_addr, e_addr);
715 *llsa = (struct sockaddr *)sdl;
720 sin6 = (struct sockaddr_in6 *)sa;
721 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
723 * An IP6 address of 0 means listen to all
724 * of the Ethernet multicast address used for IP6.
725 * (This is used for multicast routers.)
727 ifp->if_flags |= IFF_ALLMULTI;
731 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
732 return EADDRNOTAVAIL;
733 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
734 sdl->sdl_len = sizeof *sdl;
735 sdl->sdl_family = AF_LINK;
736 sdl->sdl_index = ifp->if_index;
737 sdl->sdl_type = IFT_ETHER;
738 sdl->sdl_alen = ETHER_ADDR_LEN;
739 e_addr = LLADDR(sdl);
740 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
741 *llsa = (struct sockaddr *)sdl;
747 * Well, the text isn't quite right, but it's the name
756 * This is for reference. We have a table-driven version
757 * of the little-endian crc32 generator, which is faster
758 * than the double-loop.
761 ether_crc32_le(const uint8_t *buf, size_t len)
763 uint32_t c, crc, carry;
766 crc = 0xffffffffU; /* initial value */
768 for (i = 0; i < len; i++) {
770 for (j = 0; j < 8; j++) {
771 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
775 crc = (crc ^ ETHER_CRC_POLY_LE);
783 ether_crc32_le(const uint8_t *buf, size_t len)
785 static const uint32_t crctab[] = {
786 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
787 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
788 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
789 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
794 crc = 0xffffffffU; /* initial value */
796 for (i = 0; i < len; i++) {
798 crc = (crc >> 4) ^ crctab[crc & 0xf];
799 crc = (crc >> 4) ^ crctab[crc & 0xf];
807 ether_crc32_be(const uint8_t *buf, size_t len)
809 uint32_t c, crc, carry;
812 crc = 0xffffffffU; /* initial value */
814 for (i = 0; i < len; i++) {
816 for (j = 0; j < 8; j++) {
817 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
821 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
829 * find the size of ethernet header, and call classifier
832 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
833 struct altq_pktattr *pktattr)
835 struct ether_header *eh;
837 int hlen, af, hdrsize;
839 hlen = sizeof(struct ether_header);
840 eh = mtod(m, struct ether_header *);
842 ether_type = ntohs(eh->ether_type);
843 if (ether_type < ETHERMTU) {
845 struct llc *llc = (struct llc *)(eh + 1);
848 if (m->m_len < hlen ||
849 llc->llc_dsap != LLC_SNAP_LSAP ||
850 llc->llc_ssap != LLC_SNAP_LSAP ||
851 llc->llc_control != LLC_UI)
852 goto bad; /* not snap! */
854 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
857 if (ether_type == ETHERTYPE_IP) {
859 hdrsize = 20; /* sizeof(struct ip) */
861 } else if (ether_type == ETHERTYPE_IPV6) {
863 hdrsize = 40; /* sizeof(struct ip6_hdr) */
868 while (m->m_len <= hlen) {
872 if (m->m_len < hlen + hdrsize) {
874 * ip header is not in a single mbuf. this should not
875 * happen in the current code.
876 * (todo: use m_pulldown in the future)
882 ifq_classify(ifq, m, af, pktattr);
889 pktattr->pattr_class = NULL;
890 pktattr->pattr_hdr = NULL;
891 pktattr->pattr_af = AF_UNSPEC;
895 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
896 const struct ether_header *save_eh)
898 struct mbuf *m = *m0;
903 * Prepend the header, optimize for the common case of
904 * eh pointing into the mbuf.
906 if ((const void *)(eh + 1) == (void *)m->m_data) {
907 m->m_data -= ETHER_HDR_LEN;
908 m->m_len += ETHER_HDR_LEN;
909 m->m_pkthdr.len += ETHER_HDR_LEN;
913 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
915 bcopy(save_eh, mtod(m, struct ether_header *),
923 * Upper layer processing for a received Ethernet packet.
926 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
928 struct ether_header *eh;
929 int isr, discard = 0;
931 struct ip_fw *rule = NULL;
934 KASSERT(m->m_len >= ETHER_HDR_LEN,
935 ("ether header is not contiguous!"));
937 eh = mtod(m, struct ether_header *);
939 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
942 /* Extract info from dummynet tag */
943 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
944 KKASSERT(mtag != NULL);
945 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
946 KKASSERT(rule != NULL);
948 m_tag_delete(m, mtag);
949 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
951 /* packet is passing the second time */
956 * We got a packet which was unicast to a different Ethernet
957 * address. If the driver is working properly, then this
958 * situation can only happen when the interface is in
959 * promiscuous mode. We defer the packet discarding until the
960 * vlan processing is done, so that vlan/bridge or vlan/netgraph
963 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
964 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
965 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
966 if (ether_debug & 1) {
967 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
968 "%02x:%02x:%02x:%02x:%02x:%02x "
969 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
983 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
984 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
985 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
986 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
987 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
988 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
991 if ((ether_debug & 2) == 0)
996 if (IPFW_LOADED && ether_ipfw != 0 && !discard) {
997 struct ether_header save_eh = *eh;
999 /* XXX old crufty stuff, needs to be removed */
1000 m_adj(m, sizeof(struct ether_header));
1002 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1007 ether_restore_header(&m, eh, &save_eh);
1010 eh = mtod(m, struct ether_header *);
1013 ether_type = ntohs(eh->ether_type);
1014 KKASSERT(ether_type != ETHERTYPE_VLAN);
1016 /* Handle input from a lagg(4) port */
1017 if (ifp->if_type == IFT_IEEE8023ADLAG) {
1018 KASSERT(lagg_input_p != NULL,
1019 ("%s: if_lagg not loaded!", __func__));
1020 (*lagg_input_p)(ifp, m);
1024 if (m->m_flags & M_VLANTAG) {
1025 void (*vlan_input_func)(struct mbuf *);
1027 vlan_input_func = vlan_input_p;
1028 if (vlan_input_func != NULL) {
1031 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1038 * If we have been asked to discard this packet
1039 * (e.g. not for us), drop it before entering
1048 * Clear protocol specific flags,
1049 * before entering the upper layer.
1051 m->m_flags &= ~M_ETHER_FLAGS;
1053 /* Strip ethernet header. */
1054 m_adj(m, sizeof(struct ether_header));
1056 switch (ether_type) {
1059 if ((m->m_flags & M_LENCHECKED) == 0) {
1060 if (!ip_lengthcheck(&m, 0))
1063 if (ipflow_fastforward(m))
1069 if (ifp->if_flags & IFF_NOARP) {
1070 /* Discard packet if ARP is disabled on interface */
1079 case ETHERTYPE_IPV6:
1085 case ETHERTYPE_MPLS:
1086 case ETHERTYPE_MPLS_MCAST:
1087 /* Should have been set by ether_input(). */
1088 KKASSERT(m->m_flags & M_MPLSLABELED);
1095 * The accurate msgport is not determined before
1096 * we reach here, so recharacterize packet.
1098 m->m_flags &= ~M_HASH;
1099 if (ng_ether_input_orphan_p != NULL) {
1101 * Put back the ethernet header so netgraph has a
1102 * consistent view of inbound packets.
1104 M_PREPEND(m, ETHER_HDR_LEN, MB_DONTWAIT);
1107 * M_PREPEND frees the mbuf in case of failure.
1112 * Hold BGL and recheck ng_ether_input_orphan_p
1115 if (ng_ether_input_orphan_p != NULL) {
1116 ng_ether_input_orphan_p(ifp, m);
1126 if (m->m_flags & M_HASH) {
1127 if (&curthread->td_msgport ==
1128 netisr_hashport(m->m_pkthdr.hash)) {
1129 netisr_handle(isr, m);
1133 * XXX Something is wrong,
1134 * we probably should panic here!
1136 m->m_flags &= ~M_HASH;
1137 atomic_add_long(ðer_input_wronghash, 1);
1141 atomic_add_long(ðer_input_requeue, 1);
1143 netisr_queue(isr, m);
1147 * First we perform any link layer operations, then continue to the
1148 * upper layers with ether_demux_oncpu().
1151 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1157 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1159 * Receiving interface's flags are changed, when this
1160 * packet is waiting for processing; discard it.
1167 * Tap the packet off here for a bridge. bridge_input()
1168 * will return NULL if it has consumed the packet, otherwise
1169 * it gets processed as normal. Note that bridge_input()
1170 * will always return the original packet if we need to
1171 * process it locally.
1173 if (ifp->if_bridge) {
1174 KASSERT(bridge_input_p != NULL,
1175 ("%s: if_bridge not loaded!", __func__));
1177 if(m->m_flags & M_ETHER_BRIDGED) {
1178 m->m_flags &= ~M_ETHER_BRIDGED;
1180 m = bridge_input_p(ifp, m);
1184 KASSERT(ifp == m->m_pkthdr.rcvif,
1185 ("bridge_input_p changed rcvif"));
1190 carp = ifp->if_carp;
1192 m = carp_input(carp, m);
1195 KASSERT(ifp == m->m_pkthdr.rcvif,
1196 ("carp_input changed rcvif"));
1200 /* Handle ng_ether(4) processing, if any */
1201 if (ng_ether_input_p != NULL) {
1203 * Hold BGL and recheck ng_ether_input_p
1206 if (ng_ether_input_p != NULL)
1207 ng_ether_input_p(ifp, &m);
1214 /* Continue with upper layer processing */
1215 ether_demux_oncpu(ifp, m);
1219 * Perform certain functions of ether_input():
1221 * - Update statistics
1222 * - Run bpf(4) tap if requested
1223 * Then pass the packet to ether_input_oncpu().
1225 * This function should be used by pseudo interface (e.g. vlan(4)),
1226 * when it tries to claim that the packet is received by it.
1232 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1234 /* Discard packet if interface is not up */
1235 if (!(ifp->if_flags & IFF_UP)) {
1241 * Change receiving interface. The bridge will often pass a flag to
1242 * ask that this not be done so ARPs get applied to the correct
1245 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1246 m->m_pkthdr.rcvif == NULL) {
1247 m->m_pkthdr.rcvif = ifp;
1250 /* Update statistics */
1251 IFNET_STAT_INC(ifp, ipackets, 1);
1252 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1253 if (m->m_flags & (M_MCAST | M_BCAST))
1254 IFNET_STAT_INC(ifp, imcasts, 1);
1256 if (reinput_flags & REINPUT_RUNBPF)
1259 ether_input_oncpu(ifp, m);
1262 static __inline boolean_t
1263 ether_vlancheck(struct mbuf **m0)
1265 struct mbuf *m = *m0;
1266 struct ether_header *eh;
1267 uint16_t ether_type;
1269 eh = mtod(m, struct ether_header *);
1270 ether_type = ntohs(eh->ether_type);
1272 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1274 * Extract vlan tag if hardware does not do it for us
1276 vlan_ether_decap(&m);
1280 eh = mtod(m, struct ether_header *);
1281 ether_type = ntohs(eh->ether_type);
1284 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1286 * To prevent possible dangerous recursion,
1287 * we don't do vlan-in-vlan
1289 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1292 KKASSERT(ether_type != ETHERTYPE_VLAN);
1294 m->m_flags |= M_ETHER_VLANCHECKED;
1305 ether_input_handler(netmsg_t nmsg)
1307 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1308 struct ether_header *eh;
1315 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1316 if (!ether_vlancheck(&m)) {
1317 KKASSERT(m == NULL);
1321 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1322 __predict_false(ether_input_ckhash)) {
1326 * Need to verify the hash supplied by the hardware
1327 * which could be wrong.
1329 m->m_flags &= ~(M_HASH | M_CKHASH);
1330 isr = ether_characterize(&m);
1333 KKASSERT(m->m_flags & M_HASH);
1335 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1337 * Wrong hardware supplied hash; redispatch
1339 ether_dispatch(isr, m, -1);
1340 if (__predict_false(ether_input_ckhash))
1341 atomic_add_long(ðer_input_wronghwhash, 1);
1345 ifp = m->m_pkthdr.rcvif;
1347 eh = mtod(m, struct ether_header *);
1348 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1349 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1350 ifp->if_addrlen) == 0)
1351 m->m_flags |= M_BCAST;
1353 m->m_flags |= M_MCAST;
1354 IFNET_STAT_INC(ifp, imcasts, 1);
1357 ether_input_oncpu(ifp, m);
1361 * Send the packet to the target netisr msgport
1363 * At this point the packet must be characterized (M_HASH set),
1364 * so we know which netisr to send it to.
1367 ether_dispatch(int isr, struct mbuf *m, int cpuid)
1369 struct netmsg_packet *pmsg;
1372 KKASSERT(m->m_flags & M_HASH);
1373 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash);
1375 pmsg = &m->m_hdr.mh_netmsg;
1376 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1377 0, ether_input_handler);
1378 pmsg->nm_packet = m;
1379 pmsg->base.lmsg.u.ms_result = isr;
1381 logether(disp_beg, NULL);
1382 if (target_cpuid == cpuid) {
1383 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid),
1386 lwkt_sendmsg(netisr_cpuport(target_cpuid),
1389 logether(disp_end, NULL);
1393 * Process a received Ethernet packet.
1395 * The ethernet header is assumed to be in the mbuf so the caller
1396 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1397 * bytes in the first mbuf.
1399 * If the caller knows that the current thread is stick to the current
1400 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid
1401 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should
1402 * be passed as 'cpuid' argument.
1405 ether_input(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi,
1412 /* Discard packet if interface is not up */
1413 if (!(ifp->if_flags & IFF_UP)) {
1418 if (m->m_len < sizeof(struct ether_header)) {
1419 /* XXX error in the caller. */
1424 m->m_pkthdr.rcvif = ifp;
1426 logether(pkt_beg, ifp);
1428 ETHER_BPF_MTAP(ifp, m);
1430 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1432 if (ifp->if_flags & IFF_MONITOR) {
1433 struct ether_header *eh;
1435 eh = mtod(m, struct ether_header *);
1436 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1437 IFNET_STAT_INC(ifp, imcasts, 1);
1440 * Interface marked for monitoring; discard packet.
1444 logether(pkt_end, ifp);
1449 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1450 * we can dispatch it immediately with trivial checks.
1452 if (pi != NULL && (m->m_flags & M_HASH)) {
1454 atomic_add_long(ðer_pktinfo_try, 1);
1456 netisr_hashcheck(pi->pi_netisr, m, pi);
1457 if (m->m_flags & M_HASH) {
1458 ether_dispatch(pi->pi_netisr, m, cpuid);
1460 atomic_add_long(ðer_pktinfo_hit, 1);
1462 logether(pkt_end, ifp);
1467 else if (ifp->if_capenable & IFCAP_RSS) {
1469 atomic_add_long(ðer_rss_nopi, 1);
1471 atomic_add_long(ðer_rss_nohash, 1);
1476 * Packet hash will be recalculated by software, so clear
1477 * the M_HASH and M_CKHASH flag set by the driver; the hash
1478 * value calculated by the hardware may not be exactly what
1481 m->m_flags &= ~(M_HASH | M_CKHASH);
1483 if (!ether_vlancheck(&m)) {
1484 KKASSERT(m == NULL);
1485 logether(pkt_end, ifp);
1489 isr = ether_characterize(&m);
1491 logether(pkt_end, ifp);
1496 * Finally dispatch it
1498 ether_dispatch(isr, m, cpuid);
1500 logether(pkt_end, ifp);
1504 ether_characterize(struct mbuf **m0)
1506 struct mbuf *m = *m0;
1507 struct ether_header *eh;
1508 uint16_t ether_type;
1511 eh = mtod(m, struct ether_header *);
1512 ether_type = ntohs(eh->ether_type);
1515 * Map ether type to netisr id.
1517 switch (ether_type) {
1529 case ETHERTYPE_IPV6:
1535 case ETHERTYPE_MPLS:
1536 case ETHERTYPE_MPLS_MCAST:
1537 m->m_flags |= M_MPLSLABELED;
1544 * NETISR_MAX is an invalid value; it is chosen to let
1545 * netisr_characterize() know that we have no clear
1546 * idea where this packet should go.
1553 * Ask the isr to characterize the packet since we couldn't.
1554 * This is an attempt to optimally get us onto the correct protocol
1557 netisr_characterize(isr, &m, sizeof(struct ether_header));
1564 ether_demux_handler(netmsg_t nmsg)
1566 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1572 ifp = m->m_pkthdr.rcvif;
1574 ether_demux_oncpu(ifp, m);
1578 ether_demux(struct mbuf *m)
1580 struct netmsg_packet *pmsg;
1583 isr = ether_characterize(&m);
1587 KKASSERT(m->m_flags & M_HASH);
1588 pmsg = &m->m_hdr.mh_netmsg;
1589 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1590 0, ether_demux_handler);
1591 pmsg->nm_packet = m;
1592 pmsg->base.lmsg.u.ms_result = isr;
1594 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1598 kether_aton(const char *macstr, u_char *addr)
1600 unsigned int o0, o1, o2, o3, o4, o5;
1603 if (macstr == NULL || addr == NULL)
1606 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1622 kether_ntoa(const u_char *addr, char *buf)
1624 int len = ETHER_ADDRSTRLEN + 1;
1627 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1628 addr[1], addr[2], addr[3], addr[4], addr[5]);
1636 MODULE_VERSION(ether, 1);