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/ipfw3/ip_fw.h>
79 #include <net/dummynet/ip_dummynet.h>
82 #include <netinet6/nd6.h>
86 #include <netinet/ip_carp.h>
90 #include <netproto/mpls/mpls.h>
93 /* netgraph node hooks for ng_ether(4) */
94 void (*ng_ether_input_p)(struct ifnet *ifp, struct mbuf **mp);
95 void (*ng_ether_input_orphan_p)(struct ifnet *ifp, struct mbuf *m);
96 int (*ng_ether_output_p)(struct ifnet *ifp, struct mbuf **mp);
97 void (*ng_ether_attach_p)(struct ifnet *ifp);
98 void (*ng_ether_detach_p)(struct ifnet *ifp);
100 void (*vlan_input_p)(struct mbuf *);
102 static int ether_output(struct ifnet *, struct mbuf *, struct sockaddr *,
104 static void ether_restore_header(struct mbuf **, const struct ether_header *,
105 const struct ether_header *);
106 static int ether_characterize(struct mbuf **);
107 static void ether_dispatch(int, struct mbuf *, int);
112 struct mbuf *(*bridge_input_p)(struct ifnet *, struct mbuf *);
113 int (*bridge_output_p)(struct ifnet *, struct mbuf *);
114 void (*bridge_dn_p)(struct mbuf *, struct ifnet *);
115 struct ifnet *(*bridge_interface_p)(void *if_bridge);
117 static int ether_resolvemulti(struct ifnet *, struct sockaddr **,
123 void (*lagg_input_p)(struct ifnet *, struct mbuf *);
124 int (*lagg_output_p)(struct ifnet *, struct mbuf *);
126 const uint8_t etherbroadcastaddr[ETHER_ADDR_LEN] = {
127 0xff, 0xff, 0xff, 0xff, 0xff, 0xff
130 #define gotoerr(e) do { error = (e); goto bad; } while (0)
131 #define IFP2AC(ifp) ((struct arpcom *)(ifp))
133 static boolean_t ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst,
135 const struct ether_header *eh);
137 static int ether_ipfw;
138 static u_long ether_restore_hdr;
139 static u_long ether_prepend_hdr;
140 static u_long ether_input_wronghash;
141 static int ether_debug;
144 static u_long ether_pktinfo_try;
145 static u_long ether_pktinfo_hit;
146 static u_long ether_rss_nopi;
147 static u_long ether_rss_nohash;
148 static u_long ether_input_requeue;
150 static u_long ether_input_wronghwhash;
151 static int ether_input_ckhash;
153 #define ETHER_TSOLEN_DEFAULT (4 * ETHERMTU)
155 #define ETHER_NMBCLUSTERS_DEFMIN 32
156 #define ETHER_NMBCLUSTERS_DEFAULT 256
158 static int ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
159 TUNABLE_INT("net.link.ether.tsolen", ðer_tsolen_default);
161 static int ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT;
162 TUNABLE_INT("net.link.ether.nmbclusters", ðer_nmbclusters_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), M_NOWAIT);
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);
266 case pseudo_AF_HDRCMPLT:
268 loop_copy = -1; /* if this is for us, don't do it */
269 deh = (struct ether_header *)dst->sa_data;
270 memcpy(edst, deh->ether_dhost, ETHER_ADDR_LEN);
271 eh->ether_type = deh->ether_type;
275 if_printf(ifp, "can't handle af%d\n", dst->sa_family);
276 gotoerr(EAFNOSUPPORT);
279 if (dst->sa_family == pseudo_AF_HDRCMPLT) /* unlikely */
280 memcpy(eh->ether_shost,
281 ((struct ether_header *)dst->sa_data)->ether_shost,
284 memcpy(eh->ether_shost, ac->ac_enaddr, ETHER_ADDR_LEN);
287 * Bridges require special output handling.
289 if (ifp->if_bridge) {
290 KASSERT(bridge_output_p != NULL,
291 ("%s: if_bridge not loaded!", __func__));
292 return bridge_output_p(ifp, m);
296 KASSERT(lagg_output_p != NULL,
297 ("%s: if_lagg not loaded!", __func__));
298 return lagg_output_p(ifp, m);
303 * If a simplex interface, and the packet is being sent to our
304 * Ethernet address or a broadcast address, loopback a copy.
305 * XXX To make a simplex device behave exactly like a duplex
306 * device, we should copy in the case of sending to our own
307 * ethernet address (thus letting the original actually appear
308 * on the wire). However, we don't do that here for security
309 * reasons and compatibility with the original behavior.
311 if ((ifp->if_flags & IFF_SIMPLEX) && (loop_copy != -1)) {
314 if (m->m_pkthdr.csum_flags & CSUM_IP)
315 csum_flags |= (CSUM_IP_CHECKED | CSUM_IP_VALID);
316 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA)
317 csum_flags |= (CSUM_DATA_VALID | CSUM_PSEUDO_HDR);
318 if ((m->m_flags & M_BCAST) || (loop_copy > 0)) {
321 if ((n = m_copypacket(m, M_NOWAIT)) != NULL) {
322 n->m_pkthdr.csum_flags |= csum_flags;
323 if (csum_flags & CSUM_DATA_VALID)
324 n->m_pkthdr.csum_data = 0xffff;
325 if_simloop(ifp, n, dst->sa_family, hlen);
327 IFNET_STAT_INC(ifp, iqdrops, 1);
328 } else if (bcmp(eh->ether_dhost, eh->ether_shost,
329 ETHER_ADDR_LEN) == 0) {
330 m->m_pkthdr.csum_flags |= csum_flags;
331 if (csum_flags & CSUM_DATA_VALID)
332 m->m_pkthdr.csum_data = 0xffff;
333 if_simloop(ifp, m, dst->sa_family, hlen);
334 return (0); /* XXX */
339 if (ifp->if_type == IFT_CARP) {
340 ifp = carp_parent(ifp);
342 gotoerr(ENETUNREACH);
347 * Check precondition again
349 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
351 if (ifp->if_flags & IFF_MONITOR)
353 if ((ifp->if_flags & (IFF_UP | IFF_RUNNING)) !=
354 (IFF_UP | IFF_RUNNING))
359 /* Handle ng_ether(4) processing, if any */
360 if (ng_ether_output_p != NULL) {
362 * Hold BGL and recheck ng_ether_output_p
365 if (ng_ether_output_p != NULL) {
366 if ((error = ng_ether_output_p(ifp, &m)) != 0) {
378 /* Continue with link-layer output */
379 return ether_output_frame(ifp, m);
387 * Returns the bridge interface an ifp is associated
390 * Only call if ifp->if_bridge != NULL.
393 ether_bridge_interface(struct ifnet *ifp)
395 if (bridge_interface_p)
396 return(bridge_interface_p(ifp->if_bridge));
401 * Ethernet link layer output routine to send a raw frame to the device.
403 * This assumes that the 14 byte Ethernet header is present and contiguous
407 ether_output_frame(struct ifnet *ifp, struct mbuf *m)
409 struct ip_fw *rule = NULL;
411 struct altq_pktattr pktattr;
413 ASSERT_IFNET_NOT_SERIALIZED_ALL(ifp);
415 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
418 /* Extract info from dummynet tag */
419 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
420 KKASSERT(mtag != NULL);
421 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
422 KKASSERT(rule != NULL);
424 m_tag_delete(m, mtag);
425 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
428 if (ifq_is_enabled(&ifp->if_snd))
429 altq_etherclassify(&ifp->if_snd, m, &pktattr);
431 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0) {
432 struct ether_header save_eh, *eh;
434 eh = mtod(m, struct ether_header *);
436 m_adj(m, ETHER_HDR_LEN);
437 if (!ether_ipfw_chk(&m, ifp, &rule, eh)) {
441 return ENOBUFS; /* pkt dropped */
443 return 0; /* consumed e.g. in a pipe */
446 /* packet was ok, restore the ethernet header */
447 ether_restore_header(&m, eh, &save_eh);
456 * Queue message on interface, update output statistics if
457 * successful, and start output if interface not yet active.
459 error = ifq_dispatch(ifp, m, &pktattr);
464 * ipfw processing for ethernet packets (in and out).
465 * The second parameter is NULL from ether_demux(), and ifp from
466 * ether_output_frame().
469 ether_ipfw_chk(struct mbuf **m0, struct ifnet *dst, struct ip_fw **rule,
470 const struct ether_header *eh)
472 struct ether_header save_eh = *eh; /* might be a ptr in *m0 */
473 struct ip_fw_args args;
478 if (*rule != NULL && fw_one_pass)
479 return TRUE; /* dummynet packet, already partially processed */
482 * I need some amount of data to be contiguous.
484 i = min((*m0)->m_pkthdr.len, max_protohdr);
485 if ((*m0)->m_len < i) {
486 *m0 = m_pullup(*m0, i);
494 if ((mtag = m_tag_find(*m0, PACKET_TAG_IPFW_DIVERT, NULL)) != NULL)
495 m_tag_delete(*m0, mtag);
496 if ((*m0)->m_pkthdr.fw_flags & IPFORWARD_MBUF_TAGGED) {
497 mtag = m_tag_find(*m0, PACKET_TAG_IPFORWARD, NULL);
498 KKASSERT(mtag != NULL);
499 m_tag_delete(*m0, mtag);
500 (*m0)->m_pkthdr.fw_flags &= ~IPFORWARD_MBUF_TAGGED;
503 args.m = *m0; /* the packet we are looking at */
504 args.oif = dst; /* destination, if any */
505 args.rule = *rule; /* matching rule to restart */
506 args.eh = &save_eh; /* MAC header for bridged/MAC packets */
507 i = ip_fw_chk_ptr(&args);
522 * XXX at some point add support for divert/forward actions.
523 * If none of the above matches, we have to drop the pkt.
529 * Pass the pkt to dummynet, which consumes it.
531 m = *m0; /* pass the original to dummynet */
532 *m0 = NULL; /* and nothing back to the caller */
534 ether_restore_header(&m, eh, &save_eh);
538 ip_fw_dn_io_ptr(m, args.cookie,
539 dst ? DN_TO_ETH_OUT: DN_TO_ETH_DEMUX, &args);
544 panic("unknown ipfw return value: %d", i);
549 * Perform common duties while attaching to interface list
552 ether_ifattach(struct ifnet *ifp, const uint8_t *lla,
553 lwkt_serialize_t serializer)
555 ether_ifattach_bpf(ifp, lla, DLT_EN10MB, sizeof(struct ether_header),
560 ether_ifattach_bpf(struct ifnet *ifp, const uint8_t *lla,
561 u_int dlt, u_int hdrlen, lwkt_serialize_t serializer)
563 struct sockaddr_dl *sdl;
564 char ethstr[ETHER_ADDRSTRLEN + 1];
569 * If driver does not configure # of mbuf clusters/jclusters
570 * that could sit on the device queues for quite some time,
572 * - The device queues only consume mbuf clusters.
573 * - No more than ether_nmbclusters_default (by default 256)
574 * mbuf clusters will sit on the device queues for quite
577 if (ifp->if_nmbclusters <= 0 && ifp->if_nmbjclusters <= 0) {
578 if (ether_nmbclusters_default < ETHER_NMBCLUSTERS_DEFMIN) {
579 kprintf("ether nmbclusters %d -> %d\n",
580 ether_nmbclusters_default,
581 ETHER_NMBCLUSTERS_DEFAULT);
582 ether_nmbclusters_default = ETHER_NMBCLUSTERS_DEFAULT;
584 ifp->if_nmbclusters = ether_nmbclusters_default;
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);
592 for (i = 0; i < ifq->altq_subq_cnt; ++i) {
593 struct ifaltq_subque *ifsq = ifq_get_subq(ifq, i);
595 ifsq->ifsq_maxbcnt = ifsq->ifsq_maxlen *
596 (ETHER_MAX_LEN - ETHER_CRC_LEN);
598 ifp->if_mtu = ETHERMTU;
599 if (ifp->if_tsolen <= 0) {
600 if ((ether_tsolen_default / ETHERMTU) < 2) {
601 kprintf("ether TSO maxlen %d -> %d\n",
602 ether_tsolen_default, ETHER_TSOLEN_DEFAULT);
603 ether_tsolen_default = ETHER_TSOLEN_DEFAULT;
605 ifp->if_tsolen = ether_tsolen_default;
607 if (ifp->if_baudrate == 0)
608 ifp->if_baudrate = 10000000;
609 ifp->if_output = ether_output;
610 ifp->if_input = ether_input;
611 ifp->if_resolvemulti = ether_resolvemulti;
612 ifp->if_broadcastaddr = etherbroadcastaddr;
613 sdl = IF_LLSOCKADDR(ifp);
614 sdl->sdl_type = IFT_ETHER;
615 sdl->sdl_alen = ifp->if_addrlen;
616 bcopy(lla, LLADDR(sdl), ifp->if_addrlen);
618 * XXX Keep the current drivers happy.
619 * XXX Remove once all drivers have been cleaned up
621 if (lla != IFP2AC(ifp)->ac_enaddr)
622 bcopy(lla, IFP2AC(ifp)->ac_enaddr, ifp->if_addrlen);
623 bpfattach(ifp, dlt, hdrlen);
624 if (ng_ether_attach_p != NULL)
625 (*ng_ether_attach_p)(ifp);
627 if_printf(ifp, "MAC address: %s\n", kether_ntoa(lla, ethstr));
631 * Perform common duties while detaching an Ethernet interface
634 ether_ifdetach(struct ifnet *ifp)
638 if (ng_ether_detach_p != NULL)
639 (*ng_ether_detach_p)(ifp);
645 ether_ioctl(struct ifnet *ifp, u_long command, caddr_t data)
647 struct ifaddr *ifa = (struct ifaddr *) data;
648 struct ifreq *ifr = (struct ifreq *) data;
651 #define IF_INIT(ifp) \
653 if (((ifp)->if_flags & IFF_UP) == 0) { \
654 (ifp)->if_flags |= IFF_UP; \
655 (ifp)->if_init((ifp)->if_softc); \
659 ASSERT_IFNET_SERIALIZED_ALL(ifp);
663 switch (ifa->ifa_addr->sa_family) {
666 IF_INIT(ifp); /* before arpwhohas */
667 arp_ifinit(ifp, ifa);
677 bcopy(IFP2AC(ifp)->ac_enaddr,
678 ((struct sockaddr *)ifr->ifr_data)->sa_data,
684 * Set the interface MTU.
686 if (ifr->ifr_mtu > ETHERMTU) {
689 ifp->if_mtu = ifr->ifr_mtu;
704 struct sockaddr **llsa,
707 struct sockaddr_dl *sdl;
709 struct sockaddr_in *sin;
712 struct sockaddr_in6 *sin6;
716 switch(sa->sa_family) {
719 * No mapping needed. Just check that it's a valid MC address.
721 sdl = (struct sockaddr_dl *)sa;
722 e_addr = LLADDR(sdl);
723 if ((e_addr[0] & 1) != 1)
724 return EADDRNOTAVAIL;
730 sin = (struct sockaddr_in *)sa;
731 if (!IN_MULTICAST(ntohl(sin->sin_addr.s_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_IP_MULTICAST(&sin->sin_addr, e_addr);
741 *llsa = (struct sockaddr *)sdl;
746 sin6 = (struct sockaddr_in6 *)sa;
747 if (IN6_IS_ADDR_UNSPECIFIED(&sin6->sin6_addr)) {
749 * An IP6 address of 0 means listen to all
750 * of the Ethernet multicast address used for IP6.
751 * (This is used for multicast routers.)
753 ifp->if_flags |= IFF_ALLMULTI;
757 if (!IN6_IS_ADDR_MULTICAST(&sin6->sin6_addr))
758 return EADDRNOTAVAIL;
759 sdl = kmalloc(sizeof *sdl, M_IFMADDR, M_WAITOK | M_ZERO);
760 sdl->sdl_len = sizeof *sdl;
761 sdl->sdl_family = AF_LINK;
762 sdl->sdl_index = ifp->if_index;
763 sdl->sdl_type = IFT_ETHER;
764 sdl->sdl_alen = ETHER_ADDR_LEN;
765 e_addr = LLADDR(sdl);
766 ETHER_MAP_IPV6_MULTICAST(&sin6->sin6_addr, e_addr);
767 *llsa = (struct sockaddr *)sdl;
773 * Well, the text isn't quite right, but it's the name
782 * This is for reference. We have a table-driven version
783 * of the little-endian crc32 generator, which is faster
784 * than the double-loop.
787 ether_crc32_le(const uint8_t *buf, size_t len)
789 uint32_t c, crc, carry;
792 crc = 0xffffffffU; /* initial value */
794 for (i = 0; i < len; i++) {
796 for (j = 0; j < 8; j++) {
797 carry = ((crc & 0x01) ? 1 : 0) ^ (c & 0x01);
801 crc = (crc ^ ETHER_CRC_POLY_LE);
809 ether_crc32_le(const uint8_t *buf, size_t len)
811 static const uint32_t crctab[] = {
812 0x00000000, 0x1db71064, 0x3b6e20c8, 0x26d930ac,
813 0x76dc4190, 0x6b6b51f4, 0x4db26158, 0x5005713c,
814 0xedb88320, 0xf00f9344, 0xd6d6a3e8, 0xcb61b38c,
815 0x9b64c2b0, 0x86d3d2d4, 0xa00ae278, 0xbdbdf21c
820 crc = 0xffffffffU; /* initial value */
822 for (i = 0; i < len; i++) {
824 crc = (crc >> 4) ^ crctab[crc & 0xf];
825 crc = (crc >> 4) ^ crctab[crc & 0xf];
833 ether_crc32_be(const uint8_t *buf, size_t len)
835 uint32_t c, crc, carry;
838 crc = 0xffffffffU; /* initial value */
840 for (i = 0; i < len; i++) {
842 for (j = 0; j < 8; j++) {
843 carry = ((crc & 0x80000000U) ? 1 : 0) ^ (c & 0x01);
847 crc = (crc ^ ETHER_CRC_POLY_BE) | carry;
855 * find the size of ethernet header, and call classifier
858 altq_etherclassify(struct ifaltq *ifq, struct mbuf *m,
859 struct altq_pktattr *pktattr)
861 struct ether_header *eh;
863 int hlen, af, hdrsize;
865 hlen = sizeof(struct ether_header);
866 eh = mtod(m, struct ether_header *);
868 ether_type = ntohs(eh->ether_type);
869 if (ether_type < ETHERMTU) {
871 struct llc *llc = (struct llc *)(eh + 1);
874 if (m->m_len < hlen ||
875 llc->llc_dsap != LLC_SNAP_LSAP ||
876 llc->llc_ssap != LLC_SNAP_LSAP ||
877 llc->llc_control != LLC_UI)
878 goto bad; /* not snap! */
880 ether_type = ntohs(llc->llc_un.type_snap.ether_type);
883 if (ether_type == ETHERTYPE_IP) {
885 hdrsize = 20; /* sizeof(struct ip) */
887 } else if (ether_type == ETHERTYPE_IPV6) {
889 hdrsize = 40; /* sizeof(struct ip6_hdr) */
894 while (m->m_len <= hlen) {
898 if (m->m_len < hlen + hdrsize) {
900 * ip header is not in a single mbuf. this should not
901 * happen in the current code.
902 * (todo: use m_pulldown in the future)
908 ifq_classify(ifq, m, af, pktattr);
915 pktattr->pattr_class = NULL;
916 pktattr->pattr_hdr = NULL;
917 pktattr->pattr_af = AF_UNSPEC;
921 ether_restore_header(struct mbuf **m0, const struct ether_header *eh,
922 const struct ether_header *save_eh)
924 struct mbuf *m = *m0;
929 * Prepend the header, optimize for the common case of
930 * eh pointing into the mbuf.
932 if ((const void *)(eh + 1) == (void *)m->m_data) {
933 m->m_data -= ETHER_HDR_LEN;
934 m->m_len += ETHER_HDR_LEN;
935 m->m_pkthdr.len += ETHER_HDR_LEN;
939 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
941 bcopy(save_eh, mtod(m, struct ether_header *),
949 * Upper layer processing for a received Ethernet packet.
952 ether_demux_oncpu(struct ifnet *ifp, struct mbuf *m)
954 struct ether_header *eh;
955 int isr, discard = 0;
957 struct ip_fw *rule = NULL;
960 KASSERT(m->m_len >= ETHER_HDR_LEN,
961 ("ether header is not contiguous!"));
963 eh = mtod(m, struct ether_header *);
965 if (m->m_pkthdr.fw_flags & DUMMYNET_MBUF_TAGGED) {
968 /* Extract info from dummynet tag */
969 mtag = m_tag_find(m, PACKET_TAG_DUMMYNET, NULL);
970 KKASSERT(mtag != NULL);
971 rule = ((struct dn_pkt *)m_tag_data(mtag))->dn_priv;
972 KKASSERT(rule != NULL);
974 m_tag_delete(m, mtag);
975 m->m_pkthdr.fw_flags &= ~DUMMYNET_MBUF_TAGGED;
977 /* packet is passing the second time */
982 * We got a packet which was unicast to a different Ethernet
983 * address. If the driver is working properly, then this
984 * situation can only happen when the interface is in
985 * promiscuous mode. We defer the packet discarding until the
986 * vlan processing is done, so that vlan/bridge or vlan/netgraph
989 if (((ifp->if_flags & (IFF_PROMISC | IFF_PPROMISC)) == IFF_PROMISC) &&
990 !ETHER_IS_MULTICAST(eh->ether_dhost) &&
991 bcmp(eh->ether_dhost, IFP2AC(ifp)->ac_enaddr, ETHER_ADDR_LEN)) {
992 if (ether_debug & 1) {
993 kprintf("%02x:%02x:%02x:%02x:%02x:%02x "
994 "%02x:%02x:%02x:%02x:%02x:%02x "
995 "%04x vs %02x:%02x:%02x:%02x:%02x:%02x\n",
1009 ((u_char *)IFP2AC(ifp)->ac_enaddr)[0],
1010 ((u_char *)IFP2AC(ifp)->ac_enaddr)[1],
1011 ((u_char *)IFP2AC(ifp)->ac_enaddr)[2],
1012 ((u_char *)IFP2AC(ifp)->ac_enaddr)[3],
1013 ((u_char *)IFP2AC(ifp)->ac_enaddr)[4],
1014 ((u_char *)IFP2AC(ifp)->ac_enaddr)[5]
1017 if ((ether_debug & 2) == 0)
1022 if ((IPFW_LOADED || IPFW3_LOADED) && ether_ipfw != 0 && !discard) {
1023 struct ether_header save_eh = *eh;
1025 /* XXX old crufty stuff, needs to be removed */
1026 m_adj(m, sizeof(struct ether_header));
1028 if (!ether_ipfw_chk(&m, NULL, &rule, eh)) {
1033 ether_restore_header(&m, eh, &save_eh);
1036 eh = mtod(m, struct ether_header *);
1039 ether_type = ntohs(eh->ether_type);
1040 KKASSERT(ether_type != ETHERTYPE_VLAN);
1042 /* Handle input from a lagg(4) port */
1043 if (ifp->if_type == IFT_IEEE8023ADLAG) {
1044 KASSERT(lagg_input_p != NULL,
1045 ("%s: if_lagg not loaded!", __func__));
1046 (*lagg_input_p)(ifp, m);
1050 if (m->m_flags & M_VLANTAG) {
1051 void (*vlan_input_func)(struct mbuf *);
1053 vlan_input_func = vlan_input_p;
1054 /* Make sure 'vlan_input_func' is really used. */
1056 if (vlan_input_func != NULL) {
1059 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1066 * If we have been asked to discard this packet
1067 * (e.g. not for us), drop it before entering
1076 * Clear protocol specific flags,
1077 * before entering the upper layer.
1079 m->m_flags &= ~M_ETHER_FLAGS;
1081 /* Strip ethernet header. */
1082 m_adj(m, sizeof(struct ether_header));
1084 switch (ether_type) {
1087 if ((m->m_flags & M_LENCHECKED) == 0) {
1088 if (!ip_lengthcheck(&m, 0))
1091 if (ipflow_fastforward(m))
1097 if (ifp->if_flags & IFF_NOARP) {
1098 /* Discard packet if ARP is disabled on interface */
1107 case ETHERTYPE_IPV6:
1113 case ETHERTYPE_MPLS:
1114 case ETHERTYPE_MPLS_MCAST:
1115 /* Should have been set by ether_input(). */
1116 KKASSERT(m->m_flags & M_MPLSLABELED);
1123 * The accurate msgport is not determined before
1124 * we reach here, so recharacterize packet.
1126 m->m_flags &= ~M_HASH;
1127 if (ng_ether_input_orphan_p != NULL) {
1129 * Put back the ethernet header so netgraph has a
1130 * consistent view of inbound packets.
1132 M_PREPEND(m, ETHER_HDR_LEN, M_NOWAIT);
1135 * M_PREPEND frees the mbuf in case of failure.
1140 * Hold BGL and recheck ng_ether_input_orphan_p
1143 if (ng_ether_input_orphan_p != NULL) {
1144 ng_ether_input_orphan_p(ifp, m);
1154 if (m->m_flags & M_HASH) {
1155 if (&curthread->td_msgport ==
1156 netisr_hashport(m->m_pkthdr.hash)) {
1157 netisr_handle(isr, m);
1161 * XXX Something is wrong,
1162 * we probably should panic here!
1164 m->m_flags &= ~M_HASH;
1165 atomic_add_long(ðer_input_wronghash, 1);
1169 atomic_add_long(ðer_input_requeue, 1);
1171 netisr_queue(isr, m);
1175 * First we perform any link layer operations, then continue to the
1176 * upper layers with ether_demux_oncpu().
1179 ether_input_oncpu(struct ifnet *ifp, struct mbuf *m)
1185 if ((ifp->if_flags & (IFF_UP | IFF_MONITOR)) != IFF_UP) {
1187 * Receiving interface's flags are changed, when this
1188 * packet is waiting for processing; discard it.
1195 * Tap the packet off here for a bridge. bridge_input()
1196 * will return NULL if it has consumed the packet, otherwise
1197 * it gets processed as normal. Note that bridge_input()
1198 * will always return the original packet if we need to
1199 * process it locally.
1201 if (ifp->if_bridge) {
1202 KASSERT(bridge_input_p != NULL,
1203 ("%s: if_bridge not loaded!", __func__));
1205 if(m->m_flags & M_ETHER_BRIDGED) {
1206 m->m_flags &= ~M_ETHER_BRIDGED;
1208 m = bridge_input_p(ifp, m);
1212 KASSERT(ifp == m->m_pkthdr.rcvif,
1213 ("bridge_input_p changed rcvif"));
1218 carp = ifp->if_carp;
1220 m = carp_input(carp, m);
1223 KASSERT(ifp == m->m_pkthdr.rcvif,
1224 ("carp_input changed rcvif"));
1228 /* Handle ng_ether(4) processing, if any */
1229 if (ng_ether_input_p != NULL) {
1231 * Hold BGL and recheck ng_ether_input_p
1234 if (ng_ether_input_p != NULL)
1235 ng_ether_input_p(ifp, &m);
1242 /* Continue with upper layer processing */
1243 ether_demux_oncpu(ifp, m);
1247 * Perform certain functions of ether_input():
1249 * - Update statistics
1250 * - Run bpf(4) tap if requested
1251 * Then pass the packet to ether_input_oncpu().
1253 * This function should be used by pseudo interface (e.g. vlan(4)),
1254 * when it tries to claim that the packet is received by it.
1260 ether_reinput_oncpu(struct ifnet *ifp, struct mbuf *m, int reinput_flags)
1262 /* Discard packet if interface is not up */
1263 if (!(ifp->if_flags & IFF_UP)) {
1269 * Change receiving interface. The bridge will often pass a flag to
1270 * ask that this not be done so ARPs get applied to the correct
1273 if ((reinput_flags & REINPUT_KEEPRCVIF) == 0 ||
1274 m->m_pkthdr.rcvif == NULL) {
1275 m->m_pkthdr.rcvif = ifp;
1278 /* Update statistics */
1279 IFNET_STAT_INC(ifp, ipackets, 1);
1280 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1281 if (m->m_flags & (M_MCAST | M_BCAST))
1282 IFNET_STAT_INC(ifp, imcasts, 1);
1284 if (reinput_flags & REINPUT_RUNBPF)
1287 ether_input_oncpu(ifp, m);
1290 static __inline boolean_t
1291 ether_vlancheck(struct mbuf **m0)
1293 struct mbuf *m = *m0;
1294 struct ether_header *eh;
1295 uint16_t ether_type;
1297 eh = mtod(m, struct ether_header *);
1298 ether_type = ntohs(eh->ether_type);
1300 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG) == 0) {
1302 * Extract vlan tag if hardware does not do it for us
1304 vlan_ether_decap(&m);
1308 eh = mtod(m, struct ether_header *);
1309 ether_type = ntohs(eh->ether_type);
1312 if (ether_type == ETHERTYPE_VLAN && (m->m_flags & M_VLANTAG)) {
1314 * To prevent possible dangerous recursion,
1315 * we don't do vlan-in-vlan
1317 IFNET_STAT_INC(m->m_pkthdr.rcvif, noproto, 1);
1320 KKASSERT(ether_type != ETHERTYPE_VLAN);
1322 m->m_flags |= M_ETHER_VLANCHECKED;
1333 ether_input_handler(netmsg_t nmsg)
1335 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1336 struct ether_header *eh;
1343 if ((m->m_flags & M_ETHER_VLANCHECKED) == 0) {
1344 if (!ether_vlancheck(&m)) {
1345 KKASSERT(m == NULL);
1349 if ((m->m_flags & (M_HASH | M_CKHASH)) == (M_HASH | M_CKHASH) ||
1350 __predict_false(ether_input_ckhash)) {
1354 * Need to verify the hash supplied by the hardware
1355 * which could be wrong.
1357 m->m_flags &= ~(M_HASH | M_CKHASH);
1358 isr = ether_characterize(&m);
1361 KKASSERT(m->m_flags & M_HASH);
1363 if (netisr_hashcpu(m->m_pkthdr.hash) != mycpuid) {
1365 * Wrong hardware supplied hash; redispatch
1367 ether_dispatch(isr, m, -1);
1368 if (__predict_false(ether_input_ckhash))
1369 atomic_add_long(ðer_input_wronghwhash, 1);
1373 ifp = m->m_pkthdr.rcvif;
1375 eh = mtod(m, struct ether_header *);
1376 if (ETHER_IS_MULTICAST(eh->ether_dhost)) {
1377 if (bcmp(ifp->if_broadcastaddr, eh->ether_dhost,
1378 ifp->if_addrlen) == 0)
1379 m->m_flags |= M_BCAST;
1381 m->m_flags |= M_MCAST;
1382 IFNET_STAT_INC(ifp, imcasts, 1);
1385 ether_input_oncpu(ifp, m);
1389 * Send the packet to the target netisr msgport
1391 * At this point the packet must be characterized (M_HASH set),
1392 * so we know which netisr to send it to.
1395 ether_dispatch(int isr, struct mbuf *m, int cpuid)
1397 struct netmsg_packet *pmsg;
1400 KKASSERT(m->m_flags & M_HASH);
1401 target_cpuid = netisr_hashcpu(m->m_pkthdr.hash);
1403 pmsg = &m->m_hdr.mh_netmsg;
1404 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1405 0, ether_input_handler);
1406 pmsg->nm_packet = m;
1407 pmsg->base.lmsg.u.ms_result = isr;
1409 logether(disp_beg, NULL);
1410 if (target_cpuid == cpuid) {
1411 lwkt_sendmsg_oncpu(netisr_cpuport(target_cpuid),
1414 lwkt_sendmsg(netisr_cpuport(target_cpuid),
1417 logether(disp_end, NULL);
1421 * Process a received Ethernet packet.
1423 * The ethernet header is assumed to be in the mbuf so the caller
1424 * MUST MAKE SURE that there are at least sizeof(struct ether_header)
1425 * bytes in the first mbuf.
1427 * If the caller knows that the current thread is stick to the current
1428 * cpu, e.g. the interrupt thread or the netisr thread, the current cpuid
1429 * (mycpuid) should be passed through 'cpuid' argument. Else -1 should
1430 * be passed as 'cpuid' argument.
1433 ether_input(struct ifnet *ifp, struct mbuf *m, const struct pktinfo *pi,
1440 /* Discard packet if interface is not up */
1441 if (!(ifp->if_flags & IFF_UP)) {
1446 if (m->m_len < sizeof(struct ether_header)) {
1447 /* XXX error in the caller. */
1452 m->m_pkthdr.rcvif = ifp;
1454 logether(pkt_beg, ifp);
1456 ETHER_BPF_MTAP(ifp, m);
1458 IFNET_STAT_INC(ifp, ibytes, m->m_pkthdr.len);
1460 if (ifp->if_flags & IFF_MONITOR) {
1461 struct ether_header *eh;
1463 eh = mtod(m, struct ether_header *);
1464 if (ETHER_IS_MULTICAST(eh->ether_dhost))
1465 IFNET_STAT_INC(ifp, imcasts, 1);
1468 * Interface marked for monitoring; discard packet.
1472 logether(pkt_end, ifp);
1477 * If the packet has been characterized (pi->pi_netisr / M_HASH)
1478 * we can dispatch it immediately with trivial checks.
1480 if (pi != NULL && (m->m_flags & M_HASH)) {
1482 atomic_add_long(ðer_pktinfo_try, 1);
1484 netisr_hashcheck(pi->pi_netisr, m, pi);
1485 if (m->m_flags & M_HASH) {
1486 ether_dispatch(pi->pi_netisr, m, cpuid);
1488 atomic_add_long(ðer_pktinfo_hit, 1);
1490 logether(pkt_end, ifp);
1495 else if (ifp->if_capenable & IFCAP_RSS) {
1497 atomic_add_long(ðer_rss_nopi, 1);
1499 atomic_add_long(ðer_rss_nohash, 1);
1504 * Packet hash will be recalculated by software, so clear
1505 * the M_HASH and M_CKHASH flag set by the driver; the hash
1506 * value calculated by the hardware may not be exactly what
1509 m->m_flags &= ~(M_HASH | M_CKHASH);
1511 if (!ether_vlancheck(&m)) {
1512 KKASSERT(m == NULL);
1513 logether(pkt_end, ifp);
1517 isr = ether_characterize(&m);
1519 logether(pkt_end, ifp);
1524 * Finally dispatch it
1526 ether_dispatch(isr, m, cpuid);
1528 logether(pkt_end, ifp);
1532 ether_characterize(struct mbuf **m0)
1534 struct mbuf *m = *m0;
1535 struct ether_header *eh;
1536 uint16_t ether_type;
1539 eh = mtod(m, struct ether_header *);
1540 ether_type = ntohs(eh->ether_type);
1543 * Map ether type to netisr id.
1545 switch (ether_type) {
1557 case ETHERTYPE_IPV6:
1563 case ETHERTYPE_MPLS:
1564 case ETHERTYPE_MPLS_MCAST:
1565 m->m_flags |= M_MPLSLABELED;
1572 * NETISR_MAX is an invalid value; it is chosen to let
1573 * netisr_characterize() know that we have no clear
1574 * idea where this packet should go.
1581 * Ask the isr to characterize the packet since we couldn't.
1582 * This is an attempt to optimally get us onto the correct protocol
1585 netisr_characterize(isr, &m, sizeof(struct ether_header));
1592 ether_demux_handler(netmsg_t nmsg)
1594 struct netmsg_packet *nmp = &nmsg->packet; /* actual size */
1600 ifp = m->m_pkthdr.rcvif;
1602 ether_demux_oncpu(ifp, m);
1606 ether_demux(struct mbuf *m)
1608 struct netmsg_packet *pmsg;
1611 isr = ether_characterize(&m);
1615 KKASSERT(m->m_flags & M_HASH);
1616 pmsg = &m->m_hdr.mh_netmsg;
1617 netmsg_init(&pmsg->base, NULL, &netisr_apanic_rport,
1618 0, ether_demux_handler);
1619 pmsg->nm_packet = m;
1620 pmsg->base.lmsg.u.ms_result = isr;
1622 lwkt_sendmsg(netisr_hashport(m->m_pkthdr.hash), &pmsg->base.lmsg);
1626 kether_aton(const char *macstr, u_char *addr)
1628 unsigned int o0, o1, o2, o3, o4, o5;
1631 if (macstr == NULL || addr == NULL)
1634 n = ksscanf(macstr, "%x:%x:%x:%x:%x:%x", &o0, &o1, &o2,
1650 kether_ntoa(const u_char *addr, char *buf)
1652 int len = ETHER_ADDRSTRLEN + 1;
1655 n = ksnprintf(buf, len, "%02x:%02x:%02x:%02x:%02x:%02x", addr[0],
1656 addr[1], addr[2], addr[3], addr[4], addr[5]);
1664 MODULE_VERSION(ether, 1);