1 /* $FreeBSD: src/sys/netinet6/ip6_mroute.c,v 1.2.2.9 2003/01/23 21:06:47 sam Exp $ */
2 /* $KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $ */
5 * Copyright (C) 1998 WIDE Project.
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the project nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
20 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 /* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */
36 * IP multicast forwarding procedures
38 * Written by David Waitzman, BBN Labs, August 1988.
39 * Modified by Steve Deering, Stanford, February 1989.
40 * Modified by Mark J. Steiglitz, Stanford, May, 1991
41 * Modified by Van Jacobson, LBL, January 1993
42 * Modified by Ajit Thyagarajan, PARC, August 1993
43 * Modified by Bill Fenenr, PARC, April 1994
45 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
49 #include "opt_inet6.h"
51 #include <sys/param.h>
52 #include <sys/systm.h>
53 #include <sys/callout.h>
54 #include <sys/malloc.h>
56 #include <sys/socket.h>
57 #include <sys/socketvar.h>
58 #include <sys/sockio.h>
59 #include <sys/protosw.h>
60 #include <sys/errno.h>
62 #include <sys/kernel.h>
63 #include <sys/syslog.h>
64 #include <sys/thread2.h>
67 #include <net/route.h>
68 #include <net/raw_cb.h>
70 #include <netinet/in.h>
71 #include <netinet/in_var.h>
73 #include <netinet/ip6.h>
74 #include <netinet6/ip6_var.h>
75 #include <netinet6/ip6_mroute.h>
76 #include <netinet6/pim6.h>
77 #include <netinet6/pim6_var.h>
79 #include <net/net_osdep.h>
81 static MALLOC_DEFINE(M_MRTABLE, "mf6c", "multicast forwarding cache entry");
83 #define M_HASCL(m) ((m)->m_flags & M_EXT)
85 static int ip6_mdq (struct mbuf *, struct ifnet *, struct mf6c *);
86 static void phyint_send (struct ip6_hdr *, struct mif6 *, struct mbuf *);
88 static int set_pim6 (int *);
89 static int socket_send (struct socket *, struct mbuf *,
90 struct sockaddr_in6 *);
91 static int register_send (struct ip6_hdr *, struct mif6 *,
95 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
96 * except for netstat or debugging purposes.
98 struct socket *ip6_mrouter = NULL;
99 int ip6_mrouter_ver = 0;
100 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */
101 struct mrt6stat mrt6stat;
103 #define NO_RTE_FOUND 0x1
104 #define RTE_FOUND 0x2
106 struct mf6c *mf6ctable[MF6CTBLSIZ];
107 u_char n6expire[MF6CTBLSIZ];
108 static struct mif6 mif6table[MAXMIFS];
110 u_int mrt6debug = 0; /* debug level */
111 #define DEBUG_MFC 0x02
112 #define DEBUG_FORWARD 0x04
113 #define DEBUG_EXPIRE 0x08
114 #define DEBUG_XMIT 0x10
115 #define DEBUG_REG 0x20
116 #define DEBUG_PIM 0x40
119 static void expire_upcalls (void *);
120 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
121 #define UPCALL_EXPIRE 6 /* number of timeouts */
125 extern struct socket *ip_mrouter;
130 * 'Interfaces' associated with decapsulator (so we can tell
131 * packets that went through it from ones that get reflected
132 * by a broken gateway). These interfaces are never linked into
133 * the system ifnet list & no routes point to them. I.e., packets
134 * can't be sent this way. They only exist as a placeholder for
135 * multicast source verification.
137 struct ifnet multicast_register_if;
139 #define ENCAP_HOPS 64
144 static mifi_t nummifs = 0;
145 static mifi_t reg_mif_num = (mifi_t)-1;
147 static struct pim6stat pim6stat;
151 * Hash function for a source, group entry
153 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
154 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
155 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
156 (g).s6_addr32[2] ^ (g).s6_addr32[3])
159 * Find a route for a given origin IPv6 address and Multicast group address.
160 * Quality of service parameter to be added in the future!!!
163 #define MF6CFIND(o, g, rt) do { \
164 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
166 mrt6stat.mrt6s_mfc_lookups++; \
168 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
169 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
170 (_rt->mf6c_stall == NULL)) { \
174 _rt = _rt->mf6c_next; \
177 mrt6stat.mrt6s_mfc_misses++; \
182 * Macros to compute elapsed time efficiently
183 * Borrowed from Van Jacobson's scheduling code
185 #define TV_DELTA(a, b, delta) do { \
188 delta = (a).tv_usec - (b).tv_usec; \
189 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
198 delta += (1000000 * xxs); \
203 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
204 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
207 #define UPCALL_MAX 50
208 u_long upcall_data[UPCALL_MAX + 1];
209 static void collate();
210 #endif /* UPCALL_TIMING */
212 static int get_sg_cnt (struct sioc_sg_req6 *);
213 static int get_mif6_cnt (struct sioc_mif_req6 *);
214 static int ip6_mrouter_init (struct socket *, struct mbuf *, int);
215 static int add_m6if (struct mif6ctl *);
216 static int del_m6if (mifi_t *);
217 static int add_m6fc (struct mf6cctl *);
218 static int del_m6fc (struct mf6cctl *);
220 static struct callout expire_upcalls_ch;
223 * Handle MRT setsockopt commands to modify the multicast routing tables.
226 ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
231 if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
234 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
236 soopt_to_mbuf(sopt, m); /* XXX */
238 switch (sopt->sopt_name) {
243 error = ip6_mrouter_init(so, m, sopt->sopt_name);
246 error = ip6_mrouter_done();
249 error = add_m6if(mtod(m, struct mif6ctl *));
252 error = del_m6if(mtod(m, mifi_t *));
255 error = add_m6fc(mtod(m, struct mf6cctl *));
258 error = del_m6fc(mtod(m, struct mf6cctl *));
261 error = set_pim6(mtod(m, int *));
273 * Handle MRT getsockopt commands
276 ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
280 if (so != ip6_mrouter) return EACCES;
282 switch (sopt->sopt_name) {
284 soopt_from_kbuf(sopt, &pim6, sizeof(pim6));
291 * Handle ioctl commands to obtain information from the cache
294 mrt6_ioctl(int cmd, caddr_t data)
299 case SIOCGETSGCNT_IN6:
300 return (get_sg_cnt((struct sioc_sg_req6 *)data));
301 break; /* for safety */
302 case SIOCGETMIFCNT_IN6:
303 return (get_mif6_cnt((struct sioc_mif_req6 *)data));
304 break; /* for safety */
313 * returns the packet, byte, rpf-failure count for the source group provided
316 get_sg_cnt(struct sioc_sg_req6 *req)
321 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
324 req->pktcnt = rt->mf6c_pkt_cnt;
325 req->bytecnt = rt->mf6c_byte_cnt;
326 req->wrong_if = rt->mf6c_wrong_if;
330 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
337 * returns the input and output packet and byte counts on the mif provided
340 get_mif6_cnt(struct sioc_mif_req6 *req)
342 mifi_t mifi = req->mifi;
347 req->icount = mif6table[mifi].m6_pkt_in;
348 req->ocount = mif6table[mifi].m6_pkt_out;
349 req->ibytes = mif6table[mifi].m6_bytes_in;
350 req->obytes = mif6table[mifi].m6_bytes_out;
358 if ((*i != 1) && (*i != 0))
367 * Enable multicast routing
370 ip6_mrouter_init(struct socket *so, struct mbuf *m, int cmd)
377 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
378 so->so_type, so->so_proto->pr_protocol);
381 if (so->so_type != SOCK_RAW ||
382 so->so_proto->pr_protocol != IPPROTO_ICMPV6)
385 if (!m || (m->m_len != sizeof(int *)))
392 if (ip6_mrouter != NULL) return EADDRINUSE;
395 ip6_mrouter_ver = cmd;
397 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
398 bzero((caddr_t)n6expire, sizeof(n6expire));
400 pim6 = 0;/* used for stubbing out/in pim stuff */
402 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
403 expire_upcalls, NULL);
407 log(LOG_DEBUG, "ip6_mrouter_init\n");
414 * Disable multicast routing
417 ip6_mrouter_done(void)
422 struct in6_ifreq ifr;
427 * For each phyint in use, disable promiscuous reception of all IPv6
433 * If there is still IPv4 multicast routing daemon,
434 * we remain interfaces to receive all muliticasted packets.
435 * XXX: there may be an interface in which the IPv4 multicast
436 * daemon is not interested...
442 for (mifi = 0; mifi < nummifs; mifi++) {
443 if (mif6table[mifi].m6_ifp &&
444 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
445 ifr.ifr_addr.sin6_family = AF_INET6;
446 ifr.ifr_addr.sin6_addr= kin6addr_any;
447 ifp = mif6table[mifi].m6_ifp;
448 ifnet_serialize_all(ifp);
449 ifp->if_ioctl(ifp, SIOCDELMULTI,
450 (caddr_t)&ifr, NULL);
451 ifnet_deserialize_all(ifp);
456 bzero((caddr_t)qtable, sizeof(qtable));
457 bzero((caddr_t)tbftable, sizeof(tbftable));
459 bzero((caddr_t)mif6table, sizeof(mif6table));
462 pim6 = 0; /* used to stub out/in pim specific code */
464 callout_stop(&expire_upcalls_ch);
467 * Free all multicast forwarding cache entries.
469 for (i = 0; i < MF6CTBLSIZ; i++) {
474 for (rte = rt->mf6c_stall; rte != NULL; ) {
475 struct rtdetq *n = rte->next;
478 kfree(rte, M_MRTABLE);
483 kfree(frt, M_MRTABLE);
487 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
490 * Reset de-encapsulation cache
499 log(LOG_DEBUG, "ip6_mrouter_done\n");
505 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
508 * Add a mif to the mif table
511 add_m6if(struct mif6ctl *mifcp)
517 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
520 if (mifcp->mif6c_mifi >= MAXMIFS)
522 mifp = mif6table + mifcp->mif6c_mifi;
524 return EADDRINUSE; /* XXX: is it appropriate? */
525 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
527 ifp = ifindex2ifnet[mifcp->mif6c_pifi];
529 if (mifcp->mif6c_flags & MIFF_REGISTER) {
530 if (reg_mif_num == (mifi_t)-1) {
531 strlcpy(multicast_register_if.if_xname, "register_mif",
533 multicast_register_if.if_flags |= IFF_LOOPBACK;
534 multicast_register_if.if_index = mifcp->mif6c_mifi;
535 reg_mif_num = mifcp->mif6c_mifi;
538 ifp = &multicast_register_if;
542 /* Make sure the interface supports multicast */
543 if ((ifp->if_flags & IFF_MULTICAST) == 0)
547 error = if_allmulti(ifp, 1);
554 mifp->m6_flags = mifcp->mif6c_flags;
557 /* scaling up here allows division by 1024 in critical code */
558 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
560 /* initialize per mif pkt counters */
562 mifp->m6_pkt_out = 0;
563 mifp->m6_bytes_in = 0;
564 mifp->m6_bytes_out = 0;
567 /* Adjust nummifs up if the mifi is higher than nummifs */
568 if (nummifs <= mifcp->mif6c_mifi)
569 nummifs = mifcp->mif6c_mifi + 1;
574 "add_mif #%d, phyint %s\n",
583 * Delete a mif from the mif table
586 del_m6if(mifi_t *mifip)
588 struct mif6 *mifp = mif6table + *mifip;
592 if (*mifip >= nummifs)
594 if (mifp->m6_ifp == NULL)
599 if (!(mifp->m6_flags & MIFF_REGISTER)) {
601 * XXX: what if there is yet IPv4 multicast daemon
602 * using the interface?
610 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
611 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
613 bzero((caddr_t)mifp, sizeof (*mifp));
615 /* Adjust nummifs down */
616 for (mifi = nummifs; mifi > 0; mifi--)
617 if (mif6table[mifi - 1].m6_ifp)
625 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
635 add_m6fc(struct mf6cctl *mfccp)
642 MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
643 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
645 /* If an entry already exists, just update the fields */
648 if (mrt6debug & DEBUG_MFC)
650 "add_m6fc no upcall h %d o %s g %s p %x\n",
651 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
652 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
653 mfccp->mf6cc_parent);
657 rt->mf6c_parent = mfccp->mf6cc_parent;
658 rt->mf6c_ifset = mfccp->mf6cc_ifset;
664 * Find the entry for which the upcall was made and update
667 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
668 mfccp->mf6cc_mcastgrp.sin6_addr);
669 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
670 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
671 &mfccp->mf6cc_origin.sin6_addr) &&
672 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
673 &mfccp->mf6cc_mcastgrp.sin6_addr) &&
674 (rt->mf6c_stall != NULL)) {
678 "add_m6fc: %s o %s g %s p %x dbx %p\n",
679 "multiple kernel entries",
680 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
681 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
682 mfccp->mf6cc_parent, rt->mf6c_stall);
685 if (mrt6debug & DEBUG_MFC)
687 "add_m6fc o %s g %s p %x dbg %x\n",
688 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
689 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
690 mfccp->mf6cc_parent, rt->mf6c_stall);
693 rt->mf6c_origin = mfccp->mf6cc_origin;
694 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
695 rt->mf6c_parent = mfccp->mf6cc_parent;
696 rt->mf6c_ifset = mfccp->mf6cc_ifset;
697 /* initialize pkt counters per src-grp */
698 rt->mf6c_pkt_cnt = 0;
699 rt->mf6c_byte_cnt = 0;
700 rt->mf6c_wrong_if = 0;
702 rt->mf6c_expire = 0; /* Don't clean this guy up */
705 /* free packets Qed at the end of this entry */
706 for (rte = rt->mf6c_stall; rte != NULL; ) {
707 struct rtdetq *n = rte->next;
708 ip6_mdq(rte->m, rte->ifp, rt);
712 #endif /* UPCALL_TIMING */
713 kfree(rte, M_MRTABLE);
716 rt->mf6c_stall = NULL;
721 * It is possible that an entry is being inserted without an upcall
725 if (mrt6debug & DEBUG_MFC)
726 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
728 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
729 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
730 mfccp->mf6cc_parent);
733 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
735 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
736 &mfccp->mf6cc_origin.sin6_addr)&&
737 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
738 &mfccp->mf6cc_mcastgrp.sin6_addr)) {
740 rt->mf6c_origin = mfccp->mf6cc_origin;
741 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
742 rt->mf6c_parent = mfccp->mf6cc_parent;
743 rt->mf6c_ifset = mfccp->mf6cc_ifset;
744 /* initialize pkt counters per src-grp */
745 rt->mf6c_pkt_cnt = 0;
746 rt->mf6c_byte_cnt = 0;
747 rt->mf6c_wrong_if = 0;
755 /* no upcall, so make a new entry */
756 rt = (struct mf6c *)kmalloc(sizeof(*rt), M_MRTABLE,
763 /* insert new entry at head of hash chain */
764 rt->mf6c_origin = mfccp->mf6cc_origin;
765 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
766 rt->mf6c_parent = mfccp->mf6cc_parent;
767 rt->mf6c_ifset = mfccp->mf6cc_ifset;
768 /* initialize pkt counters per src-grp */
769 rt->mf6c_pkt_cnt = 0;
770 rt->mf6c_byte_cnt = 0;
771 rt->mf6c_wrong_if = 0;
773 rt->mf6c_stall = NULL;
775 /* link into table */
776 rt->mf6c_next = mf6ctable[hash];
777 mf6ctable[hash] = rt;
786 * collect delay statistics on the upcalls
789 collate(struct timeval *t)
799 TV_DELTA(tp, *t, delta);
808 #endif /* UPCALL_TIMING */
811 * Delete an mfc entry
814 del_m6fc(struct mf6cctl *mfccp)
816 struct sockaddr_in6 origin;
817 struct sockaddr_in6 mcastgrp;
822 origin = mfccp->mf6cc_origin;
823 mcastgrp = mfccp->mf6cc_mcastgrp;
824 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
827 if (mrt6debug & DEBUG_MFC)
828 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
829 ip6_sprintf(&origin.sin6_addr),
830 ip6_sprintf(&mcastgrp.sin6_addr));
835 nptr = &mf6ctable[hash];
836 while ((rt = *nptr) != NULL) {
837 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
838 &rt->mf6c_origin.sin6_addr) &&
839 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
840 &rt->mf6c_mcastgrp.sin6_addr) &&
841 rt->mf6c_stall == NULL)
844 nptr = &rt->mf6c_next;
848 return EADDRNOTAVAIL;
851 *nptr = rt->mf6c_next;
852 kfree(rt, M_MRTABLE);
860 socket_send(struct socket *so, struct mbuf *mm, struct sockaddr_in6 *src)
863 lwkt_gettoken(&so->so_rcv.ssb_token);
864 if (ssb_appendaddr(&so->so_rcv,
865 (struct sockaddr *)src,
868 lwkt_reltoken(&so->so_rcv.ssb_token);
871 lwkt_reltoken(&so->so_rcv.ssb_token);
878 * IPv6 multicast forwarding function. This function assumes that the packet
879 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
880 * pointed to by "ifp", and the packet is to be relayed to other networks
881 * that have members of the packet's destination IPv6 multicast group.
883 * The packet is returned unscathed to the caller, unless it is
884 * erroneous, in which case a non-zero return value tells the caller to
889 ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
897 if (mrt6debug & DEBUG_FORWARD)
898 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
899 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
904 * Don't forward a packet with Hop limit of zero or one,
905 * or a packet destined to a local-only group.
907 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
908 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
913 * Source address check: do not forward packets with unspecified
914 * source. It was discussed in July 2000, on ipngwg mailing list.
915 * This is rather more serious than unicast cases, because some
916 * MLD packets can be sent with the unspecified source address
917 * (although such packets must normally set 1 to the hop limit field).
919 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
920 ip6stat.ip6s_cantforward++;
921 if (ip6_log_time + ip6_log_interval < time_uptime) {
922 ip6_log_time = time_uptime;
925 "from %s to %s nxt %d received on %s\n",
926 ip6_sprintf(&ip6->ip6_src),
927 ip6_sprintf(&ip6->ip6_dst),
929 if_name(m->m_pkthdr.rcvif));
935 * Determine forwarding mifs from the forwarding cache table
938 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
940 /* Entry exists, so forward if necessary */
943 return (ip6_mdq(m, ifp, rt));
946 * If we don't have a route for packet's origin,
947 * Make a copy of the packet &
948 * send message to routing daemon
959 #endif /* UPCALL_TIMING */
961 mrt6stat.mrt6s_no_route++;
963 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
964 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
965 ip6_sprintf(&ip6->ip6_src),
966 ip6_sprintf(&ip6->ip6_dst));
970 * Allocate mbufs early so that we don't do extra work if we
971 * are just going to fail anyway.
973 rte = (struct rtdetq *)kmalloc(sizeof(*rte), M_MRTABLE,
979 mb0 = m_copy(m, 0, M_COPYALL);
981 * Pullup packet header if needed before storing it,
982 * as other references may modify it in the meantime.
985 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
986 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
988 kfree(rte, M_MRTABLE);
993 /* is there an upcall waiting for this packet? */
994 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
995 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
996 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
997 &rt->mf6c_origin.sin6_addr) &&
998 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
999 &rt->mf6c_mcastgrp.sin6_addr) &&
1000 (rt->mf6c_stall != NULL))
1007 struct omrt6msg *oim;
1010 /* no upcall, so make a new entry */
1011 rt = (struct mf6c *)kmalloc(sizeof(*rt), M_MRTABLE,
1014 kfree(rte, M_MRTABLE);
1020 * Make a copy of the header to send to the user
1023 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1026 kfree(rte, M_MRTABLE);
1028 kfree(rt, M_MRTABLE);
1034 * Send message to routing daemon
1036 sin6.sin6_addr = ip6->ip6_src;
1042 switch (ip6_mrouter_ver) {
1045 oim = mtod(mm, struct omrt6msg *);
1046 oim->im6_msgtype = MRT6MSG_NOCACHE;
1051 im = mtod(mm, struct mrt6msg *);
1052 im->im6_msgtype = MRT6MSG_NOCACHE;
1056 kfree(rte, M_MRTABLE);
1058 kfree(rt, M_MRTABLE);
1064 if (mrt6debug & DEBUG_FORWARD)
1066 "getting the iif info in the kernel\n");
1069 for (mifp = mif6table, mifi = 0;
1070 mifi < nummifs && mifp->m6_ifp != ifp;
1074 switch (ip6_mrouter_ver) {
1077 oim->im6_mif = mifi;
1085 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1086 log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1087 "socket queue full\n");
1088 mrt6stat.mrt6s_upq_sockfull++;
1089 kfree(rte, M_MRTABLE);
1091 kfree(rt, M_MRTABLE);
1096 mrt6stat.mrt6s_upcalls++;
1098 /* insert new entry at head of hash chain */
1099 bzero(rt, sizeof(*rt));
1100 rt->mf6c_origin.sin6_family = AF_INET6;
1101 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1102 rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1103 rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1104 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1105 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1106 rt->mf6c_expire = UPCALL_EXPIRE;
1108 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1110 /* link into table */
1111 rt->mf6c_next = mf6ctable[hash];
1112 mf6ctable[hash] = rt;
1113 /* Add this entry to the end of the queue */
1114 rt->mf6c_stall = rte;
1116 /* determine if q has overflowed */
1120 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1121 if (++npkts > MAX_UPQ6) {
1122 mrt6stat.mrt6s_upq_ovflw++;
1123 kfree(rte, M_MRTABLE);
1129 /* Add this entry to the end of the queue */
1136 #ifdef UPCALL_TIMING
1138 #endif /* UPCALL_TIMING */
1147 * Clean up cache entries if upcalls are not serviced
1148 * Call from the Slow Timeout mechanism, every half second.
1151 expire_upcalls(void *unused)
1154 struct mf6c *mfc, **nptr;
1158 for (i = 0; i < MF6CTBLSIZ; i++) {
1159 if (n6expire[i] == 0)
1161 nptr = &mf6ctable[i];
1162 while ((mfc = *nptr) != NULL) {
1163 rte = mfc->mf6c_stall;
1165 * Skip real cache entries
1166 * Make sure it wasn't marked to not expire (shouldn't happen)
1170 mfc->mf6c_expire != 0 &&
1171 --mfc->mf6c_expire == 0) {
1173 if (mrt6debug & DEBUG_EXPIRE)
1174 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1175 ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1176 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1179 * drop all the packets
1180 * free the mbuf with the pkt, if, timing info
1183 struct rtdetq *n = rte->next;
1185 kfree(rte, M_MRTABLE);
1187 } while (rte != NULL);
1188 mrt6stat.mrt6s_cache_cleanups++;
1191 *nptr = mfc->mf6c_next;
1192 kfree(mfc, M_MRTABLE);
1194 nptr = &mfc->mf6c_next;
1199 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1200 expire_upcalls, NULL);
1204 * Packet forwarding routine once entry in the cache is made
1207 ip6_mdq(struct mbuf *m, struct ifnet *ifp, struct mf6c *rt)
1209 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1212 int plen = m->m_pkthdr.len;
1215 * Macro to send packet on mif. Since RSVP packets don't get counted on
1216 * input, they shouldn't get counted on output, so statistics keeping is
1220 #define MC6_SEND(ip6, mifp, m) do { \
1221 if ((mifp)->m6_flags & MIFF_REGISTER) \
1222 register_send((ip6), (mifp), (m)); \
1224 phyint_send((ip6), (mifp), (m)); \
1228 * Don't forward if it didn't arrive from the parent mif
1231 mifi = rt->mf6c_parent;
1232 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1233 /* came in the wrong interface */
1235 if (mrt6debug & DEBUG_FORWARD)
1237 "wrong if: ifid %d mifi %d mififid %x\n",
1238 ifp->if_index, mifi,
1239 mif6table[mifi].m6_ifp->if_index);
1241 mrt6stat.mrt6s_wrong_if++;
1242 rt->mf6c_wrong_if++;
1244 * If we are doing PIM processing, and we are forwarding
1245 * packets on this interface, send a message to the
1248 /* have to make sure this is a valid mif */
1249 if (mifi < nummifs && mif6table[mifi].m6_ifp)
1250 if (pim6 && (m->m_flags & M_LOOP) == 0) {
1252 * Check the M_LOOP flag to avoid an
1253 * unnecessary PIM assert.
1254 * XXX: M_LOOP is an ad-hoc hack...
1256 static struct sockaddr_in6 sin6 =
1257 { sizeof(sin6), AF_INET6 };
1262 struct omrt6msg *oim;
1265 mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1268 mm->m_len < sizeof(struct ip6_hdr)))
1269 mm = m_pullup(mm, sizeof(struct ip6_hdr));
1277 switch (ip6_mrouter_ver) {
1280 oim = mtod(mm, struct omrt6msg *);
1281 oim->im6_msgtype = MRT6MSG_WRONGMIF;
1286 im = mtod(mm, struct mrt6msg *);
1287 im->im6_msgtype = MRT6MSG_WRONGMIF;
1295 for (mifp = mif6table, iif = 0;
1296 iif < nummifs && mifp &&
1297 mifp->m6_ifp != ifp;
1301 switch (ip6_mrouter_ver) {
1305 sin6.sin6_addr = oim->im6_src;
1310 sin6.sin6_addr = im->im6_src;
1314 mrt6stat.mrt6s_upcalls++;
1316 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1319 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1321 ++mrt6stat.mrt6s_upq_sockfull;
1323 } /* if socket Q full */
1326 } /* if wrong iif */
1328 /* If I sourced this packet, it counts as output, else it was input. */
1329 if (m->m_pkthdr.rcvif == NULL) {
1330 /* XXX: is rcvif really NULL when output?? */
1331 mif6table[mifi].m6_pkt_out++;
1332 mif6table[mifi].m6_bytes_out += plen;
1334 mif6table[mifi].m6_pkt_in++;
1335 mif6table[mifi].m6_bytes_in += plen;
1338 rt->mf6c_byte_cnt += plen;
1341 * For each mif, forward a copy of the packet if there are group
1342 * members downstream on the interface.
1344 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1345 if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1347 * check if the outgoing packet is going to break
1349 * XXX For packets through PIM register tunnel
1350 * interface, we believe a routing daemon.
1352 if ((mif6table[rt->mf6c_parent].m6_flags &
1353 MIFF_REGISTER) == 0 &&
1354 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1355 (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1356 in6_addr2scopeid(mif6table[mifi].m6_ifp,
1358 in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1359 in6_addr2scopeid(mif6table[mifi].m6_ifp,
1361 ip6stat.ip6s_badscope++;
1366 mifp->m6_bytes_out += plen;
1367 MC6_SEND(ip6, mifp, m);
1373 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1375 struct mbuf *mb_copy;
1376 struct ifnet *ifp = mifp->m6_ifp;
1378 static struct route_in6 ro;
1379 struct in6_multi *in6m;
1380 struct sockaddr_in6 *dst6;
1382 crit_enter(); /* needs to protect static "ro" below. */
1385 * Make a new reference to the packet; make sure that
1386 * the IPv6 header is actually copied, not just referenced,
1387 * so that ip6_output() only scribbles on the copy.
1389 mb_copy = m_copy(m, 0, M_COPYALL);
1391 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1392 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1393 if (mb_copy == NULL) {
1397 /* set MCAST flag to the outgoing packet */
1398 mb_copy->m_flags |= M_MCAST;
1401 * If we sourced the packet, call ip6_output since we may devide
1402 * the packet into fragments when the packet is too big for the
1403 * outgoing interface.
1404 * Otherwise, we can simply send the packet to the interface
1407 if (m->m_pkthdr.rcvif == NULL) {
1408 struct ip6_moptions im6o;
1410 im6o.im6o_multicast_ifp = ifp;
1411 /* XXX: ip6_output will override ip6->ip6_hlim */
1412 im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1413 im6o.im6o_multicast_loop = 1;
1414 error = ip6_output(mb_copy, NULL, &ro,
1415 IPV6_FORWARDING, &im6o, NULL, NULL);
1418 if (mrt6debug & DEBUG_XMIT)
1419 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1420 mifp - mif6table, error);
1427 * If we belong to the destination multicast group
1428 * on the outgoing interface, loop back a copy.
1430 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1431 in6m = IN6_LOOKUP_MULTI(&ip6->ip6_dst, ifp);
1433 dst6->sin6_len = sizeof(struct sockaddr_in6);
1434 dst6->sin6_family = AF_INET6;
1435 dst6->sin6_addr = ip6->ip6_dst;
1436 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1439 * Put the packet into the sending queue of the outgoing interface
1440 * if it would fit in the MTU of the interface.
1442 if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1443 dst6->sin6_len = sizeof(struct sockaddr_in6);
1444 dst6->sin6_family = AF_INET6;
1445 dst6->sin6_addr = ip6->ip6_dst;
1447 * We just call if_output instead of nd6_output here, since
1448 * we need no ND for a multicast forwarded packet...right?
1450 error = ifp->if_output(ifp, mb_copy,
1451 (struct sockaddr *)&ro.ro_dst, NULL);
1453 if (mrt6debug & DEBUG_XMIT)
1454 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1455 mifp - mif6table, error);
1458 #ifdef MULTICAST_PMTUD
1459 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1462 if (mrt6debug & DEBUG_XMIT)
1464 "phyint_send: packet too big on %s o %s g %s"
1465 " size %d(discarded)\n",
1467 ip6_sprintf(&ip6->ip6_src),
1468 ip6_sprintf(&ip6->ip6_dst),
1469 mb_copy->m_pkthdr.len);
1470 #endif /* MRT6DEBUG */
1471 m_freem(mb_copy); /* simply discard the packet */
1479 register_send(struct ip6_hdr *ip6, struct mif6 *mif, struct mbuf *m)
1482 int i, len = m->m_pkthdr.len;
1483 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1484 struct mrt6msg *im6;
1488 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1489 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1491 ++pim6stat.pim6s_snd_registers;
1493 /* Make a copy of the packet to send to the user level process */
1494 MGETHDR(mm, M_NOWAIT, MT_HEADER);
1497 mm->m_pkthdr.rcvif = NULL;
1498 mm->m_data += max_linkhdr;
1499 mm->m_len = sizeof(struct ip6_hdr);
1501 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1505 i = MHLEN - M_LEADINGSPACE(mm);
1508 mm = m_pullup(mm, i);
1511 /* TODO: check it! */
1512 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1515 * Send message to routing daemon
1517 sin6.sin6_addr = ip6->ip6_src;
1519 im6 = mtod(mm, struct mrt6msg *);
1520 im6->im6_msgtype = MRT6MSG_WHOLEPKT;
1523 im6->im6_mif = mif - mif6table;
1525 /* iif info is not given for reg. encap.n */
1526 mrt6stat.mrt6s_upcalls++;
1528 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1532 "register_send: ip6_mrouter socket queue full\n");
1534 ++mrt6stat.mrt6s_upq_sockfull;
1541 * PIM sparse mode hook
1542 * Receives the pim control messages, and passes them up to the listening
1543 * socket, using rip6_input.
1544 * The only message processed is the REGISTER pim message; the pim header
1545 * is stripped off, and the inner packet is passed to register_mforward.
1548 pim6_input(struct mbuf **mp, int *offp, int proto)
1550 struct pim *pim; /* pointer to a pim struct */
1551 struct ip6_hdr *ip6;
1553 struct mbuf *m = *mp;
1557 ++pim6stat.pim6s_rcv_total;
1559 ip6 = mtod(m, struct ip6_hdr *);
1560 pimlen = m->m_pkthdr.len - *offp;
1565 if (pimlen < PIM_MINLEN) {
1566 ++pim6stat.pim6s_rcv_tooshort;
1568 if (mrt6debug & DEBUG_PIM)
1569 log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1572 return (IPPROTO_DONE);
1576 * if the packet is at least as big as a REGISTER, go ahead
1577 * and grab the PIM REGISTER header size, to avoid another
1578 * possible m_pullup() later.
1580 * PIM_MINLEN == pimhdr + u_int32 == 8
1581 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1583 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1586 * Make sure that the IP6 and PIM headers in contiguous memory, and
1587 * possibly the PIM REGISTER header
1589 #ifndef PULLDOWN_TEST
1590 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1591 /* adjust pointer */
1592 ip6 = mtod(m, struct ip6_hdr *);
1594 /* adjust mbuf to point to the PIM header */
1595 pim = (struct pim *)((caddr_t)ip6 + off);
1597 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1599 pim6stat.pim6s_rcv_tooshort++;
1600 return IPPROTO_DONE;
1604 #define PIM6_CHECKSUM
1605 #ifdef PIM6_CHECKSUM
1610 * Validate checksum.
1611 * If PIM REGISTER, exclude the data packet
1613 if (pim->pim_type == PIM_REGISTER)
1614 cksumlen = PIM_MINLEN;
1618 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1619 ++pim6stat.pim6s_rcv_badsum;
1621 if (mrt6debug & DEBUG_PIM)
1623 "pim6_input: invalid checksum\n");
1626 return (IPPROTO_DONE);
1629 #endif /* PIM_CHECKSUM */
1631 /* PIM version check */
1632 if (pim->pim_ver != PIM_VERSION) {
1633 ++pim6stat.pim6s_rcv_badversion;
1636 "pim6_input: incorrect version %d, expecting %d\n",
1637 pim->pim_ver, PIM_VERSION);
1640 return (IPPROTO_DONE);
1643 if (pim->pim_type == PIM_REGISTER) {
1645 * since this is a REGISTER, we'll make a copy of the register
1646 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1649 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1652 struct ip6_hdr *eip6;
1655 ++pim6stat.pim6s_rcv_registers;
1657 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1659 if (mrt6debug & DEBUG_PIM)
1661 "pim6_input: register mif not set: %d\n",
1665 return (IPPROTO_DONE);
1668 reghdr = (u_int32_t *)(pim + 1);
1670 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1671 goto pim6_input_to_daemon;
1676 if (pimlen < PIM6_REG_MINLEN) {
1677 ++pim6stat.pim6s_rcv_tooshort;
1678 ++pim6stat.pim6s_rcv_badregisters;
1681 "pim6_input: register packet size too "
1682 "small %d from %s\n",
1683 pimlen, ip6_sprintf(&ip6->ip6_src));
1686 return (IPPROTO_DONE);
1689 eip6 = (struct ip6_hdr *) (reghdr + 1);
1691 if (mrt6debug & DEBUG_PIM)
1693 "pim6_input[register], eip6: %s -> %s, "
1695 ip6_sprintf(&eip6->ip6_src),
1696 ip6_sprintf(&eip6->ip6_dst),
1697 ntohs(eip6->ip6_plen));
1700 /* verify the version number of the inner packet */
1701 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1702 ++pim6stat.pim6s_rcv_badregisters;
1704 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1705 "of the inner packet\n",
1706 (eip6->ip6_vfc & IPV6_VERSION));
1709 return (IPPROTO_NONE);
1712 /* verify the inner packet is destined to a mcast group */
1713 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1714 ++pim6stat.pim6s_rcv_badregisters;
1716 if (mrt6debug & DEBUG_PIM)
1718 "pim6_input: inner packet of register "
1719 "is not multicast %s\n",
1720 ip6_sprintf(&eip6->ip6_dst));
1723 return (IPPROTO_DONE);
1727 * make a copy of the whole header to pass to the daemon later.
1729 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1733 "pim6_input: pim register: "
1734 "could not copy register head\n");
1737 return (IPPROTO_DONE);
1741 * forward the inner ip6 packet; point m_data at the inner ip6.
1743 m_adj(m, off + PIM_MINLEN);
1745 if (mrt6debug & DEBUG_PIM) {
1747 "pim6_input: forwarding decapsulated register: "
1748 "src %s, dst %s, mif %d\n",
1749 ip6_sprintf(&eip6->ip6_src),
1750 ip6_sprintf(&eip6->ip6_dst),
1755 if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1756 dst.sin6_family, 0);
1758 /* prepare the register head to send to the mrouting daemon */
1763 * Pass the PIM message up to the daemon; if it is a register message
1764 * pass the 'head' only up to the daemon. This includes the
1765 * encapsulator ip6 header, pim header, register header and the
1766 * encapsulated ip6 header.
1768 pim6_input_to_daemon:
1769 rip6_input(&m, offp, proto);
1770 return (IPPROTO_DONE);