1 /* $FreeBSD: src/sys/netinet6/ip6_mroute.c,v 1.2.2.9 2003/01/23 21:06:47 sam Exp $ */
2 /* $DragonFly: src/sys/netinet6/ip6_mroute.c,v 1.5 2004/01/08 18:39:18 asmodai Exp $ */
3 /* $KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $ */
6 * Copyright (C) 1998 WIDE Project.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the name of the project nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 /* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */
37 * IP multicast forwarding procedures
39 * Written by David Waitzman, BBN Labs, August 1988.
40 * Modified by Steve Deering, Stanford, February 1989.
41 * Modified by Mark J. Steiglitz, Stanford, May, 1991
42 * Modified by Van Jacobson, LBL, January 1993
43 * Modified by Ajit Thyagarajan, PARC, August 1993
44 * Modified by Bill Fenenr, PARC, April 1994
46 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
50 #include "opt_inet6.h"
52 #include <sys/param.h>
53 #include <sys/systm.h>
54 #include <sys/callout.h>
55 #include <sys/malloc.h>
57 #include <sys/socket.h>
58 #include <sys/socketvar.h>
59 #include <sys/sockio.h>
60 #include <sys/protosw.h>
61 #include <sys/errno.h>
63 #include <sys/kernel.h>
64 #include <sys/syslog.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(so, sopt)
228 struct sockopt *sopt;
233 if (so != ip6_mrouter && sopt->sopt_name != MRT6_INIT)
236 if ((error = soopt_getm(sopt, &m)) != 0) /* XXX */
238 if ((error = soopt_mcopyin(sopt, m)) != 0) /* XXX */
241 switch (sopt->sopt_name) {
246 error = ip6_mrouter_init(so, m, sopt->sopt_name);
249 error = ip6_mrouter_done();
252 error = add_m6if(mtod(m, struct mif6ctl *));
255 error = del_m6if(mtod(m, mifi_t *));
258 error = add_m6fc(mtod(m, struct mf6cctl *));
261 error = del_m6fc(mtod(m, struct mf6cctl *));
264 error = set_pim6(mtod(m, int *));
276 * Handle MRT getsockopt commands
279 ip6_mrouter_get(so, sopt)
281 struct sockopt *sopt;
285 if (so != ip6_mrouter) return EACCES;
287 switch (sopt->sopt_name) {
289 error = sooptcopyout(sopt, &pim6, sizeof(pim6));
296 * Handle ioctl commands to obtain information from the cache
299 mrt6_ioctl(cmd, data)
306 case SIOCGETSGCNT_IN6:
307 return(get_sg_cnt((struct sioc_sg_req6 *)data));
308 break; /* for safety */
309 case SIOCGETMIFCNT_IN6:
310 return(get_mif6_cnt((struct sioc_mif_req6 *)data));
311 break; /* for safety */
320 * returns the packet, byte, rpf-failure count for the source group provided
324 struct sioc_sg_req6 *req;
330 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
333 req->pktcnt = rt->mf6c_pkt_cnt;
334 req->bytecnt = rt->mf6c_byte_cnt;
335 req->wrong_if = rt->mf6c_wrong_if;
339 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
346 * returns the input and output packet and byte counts on the mif provided
350 struct sioc_mif_req6 *req;
352 mifi_t mifi = req->mifi;
357 req->icount = mif6table[mifi].m6_pkt_in;
358 req->ocount = mif6table[mifi].m6_pkt_out;
359 req->ibytes = mif6table[mifi].m6_bytes_in;
360 req->obytes = mif6table[mifi].m6_bytes_out;
369 if ((*i != 1) && (*i != 0))
378 * Enable multicast routing
381 ip6_mrouter_init(so, m, cmd)
391 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
392 so->so_type, so->so_proto->pr_protocol);
395 if (so->so_type != SOCK_RAW ||
396 so->so_proto->pr_protocol != IPPROTO_ICMPV6)
399 if (!m || (m->m_len != sizeof(int *)))
406 if (ip6_mrouter != NULL) return EADDRINUSE;
409 ip6_mrouter_ver = cmd;
411 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
412 bzero((caddr_t)n6expire, sizeof(n6expire));
414 pim6 = 0;/* used for stubbing out/in pim stuff */
416 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
417 expire_upcalls, NULL);
421 log(LOG_DEBUG, "ip6_mrouter_init\n");
428 * Disable multicast routing
436 struct in6_ifreq ifr;
444 * For each phyint in use, disable promiscuous reception of all IPv6
450 * If there is still IPv4 multicast routing daemon,
451 * we remain interfaces to receive all muliticasted packets.
452 * XXX: there may be an interface in which the IPv4 multicast
453 * daemon is not interested...
459 for (mifi = 0; mifi < nummifs; mifi++) {
460 if (mif6table[mifi].m6_ifp &&
461 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
462 ifr.ifr_addr.sin6_family = AF_INET6;
463 ifr.ifr_addr.sin6_addr= in6addr_any;
464 ifp = mif6table[mifi].m6_ifp;
465 (*ifp->if_ioctl)(ifp, SIOCDELMULTI,
471 bzero((caddr_t)qtable, sizeof(qtable));
472 bzero((caddr_t)tbftable, sizeof(tbftable));
474 bzero((caddr_t)mif6table, sizeof(mif6table));
477 pim6 = 0; /* used to stub out/in pim specific code */
479 callout_stop(&expire_upcalls_ch);
482 * Free all multicast forwarding cache entries.
484 for (i = 0; i < MF6CTBLSIZ; i++) {
489 for (rte = rt->mf6c_stall; rte != NULL; ) {
490 struct rtdetq *n = rte->next;
493 free(rte, M_MRTABLE);
498 free(frt, M_MRTABLE);
502 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
505 * Reset de-encapsulation cache
516 log(LOG_DEBUG, "ip6_mrouter_done\n");
522 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
525 * Add a mif to the mif table
529 struct mif6ctl *mifcp;
535 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
538 if (mifcp->mif6c_mifi >= MAXMIFS)
540 mifp = mif6table + mifcp->mif6c_mifi;
542 return EADDRINUSE; /* XXX: is it appropriate? */
543 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
545 ifp = ifindex2ifnet[mifcp->mif6c_pifi];
547 if (mifcp->mif6c_flags & MIFF_REGISTER) {
548 if (reg_mif_num == (mifi_t)-1) {
549 strlcpy(multicast_register_if.if_xname, "register_mif",
551 multicast_register_if.if_flags |= IFF_LOOPBACK;
552 multicast_register_if.if_index = mifcp->mif6c_mifi;
553 reg_mif_num = mifcp->mif6c_mifi;
556 ifp = &multicast_register_if;
560 /* Make sure the interface supports multicast */
561 if ((ifp->if_flags & IFF_MULTICAST) == 0)
565 error = if_allmulti(ifp, 1);
572 mifp->m6_flags = mifcp->mif6c_flags;
575 /* scaling up here allows division by 1024 in critical code */
576 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
578 /* initialize per mif pkt counters */
580 mifp->m6_pkt_out = 0;
581 mifp->m6_bytes_in = 0;
582 mifp->m6_bytes_out = 0;
585 /* Adjust nummifs up if the mifi is higher than nummifs */
586 if (nummifs <= mifcp->mif6c_mifi)
587 nummifs = mifcp->mif6c_mifi + 1;
592 "add_mif #%d, phyint %s\n",
601 * Delete a mif from the mif table
607 struct mif6 *mifp = mif6table + *mifip;
612 if (*mifip >= nummifs)
614 if (mifp->m6_ifp == NULL)
619 if (!(mifp->m6_flags & MIFF_REGISTER)) {
621 * XXX: what if there is yet IPv4 multicast daemon
622 * using the interface?
630 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
631 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
633 bzero((caddr_t)mifp, sizeof (*mifp));
635 /* Adjust nummifs down */
636 for (mifi = nummifs; mifi > 0; mifi--)
637 if (mif6table[mifi - 1].m6_ifp)
645 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
656 struct mf6cctl *mfccp;
664 MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
665 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
667 /* If an entry already exists, just update the fields */
670 if (mrt6debug & DEBUG_MFC)
672 "add_m6fc no upcall h %d o %s g %s p %x\n",
673 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
674 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
675 mfccp->mf6cc_parent);
679 rt->mf6c_parent = mfccp->mf6cc_parent;
680 rt->mf6c_ifset = mfccp->mf6cc_ifset;
686 * Find the entry for which the upcall was made and update
689 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
690 mfccp->mf6cc_mcastgrp.sin6_addr);
691 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
692 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
693 &mfccp->mf6cc_origin.sin6_addr) &&
694 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
695 &mfccp->mf6cc_mcastgrp.sin6_addr) &&
696 (rt->mf6c_stall != NULL)) {
700 "add_m6fc: %s o %s g %s p %x dbx %p\n",
701 "multiple kernel entries",
702 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
703 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
704 mfccp->mf6cc_parent, rt->mf6c_stall);
707 if (mrt6debug & DEBUG_MFC)
709 "add_m6fc o %s g %s p %x dbg %x\n",
710 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
711 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
712 mfccp->mf6cc_parent, rt->mf6c_stall);
715 rt->mf6c_origin = mfccp->mf6cc_origin;
716 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
717 rt->mf6c_parent = mfccp->mf6cc_parent;
718 rt->mf6c_ifset = mfccp->mf6cc_ifset;
719 /* initialize pkt counters per src-grp */
720 rt->mf6c_pkt_cnt = 0;
721 rt->mf6c_byte_cnt = 0;
722 rt->mf6c_wrong_if = 0;
724 rt->mf6c_expire = 0; /* Don't clean this guy up */
727 /* free packets Qed at the end of this entry */
728 for (rte = rt->mf6c_stall; rte != NULL; ) {
729 struct rtdetq *n = rte->next;
730 ip6_mdq(rte->m, rte->ifp, rt);
734 #endif /* UPCALL_TIMING */
735 free(rte, M_MRTABLE);
738 rt->mf6c_stall = NULL;
743 * It is possible that an entry is being inserted without an upcall
747 if (mrt6debug & DEBUG_MFC)
748 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
750 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
751 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
752 mfccp->mf6cc_parent);
755 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
757 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
758 &mfccp->mf6cc_origin.sin6_addr)&&
759 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
760 &mfccp->mf6cc_mcastgrp.sin6_addr)) {
762 rt->mf6c_origin = mfccp->mf6cc_origin;
763 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
764 rt->mf6c_parent = mfccp->mf6cc_parent;
765 rt->mf6c_ifset = mfccp->mf6cc_ifset;
766 /* initialize pkt counters per src-grp */
767 rt->mf6c_pkt_cnt = 0;
768 rt->mf6c_byte_cnt = 0;
769 rt->mf6c_wrong_if = 0;
777 /* no upcall, so make a new entry */
778 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
785 /* insert new entry at head of hash chain */
786 rt->mf6c_origin = mfccp->mf6cc_origin;
787 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
788 rt->mf6c_parent = mfccp->mf6cc_parent;
789 rt->mf6c_ifset = mfccp->mf6cc_ifset;
790 /* initialize pkt counters per src-grp */
791 rt->mf6c_pkt_cnt = 0;
792 rt->mf6c_byte_cnt = 0;
793 rt->mf6c_wrong_if = 0;
795 rt->mf6c_stall = NULL;
797 /* link into table */
798 rt->mf6c_next = mf6ctable[hash];
799 mf6ctable[hash] = rt;
808 * collect delay statistics on the upcalls
822 TV_DELTA(tp, *t, delta);
831 #endif /* UPCALL_TIMING */
834 * Delete an mfc entry
838 struct mf6cctl *mfccp;
840 struct sockaddr_in6 origin;
841 struct sockaddr_in6 mcastgrp;
847 origin = mfccp->mf6cc_origin;
848 mcastgrp = mfccp->mf6cc_mcastgrp;
849 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
852 if (mrt6debug & DEBUG_MFC)
853 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
854 ip6_sprintf(&origin.sin6_addr),
855 ip6_sprintf(&mcastgrp.sin6_addr));
860 nptr = &mf6ctable[hash];
861 while ((rt = *nptr) != NULL) {
862 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
863 &rt->mf6c_origin.sin6_addr) &&
864 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
865 &rt->mf6c_mcastgrp.sin6_addr) &&
866 rt->mf6c_stall == NULL)
869 nptr = &rt->mf6c_next;
873 return EADDRNOTAVAIL;
876 *nptr = rt->mf6c_next;
885 socket_send(s, mm, src)
888 struct sockaddr_in6 *src;
891 if (sbappendaddr(&s->so_rcv,
892 (struct sockaddr *)src,
893 mm, (struct mbuf *)0) != 0) {
903 * IPv6 multicast forwarding function. This function assumes that the packet
904 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
905 * pointed to by "ifp", and the packet is to be relayed to other networks
906 * that have members of the packet's destination IPv6 multicast group.
908 * The packet is returned unscathed to the caller, unless it is
909 * erroneous, in which case a non-zero return value tells the caller to
914 ip6_mforward(ip6, ifp, m)
926 if (mrt6debug & DEBUG_FORWARD)
927 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
928 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
933 * Don't forward a packet with Hop limit of zero or one,
934 * or a packet destined to a local-only group.
936 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
937 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
942 * Source address check: do not forward packets with unspecified
943 * source. It was discussed in July 2000, on ipngwg mailing list.
944 * This is rather more serious than unicast cases, because some
945 * MLD packets can be sent with the unspecified source address
946 * (although such packets must normally set 1 to the hop limit field).
948 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
949 ip6stat.ip6s_cantforward++;
950 if (ip6_log_time + ip6_log_interval < time_second) {
951 ip6_log_time = time_second;
954 "from %s to %s nxt %d received on %s\n",
955 ip6_sprintf(&ip6->ip6_src),
956 ip6_sprintf(&ip6->ip6_dst),
958 if_name(m->m_pkthdr.rcvif));
964 * Determine forwarding mifs from the forwarding cache table
967 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
969 /* Entry exists, so forward if necessary */
972 return (ip6_mdq(m, ifp, rt));
975 * If we don't have a route for packet's origin,
976 * Make a copy of the packet &
977 * send message to routing daemon
988 #endif /* UPCALL_TIMING */
990 mrt6stat.mrt6s_no_route++;
992 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
993 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
994 ip6_sprintf(&ip6->ip6_src),
995 ip6_sprintf(&ip6->ip6_dst));
999 * Allocate mbufs early so that we don't do extra work if we
1000 * are just going to fail anyway.
1002 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1008 mb0 = m_copy(m, 0, M_COPYALL);
1010 * Pullup packet header if needed before storing it,
1011 * as other references may modify it in the meantime.
1014 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1015 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1017 free(rte, M_MRTABLE);
1022 /* is there an upcall waiting for this packet? */
1023 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1024 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1025 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1026 &rt->mf6c_origin.sin6_addr) &&
1027 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1028 &rt->mf6c_mcastgrp.sin6_addr) &&
1029 (rt->mf6c_stall != NULL))
1036 struct omrt6msg *oim;
1039 /* no upcall, so make a new entry */
1040 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1043 free(rte, M_MRTABLE);
1049 * Make a copy of the header to send to the user
1052 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1055 free(rte, M_MRTABLE);
1057 free(rt, M_MRTABLE);
1063 * Send message to routing daemon
1065 sin6.sin6_addr = ip6->ip6_src;
1071 switch (ip6_mrouter_ver) {
1074 oim = mtod(mm, struct omrt6msg *);
1075 oim->im6_msgtype = MRT6MSG_NOCACHE;
1080 im = mtod(mm, struct mrt6msg *);
1081 im->im6_msgtype = MRT6MSG_NOCACHE;
1085 free(rte, M_MRTABLE);
1087 free(rt, M_MRTABLE);
1093 if (mrt6debug & DEBUG_FORWARD)
1095 "getting the iif info in the kernel\n");
1098 for (mifp = mif6table, mifi = 0;
1099 mifi < nummifs && mifp->m6_ifp != ifp;
1103 switch (ip6_mrouter_ver) {
1106 oim->im6_mif = mifi;
1114 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1115 log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1116 "socket queue full\n");
1117 mrt6stat.mrt6s_upq_sockfull++;
1118 free(rte, M_MRTABLE);
1120 free(rt, M_MRTABLE);
1125 mrt6stat.mrt6s_upcalls++;
1127 /* insert new entry at head of hash chain */
1128 bzero(rt, sizeof(*rt));
1129 rt->mf6c_origin.sin6_family = AF_INET6;
1130 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1131 rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1132 rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1133 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1134 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1135 rt->mf6c_expire = UPCALL_EXPIRE;
1137 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1139 /* link into table */
1140 rt->mf6c_next = mf6ctable[hash];
1141 mf6ctable[hash] = rt;
1142 /* Add this entry to the end of the queue */
1143 rt->mf6c_stall = rte;
1145 /* determine if q has overflowed */
1149 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1150 if (++npkts > MAX_UPQ6) {
1151 mrt6stat.mrt6s_upq_ovflw++;
1152 free(rte, M_MRTABLE);
1158 /* Add this entry to the end of the queue */
1165 #ifdef UPCALL_TIMING
1167 #endif /* UPCALL_TIMING */
1176 * Clean up cache entries if upcalls are not serviced
1177 * Call from the Slow Timeout mechanism, every half second.
1180 expire_upcalls(unused)
1184 struct mf6c *mfc, **nptr;
1189 for (i = 0; i < MF6CTBLSIZ; i++) {
1190 if (n6expire[i] == 0)
1192 nptr = &mf6ctable[i];
1193 while ((mfc = *nptr) != NULL) {
1194 rte = mfc->mf6c_stall;
1196 * Skip real cache entries
1197 * Make sure it wasn't marked to not expire (shouldn't happen)
1201 mfc->mf6c_expire != 0 &&
1202 --mfc->mf6c_expire == 0) {
1204 if (mrt6debug & DEBUG_EXPIRE)
1205 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1206 ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1207 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1210 * drop all the packets
1211 * free the mbuf with the pkt, if, timing info
1214 struct rtdetq *n = rte->next;
1216 free(rte, M_MRTABLE);
1218 } while (rte != NULL);
1219 mrt6stat.mrt6s_cache_cleanups++;
1222 *nptr = mfc->mf6c_next;
1223 free(mfc, M_MRTABLE);
1225 nptr = &mfc->mf6c_next;
1230 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1231 expire_upcalls, NULL);
1235 * Packet forwarding routine once entry in the cache is made
1243 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1246 int plen = m->m_pkthdr.len;
1249 * Macro to send packet on mif. Since RSVP packets don't get counted on
1250 * input, they shouldn't get counted on output, so statistics keeping is
1254 #define MC6_SEND(ip6, mifp, m) do { \
1255 if ((mifp)->m6_flags & MIFF_REGISTER) \
1256 register_send((ip6), (mifp), (m)); \
1258 phyint_send((ip6), (mifp), (m)); \
1262 * Don't forward if it didn't arrive from the parent mif
1265 mifi = rt->mf6c_parent;
1266 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1267 /* came in the wrong interface */
1269 if (mrt6debug & DEBUG_FORWARD)
1271 "wrong if: ifid %d mifi %d mififid %x\n",
1272 ifp->if_index, mifi,
1273 mif6table[mifi].m6_ifp->if_index);
1275 mrt6stat.mrt6s_wrong_if++;
1276 rt->mf6c_wrong_if++;
1278 * If we are doing PIM processing, and we are forwarding
1279 * packets on this interface, send a message to the
1282 /* have to make sure this is a valid mif */
1283 if (mifi < nummifs && mif6table[mifi].m6_ifp)
1284 if (pim6 && (m->m_flags & M_LOOP) == 0) {
1286 * Check the M_LOOP flag to avoid an
1287 * unnecessary PIM assert.
1288 * XXX: M_LOOP is an ad-hoc hack...
1290 static struct sockaddr_in6 sin6 =
1291 { sizeof(sin6), AF_INET6 };
1296 struct omrt6msg *oim;
1299 mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1302 mm->m_len < sizeof(struct ip6_hdr)))
1303 mm = m_pullup(mm, sizeof(struct ip6_hdr));
1311 switch (ip6_mrouter_ver) {
1314 oim = mtod(mm, struct omrt6msg *);
1315 oim->im6_msgtype = MRT6MSG_WRONGMIF;
1320 im = mtod(mm, struct mrt6msg *);
1321 im->im6_msgtype = MRT6MSG_WRONGMIF;
1329 for (mifp = mif6table, iif = 0;
1330 iif < nummifs && mifp &&
1331 mifp->m6_ifp != ifp;
1335 switch (ip6_mrouter_ver) {
1339 sin6.sin6_addr = oim->im6_src;
1344 sin6.sin6_addr = im->im6_src;
1348 mrt6stat.mrt6s_upcalls++;
1350 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1353 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1355 ++mrt6stat.mrt6s_upq_sockfull;
1357 } /* if socket Q full */
1360 } /* if wrong iif */
1362 /* If I sourced this packet, it counts as output, else it was input. */
1363 if (m->m_pkthdr.rcvif == NULL) {
1364 /* XXX: is rcvif really NULL when output?? */
1365 mif6table[mifi].m6_pkt_out++;
1366 mif6table[mifi].m6_bytes_out += plen;
1368 mif6table[mifi].m6_pkt_in++;
1369 mif6table[mifi].m6_bytes_in += plen;
1372 rt->mf6c_byte_cnt += plen;
1375 * For each mif, forward a copy of the packet if there are group
1376 * members downstream on the interface.
1378 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1379 if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1381 * check if the outgoing packet is going to break
1383 * XXX For packets through PIM register tunnel
1384 * interface, we believe a routing daemon.
1386 if ((mif6table[rt->mf6c_parent].m6_flags &
1387 MIFF_REGISTER) == 0 &&
1388 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1389 (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1390 in6_addr2scopeid(mif6table[mifi].m6_ifp,
1392 in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1393 in6_addr2scopeid(mif6table[mifi].m6_ifp,
1395 ip6stat.ip6s_badscope++;
1400 mifp->m6_bytes_out += plen;
1401 MC6_SEND(ip6, mifp, m);
1407 phyint_send(ip6, mifp, m)
1408 struct ip6_hdr *ip6;
1412 struct mbuf *mb_copy;
1413 struct ifnet *ifp = mifp->m6_ifp;
1415 int s = splnet(); /* needs to protect static "ro" below. */
1416 static struct route_in6 ro;
1417 struct in6_multi *in6m;
1418 struct sockaddr_in6 *dst6;
1421 * Make a new reference to the packet; make sure that
1422 * the IPv6 header is actually copied, not just referenced,
1423 * so that ip6_output() only scribbles on the copy.
1425 mb_copy = m_copy(m, 0, M_COPYALL);
1427 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1428 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1429 if (mb_copy == NULL) {
1433 /* set MCAST flag to the outgoing packet */
1434 mb_copy->m_flags |= M_MCAST;
1437 * If we sourced the packet, call ip6_output since we may devide
1438 * the packet into fragments when the packet is too big for the
1439 * outgoing interface.
1440 * Otherwise, we can simply send the packet to the interface
1443 if (m->m_pkthdr.rcvif == NULL) {
1444 struct ip6_moptions im6o;
1446 im6o.im6o_multicast_ifp = ifp;
1447 /* XXX: ip6_output will override ip6->ip6_hlim */
1448 im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1449 im6o.im6o_multicast_loop = 1;
1450 error = ip6_output(mb_copy, NULL, &ro,
1451 IPV6_FORWARDING, &im6o, NULL, NULL);
1454 if (mrt6debug & DEBUG_XMIT)
1455 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1456 mifp - mif6table, error);
1463 * If we belong to the destination multicast group
1464 * on the outgoing interface, loop back a copy.
1466 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1467 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1469 dst6->sin6_len = sizeof(struct sockaddr_in6);
1470 dst6->sin6_family = AF_INET6;
1471 dst6->sin6_addr = ip6->ip6_dst;
1472 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1475 * Put the packet into the sending queue of the outgoing interface
1476 * if it would fit in the MTU of the interface.
1478 if (mb_copy->m_pkthdr.len <= ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1479 dst6->sin6_len = sizeof(struct sockaddr_in6);
1480 dst6->sin6_family = AF_INET6;
1481 dst6->sin6_addr = ip6->ip6_dst;
1483 * We just call if_output instead of nd6_output here, since
1484 * we need no ND for a multicast forwarded packet...right?
1486 error = (*ifp->if_output)(ifp, mb_copy,
1487 (struct sockaddr *)&ro.ro_dst, NULL);
1489 if (mrt6debug & DEBUG_XMIT)
1490 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1491 mifp - mif6table, error);
1494 #ifdef MULTICAST_PMTUD
1495 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1498 if (mrt6debug & DEBUG_XMIT)
1500 "phyint_send: packet too big on %s o %s g %s"
1501 " size %d(discarded)\n",
1503 ip6_sprintf(&ip6->ip6_src),
1504 ip6_sprintf(&ip6->ip6_dst),
1505 mb_copy->m_pkthdr.len);
1506 #endif /* MRT6DEBUG */
1507 m_freem(mb_copy); /* simply discard the packet */
1515 register_send(ip6, mif, m)
1516 struct ip6_hdr *ip6;
1521 int i, len = m->m_pkthdr.len;
1522 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1523 struct mrt6msg *im6;
1527 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1528 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1530 ++pim6stat.pim6s_snd_registers;
1532 /* Make a copy of the packet to send to the user level process */
1533 MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1536 mm->m_pkthdr.rcvif = NULL;
1537 mm->m_data += max_linkhdr;
1538 mm->m_len = sizeof(struct ip6_hdr);
1540 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1544 i = MHLEN - M_LEADINGSPACE(mm);
1547 mm = m_pullup(mm, i);
1550 /* TODO: check it! */
1551 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1554 * Send message to routing daemon
1556 sin6.sin6_addr = ip6->ip6_src;
1558 im6 = mtod(mm, struct mrt6msg *);
1559 im6->im6_msgtype = MRT6MSG_WHOLEPKT;
1562 im6->im6_mif = mif - mif6table;
1564 /* iif info is not given for reg. encap.n */
1565 mrt6stat.mrt6s_upcalls++;
1567 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1571 "register_send: ip6_mrouter socket queue full\n");
1573 ++mrt6stat.mrt6s_upq_sockfull;
1580 * PIM sparse mode hook
1581 * Receives the pim control messages, and passes them up to the listening
1582 * socket, using rip6_input.
1583 * The only message processed is the REGISTER pim message; the pim header
1584 * is stripped off, and the inner packet is passed to register_mforward.
1587 pim6_input(mp, offp, proto)
1591 struct pim *pim; /* pointer to a pim struct */
1592 struct ip6_hdr *ip6;
1594 struct mbuf *m = *mp;
1598 ++pim6stat.pim6s_rcv_total;
1600 ip6 = mtod(m, struct ip6_hdr *);
1601 pimlen = m->m_pkthdr.len - *offp;
1606 if (pimlen < PIM_MINLEN) {
1607 ++pim6stat.pim6s_rcv_tooshort;
1609 if (mrt6debug & DEBUG_PIM)
1610 log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1613 return(IPPROTO_DONE);
1617 * if the packet is at least as big as a REGISTER, go ahead
1618 * and grab the PIM REGISTER header size, to avoid another
1619 * possible m_pullup() later.
1621 * PIM_MINLEN == pimhdr + u_int32 == 8
1622 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1624 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1627 * Make sure that the IP6 and PIM headers in contiguous memory, and
1628 * possibly the PIM REGISTER header
1630 #ifndef PULLDOWN_TEST
1631 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1632 /* adjust pointer */
1633 ip6 = mtod(m, struct ip6_hdr *);
1635 /* adjust mbuf to point to the PIM header */
1636 pim = (struct pim *)((caddr_t)ip6 + off);
1638 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1640 pim6stat.pim6s_rcv_tooshort++;
1641 return IPPROTO_DONE;
1645 #define PIM6_CHECKSUM
1646 #ifdef PIM6_CHECKSUM
1651 * Validate checksum.
1652 * If PIM REGISTER, exclude the data packet
1654 if (pim->pim_type == PIM_REGISTER)
1655 cksumlen = PIM_MINLEN;
1659 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1660 ++pim6stat.pim6s_rcv_badsum;
1662 if (mrt6debug & DEBUG_PIM)
1664 "pim6_input: invalid checksum\n");
1667 return(IPPROTO_DONE);
1670 #endif /* PIM_CHECKSUM */
1672 /* PIM version check */
1673 if (pim->pim_ver != PIM_VERSION) {
1674 ++pim6stat.pim6s_rcv_badversion;
1677 "pim6_input: incorrect version %d, expecting %d\n",
1678 pim->pim_ver, PIM_VERSION);
1681 return(IPPROTO_DONE);
1684 if (pim->pim_type == PIM_REGISTER) {
1686 * since this is a REGISTER, we'll make a copy of the register
1687 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1690 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1693 struct ip6_hdr *eip6;
1697 ++pim6stat.pim6s_rcv_registers;
1699 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1701 if (mrt6debug & DEBUG_PIM)
1703 "pim6_input: register mif not set: %d\n",
1707 return(IPPROTO_DONE);
1710 reghdr = (u_int32_t *)(pim + 1);
1712 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1713 goto pim6_input_to_daemon;
1718 if (pimlen < PIM6_REG_MINLEN) {
1719 ++pim6stat.pim6s_rcv_tooshort;
1720 ++pim6stat.pim6s_rcv_badregisters;
1723 "pim6_input: register packet size too "
1724 "small %d from %s\n",
1725 pimlen, ip6_sprintf(&ip6->ip6_src));
1728 return(IPPROTO_DONE);
1731 eip6 = (struct ip6_hdr *) (reghdr + 1);
1733 if (mrt6debug & DEBUG_PIM)
1735 "pim6_input[register], eip6: %s -> %s, "
1737 ip6_sprintf(&eip6->ip6_src),
1738 ip6_sprintf(&eip6->ip6_dst),
1739 ntohs(eip6->ip6_plen));
1742 /* verify the version number of the inner packet */
1743 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1744 ++pim6stat.pim6s_rcv_badregisters;
1746 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1747 "of the inner packet\n",
1748 (eip6->ip6_vfc & IPV6_VERSION));
1751 return(IPPROTO_NONE);
1754 /* verify the inner packet is destined to a mcast group */
1755 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1756 ++pim6stat.pim6s_rcv_badregisters;
1758 if (mrt6debug & DEBUG_PIM)
1760 "pim6_input: inner packet of register "
1761 "is not multicast %s\n",
1762 ip6_sprintf(&eip6->ip6_dst));
1765 return(IPPROTO_DONE);
1769 * make a copy of the whole header to pass to the daemon later.
1771 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1775 "pim6_input: pim register: "
1776 "could not copy register head\n");
1779 return(IPPROTO_DONE);
1783 * forward the inner ip6 packet; point m_data at the inner ip6.
1785 m_adj(m, off + PIM_MINLEN);
1787 if (mrt6debug & DEBUG_PIM) {
1789 "pim6_input: forwarding decapsulated register: "
1790 "src %s, dst %s, mif %d\n",
1791 ip6_sprintf(&eip6->ip6_src),
1792 ip6_sprintf(&eip6->ip6_dst),
1797 rc = if_simloop(mif6table[reg_mif_num].m6_ifp, m,
1798 dst.sin6_family, NULL);
1800 /* prepare the register head to send to the mrouting daemon */
1805 * Pass the PIM message up to the daemon; if it is a register message
1806 * pass the 'head' only up to the daemon. This includes the
1807 * encapsulator ip6 header, pim header, register header and the
1808 * encapsulated ip6 header.
1810 pim6_input_to_daemon:
1811 rip6_input(&m, offp, proto);
1812 return(IPPROTO_DONE);