ip6_mroute.c revision 1.16 1 /* $NetBSD: ip6_mroute.c,v 1.16 2001/02/10 03:06:39 itojun Exp $ */
2 /* $KAME: ip6_mroute.c,v 1.39 2001/02/10 02:13:13 itojun Exp $ */
3
4 /*
5 * Copyright (C) 1998 WIDE Project.
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
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.
19 *
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
30 * SUCH DAMAGE.
31 */
32
33 /* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */
34
35 /*
36 * IP multicast forwarding procedures
37 *
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
44 *
45 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
46 */
47
48 #include "opt_inet.h"
49
50 #ifndef _KERNEL
51 # ifdef KERNEL
52 # define _KERNEL
53 # endif
54 #endif
55
56 #include <sys/param.h>
57 #include <sys/systm.h>
58 #include <sys/callout.h>
59 #include <sys/mbuf.h>
60 #include <sys/socket.h>
61 #include <sys/socketvar.h>
62 #include <sys/sockio.h>
63 #include <sys/protosw.h>
64 #include <sys/errno.h>
65 #include <sys/time.h>
66 #include <sys/kernel.h>
67 #include <sys/ioctl.h>
68 #include <sys/syslog.h>
69
70 #include <net/if.h>
71 #include <net/route.h>
72 #include <net/raw_cb.h>
73
74 #include <netinet/in.h>
75 #include <netinet/in_var.h>
76
77 #include <netinet/ip6.h>
78 #include <netinet6/ip6_var.h>
79 #include <netinet6/ip6_mroute.h>
80 #include <netinet6/pim6.h>
81 #include <netinet6/pim6_var.h>
82
83 #include <net/net_osdep.h>
84
85 #define M_HASCL(m) ((m)->m_flags & M_EXT)
86
87 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
88 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
89
90 static int set_pim6 __P((int *));
91 static int get_pim6 __P((struct mbuf *));
92 static int socket_send __P((struct socket *, struct mbuf *,
93 struct sockaddr_in6 *));
94 static int register_send __P((struct ip6_hdr *, struct mif6 *,
95 struct mbuf *));
96
97 /*
98 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
99 * except for netstat or debugging purposes.
100 */
101 struct socket *ip6_mrouter = NULL;
102 int ip6_mrouter_ver = 0;
103 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */
104 struct mrt6stat mrt6stat;
105
106 #define NO_RTE_FOUND 0x1
107 #define RTE_FOUND 0x2
108
109 struct mf6c *mf6ctable[MF6CTBLSIZ];
110 u_char nexpire[MF6CTBLSIZ];
111 static struct mif6 mif6table[MAXMIFS];
112 #ifdef MRT6DEBUG
113 u_int mrt6debug = 0; /* debug level */
114 #define DEBUG_MFC 0x02
115 #define DEBUG_FORWARD 0x04
116 #define DEBUG_EXPIRE 0x08
117 #define DEBUG_XMIT 0x10
118 #define DEBUG_REG 0x20
119 #define DEBUG_PIM 0x40
120 #endif
121
122 static void expire_upcalls __P((void *));
123 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
124 #define UPCALL_EXPIRE 6 /* number of timeouts */
125
126 #ifdef INET
127 #ifdef MROUTING
128 extern struct socket *ip_mrouter;
129 #endif
130 #endif
131
132 /*
133 * 'Interfaces' associated with decapsulator (so we can tell
134 * packets that went through it from ones that get reflected
135 * by a broken gateway). These interfaces are never linked into
136 * the system ifnet list & no routes point to them. I.e., packets
137 * can't be sent this way. They only exist as a placeholder for
138 * multicast source verification.
139 */
140 struct ifnet multicast_register_if;
141
142 #define ENCAP_HOPS 64
143
144 /*
145 * Private variables.
146 */
147 static mifi_t nummifs = 0;
148 static mifi_t reg_mif_num = (mifi_t)-1;
149
150 static struct pim6stat pim6stat;
151
152 /*
153 * one-back cache used by ipip_input to locate a tunnel's mif
154 * given a datagram's src ip address.
155 */
156 static int pim6;
157
158 /*
159 * Hash function for a source, group entry
160 */
161 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
162 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
163 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
164 (g).s6_addr32[2] ^ (g).s6_addr32[3])
165
166 /*
167 * Find a route for a given origin IPv6 address and Multicast group address.
168 * Quality of service parameter to be added in the future!!!
169 */
170
171 #define MF6CFIND(o, g, rt) do { \
172 struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
173 rt = NULL; \
174 mrt6stat.mrt6s_mfc_lookups++; \
175 while (_rt) { \
176 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
177 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
178 (_rt->mf6c_stall == NULL)) { \
179 rt = _rt; \
180 break; \
181 } \
182 _rt = _rt->mf6c_next; \
183 } \
184 if (rt == NULL) { \
185 mrt6stat.mrt6s_mfc_misses++; \
186 } \
187 } while (0)
188
189 /*
190 * Macros to compute elapsed time efficiently
191 * Borrowed from Van Jacobson's scheduling code
192 */
193 #define TV_DELTA(a, b, delta) do { \
194 int xxs; \
195 \
196 delta = (a).tv_usec - (b).tv_usec; \
197 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
198 switch (xxs) { \
199 case 2: \
200 delta += 1000000; \
201 /* fall through */ \
202 case 1: \
203 delta += 1000000; \
204 break; \
205 default: \
206 delta += (1000000 * xxs); \
207 } \
208 } \
209 } while (0)
210
211 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
212 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
213
214 #ifdef UPCALL_TIMING
215 #define UPCALL_MAX 50
216 u_long upcall_data[UPCALL_MAX + 1];
217 static void collate();
218 #endif /* UPCALL_TIMING */
219
220 static int get_sg_cnt __P((struct sioc_sg_req6 *));
221 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
222 static int ip6_mrouter_init __P((struct socket *, struct mbuf *, int));
223 static int add_m6if __P((struct mif6ctl *));
224 static int del_m6if __P((mifi_t *));
225 static int add_m6fc __P((struct mf6cctl *));
226 static int del_m6fc __P((struct mf6cctl *));
227
228 static struct callout expire_upcalls_ch = CALLOUT_INITIALIZER;
229
230 /*
231 * Handle MRT setsockopt commands to modify the multicast routing tables.
232 */
233 int
234 ip6_mrouter_set(cmd, so, m)
235 int cmd;
236 struct socket *so;
237 struct mbuf *m;
238 {
239 if (cmd != MRT6_INIT && so != ip6_mrouter)
240 return EACCES;
241
242 switch (cmd) {
243 case MRT6_OINIT: return ip6_mrouter_init(so, m, cmd);
244 case MRT6_INIT: return ip6_mrouter_init(so, m, cmd);
245 case MRT6_DONE: return ip6_mrouter_done();
246 case MRT6_ADD_MIF: return add_m6if(mtod(m, struct mif6ctl *));
247 case MRT6_DEL_MIF: return del_m6if(mtod(m, mifi_t *));
248 case MRT6_ADD_MFC: return add_m6fc(mtod(m, struct mf6cctl *));
249 case MRT6_DEL_MFC: return del_m6fc(mtod(m, struct mf6cctl *));
250 case MRT6_PIM: return set_pim6(mtod(m, int *));
251 default: return EOPNOTSUPP;
252 }
253 }
254
255 /*
256 * Handle MRT getsockopt commands
257 */
258 int
259 ip6_mrouter_get(cmd, so, m)
260 int cmd;
261 struct socket *so;
262 struct mbuf **m;
263 {
264 struct mbuf *mb;
265
266 if (so != ip6_mrouter) return EACCES;
267
268 *m = mb = m_get(M_WAIT, MT_SOOPTS);
269
270 switch (cmd) {
271 case MRT6_PIM: return get_pim6(mb);
272 default:
273 m_free(mb);
274 return EOPNOTSUPP;
275 }
276 }
277
278 /*
279 * Handle ioctl commands to obtain information from the cache
280 */
281 int
282 mrt6_ioctl(cmd, data)
283 int cmd;
284 caddr_t data;
285 {
286 int error = 0;
287
288 switch (cmd) {
289 case SIOCGETSGCNT_IN6:
290 return(get_sg_cnt((struct sioc_sg_req6 *)data));
291 break; /* for safety */
292 case SIOCGETMIFCNT_IN6:
293 return(get_mif6_cnt((struct sioc_mif_req6 *)data));
294 break; /* for safety */
295 default:
296 return (EINVAL);
297 break;
298 }
299 return error;
300 }
301
302 /*
303 * returns the packet, byte, rpf-failure count for the source group provided
304 */
305 static int
306 get_sg_cnt(req)
307 struct sioc_sg_req6 *req;
308 {
309 struct mf6c *rt;
310 int s;
311
312 s = splsoftnet();
313 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
314 splx(s);
315 if (rt != NULL) {
316 req->pktcnt = rt->mf6c_pkt_cnt;
317 req->bytecnt = rt->mf6c_byte_cnt;
318 req->wrong_if = rt->mf6c_wrong_if;
319 } else
320 return(ESRCH);
321 #if 0
322 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
323 #endif
324
325 return 0;
326 }
327
328 /*
329 * returns the input and output packet and byte counts on the mif provided
330 */
331 static int
332 get_mif6_cnt(req)
333 struct sioc_mif_req6 *req;
334 {
335 mifi_t mifi = req->mifi;
336
337 if (mifi >= nummifs)
338 return EINVAL;
339
340 req->icount = mif6table[mifi].m6_pkt_in;
341 req->ocount = mif6table[mifi].m6_pkt_out;
342 req->ibytes = mif6table[mifi].m6_bytes_in;
343 req->obytes = mif6table[mifi].m6_bytes_out;
344
345 return 0;
346 }
347
348 /*
349 * Get PIM processiong global
350 */
351 static int
352 get_pim6(m)
353 struct mbuf *m;
354 {
355 int *i;
356
357 i = mtod(m, int *);
358
359 *i = pim6;
360
361 return 0;
362 }
363
364 static int
365 set_pim6(i)
366 int *i;
367 {
368 if ((*i != 1) && (*i != 0))
369 return EINVAL;
370
371 pim6 = *i;
372
373 return 0;
374 }
375
376 /*
377 * Enable multicast routing
378 */
379 static int
380 ip6_mrouter_init(so, m, cmd)
381 struct socket *so;
382 struct mbuf *m;
383 int cmd;
384 {
385 int *v;
386
387 #ifdef MRT6DEBUG
388 if (mrt6debug)
389 log(LOG_DEBUG,
390 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
391 so->so_type, so->so_proto->pr_protocol);
392 #endif
393
394 if (so->so_type != SOCK_RAW ||
395 so->so_proto->pr_protocol != IPPROTO_ICMPV6)
396 return EOPNOTSUPP;
397
398 if (!m || (m->m_len != sizeof(int *)))
399 return ENOPROTOOPT;
400
401 v = mtod(m, int *);
402 if (*v != 1)
403 return ENOPROTOOPT;
404
405 if (ip6_mrouter != NULL) return EADDRINUSE;
406
407 ip6_mrouter = so;
408 ip6_mrouter_ver = cmd;
409
410 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
411 bzero((caddr_t)nexpire, sizeof(nexpire));
412
413 pim6 = 0;/* used for stubbing out/in pim stuff */
414
415 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
416 expire_upcalls, NULL);
417
418 #ifdef MRT6DEBUG
419 if (mrt6debug)
420 log(LOG_DEBUG, "ip6_mrouter_init\n");
421 #endif
422
423 return 0;
424 }
425
426 /*
427 * Disable multicast routing
428 */
429 int
430 ip6_mrouter_done()
431 {
432 mifi_t mifi;
433 int i;
434 struct ifnet *ifp;
435 struct in6_ifreq ifr;
436 struct mf6c *rt;
437 struct rtdetq *rte;
438 int s;
439
440 s = splsoftnet();
441
442 /*
443 * For each phyint in use, disable promiscuous reception of all IPv6
444 * multicasts.
445 */
446 #ifdef INET
447 #ifdef MROUTING
448 /*
449 * If there is still IPv4 multicast routing daemon,
450 * we remain interfaces to receive all muliticasted packets.
451 * XXX: there may be an interface in which the IPv4 multicast
452 * daemon is not interested...
453 */
454 if (!ip_mrouter)
455 #endif
456 #endif
457 {
458 for (mifi = 0; mifi < nummifs; mifi++) {
459 if (mif6table[mifi].m6_ifp &&
460 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
461 ifr.ifr_addr.sin6_family = AF_INET6;
462 ifr.ifr_addr.sin6_addr= in6addr_any;
463 ifp = mif6table[mifi].m6_ifp;
464 (*ifp->if_ioctl)(ifp, SIOCDELMULTI,
465 (caddr_t)&ifr);
466 }
467 }
468 }
469 #ifdef notyet
470 bzero((caddr_t)qtable, sizeof(qtable));
471 bzero((caddr_t)tbftable, sizeof(tbftable));
472 #endif
473 bzero((caddr_t)mif6table, sizeof(mif6table));
474 nummifs = 0;
475
476 pim6 = 0; /* used to stub out/in pim specific code */
477
478 callout_stop(&expire_upcalls_ch);
479
480 /*
481 * Free all multicast forwarding cache entries.
482 */
483 for (i = 0; i < MF6CTBLSIZ; i++) {
484 rt = mf6ctable[i];
485 while (rt) {
486 struct mf6c *frt;
487
488 for (rte = rt->mf6c_stall; rte != NULL; ) {
489 struct rtdetq *n = rte->next;
490
491 m_free(rte->m);
492 free(rte, M_MRTABLE);
493 rte = n;
494 }
495 frt = rt;
496 rt = rt->mf6c_next;
497 free(frt, M_MRTABLE);
498 }
499 }
500
501 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
502
503 /*
504 * Reset de-encapsulation cache
505 */
506 reg_mif_num = -1;
507
508 ip6_mrouter = NULL;
509 ip6_mrouter_ver = 0;
510
511 splx(s);
512
513 #ifdef MRT6DEBUG
514 if (mrt6debug)
515 log(LOG_DEBUG, "ip6_mrouter_done\n");
516 #endif
517
518 return 0;
519 }
520
521 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
522
523 /*
524 * Add a mif to the mif table
525 */
526 static int
527 add_m6if(mifcp)
528 struct mif6ctl *mifcp;
529 {
530 struct mif6 *mifp;
531 struct ifnet *ifp;
532 struct in6_ifreq ifr;
533 int error, s;
534 #ifdef notyet
535 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
536 #endif
537
538 if (mifcp->mif6c_mifi >= MAXMIFS)
539 return EINVAL;
540 mifp = mif6table + mifcp->mif6c_mifi;
541 if (mifp->m6_ifp)
542 return EADDRINUSE; /* XXX: is it appropriate? */
543 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
544 return ENXIO;
545 ifp = ifindex2ifnet[mifcp->mif6c_pifi];
546
547 if (mifcp->mif6c_flags & MIFF_REGISTER) {
548 if (reg_mif_num == (mifi_t)-1) {
549 strcpy(multicast_register_if.if_xname,
550 "register_mif"); /* XXX */
551 multicast_register_if.if_flags |= IFF_LOOPBACK;
552 multicast_register_if.if_index = mifcp->mif6c_mifi;
553 reg_mif_num = mifcp->mif6c_mifi;
554 }
555
556 ifp = &multicast_register_if;
557
558 } /* if REGISTER */
559 else {
560 /* Make sure the interface supports multicast */
561 if ((ifp->if_flags & IFF_MULTICAST) == 0)
562 return EOPNOTSUPP;
563
564 s = splsoftnet();
565 /*
566 * Enable promiscuous reception of all IPv6 multicasts
567 * from the interface.
568 */
569 ifr.ifr_addr.sin6_family = AF_INET6;
570 ifr.ifr_addr.sin6_addr = in6addr_any;
571 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
572 splx(s);
573 if (error)
574 return error;
575 }
576
577 s = splsoftnet();
578 mifp->m6_flags = mifcp->mif6c_flags;
579 mifp->m6_ifp = ifp;
580 #ifdef notyet
581 /* scaling up here allows division by 1024 in critical code */
582 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
583 #endif
584 /* initialize per mif pkt counters */
585 mifp->m6_pkt_in = 0;
586 mifp->m6_pkt_out = 0;
587 mifp->m6_bytes_in = 0;
588 mifp->m6_bytes_out = 0;
589 splx(s);
590
591 /* Adjust nummifs up if the mifi is higher than nummifs */
592 if (nummifs <= mifcp->mif6c_mifi)
593 nummifs = mifcp->mif6c_mifi + 1;
594
595 #ifdef MRT6DEBUG
596 if (mrt6debug)
597 log(LOG_DEBUG,
598 "add_mif #%d, phyint %s%d\n",
599 mifcp->mif6c_mifi,
600 ifp->if_name, ifp->if_unit);
601 #endif
602
603 return 0;
604 }
605
606 /*
607 * Delete a mif from the mif table
608 */
609 static int
610 del_m6if(mifip)
611 mifi_t *mifip;
612 {
613 struct mif6 *mifp = mif6table + *mifip;
614 mifi_t mifi;
615 struct ifnet *ifp;
616 struct in6_ifreq ifr;
617 int s;
618
619 if (*mifip >= nummifs)
620 return EINVAL;
621 if (mifp->m6_ifp == NULL)
622 return EINVAL;
623
624 s = splsoftnet();
625
626 if (!(mifp->m6_flags & MIFF_REGISTER)) {
627 /*
628 * XXX: what if there is yet IPv4 multicast daemon
629 * using the interface?
630 */
631 ifp = mifp->m6_ifp;
632
633 ifr.ifr_addr.sin6_family = AF_INET6;
634 ifr.ifr_addr.sin6_addr = in6addr_any;
635 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
636 }
637
638 #ifdef notyet
639 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
640 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
641 #endif
642 bzero((caddr_t)mifp, sizeof (*mifp));
643
644 /* Adjust nummifs down */
645 for (mifi = nummifs; mifi > 0; mifi--)
646 if (mif6table[mifi - 1].m6_ifp)
647 break;
648 nummifs = mifi;
649
650 splx(s);
651
652 #ifdef MRT6DEBUG
653 if (mrt6debug)
654 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
655 #endif
656
657 return 0;
658 }
659
660 /*
661 * Add an mfc entry
662 */
663 static int
664 add_m6fc(mfccp)
665 struct mf6cctl *mfccp;
666 {
667 struct mf6c *rt;
668 u_long hash;
669 struct rtdetq *rte;
670 u_short nstl;
671 int s;
672
673 MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
674 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
675
676 /* If an entry already exists, just update the fields */
677 if (rt) {
678 #ifdef MRT6DEBUG
679 if (mrt6debug & DEBUG_MFC)
680 log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
681 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
682 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
683 mfccp->mf6cc_parent);
684 #endif
685
686 s = splsoftnet();
687 rt->mf6c_parent = mfccp->mf6cc_parent;
688 rt->mf6c_ifset = mfccp->mf6cc_ifset;
689 splx(s);
690 return 0;
691 }
692
693 /*
694 * Find the entry for which the upcall was made and update
695 */
696 s = splsoftnet();
697 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
698 mfccp->mf6cc_mcastgrp.sin6_addr);
699 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
700 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
701 &mfccp->mf6cc_origin.sin6_addr) &&
702 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
703 &mfccp->mf6cc_mcastgrp.sin6_addr) &&
704 (rt->mf6c_stall != NULL)) {
705
706 if (nstl++)
707 log(LOG_ERR,
708 "add_m6fc: %s o %s g %s p %x dbx %p\n",
709 "multiple kernel entries",
710 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
711 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
712 mfccp->mf6cc_parent, rt->mf6c_stall);
713
714 #ifdef MRT6DEBUG
715 if (mrt6debug & DEBUG_MFC)
716 log(LOG_DEBUG,
717 "add_m6fc o %s g %s p %x dbg %x\n",
718 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
719 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
720 mfccp->mf6cc_parent, rt->mf6c_stall);
721 #endif
722
723 rt->mf6c_origin = mfccp->mf6cc_origin;
724 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
725 rt->mf6c_parent = mfccp->mf6cc_parent;
726 rt->mf6c_ifset = mfccp->mf6cc_ifset;
727 /* initialize pkt counters per src-grp */
728 rt->mf6c_pkt_cnt = 0;
729 rt->mf6c_byte_cnt = 0;
730 rt->mf6c_wrong_if = 0;
731
732 rt->mf6c_expire = 0; /* Don't clean this guy up */
733 nexpire[hash]--;
734
735 /* free packets Qed at the end of this entry */
736 for (rte = rt->mf6c_stall; rte != NULL; ) {
737 struct rtdetq *n = rte->next;
738 ip6_mdq(rte->m, rte->ifp, rt);
739 m_freem(rte->m);
740 #ifdef UPCALL_TIMING
741 collate(&(rte->t));
742 #endif /* UPCALL_TIMING */
743 free(rte, M_MRTABLE);
744 rte = n;
745 }
746 rt->mf6c_stall = NULL;
747 }
748 }
749
750 /*
751 * It is possible that an entry is being inserted without an upcall
752 */
753 if (nstl == 0) {
754 #ifdef MRT6DEBUG
755 if (mrt6debug & DEBUG_MFC)
756 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
757 hash,
758 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
759 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
760 mfccp->mf6cc_parent);
761 #endif
762
763 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
764
765 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
766 &mfccp->mf6cc_origin.sin6_addr)&&
767 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
768 &mfccp->mf6cc_mcastgrp.sin6_addr)) {
769
770 rt->mf6c_origin = mfccp->mf6cc_origin;
771 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
772 rt->mf6c_parent = mfccp->mf6cc_parent;
773 rt->mf6c_ifset = mfccp->mf6cc_ifset;
774 /* initialize pkt counters per src-grp */
775 rt->mf6c_pkt_cnt = 0;
776 rt->mf6c_byte_cnt = 0;
777 rt->mf6c_wrong_if = 0;
778
779 if (rt->mf6c_expire)
780 nexpire[hash]--;
781 rt->mf6c_expire = 0;
782 }
783 }
784 if (rt == NULL) {
785 /* no upcall, so make a new entry */
786 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
787 M_NOWAIT);
788 if (rt == NULL) {
789 splx(s);
790 return ENOBUFS;
791 }
792
793 /* insert new entry at head of hash chain */
794 rt->mf6c_origin = mfccp->mf6cc_origin;
795 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
796 rt->mf6c_parent = mfccp->mf6cc_parent;
797 rt->mf6c_ifset = mfccp->mf6cc_ifset;
798 /* initialize pkt counters per src-grp */
799 rt->mf6c_pkt_cnt = 0;
800 rt->mf6c_byte_cnt = 0;
801 rt->mf6c_wrong_if = 0;
802 rt->mf6c_expire = 0;
803 rt->mf6c_stall = NULL;
804
805 /* link into table */
806 rt->mf6c_next = mf6ctable[hash];
807 mf6ctable[hash] = rt;
808 }
809 }
810 splx(s);
811 return 0;
812 }
813
814 #ifdef UPCALL_TIMING
815 /*
816 * collect delay statistics on the upcalls
817 */
818 static void
819 collate(t)
820 struct timeval *t;
821 {
822 u_long d;
823 struct timeval tp;
824 u_long delta;
825
826 GET_TIME(tp);
827
828 if (TV_LT(*t, tp))
829 {
830 TV_DELTA(tp, *t, delta);
831
832 d = delta >> 10;
833 if (d > UPCALL_MAX)
834 d = UPCALL_MAX;
835
836 ++upcall_data[d];
837 }
838 }
839 #endif /* UPCALL_TIMING */
840
841 /*
842 * Delete an mfc entry
843 */
844 static int
845 del_m6fc(mfccp)
846 struct mf6cctl *mfccp;
847 {
848 struct sockaddr_in6 origin;
849 struct sockaddr_in6 mcastgrp;
850 struct mf6c *rt;
851 struct mf6c **nptr;
852 u_long hash;
853 int s;
854
855 origin = mfccp->mf6cc_origin;
856 mcastgrp = mfccp->mf6cc_mcastgrp;
857 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
858
859 #ifdef MRT6DEBUG
860 if (mrt6debug & DEBUG_MFC)
861 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
862 ip6_sprintf(&origin.sin6_addr),
863 ip6_sprintf(&mcastgrp.sin6_addr));
864 #endif
865
866 s = splsoftnet();
867
868 nptr = &mf6ctable[hash];
869 while ((rt = *nptr) != NULL) {
870 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
871 &rt->mf6c_origin.sin6_addr) &&
872 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
873 &rt->mf6c_mcastgrp.sin6_addr) &&
874 rt->mf6c_stall == NULL)
875 break;
876
877 nptr = &rt->mf6c_next;
878 }
879 if (rt == NULL) {
880 splx(s);
881 return EADDRNOTAVAIL;
882 }
883
884 *nptr = rt->mf6c_next;
885 free(rt, M_MRTABLE);
886
887 splx(s);
888
889 return 0;
890 }
891
892 static int
893 socket_send(s, mm, src)
894 struct socket *s;
895 struct mbuf *mm;
896 struct sockaddr_in6 *src;
897 {
898 if (s) {
899 if (sbappendaddr(&s->so_rcv,
900 (struct sockaddr *)src,
901 mm, (struct mbuf *)0) != 0) {
902 sorwakeup(s);
903 return 0;
904 }
905 }
906 m_freem(mm);
907 return -1;
908 }
909
910 /*
911 * IPv6 multicast forwarding function. This function assumes that the packet
912 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
913 * pointed to by "ifp", and the packet is to be relayed to other networks
914 * that have members of the packet's destination IPv6 multicast group.
915 *
916 * The packet is returned unscathed to the caller, unless it is
917 * erroneous, in which case a non-zero return value tells the caller to
918 * discard it.
919 */
920
921 int
922 ip6_mforward(ip6, ifp, m)
923 struct ip6_hdr *ip6;
924 struct ifnet *ifp;
925 struct mbuf *m;
926 {
927 struct mf6c *rt;
928 struct mif6 *mifp;
929 struct mbuf *mm;
930 int s;
931 mifi_t mifi;
932 #if !(defined(__FreeBSD__) && __FreeBSD__ >= 3)
933 long time_second = time.tv_sec;
934 #endif
935
936 #ifdef MRT6DEBUG
937 if (mrt6debug & DEBUG_FORWARD)
938 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
939 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
940 ifp->if_index);
941 #endif
942
943 /*
944 * Don't forward a packet with Hop limit of zero or one,
945 * or a packet destined to a local-only group.
946 */
947 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
948 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
949 return 0;
950 ip6->ip6_hlim--;
951
952 /*
953 * Source address check: do not forward packets with unspecified
954 * source. It was discussed in July 2000, on ipngwg mailing list.
955 * This is rather more serious than unicast cases, because some
956 * MLD packets can be sent with the unspecified source address
957 * (although such packets must normally set 1 to the hop limit field).
958 */
959 if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
960 ip6stat.ip6s_cantforward++;
961 if (ip6_log_time + ip6_log_interval < time_second) {
962 ip6_log_time = time_second;
963 log(LOG_DEBUG,
964 "cannot forward "
965 "from %s to %s nxt %d received on %s\n",
966 ip6_sprintf(&ip6->ip6_src),
967 ip6_sprintf(&ip6->ip6_dst),
968 ip6->ip6_nxt,
969 if_name(m->m_pkthdr.rcvif));
970 }
971 return 0;
972 }
973
974 /*
975 * Determine forwarding mifs from the forwarding cache table
976 */
977 s = splsoftnet();
978 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
979
980 /* Entry exists, so forward if necessary */
981 if (rt) {
982 splx(s);
983 return (ip6_mdq(m, ifp, rt));
984 } else {
985 /*
986 * If we don't have a route for packet's origin,
987 * Make a copy of the packet &
988 * send message to routing daemon
989 */
990
991 struct mbuf *mb0;
992 struct rtdetq *rte;
993 u_long hash;
994 /* int i, npkts;*/
995 #ifdef UPCALL_TIMING
996 struct timeval tp;
997
998 GET_TIME(tp);
999 #endif /* UPCALL_TIMING */
1000
1001 mrt6stat.mrt6s_no_route++;
1002 #ifdef MRT6DEBUG
1003 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
1004 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
1005 ip6_sprintf(&ip6->ip6_src),
1006 ip6_sprintf(&ip6->ip6_dst));
1007 #endif
1008
1009 /*
1010 * Allocate mbufs early so that we don't do extra work if we
1011 * are just going to fail anyway.
1012 */
1013 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
1014 M_NOWAIT);
1015 if (rte == NULL) {
1016 splx(s);
1017 return ENOBUFS;
1018 }
1019 mb0 = m_copy(m, 0, M_COPYALL);
1020 /*
1021 * Pullup packet header if needed before storing it,
1022 * as other references may modify it in the meantime.
1023 */
1024 if (mb0 &&
1025 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1026 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1027 if (mb0 == NULL) {
1028 free(rte, M_MRTABLE);
1029 splx(s);
1030 return ENOBUFS;
1031 }
1032
1033 /* is there an upcall waiting for this packet? */
1034 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1035 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
1036 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1037 &rt->mf6c_origin.sin6_addr) &&
1038 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1039 &rt->mf6c_mcastgrp.sin6_addr) &&
1040 (rt->mf6c_stall != NULL))
1041 break;
1042 }
1043
1044 if (rt == NULL) {
1045 struct mrt6msg *im;
1046 struct omrt6msg *oim;
1047
1048 /* no upcall, so make a new entry */
1049 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1050 M_NOWAIT);
1051 if (rt == NULL) {
1052 free(rte, M_MRTABLE);
1053 m_freem(mb0);
1054 splx(s);
1055 return ENOBUFS;
1056 }
1057 /*
1058 * Make a copy of the header to send to the user
1059 * level process
1060 */
1061 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1062
1063 if (mm == NULL) {
1064 free(rte, M_MRTABLE);
1065 m_freem(mb0);
1066 free(rt, M_MRTABLE);
1067 splx(s);
1068 return ENOBUFS;
1069 }
1070
1071 /*
1072 * Send message to routing daemon
1073 */
1074 sin6.sin6_addr = ip6->ip6_src;
1075
1076 im = NULL;
1077 oim = NULL;
1078 switch (ip6_mrouter_ver) {
1079 case MRT6_OINIT:
1080 oim = mtod(mm, struct omrt6msg *);
1081 oim->im6_msgtype = MRT6MSG_NOCACHE;
1082 oim->im6_mbz = 0;
1083 break;
1084 case MRT6_INIT:
1085 im = mtod(mm, struct mrt6msg *);
1086 im->im6_msgtype = MRT6MSG_NOCACHE;
1087 im->im6_mbz = 0;
1088 break;
1089 default:
1090 free(rte, M_MRTABLE);
1091 m_freem(mb0);
1092 free(rt, M_MRTABLE);
1093 splx(s);
1094 return EINVAL;
1095 }
1096
1097 #ifdef MRT6DEBUG
1098 if (mrt6debug & DEBUG_FORWARD)
1099 log(LOG_DEBUG,
1100 "getting the iif info in the kernel\n");
1101 #endif
1102
1103 for (mifp = mif6table, mifi = 0;
1104 mifi < nummifs && mifp->m6_ifp != ifp;
1105 mifp++, mifi++)
1106 ;
1107
1108 switch (ip6_mrouter_ver) {
1109 case MRT6_OINIT:
1110 oim->im6_mif = mifi;
1111 break;
1112 case MRT6_INIT:
1113 im->im6_mif = mifi;
1114 break;
1115 }
1116
1117 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1118 log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1119 "socket queue full\n");
1120 mrt6stat.mrt6s_upq_sockfull++;
1121 free(rte, M_MRTABLE);
1122 m_freem(mb0);
1123 free(rt, M_MRTABLE);
1124 splx(s);
1125 return ENOBUFS;
1126 }
1127
1128 mrt6stat.mrt6s_upcalls++;
1129
1130 /* insert new entry at head of hash chain */
1131 bzero(rt, sizeof(*rt));
1132 rt->mf6c_origin.sin6_family = AF_INET6;
1133 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1134 rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1135 rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1136 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1137 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1138 rt->mf6c_expire = UPCALL_EXPIRE;
1139 nexpire[hash]++;
1140 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1141
1142 /* link into table */
1143 rt->mf6c_next = mf6ctable[hash];
1144 mf6ctable[hash] = rt;
1145 /* Add this entry to the end of the queue */
1146 rt->mf6c_stall = rte;
1147 } else {
1148 /* determine if q has overflowed */
1149 struct rtdetq **p;
1150 int npkts = 0;
1151
1152 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1153 if (++npkts > MAX_UPQ6) {
1154 mrt6stat.mrt6s_upq_ovflw++;
1155 free(rte, M_MRTABLE);
1156 m_freem(mb0);
1157 splx(s);
1158 return 0;
1159 }
1160
1161 /* Add this entry to the end of the queue */
1162 *p = rte;
1163 }
1164
1165 rte->next = NULL;
1166 rte->m = mb0;
1167 rte->ifp = ifp;
1168 #ifdef UPCALL_TIMING
1169 rte->t = tp;
1170 #endif /* UPCALL_TIMING */
1171
1172 splx(s);
1173
1174 return 0;
1175 }
1176 }
1177
1178 /*
1179 * Clean up cache entries if upcalls are not serviced
1180 * Call from the Slow Timeout mechanism, every half second.
1181 */
1182 static void
1183 expire_upcalls(unused)
1184 void *unused;
1185 {
1186 struct rtdetq *rte;
1187 struct mf6c *mfc, **nptr;
1188 int i;
1189 int s;
1190
1191 s = splsoftnet();
1192 for (i = 0; i < MF6CTBLSIZ; i++) {
1193 if (nexpire[i] == 0)
1194 continue;
1195 nptr = &mf6ctable[i];
1196 while ((mfc = *nptr) != NULL) {
1197 rte = mfc->mf6c_stall;
1198 /*
1199 * Skip real cache entries
1200 * Make sure it wasn't marked to not expire (shouldn't happen)
1201 * If it expires now
1202 */
1203 if (rte != NULL &&
1204 mfc->mf6c_expire != 0 &&
1205 --mfc->mf6c_expire == 0) {
1206 #ifdef MRT6DEBUG
1207 if (mrt6debug & DEBUG_EXPIRE)
1208 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1209 ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1210 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1211 #endif
1212 /*
1213 * drop all the packets
1214 * free the mbuf with the pkt, if, timing info
1215 */
1216 do {
1217 struct rtdetq *n = rte->next;
1218 m_freem(rte->m);
1219 free(rte, M_MRTABLE);
1220 rte = n;
1221 } while (rte != NULL);
1222 mrt6stat.mrt6s_cache_cleanups++;
1223 nexpire[i]--;
1224
1225 *nptr = mfc->mf6c_next;
1226 free(mfc, M_MRTABLE);
1227 } else {
1228 nptr = &mfc->mf6c_next;
1229 }
1230 }
1231 }
1232 splx(s);
1233 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT,
1234 expire_upcalls, NULL);
1235 }
1236
1237 /*
1238 * Packet forwarding routine once entry in the cache is made
1239 */
1240 static int
1241 ip6_mdq(m, ifp, rt)
1242 struct mbuf *m;
1243 struct ifnet *ifp;
1244 struct mf6c *rt;
1245 {
1246 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1247 mifi_t mifi, iif;
1248 struct mif6 *mifp;
1249 int plen = m->m_pkthdr.len;
1250
1251 /*
1252 * Macro to send packet on mif. Since RSVP packets don't get counted on
1253 * input, they shouldn't get counted on output, so statistics keeping is
1254 * seperate.
1255 */
1256
1257 #define MC6_SEND(ip6, mifp, m) do { \
1258 if ((mifp)->m6_flags & MIFF_REGISTER) \
1259 register_send((ip6), (mifp), (m)); \
1260 else \
1261 phyint_send((ip6), (mifp), (m)); \
1262 } while (0)
1263
1264 /*
1265 * Don't forward if it didn't arrive from the parent mif
1266 * for its origin.
1267 */
1268 mifi = rt->mf6c_parent;
1269 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1270 /* came in the wrong interface */
1271 #ifdef MRT6DEBUG
1272 if (mrt6debug & DEBUG_FORWARD)
1273 log(LOG_DEBUG,
1274 "wrong if: ifid %d mifi %d mififid %x\n",
1275 ifp->if_index, mifi,
1276 mif6table[mifi].m6_ifp->if_index);
1277 #endif
1278 mrt6stat.mrt6s_wrong_if++;
1279 rt->mf6c_wrong_if++;
1280 /*
1281 * If we are doing PIM processing, and we are forwarding
1282 * packets on this interface, send a message to the
1283 * routing daemon.
1284 */
1285 /* have to make sure this is a valid mif */
1286 if (mifi < nummifs && mif6table[mifi].m6_ifp)
1287 if (pim6 && (m->m_flags & M_LOOP) == 0) {
1288 /*
1289 * Check the M_LOOP flag to avoid an
1290 * unnecessary PIM assert.
1291 * XXX: M_LOOP is an ad-hoc hack...
1292 */
1293 static struct sockaddr_in6 sin6 =
1294 { sizeof(sin6), AF_INET6 };
1295
1296 struct mbuf *mm;
1297 struct mrt6msg *im;
1298 struct omrt6msg *oim;
1299
1300 mm = m_copy(m, 0, sizeof(struct ip6_hdr));
1301 if (mm &&
1302 (M_HASCL(mm) ||
1303 mm->m_len < sizeof(struct ip6_hdr)))
1304 mm = m_pullup(mm, sizeof(struct ip6_hdr));
1305 if (mm == NULL)
1306 return ENOBUFS;
1307
1308 oim = NULL;
1309 im = NULL;
1310 switch (ip6_mrouter_ver) {
1311 case MRT6_OINIT:
1312 oim = mtod(mm, struct omrt6msg *);
1313 oim->im6_msgtype = MRT6MSG_WRONGMIF;
1314 oim->im6_mbz = 0;
1315 break;
1316 case MRT6_INIT:
1317 im = mtod(mm, struct mrt6msg *);
1318 im->im6_msgtype = MRT6MSG_WRONGMIF;
1319 im->im6_mbz = 0;
1320 break;
1321 default:
1322 m_freem(mm);
1323 return EINVAL;
1324 }
1325
1326 for (mifp = mif6table, iif = 0;
1327 iif < nummifs && mifp &&
1328 mifp->m6_ifp != ifp;
1329 mifp++, iif++)
1330 ;
1331
1332 switch (ip6_mrouter_ver) {
1333 case MRT6_OINIT:
1334 oim->im6_mif = iif;
1335 sin6.sin6_addr = oim->im6_src;
1336 break;
1337 case MRT6_INIT:
1338 im->im6_mif = iif;
1339 sin6.sin6_addr = im->im6_src;
1340 break;
1341 }
1342
1343 mrt6stat.mrt6s_upcalls++;
1344
1345 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1346 #ifdef MRT6DEBUG
1347 if (mrt6debug)
1348 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1349 #endif
1350 ++mrt6stat.mrt6s_upq_sockfull;
1351 return ENOBUFS;
1352 } /* if socket Q full */
1353 } /* if PIM */
1354 return 0;
1355 } /* if wrong iif */
1356
1357 /* If I sourced this packet, it counts as output, else it was input. */
1358 if (m->m_pkthdr.rcvif == NULL) {
1359 /* XXX: is rcvif really NULL when output?? */
1360 mif6table[mifi].m6_pkt_out++;
1361 mif6table[mifi].m6_bytes_out += plen;
1362 } else {
1363 mif6table[mifi].m6_pkt_in++;
1364 mif6table[mifi].m6_bytes_in += plen;
1365 }
1366 rt->mf6c_pkt_cnt++;
1367 rt->mf6c_byte_cnt += plen;
1368
1369 /*
1370 * For each mif, forward a copy of the packet if there are group
1371 * members downstream on the interface.
1372 */
1373 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1374 if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1375 #ifdef notyet
1376 /*
1377 * check if the outgoing packet is going to break
1378 * a scope boundary.
1379 * XXX For packets through PIM register tunnel
1380 * interface, we believe a routing daemon.
1381 */
1382 if ((mif6table[rt->mf6c_parent].m6_flags &
1383 MIFF_REGISTER) == 0 &&
1384 (mif6table[mifi].m6_flags & MIFF_REGISTER) == 0 &&
1385 (in6_addr2scopeid(ifp, &ip6->ip6_dst) !=
1386 in6_addr2scopeid(mif6table[mifi].m6_ifp,
1387 &ip6->ip6_dst) ||
1388 in6_addr2scopeid(ifp, &ip6->ip6_src) !=
1389 in6_addr2scopeid(mif6table[mifi].m6_ifp,
1390 &ip6->ip6_src))) {
1391 ip6stat.ip6s_badscope++;
1392 continue;
1393 }
1394 #endif
1395
1396 mifp->m6_pkt_out++;
1397 mifp->m6_bytes_out += plen;
1398 MC6_SEND(ip6, mifp, m);
1399 }
1400 return 0;
1401 }
1402
1403 static void
1404 phyint_send(ip6, mifp, m)
1405 struct ip6_hdr *ip6;
1406 struct mif6 *mifp;
1407 struct mbuf *m;
1408 {
1409 struct mbuf *mb_copy;
1410 struct ifnet *ifp = mifp->m6_ifp;
1411 int error = 0;
1412 int s = splsoftnet();
1413 static struct route_in6 ro;
1414 struct in6_multi *in6m;
1415 struct sockaddr_in6 *dst6;
1416
1417 /*
1418 * Make a new reference to the packet; make sure that
1419 * the IPv6 header is actually copied, not just referenced,
1420 * so that ip6_output() only scribbles on the copy.
1421 */
1422 mb_copy = m_copy(m, 0, M_COPYALL);
1423 if (mb_copy &&
1424 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1425 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1426 if (mb_copy == NULL)
1427 return;
1428 /* set MCAST flag to the outgoing packet */
1429 mb_copy->m_flags |= M_MCAST;
1430
1431 /*
1432 * If we sourced the packet, call ip6_output since we may devide
1433 * the packet into fragments when the packet is too big for the
1434 * outgoing interface.
1435 * Otherwise, we can simply send the packet to the interface
1436 * sending queue.
1437 */
1438 if (m->m_pkthdr.rcvif == NULL) {
1439 struct ip6_moptions im6o;
1440
1441 im6o.im6o_multicast_ifp = ifp;
1442 /* XXX: ip6_output will override ip6->ip6_hlim */
1443 im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1444 im6o.im6o_multicast_loop = 1;
1445 error = ip6_output(mb_copy, NULL, &ro,
1446 IPV6_FORWARDING, &im6o, NULL);
1447
1448 #ifdef MRT6DEBUG
1449 if (mrt6debug & DEBUG_XMIT)
1450 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1451 mifp - mif6table, error);
1452 #endif
1453 splx(s);
1454 return;
1455 }
1456
1457 /*
1458 * If we belong to the destination multicast group
1459 * on the outgoing interface, loop back a copy.
1460 */
1461 dst6 = (struct sockaddr_in6 *)&ro.ro_dst;
1462 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1463 if (in6m != NULL) {
1464 dst6->sin6_len = sizeof(struct sockaddr_in6);
1465 dst6->sin6_family = AF_INET6;
1466 dst6->sin6_addr = ip6->ip6_dst;
1467 ip6_mloopback(ifp, m, (struct sockaddr_in6 *)&ro.ro_dst);
1468 }
1469 /*
1470 * Put the packet into the sending queue of the outgoing interface
1471 * if it would fit in the MTU of the interface.
1472 */
1473 if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1474 dst6->sin6_len = sizeof(struct sockaddr_in6);
1475 dst6->sin6_family = AF_INET6;
1476 dst6->sin6_addr = ip6->ip6_dst;
1477 /*
1478 * We just call if_output instead of nd6_output here, since
1479 * we need no ND for a multicast forwarded packet...right?
1480 */
1481 error = (*ifp->if_output)(ifp, mb_copy,
1482 (struct sockaddr *)&ro.ro_dst, NULL);
1483 #ifdef MRT6DEBUG
1484 if (mrt6debug & DEBUG_XMIT)
1485 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1486 mifp - mif6table, error);
1487 #endif
1488 }
1489 else {
1490 #ifdef MULTICAST_PMTUD
1491 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1492 return;
1493 #else
1494 #ifdef MRT6DEBUG
1495 if (mrt6debug & DEBUG_DEBUG_XMIT)
1496 log(LOG_DEBUG,
1497 "phyint_send: packet too big on %s o %s g %s"
1498 " size %d(discarded)\n",
1499 ifp->if_xname,
1500 ip6_sprintf(&ip6->ip6_src),
1501 ip6_sprintf(&ip6->ip6_dst),
1502 mb_copy->m_pkthdr.len);
1503 #endif /* MRT6DEBUG */
1504 m_freem(mb_copy); /* simply discard the packet */
1505 return;
1506 #endif
1507 }
1508 }
1509
1510 static int
1511 register_send(ip6, mif, m)
1512 struct ip6_hdr *ip6;
1513 struct mif6 *mif;
1514 struct mbuf *m;
1515 {
1516 struct mbuf *mm;
1517 int i, len = m->m_pkthdr.len;
1518 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1519 struct mrt6msg *im6;
1520
1521 #ifdef MRT6DEBUG
1522 if (mrt6debug)
1523 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1524 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1525 #endif
1526 ++pim6stat.pim6s_snd_registers;
1527
1528 /* Make a copy of the packet to send to the user level process */
1529 MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1530 if (mm == NULL)
1531 return ENOBUFS;
1532 mm->m_data += max_linkhdr;
1533 mm->m_len = sizeof(struct ip6_hdr);
1534
1535 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1536 m_freem(mm);
1537 return ENOBUFS;
1538 }
1539 i = MHLEN - M_LEADINGSPACE(mm);
1540 if (i > len)
1541 i = len;
1542 mm = m_pullup(mm, i);
1543 if (mm == NULL){
1544 m_freem(mm);
1545 return ENOBUFS;
1546 }
1547 /* TODO: check it! */
1548 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1549
1550 /*
1551 * Send message to routing daemon
1552 */
1553 sin6.sin6_addr = ip6->ip6_src;
1554
1555 im6 = mtod(mm, struct mrt6msg *);
1556 im6->im6_msgtype = MRT6MSG_WHOLEPKT;
1557 im6->im6_mbz = 0;
1558
1559 im6->im6_mif = mif - mif6table;
1560
1561 /* iif info is not given for reg. encap.n */
1562 mrt6stat.mrt6s_upcalls++;
1563
1564 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1565 #ifdef MRT6DEBUG
1566 if (mrt6debug)
1567 log(LOG_WARNING,
1568 "register_send: ip_mrouter socket queue full\n");
1569 #endif
1570 ++mrt6stat.mrt6s_upq_sockfull;
1571 return ENOBUFS;
1572 }
1573 return 0;
1574 }
1575
1576 /*
1577 * PIM sparse mode hook
1578 * Receives the pim control messages, and passes them up to the listening
1579 * socket, using rip6_input.
1580 * The only message processed is the REGISTER pim message; the pim header
1581 * is stripped off, and the inner packet is passed to register_mforward.
1582 */
1583 int
1584 pim6_input(mp, offp, proto)
1585 struct mbuf **mp;
1586 int *offp, proto;
1587 {
1588 struct pim *pim; /* pointer to a pim struct */
1589 struct ip6_hdr *ip6;
1590 int pimlen;
1591 struct mbuf *m = *mp;
1592 int minlen;
1593 int off = *offp;
1594
1595 ++pim6stat.pim6s_rcv_total;
1596
1597 ip6 = mtod(m, struct ip6_hdr *);
1598 pimlen = m->m_pkthdr.len - *offp;
1599
1600 /*
1601 * Validate lengths
1602 */
1603 if (pimlen < PIM_MINLEN) {
1604 ++pim6stat.pim6s_rcv_tooshort;
1605 #ifdef MRT6DEBUG
1606 if (mrt6debug & DEBUG_PIM)
1607 log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1608 #endif
1609 m_freem(m);
1610 return(IPPROTO_DONE);
1611 }
1612
1613 /*
1614 * if the packet is at least as big as a REGISTER, go ahead
1615 * and grab the PIM REGISTER header size, to avoid another
1616 * possible m_pullup() later.
1617 *
1618 * PIM_MINLEN == pimhdr + u_int32 == 8
1619 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1620 */
1621 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1622
1623 /*
1624 * Make sure that the IP6 and PIM headers in contiguous memory, and
1625 * possibly the PIM REGISTER header
1626 */
1627 #ifndef PULLDOWN_TEST
1628 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1629 /* adjust pointer */
1630 ip6 = mtod(m, struct ip6_hdr *);
1631
1632 /* adjust mbuf to point to the PIM header */
1633 pim = (struct pim *)((caddr_t)ip6 + off);
1634 #else
1635 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1636 if (pim == NULL) {
1637 pim6stat.pim6s_rcv_tooshort++;
1638 return IPPROTO_DONE;
1639 }
1640 #endif
1641
1642 #define PIM6_CHECKSUM
1643 #ifdef PIM6_CHECKSUM
1644 {
1645 int cksumlen;
1646
1647 /*
1648 * Validate checksum.
1649 * If PIM REGISTER, exclude the data packet
1650 */
1651 if (pim->pim_type == PIM_REGISTER)
1652 cksumlen = PIM_MINLEN;
1653 else
1654 cksumlen = pimlen;
1655
1656 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1657 ++pim6stat.pim6s_rcv_badsum;
1658 #ifdef MRT6DEBUG
1659 if (mrt6debug & DEBUG_PIM)
1660 log(LOG_DEBUG,
1661 "pim6_input: invalid checksum\n");
1662 #endif
1663 m_freem(m);
1664 return(IPPROTO_DONE);
1665 }
1666 }
1667 #endif /* PIM_CHECKSUM */
1668
1669 /* PIM version check */
1670 if (pim->pim_ver != PIM_VERSION) {
1671 ++pim6stat.pim6s_rcv_badversion;
1672 #ifdef MRT6DEBUG
1673 log(LOG_ERR,
1674 "pim6_input: incorrect version %d, expecting %d\n",
1675 pim->pim_ver, PIM_VERSION);
1676 #endif
1677 m_freem(m);
1678 return(IPPROTO_DONE);
1679 }
1680
1681 if (pim->pim_type == PIM_REGISTER) {
1682 /*
1683 * since this is a REGISTER, we'll make a copy of the register
1684 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1685 * routing daemon.
1686 */
1687 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1688
1689 struct mbuf *mcp;
1690 struct ip6_hdr *eip6;
1691 u_int32_t *reghdr;
1692 int rc;
1693
1694 ++pim6stat.pim6s_rcv_registers;
1695
1696 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1697 #ifdef MRT6DEBUG
1698 if (mrt6debug & DEBUG_PIM)
1699 log(LOG_DEBUG,
1700 "pim6_input: register mif not set: %d\n",
1701 reg_mif_num);
1702 #endif
1703 m_freem(m);
1704 return(IPPROTO_DONE);
1705 }
1706
1707 reghdr = (u_int32_t *)(pim + 1);
1708
1709 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1710 goto pim6_input_to_daemon;
1711
1712 /*
1713 * Validate length
1714 */
1715 if (pimlen < PIM6_REG_MINLEN) {
1716 ++pim6stat.pim6s_rcv_tooshort;
1717 ++pim6stat.pim6s_rcv_badregisters;
1718 #ifdef MRT6DEBUG
1719 log(LOG_ERR,
1720 "pim6_input: register packet size too "
1721 "small %d from %s\n",
1722 pimlen, ip6_sprintf(&ip6->ip6_src));
1723 #endif
1724 m_freem(m);
1725 return(IPPROTO_DONE);
1726 }
1727
1728 eip6 = (struct ip6_hdr *) (reghdr + 1);
1729 #ifdef MRT6DEBUG
1730 if (mrt6debug & DEBUG_PIM)
1731 log(LOG_DEBUG,
1732 "pim6_input[register], eip6: %s -> %s, "
1733 "eip6 plen %d\n",
1734 ip6_sprintf(&eip6->ip6_src),
1735 ip6_sprintf(&eip6->ip6_dst),
1736 ntohs(eip6->ip6_plen));
1737 #endif
1738
1739 /* verify the version number of the inner packet */
1740 if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1741 ++pim6stat.pim6s_rcv_badregisters;
1742 #ifdef MRT6DEBUG
1743 log(LOG_DEBUG, "pim6_input: invalid IP version (%d) "
1744 "of the inner packet\n",
1745 (eip6->ip6_vfc & IPV6_VERSION));
1746 #endif
1747 m_freem(m);
1748 return(IPPROTO_NONE);
1749 }
1750
1751 /* verify the inner packet is destined to a mcast group */
1752 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1753 ++pim6stat.pim6s_rcv_badregisters;
1754 #ifdef MRT6DEBUG
1755 if (mrt6debug & DEBUG_PIM)
1756 log(LOG_DEBUG,
1757 "pim6_input: inner packet of register "
1758 "is not multicast %s\n",
1759 ip6_sprintf(&eip6->ip6_dst));
1760 #endif
1761 m_freem(m);
1762 return(IPPROTO_DONE);
1763 }
1764
1765 /*
1766 * make a copy of the whole header to pass to the daemon later.
1767 */
1768 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1769 if (mcp == NULL) {
1770 #ifdef MRT6DEBUG
1771 log(LOG_ERR,
1772 "pim6_input: pim register: "
1773 "could not copy register head\n");
1774 #endif
1775 m_freem(m);
1776 return(IPPROTO_DONE);
1777 }
1778
1779 /*
1780 * forward the inner ip6 packet; point m_data at the inner ip6.
1781 */
1782 m_adj(m, off + PIM_MINLEN);
1783 #ifdef MRT6DEBUG
1784 if (mrt6debug & DEBUG_PIM) {
1785 log(LOG_DEBUG,
1786 "pim6_input: forwarding decapsulated register: "
1787 "src %s, dst %s, mif %d\n",
1788 ip6_sprintf(&eip6->ip6_src),
1789 ip6_sprintf(&eip6->ip6_dst),
1790 reg_mif_num);
1791 }
1792 #endif
1793
1794 rc = looutput(mif6table[reg_mif_num].m6_ifp, m,
1795 (struct sockaddr *) &dst,
1796 (struct rtentry *) NULL);
1797
1798 /* prepare the register head to send to the mrouting daemon */
1799 m = mcp;
1800 }
1801
1802 /*
1803 * Pass the PIM message up to the daemon; if it is a register message
1804 * pass the 'head' only up to the daemon. This includes the
1805 * encapsulator ip6 header, pim header, register header and the
1806 * encapsulated ip6 header.
1807 */
1808 pim6_input_to_daemon:
1809 rip6_input(&m, offp, proto);
1810 return(IPPROTO_DONE);
1811 }
1812