ip6_mroute.c revision 1.10 1 /* $NetBSD: ip6_mroute.c,v 1.10 2000/02/06 12:49:45 itojun Exp $ */
2
3 /*
4 * Copyright (C) 1998 WIDE Project.
5 * All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the project nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32 /* BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp */
33
34 /*
35 * IP multicast forwarding procedures
36 *
37 * Written by David Waitzman, BBN Labs, August 1988.
38 * Modified by Steve Deering, Stanford, February 1989.
39 * Modified by Mark J. Steiglitz, Stanford, May, 1991
40 * Modified by Van Jacobson, LBL, January 1993
41 * Modified by Ajit Thyagarajan, PARC, August 1993
42 * Modified by Bill Fenenr, PARC, April 1994
43 *
44 * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
45 */
46
47 #include "opt_inet.h"
48
49 #ifndef _KERNEL
50 # ifdef KERNEL
51 # define _KERNEL
52 # endif
53 #endif
54
55 #include <sys/param.h>
56 #include <sys/systm.h>
57 #include <sys/mbuf.h>
58 #include <sys/socket.h>
59 #include <sys/socketvar.h>
60 #include <sys/sockio.h>
61 #include <sys/protosw.h>
62 #include <sys/errno.h>
63 #include <sys/time.h>
64 #include <sys/kernel.h>
65 #include <sys/ioctl.h>
66 #include <sys/syslog.h>
67
68 #include <net/if.h>
69 #include <net/route.h>
70 #include <net/raw_cb.h>
71
72 #include <netinet/in.h>
73 #include <netinet/in_var.h>
74
75 #include <netinet/ip6.h>
76 #include <netinet6/ip6_var.h>
77 #include <netinet6/ip6_mroute.h>
78 #include <netinet6/pim6.h>
79 #include <netinet6/pim6_var.h>
80
81 #define M_HASCL(m) ((m)->m_flags & M_EXT)
82
83 static int ip6_mdq __P((struct mbuf *, struct ifnet *, struct mf6c *));
84 static void phyint_send __P((struct ip6_hdr *, struct mif6 *, struct mbuf *));
85
86 static int set_pim6 __P((int *));
87 static int get_pim6 __P((struct mbuf *));
88 static int socket_send __P((struct socket *, struct mbuf *,
89 struct sockaddr_in6 *));
90 static int register_send __P((struct ip6_hdr *, struct mif6 *,
91 struct mbuf *));
92
93 /*
94 * Globals. All but ip6_mrouter, ip6_mrtproto and mrt6stat could be static,
95 * except for netstat or debugging purposes.
96 */
97 struct socket *ip6_mrouter = NULL;
98 int ip6_mrtproto = IPPROTO_PIM; /* for netstat only */
99 struct mrt6stat mrt6stat;
100
101 #define NO_RTE_FOUND 0x1
102 #define RTE_FOUND 0x2
103
104 struct mf6c *mf6ctable[MF6CTBLSIZ];
105 u_char nexpire[MF6CTBLSIZ];
106 static struct mif6 mif6table[MAXMIFS];
107 #ifdef MRT6DEBUG
108 u_int mrt6debug = 0; /* debug level */
109 #define DEBUG_MFC 0x02
110 #define DEBUG_FORWARD 0x04
111 #define DEBUG_EXPIRE 0x08
112 #define DEBUG_XMIT 0x10
113 #define DEBUG_REG 0x20
114 #define DEBUG_PIM 0x40
115 #endif
116
117 static void expire_upcalls __P((void *));
118 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
119 #define UPCALL_EXPIRE 6 /* number of timeouts */
120
121 #ifdef INET
122 #ifdef MROUTING
123 extern struct socket *ip_mrouter;
124 #endif
125 #endif
126
127 /*
128 * 'Interfaces' associated with decapsulator (so we can tell
129 * packets that went through it from ones that get reflected
130 * by a broken gateway). These interfaces are never linked into
131 * the system ifnet list & no routes point to them. I.e., packets
132 * can't be sent this way. They only exist as a placeholder for
133 * multicast source verification.
134 */
135 struct ifnet multicast_register_if;
136
137 #define ENCAP_HOPS 64
138
139 /*
140 * Private variables.
141 */
142 static mifi_t nummifs = 0;
143 static mifi_t reg_mif_num = (mifi_t)-1;
144
145 static struct pim6stat pim6stat;
146
147 /*
148 * one-back cache used by ipip_input to locate a tunnel's mif
149 * given a datagram's src ip address.
150 */
151 static int pim6;
152
153 /*
154 * Hash function for a source, group entry
155 */
156 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
157 (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
158 (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
159 (g).s6_addr32[2] ^ (g).s6_addr32[3])
160
161 /*
162 * Find a route for a given origin IPv6 address and Multicast group address.
163 * Quality of service parameter to be added in the future!!!
164 */
165
166 #define MF6CFIND(o, g, rt) { \
167 register struct mf6c *_rt = mf6ctable[MF6CHASH(o,g)]; \
168 rt = NULL; \
169 mrt6stat.mrt6s_mfc_lookups++; \
170 while (_rt) { \
171 if (IN6_ARE_ADDR_EQUAL(&_rt->mf6c_origin.sin6_addr, &(o)) && \
172 IN6_ARE_ADDR_EQUAL(&_rt->mf6c_mcastgrp.sin6_addr, &(g)) && \
173 (_rt->mf6c_stall == NULL)) { \
174 rt = _rt; \
175 break; \
176 } \
177 _rt = _rt->mf6c_next; \
178 } \
179 if (rt == NULL) { \
180 mrt6stat.mrt6s_mfc_misses++; \
181 } \
182 }
183
184 /*
185 * Macros to compute elapsed time efficiently
186 * Borrowed from Van Jacobson's scheduling code
187 */
188 #define TV_DELTA(a, b, delta) { \
189 register int xxs; \
190 \
191 delta = (a).tv_usec - (b).tv_usec; \
192 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
193 switch (xxs) { \
194 case 2: \
195 delta += 1000000; \
196 /* fall through */ \
197 case 1: \
198 delta += 1000000; \
199 break; \
200 default: \
201 delta += (1000000 * xxs); \
202 } \
203 } \
204 }
205
206 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
207 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
208
209 #ifdef UPCALL_TIMING
210 #define UPCALL_MAX 50
211 u_long upcall_data[UPCALL_MAX + 1];
212 static void collate();
213 #endif /* UPCALL_TIMING */
214
215 static int get_sg_cnt __P((struct sioc_sg_req6 *));
216 static int get_mif6_cnt __P((struct sioc_mif_req6 *));
217 static int ip6_mrouter_init __P((struct socket *, struct mbuf *));
218 static int add_m6if __P((struct mif6ctl *));
219 static int del_m6if __P((mifi_t *));
220 static int add_m6fc __P((struct mf6cctl *));
221 static int del_m6fc __P((struct mf6cctl *));
222
223 /*
224 * Handle MRT setsockopt commands to modify the multicast routing tables.
225 */
226 int
227 ip6_mrouter_set(cmd, so, m)
228 int cmd;
229 struct socket *so;
230 struct mbuf *m;
231 {
232 if (cmd != MRT6_INIT && so != ip6_mrouter)
233 return EACCES;
234
235 switch (cmd) {
236 case MRT6_INIT: return ip6_mrouter_init(so, m);
237 case MRT6_DONE: return ip6_mrouter_done();
238 case MRT6_ADD_MIF: return add_m6if(mtod(m, struct mif6ctl *));
239 case MRT6_DEL_MIF: return del_m6if(mtod(m, mifi_t *));
240 case MRT6_ADD_MFC: return add_m6fc(mtod(m, struct mf6cctl *));
241 case MRT6_DEL_MFC: return del_m6fc(mtod(m, struct mf6cctl *));
242 case MRT6_PIM: return set_pim6(mtod(m, int *));
243 default: return EOPNOTSUPP;
244 }
245 }
246
247 /*
248 * Handle MRT getsockopt commands
249 */
250 int
251 ip6_mrouter_get(cmd, so, m)
252 int cmd;
253 struct socket *so;
254 struct mbuf **m;
255 {
256 struct mbuf *mb;
257
258 if (so != ip6_mrouter) return EACCES;
259
260 *m = mb = m_get(M_WAIT, MT_SOOPTS);
261
262 switch (cmd) {
263 case MRT6_PIM: return get_pim6(mb);
264 default:
265 m_free(mb);
266 return EOPNOTSUPP;
267 }
268 }
269
270 /*
271 * Handle ioctl commands to obtain information from the cache
272 */
273 int
274 mrt6_ioctl(cmd, data)
275 int cmd;
276 caddr_t data;
277 {
278 int error = 0;
279
280 switch (cmd) {
281 case SIOCGETSGCNT_IN6:
282 return(get_sg_cnt((struct sioc_sg_req6 *)data));
283 break; /* for safety */
284 case SIOCGETMIFCNT_IN6:
285 return(get_mif6_cnt((struct sioc_mif_req6 *)data));
286 break; /* for safety */
287 default:
288 return (EINVAL);
289 break;
290 }
291 return error;
292 }
293
294 /*
295 * returns the packet, byte, rpf-failure count for the source group provided
296 */
297 static int
298 get_sg_cnt(req)
299 register struct sioc_sg_req6 *req;
300 {
301 register struct mf6c *rt;
302 int s;
303
304 s = splsoftnet();
305 MF6CFIND(req->src.sin6_addr, req->grp.sin6_addr, rt);
306 splx(s);
307 if (rt != NULL) {
308 req->pktcnt = rt->mf6c_pkt_cnt;
309 req->bytecnt = rt->mf6c_byte_cnt;
310 req->wrong_if = rt->mf6c_wrong_if;
311 } else
312 return(ESRCH);
313 #if 0
314 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
315 #endif
316
317 return 0;
318 }
319
320 /*
321 * returns the input and output packet and byte counts on the mif provided
322 */
323 static int
324 get_mif6_cnt(req)
325 register struct sioc_mif_req6 *req;
326 {
327 register mifi_t mifi = req->mifi;
328
329 if (mifi >= nummifs)
330 return EINVAL;
331
332 req->icount = mif6table[mifi].m6_pkt_in;
333 req->ocount = mif6table[mifi].m6_pkt_out;
334 req->ibytes = mif6table[mifi].m6_bytes_in;
335 req->obytes = mif6table[mifi].m6_bytes_out;
336
337 return 0;
338 }
339
340 /*
341 * Get PIM processiong global
342 */
343 static int
344 get_pim6(m)
345 struct mbuf *m;
346 {
347 int *i;
348
349 i = mtod(m, int *);
350
351 *i = pim6;
352
353 return 0;
354 }
355
356 static int
357 set_pim6(i)
358 int *i;
359 {
360 if ((*i != 1) && (*i != 0))
361 return EINVAL;
362
363 pim6 = *i;
364
365 return 0;
366 }
367
368 /*
369 * Enable multicast routing
370 */
371 static int
372 ip6_mrouter_init(so, m)
373 struct socket *so;
374 struct mbuf *m;
375 {
376 int *v;
377
378 #ifdef MRT6DEBUG
379 if (mrt6debug)
380 log(LOG_DEBUG,
381 "ip6_mrouter_init: so_type = %d, pr_protocol = %d\n",
382 so->so_type, so->so_proto->pr_protocol);
383 #endif
384
385 if (so->so_type != SOCK_RAW ||
386 so->so_proto->pr_protocol != IPPROTO_ICMPV6)
387 return EOPNOTSUPP;
388
389 if (!m || (m->m_len != sizeof(int *)))
390 return ENOPROTOOPT;
391
392 v = mtod(m, int *);
393 if (*v != 1)
394 return ENOPROTOOPT;
395
396 if (ip6_mrouter != NULL) return EADDRINUSE;
397
398 ip6_mrouter = so;
399
400 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
401 bzero((caddr_t)nexpire, sizeof(nexpire));
402
403 pim6 = 0;/* used for stubbing out/in pim stuff */
404
405 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
406
407 #ifdef MRT6DEBUG
408 if (mrt6debug)
409 log(LOG_DEBUG, "ip6_mrouter_init\n");
410 #endif
411
412 return 0;
413 }
414
415 /*
416 * Disable multicast routing
417 */
418 int
419 ip6_mrouter_done()
420 {
421 mifi_t mifi;
422 int i;
423 struct ifnet *ifp;
424 struct in6_ifreq ifr;
425 struct mf6c *rt;
426 struct rtdetq *rte;
427 int s;
428
429 s = splsoftnet();
430
431 /*
432 * For each phyint in use, disable promiscuous reception of all IPv6
433 * multicasts.
434 */
435 #ifdef INET
436 #ifdef MROUTING
437 /*
438 * If there is still IPv4 multicast routing daemon,
439 * we remain interfaces to receive all muliticasted packets.
440 * XXX: there may be an interface in which the IPv4 multicast
441 * daemon is not interested...
442 */
443 if (!ip_mrouter)
444 #endif
445 #endif
446 {
447 for (mifi = 0; mifi < nummifs; mifi++) {
448 if (mif6table[mifi].m6_ifp &&
449 !(mif6table[mifi].m6_flags & MIFF_REGISTER)) {
450 ifr.ifr_addr.sin6_family = AF_INET6;
451 ifr.ifr_addr.sin6_addr= in6addr_any;
452 ifp = mif6table[mifi].m6_ifp;
453 (*ifp->if_ioctl)(ifp, SIOCDELMULTI,
454 (caddr_t)&ifr);
455 }
456 }
457 }
458 #ifdef notyet
459 bzero((caddr_t)qtable, sizeof(qtable));
460 bzero((caddr_t)tbftable, sizeof(tbftable));
461 #endif
462 bzero((caddr_t)mif6table, sizeof(mif6table));
463 nummifs = 0;
464
465 pim6 = 0; /* used to stub out/in pim specific code */
466
467 untimeout(expire_upcalls, (caddr_t)NULL);
468
469 /*
470 * Free all multicast forwarding cache entries.
471 */
472 for (i = 0; i < MF6CTBLSIZ; i++) {
473 rt = mf6ctable[i];
474 while (rt) {
475 struct mf6c *frt;
476
477 for (rte = rt->mf6c_stall; rte != NULL; ) {
478 struct rtdetq *n = rte->next;
479
480 m_free(rte->m);
481 free(rte, M_MRTABLE);
482 rte = n;
483 }
484 frt = rt;
485 rt = rt->mf6c_next;
486 free(frt, M_MRTABLE);
487 }
488 }
489
490 bzero((caddr_t)mf6ctable, sizeof(mf6ctable));
491
492 /*
493 * Reset de-encapsulation cache
494 */
495 reg_mif_num = -1;
496
497 ip6_mrouter = NULL;
498
499 splx(s);
500
501 #ifdef MRT6DEBUG
502 if (mrt6debug)
503 log(LOG_DEBUG, "ip6_mrouter_done\n");
504 #endif
505
506 return 0;
507 }
508
509 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
510
511 /*
512 * Add a mif to the mif table
513 */
514 static int
515 add_m6if(mifcp)
516 register struct mif6ctl *mifcp;
517 {
518 register struct mif6 *mifp;
519 struct ifnet *ifp;
520 struct in6_ifreq ifr;
521 int error, s;
522 #ifdef notyet
523 struct tbf *m_tbf = tbftable + mifcp->mif6c_mifi;
524 #endif
525
526 if (mifcp->mif6c_mifi >= MAXMIFS)
527 return EINVAL;
528 mifp = mif6table + mifcp->mif6c_mifi;
529 if (mifp->m6_ifp)
530 return EADDRINUSE; /* XXX: is it appropriate? */
531 if (mifcp->mif6c_pifi == 0 || mifcp->mif6c_pifi > if_index)
532 return ENXIO;
533 ifp = ifindex2ifnet[mifcp->mif6c_pifi];
534
535 if (mifcp->mif6c_flags & MIFF_REGISTER) {
536 if (reg_mif_num == (mifi_t)-1) {
537 strcpy(multicast_register_if.if_xname,
538 "register_mif"); /* XXX */
539 multicast_register_if.if_flags |= IFF_LOOPBACK;
540 multicast_register_if.if_index = mifcp->mif6c_mifi;
541 reg_mif_num = mifcp->mif6c_mifi;
542 }
543
544 ifp = &multicast_register_if;
545
546 } /* if REGISTER */
547 else {
548 /* Make sure the interface supports multicast */
549 if ((ifp->if_flags & IFF_MULTICAST) == 0)
550 return EOPNOTSUPP;
551
552 s = splsoftnet();
553 /*
554 * Enable promiscuous reception of all IPv6 multicasts
555 * from the interface.
556 */
557 ifr.ifr_addr.sin6_family = AF_INET6;
558 ifr.ifr_addr.sin6_addr = in6addr_any;
559 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr);
560 splx(s);
561 if (error)
562 return error;
563 }
564
565 s = splsoftnet();
566 mifp->m6_flags = mifcp->mif6c_flags;
567 mifp->m6_ifp = ifp;
568 #ifdef notyet
569 /* scaling up here allows division by 1024 in critical code */
570 mifp->m6_rate_limit = mifcp->mif6c_rate_limit * 1024 / 1000;
571 #endif
572 /* initialize per mif pkt counters */
573 mifp->m6_pkt_in = 0;
574 mifp->m6_pkt_out = 0;
575 mifp->m6_bytes_in = 0;
576 mifp->m6_bytes_out = 0;
577 splx(s);
578
579 /* Adjust nummifs up if the mifi is higher than nummifs */
580 if (nummifs <= mifcp->mif6c_mifi)
581 nummifs = mifcp->mif6c_mifi + 1;
582
583 #ifdef MRT6DEBUG
584 if (mrt6debug)
585 log(LOG_DEBUG,
586 "add_mif #%d, phyint %s%d\n",
587 mifcp->mif6c_mifi,
588 ifp->if_name, ifp->if_unit);
589 #endif
590
591 return 0;
592 }
593
594 /*
595 * Delete a mif from the mif table
596 */
597 static int
598 del_m6if(mifip)
599 mifi_t *mifip;
600 {
601 register struct mif6 *mifp = mif6table + *mifip;
602 register mifi_t mifi;
603 struct ifnet *ifp;
604 struct in6_ifreq ifr;
605 int s;
606
607 if (*mifip >= nummifs)
608 return EINVAL;
609 if (mifp->m6_ifp == NULL)
610 return EINVAL;
611
612 s = splsoftnet();
613
614 if (!(mifp->m6_flags & MIFF_REGISTER)) {
615 /*
616 * XXX: what if there is yet IPv4 multicast daemon
617 * using the interface?
618 */
619 ifp = mifp->m6_ifp;
620
621 ifr.ifr_addr.sin6_family = AF_INET6;
622 ifr.ifr_addr.sin6_addr = in6addr_any;
623 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr);
624 }
625
626 #ifdef notyet
627 bzero((caddr_t)qtable[*mifip], sizeof(qtable[*mifip]));
628 bzero((caddr_t)mifp->m6_tbf, sizeof(*(mifp->m6_tbf)));
629 #endif
630 bzero((caddr_t)mifp, sizeof (*mifp));
631
632 /* Adjust nummifs down */
633 for (mifi = nummifs; mifi > 0; mifi--)
634 if (mif6table[mifi - 1].m6_ifp)
635 break;
636 nummifs = mifi;
637
638 splx(s);
639
640 #ifdef MRT6DEBUG
641 if (mrt6debug)
642 log(LOG_DEBUG, "del_m6if %d, nummifs %d\n", *mifip, nummifs);
643 #endif
644
645 return 0;
646 }
647
648 /*
649 * Add an mfc entry
650 */
651 static int
652 add_m6fc(mfccp)
653 struct mf6cctl *mfccp;
654 {
655 struct mf6c *rt;
656 u_long hash;
657 struct rtdetq *rte;
658 register u_short nstl;
659 int s;
660
661 MF6CFIND(mfccp->mf6cc_origin.sin6_addr,
662 mfccp->mf6cc_mcastgrp.sin6_addr, rt);
663
664 /* If an entry already exists, just update the fields */
665 if (rt) {
666 #ifdef MRT6DEBUG
667 if (mrt6debug & DEBUG_MFC)
668 log(LOG_DEBUG,"add_m6fc update o %s g %s p %x\n",
669 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
670 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
671 mfccp->mf6cc_parent);
672 #endif
673
674 s = splsoftnet();
675 rt->mf6c_parent = mfccp->mf6cc_parent;
676 rt->mf6c_ifset = mfccp->mf6cc_ifset;
677 splx(s);
678 return 0;
679 }
680
681 /*
682 * Find the entry for which the upcall was made and update
683 */
684 s = splsoftnet();
685 hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
686 mfccp->mf6cc_mcastgrp.sin6_addr);
687 for (rt = mf6ctable[hash], nstl = 0; rt; rt = rt->mf6c_next) {
688 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
689 &mfccp->mf6cc_origin.sin6_addr) &&
690 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
691 &mfccp->mf6cc_mcastgrp.sin6_addr) &&
692 (rt->mf6c_stall != NULL)) {
693
694 if (nstl++)
695 log(LOG_ERR,
696 "add_m6fc: %s o %s g %s p %x dbx %p\n",
697 "multiple kernel entries",
698 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
699 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
700 mfccp->mf6cc_parent, rt->mf6c_stall);
701
702 #ifdef MRT6DEBUG
703 if (mrt6debug & DEBUG_MFC)
704 log(LOG_DEBUG,
705 "add_m6fc o %s g %s p %x dbg %x\n",
706 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
707 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
708 mfccp->mf6cc_parent, rt->mf6c_stall);
709 #endif
710
711 rt->mf6c_origin = mfccp->mf6cc_origin;
712 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
713 rt->mf6c_parent = mfccp->mf6cc_parent;
714 rt->mf6c_ifset = mfccp->mf6cc_ifset;
715 /* initialize pkt counters per src-grp */
716 rt->mf6c_pkt_cnt = 0;
717 rt->mf6c_byte_cnt = 0;
718 rt->mf6c_wrong_if = 0;
719
720 rt->mf6c_expire = 0; /* Don't clean this guy up */
721 nexpire[hash]--;
722
723 /* free packets Qed at the end of this entry */
724 for (rte = rt->mf6c_stall; rte != NULL; ) {
725 struct rtdetq *n = rte->next;
726 ip6_mdq(rte->m, rte->ifp, rt);
727 m_freem(rte->m);
728 #ifdef UPCALL_TIMING
729 collate(&(rte->t));
730 #endif /* UPCALL_TIMING */
731 free(rte, M_MRTABLE);
732 rte = n;
733 }
734 rt->mf6c_stall = NULL;
735 }
736 }
737
738 /*
739 * It is possible that an entry is being inserted without an upcall
740 */
741 if (nstl == 0) {
742 #ifdef MRT6DEBUG
743 if (mrt6debug & DEBUG_MFC)
744 log(LOG_DEBUG,"add_mfc no upcall h %d o %s g %s p %x\n",
745 hash,
746 ip6_sprintf(&mfccp->mf6cc_origin.sin6_addr),
747 ip6_sprintf(&mfccp->mf6cc_mcastgrp.sin6_addr),
748 mfccp->mf6cc_parent);
749 #endif
750
751 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
752
753 if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
754 &mfccp->mf6cc_origin.sin6_addr)&&
755 IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
756 &mfccp->mf6cc_mcastgrp.sin6_addr)) {
757
758 rt->mf6c_origin = mfccp->mf6cc_origin;
759 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
760 rt->mf6c_parent = mfccp->mf6cc_parent;
761 /* initialize pkt counters per src-grp */
762 rt->mf6c_pkt_cnt = 0;
763 rt->mf6c_byte_cnt = 0;
764 rt->mf6c_wrong_if = 0;
765
766 if (rt->mf6c_expire)
767 nexpire[hash]--;
768 rt->mf6c_expire = 0;
769 }
770 }
771 if (rt == NULL) {
772 /* no upcall, so make a new entry */
773 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
774 M_NOWAIT);
775 if (rt == NULL) {
776 splx(s);
777 return ENOBUFS;
778 }
779
780 /* insert new entry at head of hash chain */
781 rt->mf6c_origin = mfccp->mf6cc_origin;
782 rt->mf6c_mcastgrp = mfccp->mf6cc_mcastgrp;
783 rt->mf6c_parent = mfccp->mf6cc_parent;
784 /* initialize pkt counters per src-grp */
785 rt->mf6c_pkt_cnt = 0;
786 rt->mf6c_byte_cnt = 0;
787 rt->mf6c_wrong_if = 0;
788 rt->mf6c_expire = 0;
789 rt->mf6c_stall = NULL;
790
791 /* link into table */
792 rt->mf6c_next = mf6ctable[hash];
793 mf6ctable[hash] = rt;
794 }
795 }
796 splx(s);
797 return 0;
798 }
799
800 #ifdef UPCALL_TIMING
801 /*
802 * collect delay statistics on the upcalls
803 */
804 static void
805 collate(t)
806 register struct timeval *t;
807 {
808 register u_long d;
809 register struct timeval tp;
810 register u_long delta;
811
812 GET_TIME(tp);
813
814 if (TV_LT(*t, tp))
815 {
816 TV_DELTA(tp, *t, delta);
817
818 d = delta >> 10;
819 if (d > UPCALL_MAX)
820 d = UPCALL_MAX;
821
822 ++upcall_data[d];
823 }
824 }
825 #endif /* UPCALL_TIMING */
826
827 /*
828 * Delete an mfc entry
829 */
830 static int
831 del_m6fc(mfccp)
832 struct mf6cctl *mfccp;
833 {
834 struct sockaddr_in6 origin;
835 struct sockaddr_in6 mcastgrp;
836 struct mf6c *rt;
837 struct mf6c **nptr;
838 u_long hash;
839 int s;
840
841 origin = mfccp->mf6cc_origin;
842 mcastgrp = mfccp->mf6cc_mcastgrp;
843 hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
844
845 #ifdef MRT6DEBUG
846 if (mrt6debug & DEBUG_MFC)
847 log(LOG_DEBUG,"del_m6fc orig %s mcastgrp %s\n",
848 ip6_sprintf(&origin.sin6_addr),
849 ip6_sprintf(&mcastgrp.sin6_addr));
850 #endif
851
852 s = splsoftnet();
853
854 nptr = &mf6ctable[hash];
855 while ((rt = *nptr) != NULL) {
856 if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
857 &rt->mf6c_origin.sin6_addr) &&
858 IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
859 &rt->mf6c_mcastgrp.sin6_addr) &&
860 rt->mf6c_stall == NULL)
861 break;
862
863 nptr = &rt->mf6c_next;
864 }
865 if (rt == NULL) {
866 splx(s);
867 return EADDRNOTAVAIL;
868 }
869
870 *nptr = rt->mf6c_next;
871 free(rt, M_MRTABLE);
872
873 splx(s);
874
875 return 0;
876 }
877
878 static int
879 socket_send(s, mm, src)
880 struct socket *s;
881 struct mbuf *mm;
882 struct sockaddr_in6 *src;
883 {
884 if (s) {
885 if (sbappendaddr(&s->so_rcv,
886 (struct sockaddr *)src,
887 mm, (struct mbuf *)0) != 0) {
888 sorwakeup(s);
889 return 0;
890 }
891 }
892 m_freem(mm);
893 return -1;
894 }
895
896 /*
897 * IPv6 multicast forwarding function. This function assumes that the packet
898 * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
899 * pointed to by "ifp", and the packet is to be relayed to other networks
900 * that have members of the packet's destination IPv6 multicast group.
901 *
902 * The packet is returned unscathed to the caller, unless it is
903 * erroneous, in which case a non-zero return value tells the caller to
904 * discard it.
905 */
906
907 int
908 ip6_mforward(ip6, ifp, m)
909 register struct ip6_hdr *ip6;
910 struct ifnet *ifp;
911 struct mbuf *m;
912 {
913 register struct mf6c *rt;
914 register struct mif6 *mifp;
915 register struct mbuf *mm;
916 int s;
917 mifi_t mifi;
918
919 #ifdef MRT6DEBUG
920 if (mrt6debug & DEBUG_FORWARD)
921 log(LOG_DEBUG, "ip6_mforward: src %s, dst %s, ifindex %d\n",
922 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst),
923 ifp->if_index);
924 #endif
925
926 /*
927 * Don't forward a packet with Hop limit of zero or one,
928 * or a packet destined to a local-only group.
929 */
930 if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_NODELOCAL(&ip6->ip6_dst) ||
931 IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
932 return 0;
933 ip6->ip6_hlim--;
934
935 /*
936 * Determine forwarding mifs from the forwarding cache table
937 */
938 s = splsoftnet();
939 MF6CFIND(ip6->ip6_src, ip6->ip6_dst, rt);
940
941 /* Entry exists, so forward if necessary */
942 if (rt) {
943 splx(s);
944 return (ip6_mdq(m, ifp, rt));
945 } else {
946 /*
947 * If we don't have a route for packet's origin,
948 * Make a copy of the packet &
949 * send message to routing daemon
950 */
951
952 register struct mbuf *mb0;
953 register struct rtdetq *rte;
954 register u_long hash;
955 /* register int i, npkts;*/
956 #ifdef UPCALL_TIMING
957 struct timeval tp;
958
959 GET_TIME(tp);
960 #endif /* UPCALL_TIMING */
961
962 mrt6stat.mrt6s_no_route++;
963 #ifdef MRT6DEBUG
964 if (mrt6debug & (DEBUG_FORWARD | DEBUG_MFC))
965 log(LOG_DEBUG, "ip6_mforward: no rte s %s g %s\n",
966 ip6_sprintf(&ip6->ip6_src),
967 ip6_sprintf(&ip6->ip6_dst));
968 #endif
969
970 /*
971 * Allocate mbufs early so that we don't do extra work if we
972 * are just going to fail anyway.
973 */
974 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE,
975 M_NOWAIT);
976 if (rte == NULL) {
977 splx(s);
978 return ENOBUFS;
979 }
980 mb0 = m_copy(m, 0, M_COPYALL);
981 /*
982 * Pullup packet header if needed before storing it,
983 * as other references may modify it in the meantime.
984 */
985 if (mb0 &&
986 (M_HASCL(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
987 mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
988 if (mb0 == NULL) {
989 free(rte, M_MRTABLE);
990 splx(s);
991 return ENOBUFS;
992 }
993
994 /* is there an upcall waiting for this packet? */
995 hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
996 for (rt = mf6ctable[hash]; rt; rt = rt->mf6c_next) {
997 if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
998 &rt->mf6c_origin.sin6_addr) &&
999 IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1000 &rt->mf6c_mcastgrp.sin6_addr) &&
1001 (rt->mf6c_stall != NULL))
1002 break;
1003 }
1004
1005 if (rt == NULL) {
1006 struct mrt6msg *im;
1007
1008 /* no upcall, so make a new entry */
1009 rt = (struct mf6c *)malloc(sizeof(*rt), M_MRTABLE,
1010 M_NOWAIT);
1011 if (rt == NULL) {
1012 free(rte, M_MRTABLE);
1013 m_freem(mb0);
1014 splx(s);
1015 return ENOBUFS;
1016 }
1017 /*
1018 * Make a copy of the header to send to the user
1019 * level process
1020 */
1021 mm = m_copy(mb0, 0, sizeof(struct ip6_hdr));
1022
1023 if (mm == NULL) {
1024 free(rte, M_MRTABLE);
1025 m_freem(mb0);
1026 free(rt, M_MRTABLE);
1027 splx(s);
1028 return ENOBUFS;
1029 }
1030
1031 /*
1032 * Send message to routing daemon
1033 */
1034 sin6.sin6_addr = ip6->ip6_src;
1035
1036 im = mtod(mm, struct mrt6msg *);
1037 im->im6_msgtype = MRT6MSG_NOCACHE;
1038 im->im6_mbz = 0;
1039
1040 #ifdef MRT6DEBUG
1041 if (mrt6debug & DEBUG_FORWARD)
1042 log(LOG_DEBUG,
1043 "getting the iif info in the kernel\n");
1044 #endif
1045
1046 for (mifp = mif6table, mifi = 0;
1047 mifi < nummifs && mifp->m6_ifp != ifp;
1048 mifp++, mifi++)
1049 ;
1050
1051 im->im6_mif = mifi;
1052
1053 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1054 log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1055 "socket queue full\n");
1056 mrt6stat.mrt6s_upq_sockfull++;
1057 free(rte, M_MRTABLE);
1058 m_freem(mb0);
1059 free(rt, M_MRTABLE);
1060 splx(s);
1061 return ENOBUFS;
1062 }
1063
1064 mrt6stat.mrt6s_upcalls++;
1065
1066 /* insert new entry at head of hash chain */
1067 bzero(rt, sizeof(*rt));
1068 rt->mf6c_origin.sin6_family = AF_INET6;
1069 rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1070 rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1071 rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1072 rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1073 rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1074 rt->mf6c_expire = UPCALL_EXPIRE;
1075 nexpire[hash]++;
1076 rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1077
1078 /* link into table */
1079 rt->mf6c_next = mf6ctable[hash];
1080 mf6ctable[hash] = rt;
1081 /* Add this entry to the end of the queue */
1082 rt->mf6c_stall = rte;
1083 } else {
1084 /* determine if q has overflowed */
1085 struct rtdetq **p;
1086 register int npkts = 0;
1087
1088 for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1089 if (++npkts > MAX_UPQ6) {
1090 mrt6stat.mrt6s_upq_ovflw++;
1091 free(rte, M_MRTABLE);
1092 m_freem(mb0);
1093 splx(s);
1094 return 0;
1095 }
1096
1097 /* Add this entry to the end of the queue */
1098 *p = rte;
1099 }
1100
1101 rte->next = NULL;
1102 rte->m = mb0;
1103 rte->ifp = ifp;
1104 #ifdef UPCALL_TIMING
1105 rte->t = tp;
1106 #endif /* UPCALL_TIMING */
1107
1108 splx(s);
1109
1110 return 0;
1111 }
1112 }
1113
1114 /*
1115 * Clean up cache entries if upcalls are not serviced
1116 * Call from the Slow Timeout mechanism, every half second.
1117 */
1118 static void
1119 expire_upcalls(unused)
1120 void *unused;
1121 {
1122 struct rtdetq *rte;
1123 struct mf6c *mfc, **nptr;
1124 int i;
1125 int s;
1126
1127 s = splsoftnet();
1128 for (i = 0; i < MF6CTBLSIZ; i++) {
1129 if (nexpire[i] == 0)
1130 continue;
1131 nptr = &mf6ctable[i];
1132 while ((mfc = *nptr) != NULL) {
1133 rte = mfc->mf6c_stall;
1134 /*
1135 * Skip real cache entries
1136 * Make sure it wasn't marked to not expire (shouldn't happen)
1137 * If it expires now
1138 */
1139 if (rte != NULL &&
1140 mfc->mf6c_expire != 0 &&
1141 --mfc->mf6c_expire == 0) {
1142 #ifdef MRT6DEBUG
1143 if (mrt6debug & DEBUG_EXPIRE)
1144 log(LOG_DEBUG, "expire_upcalls: expiring (%s %s)\n",
1145 ip6_sprintf(&mfc->mf6c_origin.sin6_addr),
1146 ip6_sprintf(&mfc->mf6c_mcastgrp.sin6_addr));
1147 #endif
1148 /*
1149 * drop all the packets
1150 * free the mbuf with the pkt, if, timing info
1151 */
1152 do {
1153 struct rtdetq *n = rte->next;
1154 m_freem(rte->m);
1155 free(rte, M_MRTABLE);
1156 rte = n;
1157 } while (rte != NULL);
1158 mrt6stat.mrt6s_cache_cleanups++;
1159 nexpire[i]--;
1160
1161 *nptr = mfc->mf6c_next;
1162 free(mfc, M_MRTABLE);
1163 } else {
1164 nptr = &mfc->mf6c_next;
1165 }
1166 }
1167 }
1168 splx(s);
1169 timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT);
1170 }
1171
1172 /*
1173 * Packet forwarding routine once entry in the cache is made
1174 */
1175 static int
1176 ip6_mdq(m, ifp, rt)
1177 register struct mbuf *m;
1178 register struct ifnet *ifp;
1179 register struct mf6c *rt;
1180 {
1181 register struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1182 register mifi_t mifi, iif;
1183 register struct mif6 *mifp;
1184 register int plen = m->m_pkthdr.len;
1185
1186 /*
1187 * Macro to send packet on mif. Since RSVP packets don't get counted on
1188 * input, they shouldn't get counted on output, so statistics keeping is
1189 * seperate.
1190 */
1191
1192 #define MC6_SEND(ip6,mifp,m) { \
1193 if ((mifp)->m6_flags & MIFF_REGISTER) \
1194 register_send((ip6), (mifp), (m)); \
1195 else \
1196 phyint_send((ip6), (mifp), (m)); \
1197 }
1198
1199 /*
1200 * Don't forward if it didn't arrive from the parent mif
1201 * for its origin.
1202 */
1203 mifi = rt->mf6c_parent;
1204 if ((mifi >= nummifs) || (mif6table[mifi].m6_ifp != ifp)) {
1205 /* came in the wrong interface */
1206 #ifdef MRT6DEBUG
1207 if (mrt6debug & DEBUG_FORWARD)
1208 log(LOG_DEBUG,
1209 "wrong if: ifid %d mifi %d mififid %x\n",
1210 ifp->if_index, mifi,
1211 mif6table[mifi].m6_ifp->if_index);
1212 #endif
1213 mrt6stat.mrt6s_wrong_if++;
1214 rt->mf6c_wrong_if++;
1215 /*
1216 * If we are doing PIM processing, and we are forwarding
1217 * packets on this interface, send a message to the
1218 * routing daemon.
1219 */
1220 if(mifi < nummifs) /* have to make sure this is a valid mif */
1221 if(mif6table[mifi].m6_ifp)
1222
1223 if (pim6 && (m->m_flags & M_LOOP) == 0) {
1224 /*
1225 * Check the M_LOOP flag to avoid an
1226 * unnecessary PIM assert.
1227 * XXX: M_LOOP is an ad-hoc hack...
1228 */
1229 static struct sockaddr_in6 sin6 =
1230 { sizeof(sin6), AF_INET6 };
1231
1232 register struct mbuf *mm;
1233 struct mrt6msg *im;
1234
1235 mm = m_copy(m, 0,
1236 sizeof(struct ip6_hdr));
1237 if (mm &&
1238 (M_HASCL(mm) ||
1239 mm->m_len < sizeof(struct ip6_hdr)))
1240 mm = m_pullup(mm, sizeof(struct ip6_hdr));
1241 if (mm == NULL)
1242 return ENOBUFS;
1243
1244 im = mtod(mm, struct mrt6msg *);
1245 im->im6_msgtype = MRT6MSG_WRONGMIF;
1246 im->im6_mbz = 0;
1247
1248 for (mifp = mif6table, iif = 0;
1249 iif < nummifs && mifp &&
1250 mifp->m6_ifp != ifp;
1251 mifp++, iif++);
1252
1253 im->im6_mif = iif;
1254
1255 sin6.sin6_addr = im->im6_src;
1256
1257 mrt6stat.mrt6s_upcalls++;
1258
1259 if (socket_send(ip6_mrouter, mm,
1260 &sin6) < 0) {
1261 #ifdef MRT6DEBUG
1262 if (mrt6debug)
1263 log(LOG_WARNING, "mdq, ip6_mrouter socket queue full\n");
1264 #endif
1265 ++mrt6stat.mrt6s_upq_sockfull;
1266 return ENOBUFS;
1267 } /* if socket Q full */
1268 } /* if PIM */
1269 return 0;
1270 } /* if wrong iif */
1271
1272 /* If I sourced this packet, it counts as output, else it was input. */
1273 if (m->m_pkthdr.rcvif == NULL) {
1274 /* XXX: is rcvif really NULL when output?? */
1275 mif6table[mifi].m6_pkt_out++;
1276 mif6table[mifi].m6_bytes_out += plen;
1277 } else {
1278 mif6table[mifi].m6_pkt_in++;
1279 mif6table[mifi].m6_bytes_in += plen;
1280 }
1281 rt->mf6c_pkt_cnt++;
1282 rt->mf6c_byte_cnt += plen;
1283
1284 /*
1285 * For each mif, forward a copy of the packet if there are group
1286 * members downstream on the interface.
1287 */
1288 for (mifp = mif6table, mifi = 0; mifi < nummifs; mifp++, mifi++)
1289 if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1290 mifp->m6_pkt_out++;
1291 mifp->m6_bytes_out += plen;
1292 MC6_SEND(ip6, mifp, m);
1293 }
1294 return 0;
1295 }
1296
1297 static void
1298 phyint_send(ip6, mifp, m)
1299 struct ip6_hdr *ip6;
1300 struct mif6 *mifp;
1301 struct mbuf *m;
1302 {
1303 register struct mbuf *mb_copy;
1304 struct ifnet *ifp = mifp->m6_ifp;
1305 int error = 0;
1306 int s = splsoftnet();
1307 static struct route_in6 ro6;
1308 struct in6_multi *in6m;
1309
1310 /*
1311 * Make a new reference to the packet; make sure that
1312 * the IPv6 header is actually copied, not just referenced,
1313 * so that ip6_output() only scribbles on the copy.
1314 */
1315 mb_copy = m_copy(m, 0, M_COPYALL);
1316 if (mb_copy &&
1317 (M_HASCL(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1318 mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1319 if (mb_copy == NULL)
1320 return;
1321 /* set MCAST flag to the outgoing packet */
1322 mb_copy->m_flags |= M_MCAST;
1323
1324 /*
1325 * If we sourced the packet, call ip6_output since we may devide
1326 * the packet into fragments when the packet is too big for the
1327 * outgoing interface.
1328 * Otherwise, we can simply send the packet to the interface
1329 * sending queue.
1330 */
1331 if (m->m_pkthdr.rcvif == NULL) {
1332 struct ip6_moptions im6o;
1333
1334 im6o.im6o_multicast_ifp = ifp;
1335 /* XXX: ip6_output will override ip6->ip6_hlim */
1336 im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1337 im6o.im6o_multicast_loop = 1;
1338 error = ip6_output(mb_copy, NULL, &ro6,
1339 IPV6_FORWARDING, &im6o, NULL);
1340
1341 #ifdef MRT6DEBUG
1342 if (mrt6debug & DEBUG_XMIT)
1343 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1344 mifp - mif6table, error);
1345 #endif
1346 splx(s);
1347 return;
1348 }
1349
1350 /*
1351 * If we belong to the destination multicast group
1352 * on the outgoing interface, loop back a copy.
1353 */
1354 IN6_LOOKUP_MULTI(ip6->ip6_dst, ifp, in6m);
1355 if (in6m != NULL) {
1356 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
1357 ro6.ro_dst.sin6_family = AF_INET6;
1358 ro6.ro_dst.sin6_addr = ip6->ip6_dst;
1359 ip6_mloopback(ifp, m, &ro6.ro_dst);
1360 }
1361 /*
1362 * Put the packet into the sending queue of the outgoing interface
1363 * if it would fit in the MTU of the interface.
1364 */
1365 if (mb_copy->m_pkthdr.len < ifp->if_mtu || ifp->if_mtu < IPV6_MMTU) {
1366 ro6.ro_dst.sin6_len = sizeof(struct sockaddr_in6);
1367 ro6.ro_dst.sin6_family = AF_INET6;
1368 ro6.ro_dst.sin6_addr = ip6->ip6_dst;
1369 /*
1370 * We just call if_output instead of nd6_output here, since
1371 * we need no ND for a multicast forwarded packet...right?
1372 */
1373 error = (*ifp->if_output)(ifp, mb_copy,
1374 (struct sockaddr *)&ro6.ro_dst,
1375 NULL);
1376 #ifdef MRT6DEBUG
1377 if (mrt6debug & DEBUG_XMIT)
1378 log(LOG_DEBUG, "phyint_send on mif %d err %d\n",
1379 mifp - mif6table, error);
1380 #endif
1381 }
1382 else {
1383 #ifdef MULTICAST_PMTUD
1384 icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, ifp->if_mtu);
1385 return;
1386 #else
1387 #ifdef MRT6DEBUG
1388 if (mrt6debug & DEBUG_DEBUG_XMIT)
1389 log(LOG_DEBUG,
1390 "phyint_send: packet too big on %s o %s g %s"
1391 " size %d(discarded)\n",
1392 ifp->if_xname,
1393 ip6_sprintf(&ip6->ip6_src),
1394 ip6_sprintf(&ip6->ip6_dst),
1395 mb_copy->m_pkthdr.len);
1396 #endif /* MRT6DEBUG */
1397 m_freem(mb_copy); /* simply discard the packet */
1398 return;
1399 #endif
1400 }
1401 }
1402
1403 static int
1404 register_send(ip6, mif, m)
1405 register struct ip6_hdr *ip6;
1406 struct mif6 *mif;
1407 register struct mbuf *m;
1408 {
1409 register struct mbuf *mm;
1410 register int i, len = m->m_pkthdr.len;
1411 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1412 struct mrt6msg *im6;
1413
1414 #ifdef MRT6DEBUG
1415 if (mrt6debug)
1416 log(LOG_DEBUG, "** IPv6 register_send **\n src %s dst %s\n",
1417 ip6_sprintf(&ip6->ip6_src), ip6_sprintf(&ip6->ip6_dst));
1418 #endif
1419 ++pim6stat.pim6s_snd_registers;
1420
1421 /* Make a copy of the packet to send to the user level process */
1422 MGETHDR(mm, M_DONTWAIT, MT_HEADER);
1423 if (mm == NULL)
1424 return ENOBUFS;
1425 mm->m_data += max_linkhdr;
1426 mm->m_len = sizeof(struct ip6_hdr);
1427
1428 if ((mm->m_next = m_copy(m, 0, M_COPYALL)) == NULL) {
1429 m_freem(mm);
1430 return ENOBUFS;
1431 }
1432 i = MHLEN - M_LEADINGSPACE(mm);
1433 if (i > len)
1434 i = len;
1435 mm = m_pullup(mm, i);
1436 if (mm == NULL){
1437 m_freem(mm);
1438 return ENOBUFS;
1439 }
1440 /* TODO: check it! */
1441 mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1442
1443 /*
1444 * Send message to routing daemon
1445 */
1446 sin6.sin6_addr = ip6->ip6_src;
1447
1448 im6 = mtod(mm, struct mrt6msg *);
1449 im6->im6_msgtype = MRT6MSG_WHOLEPKT;
1450 im6->im6_mbz = 0;
1451
1452 im6->im6_mif = mif - mif6table;
1453
1454 /* iif info is not given for reg. encap.n */
1455 mrt6stat.mrt6s_upcalls++;
1456
1457 if (socket_send(ip6_mrouter, mm, &sin6) < 0) {
1458 #ifdef MRT6DEBUG
1459 if (mrt6debug)
1460 log(LOG_WARNING,
1461 "register_send: ip_mrouter socket queue full\n");
1462 #endif
1463 ++mrt6stat.mrt6s_upq_sockfull;
1464 return ENOBUFS;
1465 }
1466 return 0;
1467 }
1468
1469 /*
1470 * PIM sparse mode hook
1471 * Receives the pim control messages, and passes them up to the listening
1472 * socket, using rip6_input.
1473 * The only message processed is the REGISTER pim message; the pim header
1474 * is stripped off, and the inner packet is passed to register_mforward.
1475 */
1476 int
1477 pim6_input(mp, offp, proto)
1478 struct mbuf **mp;
1479 int *offp, proto;
1480 {
1481 register struct pim *pim; /* pointer to a pim struct */
1482 register struct ip6_hdr *ip6;
1483 register int pimlen;
1484 struct mbuf *m = *mp;
1485 int minlen;
1486 int off = *offp;
1487
1488 ++pim6stat.pim6s_rcv_total;
1489
1490 ip6 = mtod(m, struct ip6_hdr *);
1491 pimlen = m->m_pkthdr.len - *offp;
1492
1493 /*
1494 * Validate lengths
1495 */
1496 if (pimlen < PIM_MINLEN) {
1497 ++pim6stat.pim6s_rcv_tooshort;
1498 #ifdef MRT6DEBUG
1499 if (mrt6debug & DEBUG_PIM)
1500 log(LOG_DEBUG,"pim6_input: PIM packet too short\n");
1501 #endif
1502 m_freem(m);
1503 return(IPPROTO_DONE);
1504 }
1505
1506 /*
1507 * if the packet is at least as big as a REGISTER, go ahead
1508 * and grab the PIM REGISTER header size, to avoid another
1509 * possible m_pullup() later.
1510 *
1511 * PIM_MINLEN == pimhdr + u_int32 == 8
1512 * PIM6_REG_MINLEN == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1513 */
1514 minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1515
1516 /*
1517 * Make sure that the IP6 and PIM headers in contiguous memory, and
1518 * possibly the PIM REGISTER header
1519 */
1520 #ifndef PULLDOWN_TEST
1521 IP6_EXTHDR_CHECK(m, off, minlen, IPPROTO_DONE);
1522 /* adjust pointer */
1523 ip6 = mtod(m, struct ip6_hdr *);
1524
1525 /* adjust mbuf to point to the PIM header */
1526 pim = (struct pim *)((caddr_t)ip6 + off);
1527 #else
1528 IP6_EXTHDR_GET(pim, struct pim *, m, off, minlen);
1529 if (pim == NULL) {
1530 pim6stat.pim6s_rcv_tooshort++;
1531 return IPPROTO_DONE;
1532 }
1533 #endif
1534
1535 #define PIM6_CHECKSUM
1536 #ifdef PIM6_CHECKSUM
1537 {
1538 int cksumlen;
1539
1540 /*
1541 * Validate checksum.
1542 * If PIM REGISTER, exclude the data packet
1543 */
1544 if (pim->pim_type == PIM_REGISTER)
1545 cksumlen = PIM_MINLEN;
1546 else
1547 cksumlen = pimlen;
1548
1549 if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1550 ++pim6stat.pim6s_rcv_badsum;
1551 #ifdef MRT6DEBUG
1552 if (mrt6debug & DEBUG_PIM)
1553 log(LOG_DEBUG,
1554 "pim6_input: invalid checksum\n");
1555 #endif
1556 m_freem(m);
1557 return(IPPROTO_DONE);
1558 }
1559 }
1560 #endif /* PIM_CHECKSUM */
1561
1562 /* PIM version check */
1563 if (pim->pim_ver != PIM_VERSION) {
1564 ++pim6stat.pim6s_rcv_badversion;
1565 #ifdef MRT6DEBUG
1566 log(LOG_ERR,
1567 "pim6_input: incorrect version %d, expecting %d\n",
1568 pim->pim_ver, PIM_VERSION);
1569 #endif
1570 m_freem(m);
1571 return(IPPROTO_DONE);
1572 }
1573
1574 if (pim->pim_type == PIM_REGISTER) {
1575 /*
1576 * since this is a REGISTER, we'll make a copy of the register
1577 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1578 * routing daemon.
1579 */
1580 static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1581
1582 struct mbuf *mcp;
1583 struct ip6_hdr *eip6;
1584 u_int32_t *reghdr;
1585 int rc;
1586
1587 ++pim6stat.pim6s_rcv_registers;
1588
1589 if ((reg_mif_num >= nummifs) || (reg_mif_num == (mifi_t) -1)) {
1590 #ifdef MRT6DEBUG
1591 if (mrt6debug & DEBUG_PIM)
1592 log(LOG_DEBUG,
1593 "pim6_input: register mif not set: %d\n",
1594 reg_mif_num);
1595 #endif
1596 m_freem(m);
1597 return(IPPROTO_DONE);
1598 }
1599
1600 reghdr = (u_int32_t *)(pim + 1);
1601
1602 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1603 goto pim6_input_to_daemon;
1604
1605 /*
1606 * Validate length
1607 */
1608 if (pimlen < PIM6_REG_MINLEN) {
1609 ++pim6stat.pim6s_rcv_tooshort;
1610 ++pim6stat.pim6s_rcv_badregisters;
1611 #ifdef MRT6DEBUG
1612 log(LOG_ERR,
1613 "pim6_input: register packet size too "
1614 "small %d from %s\n",
1615 pimlen, ip6_sprintf(&ip6->ip6_src));
1616 #endif
1617 m_freem(m);
1618 return(IPPROTO_DONE);
1619 }
1620
1621 eip6 = (struct ip6_hdr *) (reghdr + 1);
1622 #ifdef MRT6DEBUG
1623 if (mrt6debug & DEBUG_PIM)
1624 log(LOG_DEBUG,
1625 "pim6_input[register], eip6: %s -> %s, "
1626 "eip6 plen %d\n",
1627 ip6_sprintf(&eip6->ip6_src),
1628 ip6_sprintf(&eip6->ip6_dst),
1629 ntohs(eip6->ip6_plen));
1630 #endif
1631
1632 /* verify the inner packet is destined to a mcast group */
1633 if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1634 ++pim6stat.pim6s_rcv_badregisters;
1635 #ifdef MRT6DEBUG
1636 if (mrt6debug & DEBUG_PIM)
1637 log(LOG_DEBUG,
1638 "pim6_input: inner packet of register "
1639 "is not multicast %s\n",
1640 ip6_sprintf(&eip6->ip6_dst));
1641 #endif
1642 m_freem(m);
1643 return(IPPROTO_DONE);
1644 }
1645
1646 /*
1647 * make a copy of the whole header to pass to the daemon later.
1648 */
1649 mcp = m_copy(m, 0, off + PIM6_REG_MINLEN);
1650 if (mcp == NULL) {
1651 #ifdef MRT6DEBUG
1652 log(LOG_ERR,
1653 "pim6_input: pim register: "
1654 "could not copy register head\n");
1655 #endif
1656 m_freem(m);
1657 return(IPPROTO_DONE);
1658 }
1659
1660 /*
1661 * forward the inner ip6 packet; point m_data at the inner ip6.
1662 */
1663 m_adj(m, off + PIM_MINLEN);
1664 #ifdef MRT6DEBUG
1665 if (mrt6debug & DEBUG_PIM) {
1666 log(LOG_DEBUG,
1667 "pim6_input: forwarding decapsulated register: "
1668 "src %s, dst %s, mif %d\n",
1669 ip6_sprintf(&eip6->ip6_src),
1670 ip6_sprintf(&eip6->ip6_dst),
1671 reg_mif_num);
1672 }
1673 #endif
1674
1675 rc = looutput(mif6table[reg_mif_num].m6_ifp, m,
1676 (struct sockaddr *) &dst,
1677 (struct rtentry *) NULL);
1678
1679 /* prepare the register head to send to the mrouting daemon */
1680 m = mcp;
1681 }
1682
1683 /*
1684 * Pass the PIM message up to the daemon; if it is a register message
1685 * pass the 'head' only up to the daemon. This includes the
1686 * encapsulator ip6 header, pim header, register header and the
1687 * encapsulated ip6 header.
1688 */
1689 pim6_input_to_daemon:
1690 rip6_input(&m, offp, proto);
1691 return(IPPROTO_DONE);
1692 }
1693