xref: /freebsd/sys/netinet6/ip6_mroute.c (revision 18b7115cba2f698909a4801dc2cc1b04b1f4f210)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
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  *	$KAME: ip6_mroute.c,v 1.58 2001/12/18 02:36:31 itojun Exp $
32  */
33 
34 /*-
35  * Copyright (c) 1989 Stephen Deering
36  * Copyright (c) 1992, 1993
37  *      The Regents of the University of California.  All rights reserved.
38  *
39  * This code is derived from software contributed to Berkeley by
40  * Stephen Deering of Stanford University.
41  *
42  * Redistribution and use in source and binary forms, with or without
43  * modification, are permitted provided that the following conditions
44  * are met:
45  * 1. Redistributions of source code must retain the above copyright
46  *    notice, this list of conditions and the following disclaimer.
47  * 2. Redistributions in binary form must reproduce the above copyright
48  *    notice, this list of conditions and the following disclaimer in the
49  *    documentation and/or other materials provided with the distribution.
50  * 3. Neither the name of the University nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *	BSDI ip_mroute.c,v 2.10 1996/11/14 00:29:52 jch Exp
66  */
67 
68 /*
69  * IP multicast forwarding procedures
70  *
71  * Written by David Waitzman, BBN Labs, August 1988.
72  * Modified by Steve Deering, Stanford, February 1989.
73  * Modified by Mark J. Steiglitz, Stanford, May, 1991
74  * Modified by Van Jacobson, LBL, January 1993
75  * Modified by Ajit Thyagarajan, PARC, August 1993
76  * Modified by Bill Fenner, PARC, April 1994
77  *
78  * MROUTING Revision: 3.5.1.2 + PIM-SMv2 (pimd) Support
79  */
80 
81 #include "opt_inet6.h"
82 
83 #include <sys/param.h>
84 #include <sys/callout.h>
85 #include <sys/errno.h>
86 #include <sys/eventhandler.h>
87 #include <sys/kernel.h>
88 #include <sys/lock.h>
89 #include <sys/malloc.h>
90 #include <sys/mbuf.h>
91 #include <sys/module.h>
92 #include <sys/domain.h>
93 #include <sys/priv.h>
94 #include <sys/proc.h>
95 #include <sys/protosw.h>
96 #include <sys/sdt.h>
97 #include <sys/signalvar.h>
98 #include <sys/socket.h>
99 #include <sys/socketvar.h>
100 #include <sys/sockio.h>
101 #include <sys/sx.h>
102 #include <sys/sysctl.h>
103 #include <sys/syslog.h>
104 #include <sys/systm.h>
105 #include <sys/time.h>
106 
107 #include <net/if.h>
108 #include <net/if_var.h>
109 #include <net/if_private.h>
110 #include <net/if_types.h>
111 #include <net/route.h>
112 #include <net/vnet.h>
113 
114 #include <netinet/in.h>
115 #include <netinet/in_var.h>
116 #include <netinet/icmp6.h>
117 #include <netinet/ip_encap.h>
118 
119 #include <netinet/ip6.h>
120 #include <netinet/in_kdtrace.h>
121 #include <netinet6/ip6_var.h>
122 #include <netinet6/scope6_var.h>
123 #include <netinet6/nd6.h>
124 #include <netinet6/ip6_mroute.h>
125 #include <netinet6/pim6.h>
126 #include <netinet6/pim6_var.h>
127 
128 static MALLOC_DEFINE(M_MRTABLE6, "mf6c", "multicast forwarding cache entry");
129 
130 struct mf6ctable;
131 
132 static int	ip6_mdq(struct mf6ctable *, struct mbuf *, struct ifnet *,
133 		    struct mf6c *);
134 static void	phyint_send(struct ip6_hdr *, struct mif6 *, struct mbuf *);
135 static int	register_send(struct mf6ctable *, struct ip6_hdr *, mifi_t,
136 		    struct mbuf *);
137 static int	set_pim6(int *);
138 static int	socket_send(struct socket *, struct mbuf *,
139 		    struct sockaddr_in6 *);
140 
141 extern int in6_mcast_loop;
142 extern struct domain inet6domain;
143 
144 static const struct encaptab *pim6_encap_cookie;
145 static int pim6_encapcheck(const struct mbuf *, int, int, void *);
146 static int pim6_input(struct mbuf *, int, int, void *);
147 
148 static const struct encap_config ipv6_encap_cfg = {
149 	.proto = IPPROTO_PIM,
150 	.min_length = sizeof(struct ip6_hdr) + PIM_MINLEN,
151 	.exact_match = 8,
152 	.check = pim6_encapcheck,
153 	.input = pim6_input
154 };
155 
156 SYSCTL_DECL(_net_inet6);
157 SYSCTL_DECL(_net_inet6_ip6);
158 static SYSCTL_NODE(_net_inet6, IPPROTO_PIM, pim,
159     CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
160     "PIM");
161 
162 VNET_DEFINE_STATIC(struct mrt6stat, mrt6stat);
163 #define	V_mrt6stat	VNET(mrt6stat)
164 SYSCTL_STRUCT(_net_inet6_ip6, OID_AUTO, mrt6stat, CTLFLAG_VNET | CTLFLAG_RW,
165     &VNET_NAME(mrt6stat), mrt6stat,
166     "Multicast Routing Statistics (struct mrt6stat, netinet6/ip6_mroute.h)");
167 
168 #define	MRT6STAT_INC(name)	V_mrt6stat.name += 1
169 #define NO_RTE_FOUND	0x1
170 #define RTE_FOUND	0x2
171 
172 static struct sx mrouter6_mtx;
173 #define	MROUTER6_LOCKPTR()	(&mrouter6_mtx)
174 #define	MROUTER6_LOCK()		sx_xlock(MROUTER6_LOCKPTR())
175 #define	MROUTER6_UNLOCK()	sx_xunlock(MROUTER6_LOCKPTR())
176 #define	MROUTER6_LOCK_ASSERT()	sx_assert(MROUTER6_LOCKPTR(), SA_XLOCKED
177 #define	MROUTER6_LOCK_INIT()	sx_init(MROUTER6_LOCKPTR(), "mrouter6")
178 #define	MROUTER6_LOCK_DESTROY()	sx_destroy(MROUTER6_LOCKPTR())
179 
180 static struct mtx mfc6_mtx;
181 #define	MFC6_LOCKPTR()		(&mfc6_mtx)
182 #define	MFC6_LOCK()		mtx_lock(MFC6_LOCKPTR())
183 #define	MFC6_UNLOCK()		mtx_unlock(MFC6_LOCKPTR())
184 #define	MFC6_LOCK_ASSERT()	mtx_assert(MFC6_LOCKPTR(), MA_OWNED)
185 #define	MFC6_LOCK_INIT()	mtx_init(MFC6_LOCKPTR(),		\
186 				    "IPv6 multicast forwarding cache",	\
187 				    NULL, MTX_DEF)
188 #define	MFC6_LOCK_DESTROY()	mtx_destroy(MFC6_LOCKPTR())
189 
190 struct mf6ctable {
191 	struct socket	*router;
192 	int		router_ver;
193 	struct mf6c	*mfchashtbl[MF6CTBLSIZ];
194 	u_char		nexpire[MF6CTBLSIZ];
195 	int		nummifs;
196 	struct mif6	miftable[MAXMIFS];
197 
198 	/*
199 	 * 'Interfaces' associated with decapsulator (so we can tell packets
200 	 * that went through it from ones that get reflected by a broken
201 	 * gateway).  Different from IPv4 register_if, these interfaces are
202 	 * linked into the system ifnet list, because per-interface IPv6
203 	 * statistics are maintained in ifp->if_afdata.  But it does not have
204 	 * any routes point to them.  I.e., packets can't be sent this way.
205 	 * They only exist as a placeholder for multicast source verification.
206 	 */
207 	struct ifnet	*register_if;
208 	mifi_t		register_mif;
209 };
210 
211 VNET_DEFINE_STATIC(struct mf6ctable *, mfctables);
212 #define	V_mfctables		VNET(mfctables)
213 VNET_DEFINE_STATIC(uint32_t, nmfctables);
214 #define	V_nmfctables		VNET(nmfctables)
215 
216 static eventhandler_tag ifdetach_tag, rtnumfibs_change_tag;
217 
218 static int
sysctl_mfctable(SYSCTL_HANDLER_ARGS)219 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
220 {
221 	int fibnum;
222 
223 	fibnum = curthread->td_proc->p_fibnum;
224 	return (SYSCTL_OUT(req, &V_mfctables[fibnum].mfchashtbl,
225 	    sizeof(struct mf6c *) * MF6CTBLSIZ));
226 }
227 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, mf6ctable,
228     CTLTYPE_OPAQUE | CTLFLAG_RD,
229     NULL, 0, sysctl_mfctable, "S,*mf6c[MF6CTBLSIZ]",
230     "IPv6 Multicast Forwarding Table (struct mf6c *[MF6CTBLSIZ], "
231     "netinet6/ip6_mroute.h)");
232 
233 static int
sysctl_mif6table(SYSCTL_HANDLER_ARGS)234 sysctl_mif6table(SYSCTL_HANDLER_ARGS)
235 {
236 	struct mif6_sctl *out;
237 	struct mf6ctable *mfct;
238 	int error;
239 
240 	mfct = &V_mfctables[curthread->td_proc->p_fibnum];
241 	out = malloc(sizeof(struct mif6_sctl) * MAXMIFS, M_TEMP,
242 	    M_WAITOK | M_ZERO);
243 	for (int i = 0; i < MAXMIFS; i++) {
244 		struct mif6_sctl *outp = &out[i];
245 		struct mif6 *mifp = &mfct->miftable[i];
246 
247 		outp->m6_flags = mifp->m6_flags;
248 		outp->m6_rate_limit = mifp->m6_rate_limit;
249 		outp->m6_lcl_addr = mifp->m6_lcl_addr;
250 		if (mifp->m6_ifp != NULL)
251 			outp->m6_ifp = mifp->m6_ifp->if_index;
252 		else
253 			outp->m6_ifp = 0;
254 		outp->m6_pkt_in	= mifp->m6_pkt_in;
255 		outp->m6_pkt_out = mifp->m6_pkt_out;
256 		outp->m6_bytes_in = mifp->m6_bytes_in;
257 		outp->m6_bytes_out = mifp->m6_bytes_out;
258 	}
259 	error = SYSCTL_OUT(req, out, sizeof(struct mif6_sctl) * MAXMIFS);
260 	free(out, M_TEMP);
261 	return (error);
262 }
263 SYSCTL_PROC(_net_inet6_ip6, OID_AUTO, mif6table,
264     CTLTYPE_OPAQUE | CTLFLAG_RD | CTLFLAG_NEEDGIANT,
265     NULL, 0, sysctl_mif6table, "S,mif6_sctl[MAXMIFS]",
266     "IPv6 Multicast Interfaces (struct mif6_sctl[MAXMIFS], "
267     "netinet6/ip6_mroute.h)");
268 
269 static struct mtx mif6_mtx;
270 #define	MIF6_LOCKPTR()		(&mif6_mtx)
271 #define	MIF6_LOCK()		mtx_lock(MIF6_LOCKPTR())
272 #define	MIF6_UNLOCK()		mtx_unlock(MIF6_LOCKPTR())
273 #define	MIF6_LOCK_ASSERT()	mtx_assert(MIF6_LOCKPTR(), MA_OWNED)
274 #define	MIF6_LOCK_INIT()	\
275 	mtx_init(MIF6_LOCKPTR(), "IPv6 multicast interfaces", NULL, MTX_DEF)
276 #define	MIF6_LOCK_DESTROY()	mtx_destroy(MIF6_LOCKPTR())
277 
278 #ifdef MRT6DEBUG
279 VNET_DEFINE_STATIC(u_int, mrt6debug) = 0;	/* debug level */
280 #define	V_mrt6debug		VNET(mrt6debug)
281 #define DEBUG_MFC	0x02
282 #define DEBUG_FORWARD	0x04
283 #define DEBUG_EXPIRE	0x08
284 #define DEBUG_XMIT	0x10
285 #define DEBUG_REG	0x20
286 #define DEBUG_PIM	0x40
287 #define	DEBUG_ERR	0x80
288 #define	DEBUG_ANY	0x7f
289 #define	MRT6_DLOG(m, fmt, ...)	\
290 	if (V_mrt6debug & (m))	\
291 		log(((m) & DEBUG_ERR) ? LOG_ERR: LOG_DEBUG, \
292 		    "%s: " fmt "\n", __func__, ##__VA_ARGS__)
293 #else
294 #define	MRT6_DLOG(m, fmt, ...)
295 #endif
296 
297 static void	expire_upcalls(struct mf6ctable *);
298 static void	expire_upcalls_all(void *);
299 #define	EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second */
300 #define	UPCALL_EXPIRE	6		/* number of timeouts */
301 
302 VNET_DEFINE_STATIC(struct pim6stat, pim6stat);
303 #define	V_pim6stat	VNET(pim6stat)
304 SYSCTL_STRUCT(_net_inet6_pim, PIM6CTL_STATS, stats, CTLFLAG_VNET | CTLFLAG_RW,
305     &VNET_NAME(pim6stat), pim6stat,
306     "PIM Statistics (struct pim6stat, netinet6/pim6_var.h)");
307 
308 #define	PIM6STAT_INC(name)	V_pim6stat.name += 1
309 VNET_DEFINE_STATIC(int, pim6);
310 #define	V_pim6		VNET(pim6)
311 
312 /*
313  * Hash function for a source, group entry
314  */
315 #define MF6CHASH(a, g) MF6CHASHMOD((a).s6_addr32[0] ^ (a).s6_addr32[1] ^ \
316 				   (a).s6_addr32[2] ^ (a).s6_addr32[3] ^ \
317 				   (g).s6_addr32[0] ^ (g).s6_addr32[1] ^ \
318 				   (g).s6_addr32[2] ^ (g).s6_addr32[3])
319 
320 /*
321  * Macros to compute elapsed time efficiently
322  * Borrowed from Van Jacobson's scheduling code
323  * XXX: replace with timersub() ?
324  */
325 #define TV_DELTA(a, b, delta) do { \
326 	    int xxs; \
327 		\
328 	    delta = (a).tv_usec - (b).tv_usec; \
329 	    if ((xxs = (a).tv_sec - (b).tv_sec)) { \
330 	       switch (xxs) { \
331 		      case 2: \
332 			  delta += 1000000; \
333 			      /* FALLTHROUGH */ \
334 		      case 1: \
335 			  delta += 1000000; \
336 			  break; \
337 		      default: \
338 			  delta += (1000000 * xxs); \
339 	       } \
340 	    } \
341 } while (/*CONSTCOND*/ 0)
342 
343 /* XXX: replace with timercmp(a, b, <) ? */
344 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
345 	      (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
346 
347 #ifdef UPCALL_TIMING
348 #define UPCALL_MAX	50
349 static u_long upcall_data[UPCALL_MAX + 1];
350 static void collate(struct timeval *);
351 #endif /* UPCALL_TIMING */
352 
353 static int ip6_mrouter_init(struct socket *, int, int);
354 static int add_m6fc(struct mf6ctable *, struct mf6cctl *);
355 static int add_m6if(struct mf6ctable *, int, struct mif6ctl *);
356 static int del_m6fc(struct mf6ctable *, struct mf6cctl *);
357 static int del_m6if(struct mf6ctable *, mifi_t);
358 static int del_m6if_locked(struct mf6ctable *, mifi_t);
359 static int get_mif6_cnt(struct mf6ctable *, struct sioc_mif_req6 *);
360 static int get_sg_cnt(struct mf6ctable *, struct sioc_sg_req6 *);
361 
362 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
363 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
364 
365 static int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *);
366 static void X_ip6_mrouter_done(struct socket *);
367 static int X_ip6_mrouter_set(struct socket *, struct sockopt *);
368 static int X_ip6_mrouter_get(struct socket *, struct sockopt *);
369 static int X_mrt6_ioctl(u_long, caddr_t, int);
370 
371 static struct mf6c *
mf6c_find(const struct mf6ctable * mfct,const struct in6_addr * origin,const struct in6_addr * group)372 mf6c_find(const struct mf6ctable *mfct, const struct in6_addr *origin,
373     const struct in6_addr *group)
374 {
375 	MFC6_LOCK_ASSERT();
376 
377 	for (struct mf6c *rt = mfct->mfchashtbl[MF6CHASH(*origin, *group)];
378 	    rt != NULL; rt = rt->mf6c_next) {
379 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr, origin) &&
380 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr, group) &&
381 		    rt->mf6c_stall == NULL)
382 			return (rt);
383 	}
384 	MRT6STAT_INC(mrt6s_mfc_misses);
385 	return (NULL);
386 }
387 
388 static struct mf6ctable *
somfctable(struct socket * so)389 somfctable(struct socket *so)
390 {
391 	int fib;
392 
393 	fib = atomic_load_int(&so->so_fibnum);
394 	KASSERT(fib >= 0 && fib < V_nmfctables,
395 	    ("%s: so_fibnum %d out of range", __func__, fib));
396 	return (&V_mfctables[fib]);
397 }
398 
399 /*
400  * Handle MRT setsockopt commands to modify the multicast routing tables.
401  */
402 static int
X_ip6_mrouter_set(struct socket * so,struct sockopt * sopt)403 X_ip6_mrouter_set(struct socket *so, struct sockopt *sopt)
404 {
405 	struct mf6ctable *mfct;
406 	int error = 0;
407 	int optval;
408 	struct mif6ctl mifc;
409 	struct mf6cctl mfcc;
410 	mifi_t mifi;
411 
412 	mfct = somfctable(so);
413 	if (so != mfct->router && sopt->sopt_name != MRT6_INIT)
414 		return (EPERM);
415 
416 	switch (sopt->sopt_name) {
417 	case MRT6_INIT:
418 #ifdef MRT6_OINIT
419 	case MRT6_OINIT:
420 #endif
421 		error = sooptcopyin(sopt, &optval, sizeof(optval),
422 		    sizeof(optval));
423 		if (error)
424 			break;
425 		error = ip6_mrouter_init(so, optval, sopt->sopt_name);
426 		break;
427 	case MRT6_DONE:
428 		X_ip6_mrouter_done(so);
429 		break;
430 	case MRT6_ADD_MIF:
431 		error = sooptcopyin(sopt, &mifc, sizeof(mifc), sizeof(mifc));
432 		if (error)
433 			break;
434 		error = add_m6if(mfct, so->so_fibnum, &mifc);
435 		break;
436 	case MRT6_ADD_MFC:
437 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
438 		if (error)
439 			break;
440 		error = add_m6fc(mfct, &mfcc);
441 		break;
442 	case MRT6_DEL_MFC:
443 		error = sooptcopyin(sopt, &mfcc, sizeof(mfcc), sizeof(mfcc));
444 		if (error)
445 			break;
446 		error = del_m6fc(mfct, &mfcc);
447 		break;
448 	case MRT6_DEL_MIF:
449 		error = sooptcopyin(sopt, &mifi, sizeof(mifi), sizeof(mifi));
450 		if (error)
451 			break;
452 		error = del_m6if(mfct, mifi);
453 		break;
454 	case MRT6_PIM:
455 		error = sooptcopyin(sopt, &optval, sizeof(optval),
456 		    sizeof(optval));
457 		if (error)
458 			break;
459 		error = set_pim6(&optval);
460 		break;
461 	default:
462 		error = EOPNOTSUPP;
463 		break;
464 	}
465 
466 	return (error);
467 }
468 
469 /*
470  * Handle MRT getsockopt commands
471  */
472 static int
X_ip6_mrouter_get(struct socket * so,struct sockopt * sopt)473 X_ip6_mrouter_get(struct socket *so, struct sockopt *sopt)
474 {
475 	struct mf6ctable *mfct;
476 	int error = 0;
477 
478 	mfct = somfctable(so);
479 	if (so != mfct->router)
480 		return (EACCES);
481 
482 	switch (sopt->sopt_name) {
483 		case MRT6_PIM:
484 			error = sooptcopyout(sopt, &V_pim6, sizeof(V_pim6));
485 			break;
486 	}
487 	return (error);
488 }
489 
490 /*
491  * Handle ioctl commands to obtain information from the cache
492  */
493 static int
X_mrt6_ioctl(u_long cmd,caddr_t data,int fibnum)494 X_mrt6_ioctl(u_long cmd, caddr_t data, int fibnum)
495 {
496 	struct mf6ctable *mfct;
497 	int error;
498 
499 	error = priv_check(curthread, PRIV_NETINET_MROUTE);
500 	if (error)
501 		return (error);
502 
503 	mfct = &V_mfctables[fibnum];
504 	switch (cmd) {
505 	case SIOCGETSGCNT_IN6:
506 		error = get_sg_cnt(mfct, (struct sioc_sg_req6 *)data);
507 		break;
508 
509 	case SIOCGETMIFCNT_IN6:
510 		error = get_mif6_cnt(mfct, (struct sioc_mif_req6 *)data);
511 		break;
512 
513 	default:
514 		error = EINVAL;
515 		break;
516 	}
517 
518 	return (error);
519 }
520 
521 /*
522  * returns the packet, byte, rpf-failure count for the source group provided
523  */
524 static int
get_sg_cnt(struct mf6ctable * mfct,struct sioc_sg_req6 * req)525 get_sg_cnt(struct mf6ctable *mfct, struct sioc_sg_req6 *req)
526 {
527 	struct mf6c *rt;
528 	int ret;
529 
530 	ret = 0;
531 
532 	MFC6_LOCK();
533 	rt = mf6c_find(mfct, &req->src.sin6_addr, &req->grp.sin6_addr);
534 	if (rt == NULL) {
535 		ret = ESRCH;
536 	} else {
537 		req->pktcnt = rt->mf6c_pkt_cnt;
538 		req->bytecnt = rt->mf6c_byte_cnt;
539 		req->wrong_if = rt->mf6c_wrong_if;
540 	}
541 	MFC6_UNLOCK();
542 
543 	return (ret);
544 }
545 
546 /*
547  * returns the input and output packet and byte counts on the mif provided
548  */
549 static int
get_mif6_cnt(struct mf6ctable * mfct,struct sioc_mif_req6 * req)550 get_mif6_cnt(struct mf6ctable *mfct, struct sioc_mif_req6 *req)
551 {
552 	mifi_t mifi;
553 	int ret;
554 
555 	ret = 0;
556 	mifi = req->mifi;
557 
558 	MIF6_LOCK();
559 
560 	if (mifi >= mfct->nummifs) {
561 		ret = EINVAL;
562 	} else {
563 		struct mif6 *mif = &mfct->miftable[mifi];
564 
565 		req->icount = mif->m6_pkt_in;
566 		req->ocount = mif->m6_pkt_out;
567 		req->ibytes = mif->m6_bytes_in;
568 		req->obytes = mif->m6_bytes_out;
569 	}
570 
571 	MIF6_UNLOCK();
572 
573 	return (ret);
574 }
575 
576 static int
set_pim6(int * i)577 set_pim6(int *i)
578 {
579 	if ((*i != 1) && (*i != 0))
580 		return (EINVAL);
581 
582 	/* XXX-MJ */
583 	V_pim6 = *i;
584 
585 	return (0);
586 }
587 
588 /*
589  * Enable multicast routing
590  */
591 static int
ip6_mrouter_init(struct socket * so,int v,int cmd)592 ip6_mrouter_init(struct socket *so, int v, int cmd)
593 {
594 	struct mf6ctable *mfct;
595 
596 	MRT6_DLOG(DEBUG_ANY, "%s: socket %p", __func__, so);
597 
598 	if (v != 1)
599 		return (ENOPROTOOPT);
600 
601 	mfct = somfctable(so);
602 	MROUTER6_LOCK();
603 
604 	if (mfct->router != NULL) {
605 		MROUTER6_UNLOCK();
606 		return (EADDRINUSE);
607 	}
608 
609 	MFC6_LOCK();
610 	V_ip6_mrouting_enabled = true;
611 	mfct->router = so;
612 	mfct->router_ver = cmd;
613 
614 	bzero(&mfct->mfchashtbl, sizeof(mfct->mfchashtbl));
615 	bzero(&mfct->nexpire, sizeof(mfct->nexpire));
616 
617 	V_pim6 = 0;/* used for stubbing out/in pim stuff */
618 
619 	callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls_all,
620 	    curvnet);
621 
622 	MFC6_UNLOCK();
623 	MROUTER6_UNLOCK();
624 
625 	MRT6_DLOG(DEBUG_ANY, "finished");
626 
627 	return (0);
628 }
629 
630 /*
631  * Disable IPv6 multicast forwarding.
632  */
633 static void
X_ip6_mrouter_done(struct socket * so)634 X_ip6_mrouter_done(struct socket *so)
635 {
636 	struct mf6ctable *mfct;
637 	mifi_t mifi;
638 	u_long i;
639 	struct mf6c *rt;
640 	struct rtdetq *rte;
641 
642 	mfct = somfctable(so);
643 	MROUTER6_LOCK();
644 
645 	if (mfct->router != so) {
646 		MROUTER6_UNLOCK();
647 		return;
648 	}
649 
650 	/*
651 	 * For each phyint in use, disable promiscuous reception of all IPv6
652 	 * multicasts.
653 	 */
654 	for (mifi = 0; mifi < mfct->nummifs; mifi++) {
655 		struct mif6 *mif = &mfct->miftable[mifi];
656 
657 		if (mif->m6_ifp && !(mif->m6_flags & MIFF_REGISTER)) {
658 			if_allmulti(mif->m6_ifp, 0);
659 		}
660 	}
661 	MFC6_LOCK();
662 	bzero(mfct->miftable, sizeof(mfct->miftable));
663 	mfct->nummifs = 0;
664 
665 	V_pim6 = 0; /* used to stub out/in pim specific code */
666 
667 	/*
668 	 * Free all multicast forwarding cache entries.
669 	 */
670 	for (i = 0; i < MF6CTBLSIZ; i++) {
671 		rt = mfct->mfchashtbl[i];
672 		while (rt) {
673 			struct mf6c *frt;
674 
675 			for (rte = rt->mf6c_stall; rte != NULL; ) {
676 				struct rtdetq *n = rte->next;
677 
678 				m_freem(rte->m);
679 				free(rte, M_MRTABLE6);
680 				rte = n;
681 			}
682 			frt = rt;
683 			rt = rt->mf6c_next;
684 			free(frt, M_MRTABLE6);
685 		}
686 	}
687 	mfct->router = NULL;
688 	mfct->router_ver = 0;
689 	V_ip6_mrouting_enabled = false;
690 
691 	bzero(mfct->mfchashtbl, sizeof(mfct->mfchashtbl));
692 	MFC6_UNLOCK();
693 
694 	/*
695 	 * Reset register interface
696 	 */
697 	if (mfct->register_mif != (mifi_t)-1 && mfct->register_if != NULL) {
698 		if_detach(mfct->register_if);
699 		if_free(mfct->register_if);
700 		mfct->register_mif = (mifi_t)-1;
701 		mfct->register_if = NULL;
702 	}
703 
704 	MROUTER6_UNLOCK();
705 	MRT6_DLOG(DEBUG_ANY, "finished");
706 }
707 
708 static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
709 
710 /*
711  * Add a mif to the mif table
712  */
713 static int
add_m6if(struct mf6ctable * mfct,int fibnum,struct mif6ctl * mifcp)714 add_m6if(struct mf6ctable *mfct, int fibnum, struct mif6ctl *mifcp)
715 {
716 	struct epoch_tracker et;
717 	struct mif6 *mifp;
718 	struct ifnet *ifp;
719 	int error;
720 
721 	MIF6_LOCK();
722 
723 	if (mifcp->mif6c_mifi >= MAXMIFS) {
724 		MIF6_UNLOCK();
725 		return (EINVAL);
726 	}
727 	mifp = &mfct->miftable[mifcp->mif6c_mifi];
728 	if (mifp->m6_ifp != NULL) {
729 		MIF6_UNLOCK();
730 		return (EADDRINUSE); /* XXX: is it appropriate? */
731 	}
732 
733 	NET_EPOCH_ENTER(et);
734 	if ((ifp = ifnet_byindex(mifcp->mif6c_pifi)) == NULL) {
735 		NET_EPOCH_EXIT(et);
736 		MIF6_UNLOCK();
737 		return (ENXIO);
738 	}
739 	NET_EPOCH_EXIT(et);	/* XXXGL: unsafe ifp */
740 
741 	if (mifcp->mif6c_flags & MIFF_REGISTER) {
742 		if (mfct->register_mif == (mifi_t)-1) {
743 			ifp = if_alloc(IFT_OTHER);
744 
745 			if_initname(ifp, "register_mif", 0);
746 			ifp->if_flags |= IFF_LOOPBACK;
747 			if_attach(ifp);
748 			mfct->register_if = ifp;
749 			mfct->register_mif = mifcp->mif6c_mifi;
750 			/*
751 			 * it is impossible to guess the ifindex of the
752 			 * register interface.  So mif6c_pifi is automatically
753 			 * calculated.
754 			 */
755 			mifcp->mif6c_pifi = ifp->if_index;
756 		} else {
757 			ifp = mfct->register_if;
758 		}
759 	} else {
760 		/* Make sure the interface supports multicast */
761 		if ((ifp->if_flags & IFF_MULTICAST) == 0) {
762 			MIF6_UNLOCK();
763 			return (EOPNOTSUPP);
764 		}
765 		if (ifp->if_fib != fibnum) {
766 			MIF6_UNLOCK();
767 			return (EADDRNOTAVAIL);
768 		}
769 
770 		error = if_allmulti(ifp, 1);
771 		if (error) {
772 			MIF6_UNLOCK();
773 			return (error);
774 		}
775 	}
776 
777 	mifp->m6_flags     = mifcp->mif6c_flags;
778 	mifp->m6_ifp       = ifp;
779 
780 	/* initialize per mif pkt counters */
781 	mifp->m6_pkt_in    = 0;
782 	mifp->m6_pkt_out   = 0;
783 	mifp->m6_bytes_in  = 0;
784 	mifp->m6_bytes_out = 0;
785 
786 	/* Adjust nummifs up if the mifi is higher than nummifs */
787 	if (mfct->nummifs <= mifcp->mif6c_mifi)
788 		mfct->nummifs = mifcp->mif6c_mifi + 1;
789 
790 	MIF6_UNLOCK();
791 	MRT6_DLOG(DEBUG_ANY, "mif #%d, phyint %s", mifcp->mif6c_mifi,
792 	    if_name(ifp));
793 
794 	return (0);
795 }
796 
797 static void
expire_mf6c(struct mf6c * mfc)798 expire_mf6c(struct mf6c *mfc)
799 {
800 	struct rtdetq *rte;
801 
802 	while ((rte = mfc->mf6c_stall) != NULL) {
803 		mfc->mf6c_stall = rte->next;
804 		m_freem(rte->m);
805 		free(rte, M_MRTABLE6);
806 	}
807 
808 	free(mfc, M_MRTABLE6);
809 }
810 
811 /*
812  * Delete a mif from the mif table
813  */
814 static int
del_m6if_locked(struct mf6ctable * mfct,mifi_t mifi)815 del_m6if_locked(struct mf6ctable *mfct, mifi_t mifi)
816 {
817 	struct mif6 *mifp;
818 	mifi_t tmp;
819 	struct ifnet *ifp;
820 
821 	MIF6_LOCK_ASSERT();
822 
823 	if (mifi >= mfct->nummifs)
824 		return (EINVAL);
825 	mifp = &mfct->miftable[mifi];
826 	if (mifp->m6_ifp == NULL)
827 		return (EINVAL);
828 
829 	if (!(mifp->m6_flags & MIFF_REGISTER)) {
830 		ifp = mifp->m6_ifp;
831 		if_allmulti(ifp, 0);
832 
833 		MFC6_LOCK();
834 		for (int i = 0; i < MF6CTBLSIZ; i++) {
835 			struct mf6c *mfc, **nmfc;
836 
837 			nmfc = &mfct->mfchashtbl[i];
838 			while ((mfc = *nmfc) != NULL) {
839 				if (mfc->mf6c_parent == mifi) {
840 					*nmfc = mfc->mf6c_next;
841 					if (mfc->mf6c_expire)
842 						mfct->nexpire[i]--;
843 					expire_mf6c(mfc);
844 				} else {
845 					/* Remove this mif from the ifset */
846 					if (IF_ISSET(mifi, &mfc->mf6c_ifset))
847 						IF_CLR(mifi, &mfc->mf6c_ifset);
848 					nmfc = &mfc->mf6c_next;
849 				}
850 			}
851 		}
852 		MFC6_UNLOCK();
853 	} else {
854 		if (mfct->register_mif != (mifi_t)-1 &&
855 		    mfct->register_if != NULL) {
856 			if_detach(mfct->register_if);
857 			if_free(mfct->register_if);
858 			mfct->register_mif = (mifi_t)-1;
859 			mfct->register_if = NULL;
860 		}
861 	}
862 
863 	bzero(mifp, sizeof(*mifp));
864 
865 	/* Adjust nummifs down */
866 	for (tmp = mfct->nummifs; tmp > 0; tmp--)
867 		if (mfct->miftable[tmp - 1].m6_ifp != NULL)
868 			break;
869 	mfct->nummifs = tmp;
870 	MRT6_DLOG(DEBUG_ANY, "mif %d, nummifs %d", mifi, mfct->nummifs);
871 
872 	return (0);
873 }
874 
875 static int
del_m6if(struct mf6ctable * mfct,mifi_t mifi)876 del_m6if(struct mf6ctable *mfct, mifi_t mifi)
877 {
878 	int cc;
879 
880 	MIF6_LOCK();
881 	cc = del_m6if_locked(mfct, mifi);
882 	MIF6_UNLOCK();
883 
884 	return (cc);
885 }
886 
887 /*
888  * Add an mfc entry
889  */
890 static int
add_m6fc(struct mf6ctable * mfct,struct mf6cctl * mfccp)891 add_m6fc(struct mf6ctable *mfct, struct mf6cctl *mfccp)
892 {
893 	struct mf6c *rt;
894 	u_long hash;
895 	struct rtdetq *rte;
896 	u_short nstl;
897 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
898 
899 	MFC6_LOCK();
900 	rt = mf6c_find(mfct, &mfccp->mf6cc_origin.sin6_addr,
901 	    &mfccp->mf6cc_mcastgrp.sin6_addr);
902 	/* If an entry already exists, just update the fields */
903 	if (rt) {
904 		MRT6_DLOG(DEBUG_MFC, "no upcall o %s g %s p %x",
905 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
906 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
907 		    mfccp->mf6cc_parent);
908 
909 		rt->mf6c_parent = mfccp->mf6cc_parent;
910 		rt->mf6c_ifset = mfccp->mf6cc_ifset;
911 
912 		MFC6_UNLOCK();
913 		return (0);
914 	}
915 
916 	/*
917 	 * Find the entry for which the upcall was made and update
918 	 */
919 	hash = MF6CHASH(mfccp->mf6cc_origin.sin6_addr,
920 			mfccp->mf6cc_mcastgrp.sin6_addr);
921 	for (rt = mfct->mfchashtbl[hash], nstl = 0; rt; rt = rt->mf6c_next) {
922 		if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
923 				       &mfccp->mf6cc_origin.sin6_addr) &&
924 		    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
925 				       &mfccp->mf6cc_mcastgrp.sin6_addr) &&
926 		    (rt->mf6c_stall != NULL)) {
927 			if (nstl++)
928 				log(LOG_ERR,
929 				    "add_m6fc: %s o %s g %s p %x dbx %p\n",
930 				    "multiple kernel entries",
931 				    ip6_sprintf(ip6bufo,
932 					    &mfccp->mf6cc_origin.sin6_addr),
933 				    ip6_sprintf(ip6bufg,
934 					    &mfccp->mf6cc_mcastgrp.sin6_addr),
935 				    mfccp->mf6cc_parent, rt->mf6c_stall);
936 
937 			MRT6_DLOG(DEBUG_MFC, "o %s g %s p %x dbg %p",
938 			    ip6_sprintf(ip6bufo,
939 			    &mfccp->mf6cc_origin.sin6_addr),
940 			    ip6_sprintf(ip6bufg,
941 				&mfccp->mf6cc_mcastgrp.sin6_addr),
942 			    mfccp->mf6cc_parent, rt->mf6c_stall);
943 
944 			rt->mf6c_origin     = mfccp->mf6cc_origin;
945 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
946 			rt->mf6c_parent     = mfccp->mf6cc_parent;
947 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
948 			/* initialize pkt counters per src-grp */
949 			rt->mf6c_pkt_cnt    = 0;
950 			rt->mf6c_byte_cnt   = 0;
951 			rt->mf6c_wrong_if   = 0;
952 
953 			rt->mf6c_expire = 0;	/* Don't clean this guy up */
954 			mfct->nexpire[hash]--;
955 
956 			/* free packets Qed at the end of this entry */
957 			for (rte = rt->mf6c_stall; rte != NULL; ) {
958 				struct rtdetq *n = rte->next;
959 				ip6_mdq(mfct, rte->m, rte->ifp, rt);
960 				m_freem(rte->m);
961 #ifdef UPCALL_TIMING
962 				collate(&(rte->t));
963 #endif /* UPCALL_TIMING */
964 				free(rte, M_MRTABLE6);
965 				rte = n;
966 			}
967 			rt->mf6c_stall = NULL;
968 		}
969 	}
970 
971 	/*
972 	 * It is possible that an entry is being inserted without an upcall
973 	 */
974 	if (nstl == 0) {
975 		MRT6_DLOG(DEBUG_MFC, "no upcall h %lu o %s g %s p %x", hash,
976 		    ip6_sprintf(ip6bufo, &mfccp->mf6cc_origin.sin6_addr),
977 		    ip6_sprintf(ip6bufg, &mfccp->mf6cc_mcastgrp.sin6_addr),
978 		    mfccp->mf6cc_parent);
979 
980 		for (rt = mfct->mfchashtbl[hash]; rt; rt = rt->mf6c_next) {
981 			if (IN6_ARE_ADDR_EQUAL(&rt->mf6c_origin.sin6_addr,
982 					       &mfccp->mf6cc_origin.sin6_addr)&&
983 			    IN6_ARE_ADDR_EQUAL(&rt->mf6c_mcastgrp.sin6_addr,
984 					       &mfccp->mf6cc_mcastgrp.sin6_addr)) {
985 				rt->mf6c_origin     = mfccp->mf6cc_origin;
986 				rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
987 				rt->mf6c_parent     = mfccp->mf6cc_parent;
988 				rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
989 				/* initialize pkt counters per src-grp */
990 				rt->mf6c_pkt_cnt    = 0;
991 				rt->mf6c_byte_cnt   = 0;
992 				rt->mf6c_wrong_if   = 0;
993 
994 				if (rt->mf6c_expire)
995 					mfct->nexpire[hash]--;
996 				rt->mf6c_expire	   = 0;
997 			}
998 		}
999 		if (rt == NULL) {
1000 			/* no upcall, so make a new entry */
1001 			rt = malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1002 			if (rt == NULL) {
1003 				MFC6_UNLOCK();
1004 				return (ENOBUFS);
1005 			}
1006 
1007 			/* insert new entry at head of hash chain */
1008 			rt->mf6c_origin     = mfccp->mf6cc_origin;
1009 			rt->mf6c_mcastgrp   = mfccp->mf6cc_mcastgrp;
1010 			rt->mf6c_parent     = mfccp->mf6cc_parent;
1011 			rt->mf6c_ifset	    = mfccp->mf6cc_ifset;
1012 			/* initialize pkt counters per src-grp */
1013 			rt->mf6c_pkt_cnt    = 0;
1014 			rt->mf6c_byte_cnt   = 0;
1015 			rt->mf6c_wrong_if   = 0;
1016 			rt->mf6c_expire     = 0;
1017 			rt->mf6c_stall = NULL;
1018 
1019 			/* link into table */
1020 			rt->mf6c_next  = mfct->mfchashtbl[hash];
1021 			mfct->mfchashtbl[hash] = rt;
1022 		}
1023 	}
1024 
1025 	MFC6_UNLOCK();
1026 	return (0);
1027 }
1028 
1029 #ifdef UPCALL_TIMING
1030 /*
1031  * collect delay statistics on the upcalls
1032  */
1033 static void
collate(struct timeval * t)1034 collate(struct timeval *t)
1035 {
1036 	u_long d;
1037 	struct timeval tp;
1038 	u_long delta;
1039 
1040 	GET_TIME(tp);
1041 
1042 	if (TV_LT(*t, tp))
1043 	{
1044 		TV_DELTA(tp, *t, delta);
1045 
1046 		d = delta >> 10;
1047 		if (d > UPCALL_MAX)
1048 			d = UPCALL_MAX;
1049 
1050 		++upcall_data[d];
1051 	}
1052 }
1053 #endif /* UPCALL_TIMING */
1054 
1055 /*
1056  * Delete an mfc entry
1057  */
1058 static int
del_m6fc(struct mf6ctable * mfct,struct mf6cctl * mfccp)1059 del_m6fc(struct mf6ctable *mfct, struct mf6cctl *mfccp)
1060 {
1061 #ifdef MRT6DEBUG
1062 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1063 #endif
1064 	struct sockaddr_in6	origin;
1065 	struct sockaddr_in6	mcastgrp;
1066 	struct mf6c		*rt;
1067 	struct mf6c		**nptr;
1068 	u_long		hash;
1069 
1070 	origin = mfccp->mf6cc_origin;
1071 	mcastgrp = mfccp->mf6cc_mcastgrp;
1072 	hash = MF6CHASH(origin.sin6_addr, mcastgrp.sin6_addr);
1073 
1074 	MRT6_DLOG(DEBUG_MFC, "orig %s mcastgrp %s",
1075 	    ip6_sprintf(ip6bufo, &origin.sin6_addr),
1076 	    ip6_sprintf(ip6bufg, &mcastgrp.sin6_addr));
1077 
1078 	MFC6_LOCK();
1079 
1080 	nptr = &mfct->mfchashtbl[hash];
1081 	while ((rt = *nptr) != NULL) {
1082 		if (IN6_ARE_ADDR_EQUAL(&origin.sin6_addr,
1083 				       &rt->mf6c_origin.sin6_addr) &&
1084 		    IN6_ARE_ADDR_EQUAL(&mcastgrp.sin6_addr,
1085 				       &rt->mf6c_mcastgrp.sin6_addr) &&
1086 		    rt->mf6c_stall == NULL)
1087 			break;
1088 
1089 		nptr = &rt->mf6c_next;
1090 	}
1091 	if (rt == NULL) {
1092 		MFC6_UNLOCK();
1093 		return (EADDRNOTAVAIL);
1094 	}
1095 
1096 	*nptr = rt->mf6c_next;
1097 	free(rt, M_MRTABLE6);
1098 
1099 	MFC6_UNLOCK();
1100 
1101 	return (0);
1102 }
1103 
1104 static int
socket_send(struct socket * s,struct mbuf * mm,struct sockaddr_in6 * src)1105 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in6 *src)
1106 {
1107 
1108 	if (s) {
1109 		if (sbappendaddr(&s->so_rcv,
1110 				 (struct sockaddr *)src,
1111 				 mm, (struct mbuf *)0) != 0) {
1112 			sorwakeup(s);
1113 			return (0);
1114 		} else
1115 			soroverflow(s);
1116 	}
1117 	m_freem(mm);
1118 	return (-1);
1119 }
1120 
1121 /*
1122  * IPv6 multicast forwarding function. This function assumes that the packet
1123  * pointed to by "ip6" has arrived on (or is about to be sent to) the interface
1124  * pointed to by "ifp", and the packet is to be relayed to other networks
1125  * that have members of the packet's destination IPv6 multicast group.
1126  *
1127  * The packet is returned unscathed to the caller, unless it is
1128  * erroneous, in which case a non-zero return value tells the caller to
1129  * discard it.
1130  *
1131  * NOTE: this implementation assumes that m->m_pkthdr.rcvif is NULL iff
1132  * this function is called in the originating context (i.e., not when
1133  * forwarding a packet from other node).  ip6_output(), which is currently the
1134  * only function that calls this function is called in the originating context,
1135  * explicitly ensures this condition.  It is caller's responsibility to ensure
1136  * that if this function is called from somewhere else in the originating
1137  * context in the future.
1138  */
1139 static int
X_ip6_mforward(struct ip6_hdr * ip6,struct ifnet * ifp,struct mbuf * m)1140 X_ip6_mforward(struct ip6_hdr *ip6, struct ifnet *ifp, struct mbuf *m)
1141 {
1142 	struct mf6ctable *mfct;
1143 	struct rtdetq *rte;
1144 	struct mbuf *mb0;
1145 	struct mf6c *rt;
1146 	struct mif6 *mifp;
1147 	struct mbuf *mm;
1148 	u_long hash;
1149 	mifi_t mifi;
1150 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1151 #ifdef UPCALL_TIMING
1152 	struct timeval tp;
1153 
1154 	GET_TIME(tp);
1155 #endif /* UPCALL_TIMING */
1156 
1157 	M_ASSERTMAPPED(m);
1158 	MRT6_DLOG(DEBUG_FORWARD, "src %s, dst %s, ifindex %d",
1159 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1160 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst), ifp->if_index);
1161 
1162 	/*
1163 	 * Don't forward a packet with Hop limit of zero or one,
1164 	 * or a packet destined to a local-only group.
1165 	 */
1166 	if (ip6->ip6_hlim <= 1 || IN6_IS_ADDR_MC_INTFACELOCAL(&ip6->ip6_dst) ||
1167 	    IN6_IS_ADDR_MC_LINKLOCAL(&ip6->ip6_dst))
1168 		return (0);
1169 	ip6->ip6_hlim--;
1170 
1171 	/*
1172 	 * Source address check: do not forward packets with unspecified
1173 	 * source. It was discussed in July 2000, on ipngwg mailing list.
1174 	 * This is rather more serious than unicast cases, because some
1175 	 * MLD packets can be sent with the unspecified source address
1176 	 * (although such packets must normally set 1 to the hop limit field).
1177 	 */
1178 	if (IN6_IS_ADDR_UNSPECIFIED(&ip6->ip6_src)) {
1179 		IP6STAT_INC(ip6s_cantforward);
1180 		if (V_ip6_log_cannot_forward && ip6_log_ratelimit()) {
1181 			log(LOG_DEBUG,
1182 			    "cannot forward "
1183 			    "from %s to %s nxt %d received on %s\n",
1184 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1185 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1186 			    ip6->ip6_nxt,
1187 			    if_name(m->m_pkthdr.rcvif));
1188 		}
1189 		return (0);
1190 	}
1191 
1192 	mfct = &V_mfctables[M_GETFIB(m)];
1193 	MFC6_LOCK();
1194 	if (__predict_false(mfct->router == NULL)) {
1195 		MFC6_UNLOCK();
1196 		return (EADDRNOTAVAIL);
1197 	}
1198 
1199 	/*
1200 	 * Determine forwarding mifs from the forwarding cache table
1201 	 */
1202 	rt = mf6c_find(mfct, &ip6->ip6_src, &ip6->ip6_dst);
1203 	MRT6STAT_INC(mrt6s_mfc_lookups);
1204 
1205 	/* Entry exists, so forward if necessary */
1206 	if (rt) {
1207 		MFC6_UNLOCK();
1208 		return (ip6_mdq(mfct, m, ifp, rt));
1209 	}
1210 
1211 	/*
1212 	 * If we don't have a route for packet's origin,
1213 	 * Make a copy of the packet & send message to routing daemon.
1214 	 */
1215 	MRT6STAT_INC(mrt6s_no_route);
1216 	MRT6_DLOG(DEBUG_FORWARD | DEBUG_MFC, "no rte s %s g %s",
1217 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1218 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1219 
1220 	/*
1221 	 * Allocate mbufs early so that we don't do extra work if we
1222 	 * are just going to fail anyway.
1223 	 */
1224 	rte = malloc(sizeof(*rte), M_MRTABLE6, M_NOWAIT);
1225 	if (rte == NULL) {
1226 		MFC6_UNLOCK();
1227 		return (ENOBUFS);
1228 	}
1229 	mb0 = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1230 	/*
1231 	 * Pullup packet header if needed before storing it,
1232 	 * as other references may modify it in the meantime.
1233 	 */
1234 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < sizeof(struct ip6_hdr)))
1235 		mb0 = m_pullup(mb0, sizeof(struct ip6_hdr));
1236 	if (mb0 == NULL) {
1237 		free(rte, M_MRTABLE6);
1238 		MFC6_UNLOCK();
1239 		return (ENOBUFS);
1240 	}
1241 
1242 	/* is there an upcall waiting for this packet? */
1243 	hash = MF6CHASH(ip6->ip6_src, ip6->ip6_dst);
1244 	for (rt = mfct->mfchashtbl[hash]; rt; rt = rt->mf6c_next) {
1245 		if (IN6_ARE_ADDR_EQUAL(&ip6->ip6_src,
1246 		    &rt->mf6c_origin.sin6_addr) &&
1247 		    IN6_ARE_ADDR_EQUAL(&ip6->ip6_dst,
1248 		    &rt->mf6c_mcastgrp.sin6_addr) && (rt->mf6c_stall != NULL))
1249 			break;
1250 	}
1251 
1252 	if (rt == NULL) {
1253 		struct mrt6msg *im;
1254 #ifdef MRT6_OINIT
1255 		struct omrt6msg *oim;
1256 #endif
1257 		/* no upcall, so make a new entry */
1258 		rt = malloc(sizeof(*rt), M_MRTABLE6, M_NOWAIT);
1259 		if (rt == NULL) {
1260 			free(rte, M_MRTABLE6);
1261 			m_freem(mb0);
1262 			MFC6_UNLOCK();
1263 			return (ENOBUFS);
1264 		}
1265 		/*
1266 		 * Make a copy of the header to send to the user
1267 		 * level process
1268 		 */
1269 		mm = m_copym(mb0, 0, sizeof(struct ip6_hdr), M_NOWAIT);
1270 		if (mm == NULL) {
1271 			free(rte, M_MRTABLE6);
1272 			m_freem(mb0);
1273 			free(rt, M_MRTABLE6);
1274 			MFC6_UNLOCK();
1275 			return (ENOBUFS);
1276 		}
1277 
1278 		/*
1279 		 * Send message to routing daemon
1280 		 */
1281 		sin6.sin6_addr = ip6->ip6_src;
1282 		im = NULL;
1283 #ifdef MRT6_OINIT
1284 		oim = NULL;
1285 #endif
1286 		switch (mfct->router_ver) {
1287 #ifdef MRT6_OINIT
1288 		case MRT6_OINIT:
1289 			oim = mtod(mm, struct omrt6msg *);
1290 			oim->im6_msgtype = MRT6MSG_NOCACHE;
1291 			oim->im6_mbz = 0;
1292 			break;
1293 #endif
1294 		case MRT6_INIT:
1295 			im = mtod(mm, struct mrt6msg *);
1296 			im->im6_msgtype = MRT6MSG_NOCACHE;
1297 			im->im6_mbz = 0;
1298 			break;
1299 		default:
1300 			free(rte, M_MRTABLE6);
1301 			m_freem(mb0);
1302 			free(rt, M_MRTABLE6);
1303 			MFC6_UNLOCK();
1304 			return (EINVAL);
1305 		}
1306 
1307 		MRT6_DLOG(DEBUG_FORWARD, "getting the iif info in the kernel");
1308 		for (mifp = mfct->miftable, mifi = 0;
1309 		    mifi < mfct->nummifs && mifp->m6_ifp != ifp; mifp++, mifi++)
1310 			;
1311 
1312 		switch (mfct->router_ver) {
1313 #ifdef MRT6_OINIT
1314 		case MRT6_OINIT:
1315 			oim->im6_mif = mifi;
1316 			break;
1317 #endif
1318 		case MRT6_INIT:
1319 			im->im6_mif = mifi;
1320 			break;
1321 		}
1322 
1323 		if (socket_send(mfct->router, mm, &sin6) < 0) {
1324 			log(LOG_WARNING, "ip6_mforward: ip6_mrouter "
1325 			    "socket queue full\n");
1326 			MRT6STAT_INC(mrt6s_upq_sockfull);
1327 			free(rte, M_MRTABLE6);
1328 			m_freem(mb0);
1329 			free(rt, M_MRTABLE6);
1330 			MFC6_UNLOCK();
1331 			return (ENOBUFS);
1332 		}
1333 
1334 		MRT6STAT_INC(mrt6s_upcalls);
1335 
1336 		/* insert new entry at head of hash chain */
1337 		bzero(rt, sizeof(*rt));
1338 		rt->mf6c_origin.sin6_family = AF_INET6;
1339 		rt->mf6c_origin.sin6_len = sizeof(struct sockaddr_in6);
1340 		rt->mf6c_origin.sin6_addr = ip6->ip6_src;
1341 		rt->mf6c_mcastgrp.sin6_family = AF_INET6;
1342 		rt->mf6c_mcastgrp.sin6_len = sizeof(struct sockaddr_in6);
1343 		rt->mf6c_mcastgrp.sin6_addr = ip6->ip6_dst;
1344 		rt->mf6c_expire = UPCALL_EXPIRE;
1345 		mfct->nexpire[hash]++;
1346 		rt->mf6c_parent = MF6C_INCOMPLETE_PARENT;
1347 
1348 		/* link into table */
1349 		rt->mf6c_next = mfct->mfchashtbl[hash];
1350 		mfct->mfchashtbl[hash] = rt;
1351 		/* Add this entry to the end of the queue */
1352 		rt->mf6c_stall = rte;
1353 	} else {
1354 		/* determine if q has overflowed */
1355 		struct rtdetq **p;
1356 		int npkts = 0;
1357 
1358 		for (p = &rt->mf6c_stall; *p != NULL; p = &(*p)->next)
1359 			if (++npkts > MAX_UPQ6) {
1360 				MRT6STAT_INC(mrt6s_upq_ovflw);
1361 				free(rte, M_MRTABLE6);
1362 				m_freem(mb0);
1363 				MFC6_UNLOCK();
1364 				return (0);
1365 			}
1366 
1367 		/* Add this entry to the end of the queue */
1368 		*p = rte;
1369 	}
1370 
1371 	rte->next = NULL;
1372 	rte->m = mb0;
1373 	rte->ifp = ifp;
1374 #ifdef UPCALL_TIMING
1375 	rte->t = tp;
1376 #endif /* UPCALL_TIMING */
1377 
1378 	MFC6_UNLOCK();
1379 
1380 	return (0);
1381 }
1382 
1383 /*
1384  * Clean up cache entries if upcalls are not serviced
1385  * Call from the Slow Timeout mechanism, every half second.
1386  */
1387 static void
expire_upcalls(struct mf6ctable * mfct)1388 expire_upcalls(struct mf6ctable *mfct)
1389 {
1390 #ifdef MRT6DEBUG
1391 	char ip6bufo[INET6_ADDRSTRLEN], ip6bufg[INET6_ADDRSTRLEN];
1392 #endif
1393 	struct rtdetq *rte;
1394 	struct mf6c *mfc, **nptr;
1395 	u_long i;
1396 
1397 	MFC6_LOCK_ASSERT();
1398 
1399 	for (i = 0; i < MF6CTBLSIZ; i++) {
1400 		if (mfct->nexpire[i] == 0)
1401 			continue;
1402 		nptr = &mfct->mfchashtbl[i];
1403 		while ((mfc = *nptr) != NULL) {
1404 			rte = mfc->mf6c_stall;
1405 			/*
1406 			 * Skip real cache entries
1407 			 * Make sure it wasn't marked to not expire (shouldn't happen)
1408 			 * If it expires now
1409 			 */
1410 			if (rte != NULL &&
1411 			    mfc->mf6c_expire != 0 &&
1412 			    --mfc->mf6c_expire == 0) {
1413 				MRT6_DLOG(DEBUG_EXPIRE, "expiring (%s %s)",
1414 				    ip6_sprintf(ip6bufo, &mfc->mf6c_origin.sin6_addr),
1415 				    ip6_sprintf(ip6bufg, &mfc->mf6c_mcastgrp.sin6_addr));
1416 				/*
1417 				 * drop all the packets
1418 				 * free the mbuf with the pkt, if, timing info
1419 				 */
1420 				do {
1421 					struct rtdetq *n = rte->next;
1422 					m_freem(rte->m);
1423 					free(rte, M_MRTABLE6);
1424 					rte = n;
1425 				} while (rte != NULL);
1426 				MRT6STAT_INC(mrt6s_cache_cleanups);
1427 				mfct->nexpire[i]--;
1428 
1429 				*nptr = mfc->mf6c_next;
1430 				free(mfc, M_MRTABLE6);
1431 			} else {
1432 				nptr = &mfc->mf6c_next;
1433 			}
1434 		}
1435 	}
1436 }
1437 
1438 /*
1439  * Clean up the cache entry if upcall is not serviced
1440  */
1441 static void
expire_upcalls_all(void * arg)1442 expire_upcalls_all(void *arg)
1443 {
1444 	CURVNET_SET((struct vnet *)arg);
1445 
1446 	for (int i = 0; i < V_nmfctables; i++)
1447 		expire_upcalls(&V_mfctables[i]);
1448 
1449 	callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls_all,
1450 	    curvnet);
1451 
1452 	CURVNET_RESTORE();
1453 }
1454 
1455 /*
1456  * Packet forwarding routine once entry in the cache is made
1457  */
1458 static int
ip6_mdq(struct mf6ctable * mfct,struct mbuf * m,struct ifnet * ifp,struct mf6c * rt)1459 ip6_mdq(struct mf6ctable *mfct, struct mbuf *m, struct ifnet *ifp,
1460     struct mf6c *rt)
1461 {
1462 	struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *);
1463 	mifi_t mifi, iif;
1464 	struct mif6 *mifp;
1465 	int plen = m->m_pkthdr.len;
1466 	struct in6_addr src0, dst0; /* copies for local work */
1467 	u_int32_t iszone, idzone, oszone, odzone;
1468 	int error = 0;
1469 
1470 	M_ASSERTMAPPED(m);
1471 
1472 	/*
1473 	 * Don't forward if it didn't arrive from the parent mif
1474 	 * for its origin.
1475 	 */
1476 	mifi = rt->mf6c_parent;
1477 	if (mifi >= mfct->nummifs || mfct->miftable[mifi].m6_ifp != ifp) {
1478 		MRT6STAT_INC(mrt6s_wrong_if);
1479 		rt->mf6c_wrong_if++;
1480 		if (mifi >= mfct->nummifs)
1481 			return (0);
1482 
1483 		mifp = &mfct->miftable[mifi];
1484 		MRT6_DLOG(DEBUG_FORWARD,
1485 		    "wrong if: ifid %d mifi %d mififid %x", ifp->if_index,
1486 		    mifi, mifp->m6_ifp->if_index);
1487 
1488 		/*
1489 		 * If we are doing PIM processing, and we are forwarding
1490 		 * packets on this interface, send a message to the
1491 		 * routing daemon.
1492 		 */
1493 		/* have to make sure this is a valid mif */
1494 		if (mifp->m6_ifp && V_pim6 && (m->m_flags & M_LOOP) == 0) {
1495 			/*
1496 			 * Check the M_LOOP flag to avoid an
1497 			 * unnecessary PIM assert.
1498 			 * XXX: M_LOOP is an ad-hoc hack...
1499 			 */
1500 			static struct sockaddr_in6 sin6 =
1501 			{ sizeof(sin6), AF_INET6 };
1502 
1503 			struct mbuf *mm;
1504 			struct mrt6msg *im;
1505 #ifdef MRT6_OINIT
1506 			struct omrt6msg *oim;
1507 #endif
1508 
1509 			mm = m_copym(m, 0, sizeof(struct ip6_hdr),
1510 			    M_NOWAIT);
1511 			if (mm &&
1512 			    (!M_WRITABLE(mm) ||
1513 			     mm->m_len < sizeof(struct ip6_hdr)))
1514 				mm = m_pullup(mm, sizeof(struct ip6_hdr));
1515 			if (mm == NULL)
1516 				return (ENOBUFS);
1517 
1518 #ifdef MRT6_OINIT
1519 			oim = NULL;
1520 #endif
1521 			im = NULL;
1522 			switch (mfct->router_ver) {
1523 #ifdef MRT6_OINIT
1524 			case MRT6_OINIT:
1525 				oim = mtod(mm, struct omrt6msg *);
1526 				oim->im6_msgtype = MRT6MSG_WRONGMIF;
1527 				oim->im6_mbz = 0;
1528 				break;
1529 #endif
1530 			case MRT6_INIT:
1531 				im = mtod(mm, struct mrt6msg *);
1532 				im->im6_msgtype = MRT6MSG_WRONGMIF;
1533 				im->im6_mbz = 0;
1534 				break;
1535 			default:
1536 				m_freem(mm);
1537 				return (EINVAL);
1538 			}
1539 
1540 			for (mifp = mfct->miftable, iif = 0;
1541 			     iif < mfct->nummifs && mifp->m6_ifp != ifp;
1542 			     mifp++, iif++)
1543 				;
1544 
1545 			switch (mfct->router_ver) {
1546 #ifdef MRT6_OINIT
1547 			case MRT6_OINIT:
1548 				oim->im6_mif = iif;
1549 				sin6.sin6_addr = oim->im6_src;
1550 				break;
1551 #endif
1552 			case MRT6_INIT:
1553 				im->im6_mif = iif;
1554 				sin6.sin6_addr = im->im6_src;
1555 				break;
1556 			}
1557 
1558 			MRT6STAT_INC(mrt6s_upcalls);
1559 
1560 			if (socket_send(mfct->router, mm, &sin6) < 0) {
1561 				MRT6_DLOG(DEBUG_ANY,
1562 				    "ip6_mrouter socket queue full");
1563 				MRT6STAT_INC(mrt6s_upq_sockfull);
1564 				return (ENOBUFS);
1565 			}
1566 		}
1567 		return (0);
1568 	}
1569 
1570 	mifp = &mfct->miftable[mifi];
1571 
1572 	/* If I sourced this packet, it counts as output, else it was input. */
1573 	if (m->m_pkthdr.rcvif == NULL) {
1574 		/* XXX: is rcvif really NULL when output?? */
1575 		mifp->m6_pkt_out++;
1576 		mifp->m6_bytes_out += plen;
1577 	} else {
1578 		mifp->m6_pkt_in++;
1579 		mifp->m6_bytes_in += plen;
1580 	}
1581 	rt->mf6c_pkt_cnt++;
1582 	rt->mf6c_byte_cnt += plen;
1583 
1584 	/*
1585 	 * For each mif, forward a copy of the packet if there are group
1586 	 * members downstream on the interface.
1587 	 */
1588 	src0 = ip6->ip6_src;
1589 	dst0 = ip6->ip6_dst;
1590 	if ((error = in6_setscope(&src0, ifp, &iszone)) != 0 ||
1591 	    (error = in6_setscope(&dst0, ifp, &idzone)) != 0) {
1592 		IP6STAT_INC(ip6s_badscope);
1593 		return (error);
1594 	}
1595 	for (mifp = mfct->miftable, mifi = 0; mifi < mfct->nummifs;
1596 	    mifp++, mifi++) {
1597 		if (IF_ISSET(mifi, &rt->mf6c_ifset)) {
1598 			/*
1599 			 * check if the outgoing packet is going to break
1600 			 * a scope boundary.
1601 			 * XXX For packets through PIM register tunnel
1602 			 * interface, we believe a routing daemon.
1603 			 */
1604 			if (!(mfct->miftable[rt->mf6c_parent].m6_flags &
1605 			      MIFF_REGISTER) &&
1606 			    !(mifp->m6_flags & MIFF_REGISTER)) {
1607 				if (in6_setscope(&src0, mifp->m6_ifp,
1608 				    &oszone) ||
1609 				    in6_setscope(&dst0, mifp->m6_ifp,
1610 				    &odzone) ||
1611 				    iszone != oszone ||
1612 				    idzone != odzone) {
1613 					IP6STAT_INC(ip6s_badscope);
1614 					continue;
1615 				}
1616 			}
1617 
1618 			mifp->m6_pkt_out++;
1619 			mifp->m6_bytes_out += plen;
1620 			if (mifp->m6_flags & MIFF_REGISTER)
1621 				register_send(mfct, ip6, mifi, m);
1622 			else
1623 				phyint_send(ip6, mifp, m);
1624 		}
1625 	}
1626 	return (0);
1627 }
1628 
1629 static void
phyint_send(struct ip6_hdr * ip6,struct mif6 * mifp,struct mbuf * m)1630 phyint_send(struct ip6_hdr *ip6, struct mif6 *mifp, struct mbuf *m)
1631 {
1632 #ifdef MRT6DEBUG
1633 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1634 #endif
1635 	struct mbuf *mb_copy;
1636 	struct ifnet *ifp = mifp->m6_ifp;
1637 	int error __unused = 0;
1638 	u_long linkmtu;
1639 
1640 	M_ASSERTMAPPED(m);
1641 
1642 	/*
1643 	 * Make a new reference to the packet; make sure that
1644 	 * the IPv6 header is actually copied, not just referenced,
1645 	 * so that ip6_output() only scribbles on the copy.
1646 	 */
1647 	mb_copy = m_copym(m, 0, M_COPYALL, M_NOWAIT);
1648 	if (mb_copy &&
1649 	    (!M_WRITABLE(mb_copy) || mb_copy->m_len < sizeof(struct ip6_hdr)))
1650 		mb_copy = m_pullup(mb_copy, sizeof(struct ip6_hdr));
1651 	if (mb_copy == NULL) {
1652 		return;
1653 	}
1654 	/* set MCAST flag to the outgoing packet */
1655 	mb_copy->m_flags |= M_MCAST;
1656 
1657 	/*
1658 	 * If we sourced the packet, call ip6_output since we may devide
1659 	 * the packet into fragments when the packet is too big for the
1660 	 * outgoing interface.
1661 	 * Otherwise, we can simply send the packet to the interface
1662 	 * sending queue.
1663 	 */
1664 	if (m->m_pkthdr.rcvif == NULL) {
1665 		struct ip6_moptions im6o;
1666 		struct epoch_tracker et;
1667 
1668 		im6o.im6o_multicast_ifp = ifp;
1669 		/* XXX: ip6_output will override ip6->ip6_hlim */
1670 		im6o.im6o_multicast_hlim = ip6->ip6_hlim;
1671 		im6o.im6o_multicast_loop = 1;
1672 		NET_EPOCH_ENTER(et);
1673 		error = ip6_output(mb_copy, NULL, NULL, IPV6_FORWARDING, &im6o,
1674 		    NULL, NULL);
1675 		NET_EPOCH_EXIT(et);
1676 
1677 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1678 		    (uint16_t)(mifp - mif6table), error);
1679 		return;
1680 	}
1681 
1682 	/*
1683 	 * If configured to loop back multicasts by default,
1684 	 * loop back a copy now.
1685 	 */
1686 	if (in6_mcast_loop)
1687 		ip6_mloopback(ifp, m);
1688 
1689 	/*
1690 	 * Put the packet into the sending queue of the outgoing interface
1691 	 * if it would fit in the MTU of the interface.
1692 	 */
1693 	linkmtu = in6_ifmtu(ifp);
1694 	if (mb_copy->m_pkthdr.len <= linkmtu || linkmtu < IPV6_MMTU) {
1695 		struct sockaddr_in6 dst6;
1696 
1697 		bzero(&dst6, sizeof(dst6));
1698 		dst6.sin6_len = sizeof(struct sockaddr_in6);
1699 		dst6.sin6_family = AF_INET6;
1700 		dst6.sin6_addr = ip6->ip6_dst;
1701 
1702 		IP_PROBE(send, NULL, NULL, ip6, ifp, NULL, ip6);
1703 		/*
1704 		 * We just call if_output instead of nd6_output here, since
1705 		 * we need no ND for a multicast forwarded packet...right?
1706 		 */
1707 		m_clrprotoflags(m);	/* Avoid confusing lower layers. */
1708 		error = (*ifp->if_output)(ifp, mb_copy,
1709 		    (struct sockaddr *)&dst6, NULL);
1710 		MRT6_DLOG(DEBUG_XMIT, "mif %u err %d",
1711 		    (uint16_t)(mifp - mif6table), error);
1712 	} else {
1713 		/*
1714 		 * pMTU discovery is intentionally disabled by default, since
1715 		 * various router may notify pMTU in multicast, which can be
1716 		 * a DDoS to a router
1717 		 */
1718 		if (V_ip6_mcast_pmtu)
1719 			icmp6_error(mb_copy, ICMP6_PACKET_TOO_BIG, 0, linkmtu);
1720 		else {
1721 			MRT6_DLOG(DEBUG_XMIT, " packet too big on %s o %s "
1722 			    "g %s size %d (discarded)", if_name(ifp),
1723 			    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1724 			    ip6_sprintf(ip6bufd, &ip6->ip6_dst),
1725 			    mb_copy->m_pkthdr.len);
1726 			m_freem(mb_copy); /* simply discard the packet */
1727 		}
1728 	}
1729 }
1730 
1731 static int
register_send(struct mf6ctable * mfct,struct ip6_hdr * ip6,mifi_t mifi,struct mbuf * m)1732 register_send(struct mf6ctable *mfct, struct ip6_hdr *ip6, mifi_t mifi,
1733     struct mbuf *m)
1734 {
1735 #ifdef MRT6DEBUG
1736 	char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1737 #endif
1738 	struct mbuf *mm;
1739 	int i, len = m->m_pkthdr.len;
1740 	static struct sockaddr_in6 sin6 = { sizeof(sin6), AF_INET6 };
1741 	struct mrt6msg *im6;
1742 
1743 	MRT6_DLOG(DEBUG_ANY, "src %s dst %s",
1744 	    ip6_sprintf(ip6bufs, &ip6->ip6_src),
1745 	    ip6_sprintf(ip6bufd, &ip6->ip6_dst));
1746 	PIM6STAT_INC(pim6s_snd_registers);
1747 
1748 	/* Make a copy of the packet to send to the user level process. */
1749 	mm = m_gethdr(M_NOWAIT, MT_DATA);
1750 	if (mm == NULL)
1751 		return (ENOBUFS);
1752 	mm->m_data += max_linkhdr;
1753 	mm->m_len = sizeof(struct ip6_hdr);
1754 
1755 	if ((mm->m_next = m_copym(m, 0, M_COPYALL, M_NOWAIT)) == NULL) {
1756 		m_freem(mm);
1757 		return (ENOBUFS);
1758 	}
1759 	i = MHLEN - M_LEADINGSPACE(mm);
1760 	if (i > len)
1761 		i = len;
1762 	mm = m_pullup(mm, i);
1763 	if (mm == NULL)
1764 		return (ENOBUFS);
1765 /* TODO: check it! */
1766 	mm->m_pkthdr.len = len + sizeof(struct ip6_hdr);
1767 
1768 	/*
1769 	 * Send message to routing daemon
1770 	 */
1771 	sin6.sin6_addr = ip6->ip6_src;
1772 
1773 	im6 = mtod(mm, struct mrt6msg *);
1774 	im6->im6_msgtype      = MRT6MSG_WHOLEPKT;
1775 	im6->im6_mbz          = 0;
1776 
1777 	im6->im6_mif = mifi;
1778 
1779 	/* iif info is not given for reg. encap.n */
1780 	MRT6STAT_INC(mrt6s_upcalls);
1781 
1782 	if (socket_send(mfct->router, mm, &sin6) < 0) {
1783 		MRT6_DLOG(DEBUG_ANY, "ip6_mrouter socket queue full");
1784 		MRT6STAT_INC(mrt6s_upq_sockfull);
1785 		return (ENOBUFS);
1786 	}
1787 	return (0);
1788 }
1789 
1790 /*
1791  * pim6_encapcheck() is called by the encap6_input() path at runtime to
1792  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
1793  * into the kernel.
1794  */
1795 static int
pim6_encapcheck(const struct mbuf * m __unused,int off __unused,int proto __unused,void * arg __unused)1796 pim6_encapcheck(const struct mbuf *m __unused, int off __unused,
1797     int proto __unused, void *arg __unused)
1798 {
1799 
1800     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
1801     return (8);		/* claim the datagram. */
1802 }
1803 
1804 /*
1805  * PIM sparse mode hook
1806  * Receives the pim control messages, and passes them up to the listening
1807  * socket, using rip6_input.
1808  * The only message processed is the REGISTER pim message; the pim header
1809  * is stripped off, and the inner packet is passed to register_mforward.
1810  */
1811 static int
pim6_input(struct mbuf * m,int off,int proto,void * arg __unused)1812 pim6_input(struct mbuf *m, int off, int proto, void *arg __unused)
1813 {
1814 	struct mf6ctable *mfct;
1815 	struct pim *pim;
1816 	struct ip6_hdr *ip6;
1817 	int pimlen;
1818 	int minlen;
1819 
1820 	mfct = &V_mfctables[M_GETFIB(m)];
1821 
1822 	PIM6STAT_INC(pim6s_rcv_total);
1823 
1824 	/*
1825 	 * Validate lengths
1826 	 */
1827 	pimlen = m->m_pkthdr.len - off;
1828 	if (pimlen < PIM_MINLEN) {
1829 		PIM6STAT_INC(pim6s_rcv_tooshort);
1830 		MRT6_DLOG(DEBUG_PIM, "PIM packet too short");
1831 		m_freem(m);
1832 		return (IPPROTO_DONE);
1833 	}
1834 
1835 	/*
1836 	 * if the packet is at least as big as a REGISTER, go ahead
1837 	 * and grab the PIM REGISTER header size, to avoid another
1838 	 * possible m_pullup() later.
1839 	 *
1840 	 * PIM_MINLEN       == pimhdr + u_int32 == 8
1841 	 * PIM6_REG_MINLEN   == pimhdr + reghdr + eip6hdr == 4 + 4 + 40
1842 	 */
1843 	minlen = (pimlen >= PIM6_REG_MINLEN) ? PIM6_REG_MINLEN : PIM_MINLEN;
1844 
1845 	/*
1846 	 * Make sure that the IP6 and PIM headers in contiguous memory, and
1847 	 * possibly the PIM REGISTER header
1848 	 */
1849 	if (m->m_len < off + minlen) {
1850 		m = m_pullup(m, off + minlen);
1851 		if (m == NULL) {
1852 			IP6STAT_INC(ip6s_exthdrtoolong);
1853 			return (IPPROTO_DONE);
1854 		}
1855 	}
1856 	ip6 = mtod(m, struct ip6_hdr *);
1857 	pim = (struct pim *)((caddr_t)ip6 + off);
1858 
1859 #define PIM6_CHECKSUM
1860 #ifdef PIM6_CHECKSUM
1861 	{
1862 		int cksumlen;
1863 
1864 		/*
1865 		 * Validate checksum.
1866 		 * If PIM REGISTER, exclude the data packet
1867 		 */
1868 		if (pim->pim_type == PIM_REGISTER)
1869 			cksumlen = PIM_MINLEN;
1870 		else
1871 			cksumlen = pimlen;
1872 
1873 		if (in6_cksum(m, IPPROTO_PIM, off, cksumlen)) {
1874 			PIM6STAT_INC(pim6s_rcv_badsum);
1875 			MRT6_DLOG(DEBUG_PIM, "invalid checksum");
1876 			m_freem(m);
1877 			return (IPPROTO_DONE);
1878 		}
1879 	}
1880 #endif /* PIM_CHECKSUM */
1881 
1882 	/* PIM version check */
1883 	if (pim->pim_ver != PIM_VERSION) {
1884 		PIM6STAT_INC(pim6s_rcv_badversion);
1885 		MRT6_DLOG(DEBUG_ANY | DEBUG_ERR,
1886 		    "incorrect version %d, expecting %d",
1887 		    pim->pim_ver, PIM_VERSION);
1888 		m_freem(m);
1889 		return (IPPROTO_DONE);
1890 	}
1891 
1892 	if (pim->pim_type == PIM_REGISTER) {
1893 		/*
1894 		 * since this is a REGISTER, we'll make a copy of the register
1895 		 * headers ip6+pim+u_int32_t+encap_ip6, to be passed up to the
1896 		 * routing daemon.
1897 		 */
1898 		static struct sockaddr_in6 dst = { sizeof(dst), AF_INET6 };
1899 
1900 		struct mbuf *mcp;
1901 		struct ip6_hdr *eip6;
1902 		u_int32_t *reghdr;
1903 #ifdef MRT6DEBUG
1904 		char ip6bufs[INET6_ADDRSTRLEN], ip6bufd[INET6_ADDRSTRLEN];
1905 #endif
1906 
1907 		PIM6STAT_INC(pim6s_rcv_registers);
1908 
1909 		if (mfct->register_mif >= mfct->nummifs ||
1910 		    mfct->register_mif == (mifi_t)-1) {
1911 			MRT6_DLOG(DEBUG_PIM, "register mif not set: %d",
1912 			    mfct->register_mif);
1913 			m_freem(m);
1914 			return (IPPROTO_DONE);
1915 		}
1916 
1917 		reghdr = (u_int32_t *)(pim + 1);
1918 
1919 		if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
1920 			goto pim6_input_to_daemon;
1921 
1922 		/*
1923 		 * Validate length
1924 		 */
1925 		if (pimlen < PIM6_REG_MINLEN) {
1926 			PIM6STAT_INC(pim6s_rcv_tooshort);
1927 			PIM6STAT_INC(pim6s_rcv_badregisters);
1928 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "register packet "
1929 			    "size too small %d from %s",
1930 			    pimlen, ip6_sprintf(ip6bufs, &ip6->ip6_src));
1931 			m_freem(m);
1932 			return (IPPROTO_DONE);
1933 		}
1934 
1935 		eip6 = (struct ip6_hdr *) (reghdr + 1);
1936 		MRT6_DLOG(DEBUG_PIM, "eip6: %s -> %s, eip6 plen %d",
1937 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1938 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst),
1939 		    ntohs(eip6->ip6_plen));
1940 
1941 		/* verify the version number of the inner packet */
1942 		if ((eip6->ip6_vfc & IPV6_VERSION_MASK) != IPV6_VERSION) {
1943 			PIM6STAT_INC(pim6s_rcv_badregisters);
1944 			MRT6_DLOG(DEBUG_ANY, "invalid IP version (%d) "
1945 			    "of the inner packet",
1946 			    (eip6->ip6_vfc & IPV6_VERSION));
1947 			m_freem(m);
1948 			return (IPPROTO_DONE);
1949 		}
1950 
1951 		/* verify the inner packet is destined to a mcast group */
1952 		if (!IN6_IS_ADDR_MULTICAST(&eip6->ip6_dst)) {
1953 			PIM6STAT_INC(pim6s_rcv_badregisters);
1954 			MRT6_DLOG(DEBUG_PIM, "inner packet of register "
1955 			    "is not multicast %s",
1956 			    ip6_sprintf(ip6bufd, &eip6->ip6_dst));
1957 			m_freem(m);
1958 			return (IPPROTO_DONE);
1959 		}
1960 
1961 		/*
1962 		 * make a copy of the whole header to pass to the daemon later.
1963 		 */
1964 		mcp = m_copym(m, 0, off + PIM6_REG_MINLEN, M_NOWAIT);
1965 		if (mcp == NULL) {
1966 			MRT6_DLOG(DEBUG_ANY | DEBUG_ERR, "pim register: "
1967 			    "could not copy register head");
1968 			m_freem(m);
1969 			return (IPPROTO_DONE);
1970 		}
1971 
1972 		/*
1973 		 * forward the inner ip6 packet; point m_data at the inner ip6.
1974 		 */
1975 		m_adj(m, off + PIM_MINLEN);
1976 		MRT6_DLOG(DEBUG_PIM, "forwarding decapsulated register: "
1977 		    "src %s, dst %s, mif %d",
1978 		    ip6_sprintf(ip6bufs, &eip6->ip6_src),
1979 		    ip6_sprintf(ip6bufd, &eip6->ip6_dst), mfct->register_mif);
1980 
1981 		if_simloop(mfct->miftable[mfct->register_mif].m6_ifp, m,
1982 		    dst.sin6_family, 0);
1983 
1984 		/* prepare the register head to send to the mrouting daemon */
1985 		m = mcp;
1986 	}
1987 
1988 	/*
1989 	 * Pass the PIM message up to the daemon; if it is a register message
1990 	 * pass the 'head' only up to the daemon. This includes the
1991 	 * encapsulator ip6 header, pim header, register header and the
1992 	 * encapsulated ip6 header.
1993 	 */
1994   pim6_input_to_daemon:
1995 	return (rip6_input(&m, &off, proto));
1996 }
1997 
1998 static void
ip6_mrouter_ifdetach(void * arg __unused,struct ifnet * ifp)1999 ip6_mrouter_ifdetach(void *arg __unused, struct ifnet *ifp)
2000 {
2001 	struct mf6ctable *mfct;
2002 
2003 	if (!V_ip6_mrouting_enabled)
2004 		return;
2005 	for (int i = 0; i < V_nmfctables; i++) {
2006 		mfct = &V_mfctables[i];
2007 
2008 		MIF6_LOCK();
2009 restart:
2010 		for (mifi_t mifi = 0; mifi < mfct->nummifs; mifi++) {
2011 			int error __diagused;
2012 
2013 			if (mfct->miftable[mifi].m6_ifp != ifp)
2014 				continue;
2015 			error = del_m6if_locked(mfct, mifi);
2016 			KASSERT(error == 0,
2017 			    ("del_m6if_locked(%s) %d", ifp->if_xname, error));
2018 			goto restart;
2019 		}
2020 		MIF6_UNLOCK();
2021 	}
2022 }
2023 
2024 static void
ip6_mroute_rtnumfibs_change(void * arg __unused,uint32_t ntables)2025 ip6_mroute_rtnumfibs_change(void *arg __unused, uint32_t ntables)
2026 {
2027 	struct mf6ctable *mfctables, *omfctables;
2028 
2029 	KASSERT(ntables >= V_nmfctables,
2030 	    ("%s: ntables %u nmfctables %u", __func__, ntables, V_nmfctables));
2031 
2032 	mfctables = mallocarray(ntables, sizeof(*mfctables), M_MRTABLE6,
2033 	    M_WAITOK | M_ZERO);
2034 	omfctables = V_mfctables;
2035 
2036 	MROUTER6_LOCK();
2037 	MFC6_LOCK();
2038 	for (int i = 0; i < V_nmfctables; i++)
2039 		memcpy(&mfctables[i], &omfctables[i], sizeof(*mfctables));
2040 	atomic_store_rel_ptr((uintptr_t *)&V_mfctables, (uintptr_t)mfctables);
2041 	MFC6_UNLOCK();
2042 	MROUTER6_UNLOCK();
2043 
2044 	NET_EPOCH_WAIT();
2045 
2046 	V_nmfctables = ntables;
2047 	free(omfctables, M_MRTABLE6);
2048 }
2049 
2050 static void
vnet_mroute_init(const void * unused __unused)2051 vnet_mroute_init(const void *unused __unused)
2052 {
2053 	ip6_mroute_rtnumfibs_change(NULL, V_rt_numfibs);
2054 
2055 	callout_init_mtx(&V_expire_upcalls_ch, MFC6_LOCKPTR(), 0);
2056 }
2057 VNET_SYSINIT(vnet_mroute6_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2058     NULL);
2059 
2060 static void
vnet_mroute_uninit(const void * unused __unused)2061 vnet_mroute_uninit(const void *unused __unused)
2062 {
2063 	callout_drain(&V_expire_upcalls_ch);
2064 	free(V_mfctables, M_MRTABLE6);
2065 	V_mfctables = NULL;
2066 }
2067 VNET_SYSUNINIT(vnet_mroute6_uninit, SI_SUB_PROTO_MC, SI_ORDER_ANY,
2068     vnet_mroute_uninit, NULL);
2069 
2070 static int
ip6_mroute_modevent(module_t mod,int type,void * unused)2071 ip6_mroute_modevent(module_t mod, int type, void *unused)
2072 {
2073 
2074 	switch (type) {
2075 	case MOD_LOAD:
2076 		MROUTER6_LOCK_INIT();
2077 		MFC6_LOCK_INIT();
2078 		MIF6_LOCK_INIT();
2079 
2080 		ifdetach_tag = EVENTHANDLER_REGISTER(
2081 		    ifnet_departure_event, ip6_mrouter_ifdetach,
2082 		    NULL, EVENTHANDLER_PRI_ANY);
2083 		rtnumfibs_change_tag = EVENTHANDLER_REGISTER(
2084 		    rtnumfibs_change, ip6_mroute_rtnumfibs_change,
2085 		    NULL, EVENTHANDLER_PRI_ANY);
2086 
2087 		pim6_encap_cookie = ip6_encap_attach(&ipv6_encap_cfg,
2088 		    NULL, M_WAITOK);
2089 		if (pim6_encap_cookie == NULL) {
2090 			printf("ip6_mroute: unable to attach pim6 encap\n");
2091 			MIF6_LOCK_DESTROY();
2092 			MFC6_LOCK_DESTROY();
2093 			MROUTER6_LOCK_DESTROY();
2094 			return (EINVAL);
2095 		}
2096 
2097 		ip6_mforward = X_ip6_mforward;
2098 		ip6_mrouter_done = X_ip6_mrouter_done;
2099 		ip6_mrouter_get = X_ip6_mrouter_get;
2100 		ip6_mrouter_set = X_ip6_mrouter_set;
2101 		mrt6_ioctl = X_mrt6_ioctl;
2102 		break;
2103 
2104 	case MOD_UNLOAD:
2105 		if (V_ip6_mrouting_enabled)
2106 			return (EBUSY);
2107 
2108 		EVENTHANDLER_DEREGISTER(ifnet_departure_event,
2109 		    ifdetach_tag);
2110 		EVENTHANDLER_DEREGISTER(rtnumfibs_change,
2111 		    rtnumfibs_change_tag);
2112 
2113 		if (pim6_encap_cookie) {
2114 			ip6_encap_detach(pim6_encap_cookie);
2115 			pim6_encap_cookie = NULL;
2116 		}
2117 
2118 		ip6_mforward = NULL;
2119 		ip6_mrouter_done = NULL;
2120 		ip6_mrouter_get = NULL;
2121 		ip6_mrouter_set = NULL;
2122 		mrt6_ioctl = NULL;
2123 
2124 		MIF6_LOCK_DESTROY();
2125 		MFC6_LOCK_DESTROY();
2126 		MROUTER6_LOCK_DESTROY();
2127 		break;
2128 
2129 	default:
2130 		return (EOPNOTSUPP);
2131 	}
2132 
2133 	return (0);
2134 }
2135 
2136 static moduledata_t ip6_mroutemod = {
2137 	"ip6_mroute",
2138 	ip6_mroute_modevent,
2139 	0
2140 };
2141 
2142 DECLARE_MODULE(ip6_mroute, ip6_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2143