xref: /freebsd/sys/netinet/ip_mroute.c (revision e92ffd9b626833ebdbf2742c8ffddc6cd94b963e)
1 /*-
2  * SPDX-License-Identifier: BSD-3-Clause
3  *
4  * Copyright (c) 1989 Stephen Deering
5  * Copyright (c) 1992, 1993
6  *      The Regents of the University of California.  All rights reserved.
7  *
8  * This code is derived from software contributed to Berkeley by
9  * Stephen Deering of Stanford University.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 3. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *      @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
36  */
37 
38 /*
39  * IP multicast forwarding procedures
40  *
41  * Written by David Waitzman, BBN Labs, August 1988.
42  * Modified by Steve Deering, Stanford, February 1989.
43  * Modified by Mark J. Steiglitz, Stanford, May, 1991
44  * Modified by Van Jacobson, LBL, January 1993
45  * Modified by Ajit Thyagarajan, PARC, August 1993
46  * Modified by Bill Fenner, PARC, April 1995
47  * Modified by Ahmed Helmy, SGI, June 1996
48  * Modified by George Edmond Eddy (Rusty), ISI, February 1998
49  * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
50  * Modified by Hitoshi Asaeda, WIDE, August 2000
51  * Modified by Pavlin Radoslavov, ICSI, October 2002
52  * Modified by Wojciech Macek, Semihalf, May 2021
53  *
54  * MROUTING Revision: 3.5
55  * and PIM-SMv2 and PIM-DM support, advanced API support,
56  * bandwidth metering and signaling
57  */
58 
59 /*
60  * TODO: Prefix functions with ipmf_.
61  * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
62  * domain attachment (if_afdata) so we can track consumers of that service.
63  * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
64  * move it to socket options.
65  * TODO: Cleanup LSRR removal further.
66  * TODO: Push RSVP stubs into raw_ip.c.
67  * TODO: Use bitstring.h for vif set.
68  * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
69  * TODO: Sync ip6_mroute.c with this file.
70  */
71 
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
74 
75 #include "opt_inet.h"
76 #include "opt_mrouting.h"
77 
78 #define _PIM_VT 1
79 
80 #include <sys/types.h>
81 #include <sys/param.h>
82 #include <sys/kernel.h>
83 #include <sys/stddef.h>
84 #include <sys/condvar.h>
85 #include <sys/eventhandler.h>
86 #include <sys/lock.h>
87 #include <sys/kthread.h>
88 #include <sys/ktr.h>
89 #include <sys/malloc.h>
90 #include <sys/mbuf.h>
91 #include <sys/module.h>
92 #include <sys/priv.h>
93 #include <sys/protosw.h>
94 #include <sys/signalvar.h>
95 #include <sys/socket.h>
96 #include <sys/socketvar.h>
97 #include <sys/sockio.h>
98 #include <sys/sx.h>
99 #include <sys/sysctl.h>
100 #include <sys/syslog.h>
101 #include <sys/systm.h>
102 #include <sys/taskqueue.h>
103 #include <sys/time.h>
104 #include <sys/counter.h>
105 #include <machine/atomic.h>
106 
107 #include <net/if.h>
108 #include <net/if_var.h>
109 #include <net/if_types.h>
110 #include <net/netisr.h>
111 #include <net/route.h>
112 #include <net/vnet.h>
113 
114 #include <netinet/in.h>
115 #include <netinet/igmp.h>
116 #include <netinet/in_systm.h>
117 #include <netinet/in_var.h>
118 #include <netinet/ip.h>
119 #include <netinet/ip_encap.h>
120 #include <netinet/ip_mroute.h>
121 #include <netinet/ip_var.h>
122 #include <netinet/ip_options.h>
123 #include <netinet/pim.h>
124 #include <netinet/pim_var.h>
125 #include <netinet/udp.h>
126 
127 #include <machine/in_cksum.h>
128 
129 #ifndef KTR_IPMF
130 #define KTR_IPMF KTR_INET
131 #endif
132 
133 #define		VIFI_INVALID	((vifi_t) -1)
134 
135 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
136 
137 /*
138  * Locking.  We use two locks: one for the virtual interface table and
139  * one for the forwarding table.  These locks may be nested in which case
140  * the VIF lock must always be taken first.  Note that each lock is used
141  * to cover not only the specific data structure but also related data
142  * structures.
143  */
144 
145 static struct rwlock mrouter_mtx;
146 #define	MRW_RLOCK()		rw_rlock(&mrouter_mtx)
147 #define	MRW_WLOCK()		rw_wlock(&mrouter_mtx)
148 #define	MRW_RUNLOCK()	rw_runlock(&mrouter_mtx)
149 #define	MRW_WUNLOCK()	rw_wunlock(&mrouter_mtx)
150 #define	MRW_UNLOCK()	rw_unlock(&mrouter_mtx)
151 #define	MRW_LOCK_ASSERT()	rw_assert(&mrouter_mtx, RA_LOCKED)
152 #define	MRW_WLOCK_ASSERT()	rw_assert(&mrouter_mtx, RA_WLOCKED)
153 #define	MRW_LOCK_TRY_UPGRADE()	rw_try_upgrade(&mrouter_mtx)
154 #define	MRW_WOWNED()	rw_wowned(&mrouter_mtx)
155 #define	MRW_LOCK_INIT()						\
156 	rw_init(&mrouter_mtx, "IPv4 multicast forwarding")
157 #define	MRW_LOCK_DESTROY()	rw_destroy(&mrouter_mtx)
158 
159 static int ip_mrouter_cnt;	/* # of vnets with active mrouters */
160 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
161 
162 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
163 VNET_PCPUSTAT_SYSINIT(mrtstat);
164 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
165 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
166     mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
167     "netinet/ip_mroute.h)");
168 
169 VNET_DEFINE_STATIC(u_long, mfchash);
170 #define	V_mfchash		VNET(mfchash)
171 #define	MFCHASH(a, g)							\
172 	((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
173 	  ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
174 #define	MFCHASHSIZE	256
175 
176 static u_long mfchashsize;			/* Hash size */
177 VNET_DEFINE_STATIC(u_char *, nexpire);		/* 0..mfchashsize-1 */
178 #define	V_nexpire		VNET(nexpire)
179 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
180 #define	V_mfchashtbl		VNET(mfchashtbl)
181 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
182 #define	V_task_queue		VNET(task_queue)
183 VNET_DEFINE_STATIC(struct task, task);
184 #define	V_task		VNET(task)
185 
186 VNET_DEFINE_STATIC(vifi_t, numvifs);
187 #define	V_numvifs		VNET(numvifs)
188 VNET_DEFINE_STATIC(struct vif *, viftable);
189 #define	V_viftable		VNET(viftable)
190 
191 static eventhandler_tag if_detach_event_tag = NULL;
192 
193 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
194 #define	V_expire_upcalls_ch	VNET(expire_upcalls_ch)
195 
196 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
197 #define	V_buf_ring_mtx	VNET(buf_ring_mtx)
198 
199 #define		EXPIRE_TIMEOUT	(hz / 4)	/* 4x / second		*/
200 #define		UPCALL_EXPIRE	6		/* number of timeouts	*/
201 
202 /*
203  * Bandwidth meter variables and constants
204  */
205 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 
207 /*
208  * Pending upcalls are stored in a ring which is flushed when
209  * full, or periodically
210  */
211 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
212 #define	V_bw_upcalls_ch		VNET(bw_upcalls_ch)
213 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
214 #define	V_bw_upcalls_ring    	VNET(bw_upcalls_ring)
215 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
216 #define	V_bw_upcalls_ring_mtx    	VNET(bw_upcalls_ring_mtx)
217 
218 #define BW_UPCALLS_PERIOD (hz)		/* periodical flush of bw upcalls */
219 
220 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
221 VNET_PCPUSTAT_SYSINIT(pimstat);
222 VNET_PCPUSTAT_SYSUNINIT(pimstat);
223 
224 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
225     "PIM");
226 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
227     pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
228 
229 static u_long	pim_squelch_wholepkt = 0;
230 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
231     &pim_squelch_wholepkt, 0,
232     "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
233 
234 static const struct encaptab *pim_encap_cookie;
235 static int pim_encapcheck(const struct mbuf *, int, int, void *);
236 static int pim_input(struct mbuf *, int, int, void *);
237 
238 extern int in_mcast_loop;
239 
240 static const struct encap_config ipv4_encap_cfg = {
241 	.proto = IPPROTO_PIM,
242 	.min_length = sizeof(struct ip) + PIM_MINLEN,
243 	.exact_match = 8,
244 	.check = pim_encapcheck,
245 	.input = pim_input
246 };
247 
248 /*
249  * Note: the PIM Register encapsulation adds the following in front of a
250  * data packet:
251  *
252  * struct pim_encap_hdr {
253  *    struct ip ip;
254  *    struct pim_encap_pimhdr  pim;
255  * }
256  *
257  */
258 
259 struct pim_encap_pimhdr {
260 	struct pim pim;
261 	uint32_t   flags;
262 };
263 #define		PIM_ENCAP_TTL	64
264 
265 static struct ip pim_encap_iphdr = {
266 #if BYTE_ORDER == LITTLE_ENDIAN
267 	sizeof(struct ip) >> 2,
268 	IPVERSION,
269 #else
270 	IPVERSION,
271 	sizeof(struct ip) >> 2,
272 #endif
273 	0,			/* tos */
274 	sizeof(struct ip),	/* total length */
275 	0,			/* id */
276 	0,			/* frag offset */
277 	PIM_ENCAP_TTL,
278 	IPPROTO_PIM,
279 	0,			/* checksum */
280 };
281 
282 static struct pim_encap_pimhdr pim_encap_pimhdr = {
283     {
284 	PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
285 	0,			/* reserved */
286 	0,			/* checksum */
287     },
288     0				/* flags */
289 };
290 
291 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
292 #define	V_reg_vif_num		VNET(reg_vif_num)
293 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
294 #define	V_multicast_register_if	VNET(multicast_register_if)
295 
296 /*
297  * Private variables.
298  */
299 
300 static u_long	X_ip_mcast_src(int);
301 static int	X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
302 		    struct ip_moptions *);
303 static int	X_ip_mrouter_done(void *);
304 static int	X_ip_mrouter_get(struct socket *, struct sockopt *);
305 static int	X_ip_mrouter_set(struct socket *, struct sockopt *);
306 static int	X_legal_vif_num(int);
307 static int	X_mrt_ioctl(u_long, caddr_t, int);
308 
309 static int	add_bw_upcall(struct bw_upcall *);
310 static int	add_mfc(struct mfcctl2 *);
311 static int	add_vif(struct vifctl *);
312 static void	bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
313 static void	bw_meter_geq_receive_packet(struct bw_meter *, int,
314 		    struct timeval *);
315 static void	bw_upcalls_send(void);
316 static int	del_bw_upcall(struct bw_upcall *);
317 static int	del_mfc(struct mfcctl2 *);
318 static int	del_vif(vifi_t);
319 static int	del_vif_locked(vifi_t);
320 static void	expire_bw_upcalls_send(void *);
321 static void	expire_mfc(struct mfc *);
322 static void	expire_upcalls(void *);
323 static void	free_bw_list(struct bw_meter *);
324 static int	get_sg_cnt(struct sioc_sg_req *);
325 static int	get_vif_cnt(struct sioc_vif_req *);
326 static void	if_detached_event(void *, struct ifnet *);
327 static int	ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
328 static int	ip_mrouter_init(struct socket *, int);
329 static __inline struct mfc *
330 		mfc_find(struct in_addr *, struct in_addr *);
331 static void	phyint_send(struct ip *, struct vif *, struct mbuf *);
332 static struct mbuf *
333 		pim_register_prepare(struct ip *, struct mbuf *);
334 static int	pim_register_send(struct ip *, struct vif *,
335 		    struct mbuf *, struct mfc *);
336 static int	pim_register_send_rp(struct ip *, struct vif *,
337 		    struct mbuf *, struct mfc *);
338 static int	pim_register_send_upcall(struct ip *, struct vif *,
339 		    struct mbuf *, struct mfc *);
340 static void	send_packet(struct vif *, struct mbuf *);
341 static int	set_api_config(uint32_t *);
342 static int	set_assert(int);
343 static int	socket_send(struct socket *, struct mbuf *,
344 		    struct sockaddr_in *);
345 
346 /*
347  * Kernel multicast forwarding API capabilities and setup.
348  * If more API capabilities are added to the kernel, they should be
349  * recorded in `mrt_api_support'.
350  */
351 #define MRT_API_VERSION		0x0305
352 
353 static const int mrt_api_version = MRT_API_VERSION;
354 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
355 					 MRT_MFC_FLAGS_BORDER_VIF |
356 					 MRT_MFC_RP |
357 					 MRT_MFC_BW_UPCALL);
358 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
359 #define	V_mrt_api_config	VNET(mrt_api_config)
360 VNET_DEFINE_STATIC(int, pim_assert_enabled);
361 #define	V_pim_assert_enabled	VNET(pim_assert_enabled)
362 static struct timeval pim_assert_interval = { 3, 0 };	/* Rate limit */
363 
364 /*
365  * Find a route for a given origin IP address and multicast group address.
366  * Statistics must be updated by the caller.
367  */
368 static __inline struct mfc *
369 mfc_find(struct in_addr *o, struct in_addr *g)
370 {
371 	struct mfc *rt;
372 
373 	/*
374 	 * Might be called both RLOCK and WLOCK.
375 	 * Check if any, it's caller responsibility
376 	 * to choose correct option.
377 	 */
378 	MRW_LOCK_ASSERT();
379 
380 	LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
381 		if (in_hosteq(rt->mfc_origin, *o) &&
382 		    in_hosteq(rt->mfc_mcastgrp, *g) &&
383 		    buf_ring_empty(rt->mfc_stall_ring))
384 			break;
385 	}
386 
387 	return (rt);
388 }
389 
390 static __inline struct mfc *
391 mfc_alloc(void)
392 {
393 	struct mfc *rt;
394 	rt = (struct mfc*) malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
395 	if (rt == NULL)
396 		return rt;
397 
398 	rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
399 	    M_NOWAIT, &V_buf_ring_mtx);
400 	if (rt->mfc_stall_ring == NULL) {
401 		free(rt, M_MRTABLE);
402 		return NULL;
403 	}
404 
405 	return rt;
406 }
407 
408 /*
409  * Handle MRT setsockopt commands to modify the multicast forwarding tables.
410  */
411 static int
412 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
413 {
414     int	error, optval;
415     vifi_t	vifi;
416     struct	vifctl vifc;
417     struct	mfcctl2 mfc;
418     struct	bw_upcall bw_upcall;
419     uint32_t	i;
420 
421     if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
422 	return EPERM;
423 
424     error = 0;
425     switch (sopt->sopt_name) {
426     case MRT_INIT:
427 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
428 	if (error)
429 	    break;
430 	error = ip_mrouter_init(so, optval);
431 	break;
432 
433     case MRT_DONE:
434 	error = ip_mrouter_done(NULL);
435 	break;
436 
437     case MRT_ADD_VIF:
438 	error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
439 	if (error)
440 	    break;
441 	error = add_vif(&vifc);
442 	break;
443 
444     case MRT_DEL_VIF:
445 	error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
446 	if (error)
447 	    break;
448 	error = del_vif(vifi);
449 	break;
450 
451     case MRT_ADD_MFC:
452     case MRT_DEL_MFC:
453 	/*
454 	 * select data size depending on API version.
455 	 */
456 	if (sopt->sopt_name == MRT_ADD_MFC &&
457 		V_mrt_api_config & MRT_API_FLAGS_ALL) {
458 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
459 				sizeof(struct mfcctl2));
460 	} else {
461 	    error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
462 				sizeof(struct mfcctl));
463 	    bzero((caddr_t)&mfc + sizeof(struct mfcctl),
464 			sizeof(mfc) - sizeof(struct mfcctl));
465 	}
466 	if (error)
467 	    break;
468 	if (sopt->sopt_name == MRT_ADD_MFC)
469 	    error = add_mfc(&mfc);
470 	else
471 	    error = del_mfc(&mfc);
472 	break;
473 
474     case MRT_ASSERT:
475 	error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
476 	if (error)
477 	    break;
478 	set_assert(optval);
479 	break;
480 
481     case MRT_API_CONFIG:
482 	error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
483 	if (!error)
484 	    error = set_api_config(&i);
485 	if (!error)
486 	    error = sooptcopyout(sopt, &i, sizeof i);
487 	break;
488 
489     case MRT_ADD_BW_UPCALL:
490     case MRT_DEL_BW_UPCALL:
491 	error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
492 				sizeof bw_upcall);
493 	if (error)
494 	    break;
495 	if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
496 	    error = add_bw_upcall(&bw_upcall);
497 	else
498 	    error = del_bw_upcall(&bw_upcall);
499 	break;
500 
501     default:
502 	error = EOPNOTSUPP;
503 	break;
504     }
505     return error;
506 }
507 
508 /*
509  * Handle MRT getsockopt commands
510  */
511 static int
512 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
513 {
514     int error;
515 
516     switch (sopt->sopt_name) {
517     case MRT_VERSION:
518 	error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
519 	break;
520 
521     case MRT_ASSERT:
522 	error = sooptcopyout(sopt, &V_pim_assert_enabled,
523 	    sizeof V_pim_assert_enabled);
524 	break;
525 
526     case MRT_API_SUPPORT:
527 	error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
528 	break;
529 
530     case MRT_API_CONFIG:
531 	error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
532 	break;
533 
534     default:
535 	error = EOPNOTSUPP;
536 	break;
537     }
538     return error;
539 }
540 
541 /*
542  * Handle ioctl commands to obtain information from the cache
543  */
544 static int
545 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
546 {
547     int error = 0;
548 
549     /*
550      * Currently the only function calling this ioctl routine is rtioctl_fib().
551      * Typically, only root can create the raw socket in order to execute
552      * this ioctl method, however the request might be coming from a prison
553      */
554     error = priv_check(curthread, PRIV_NETINET_MROUTE);
555     if (error)
556 	return (error);
557     switch (cmd) {
558     case (SIOCGETVIFCNT):
559 	error = get_vif_cnt((struct sioc_vif_req *)data);
560 	break;
561 
562     case (SIOCGETSGCNT):
563 	error = get_sg_cnt((struct sioc_sg_req *)data);
564 	break;
565 
566     default:
567 	error = EINVAL;
568 	break;
569     }
570     return error;
571 }
572 
573 /*
574  * returns the packet, byte, rpf-failure count for the source group provided
575  */
576 static int
577 get_sg_cnt(struct sioc_sg_req *req)
578 {
579     struct mfc *rt;
580 
581     MRW_RLOCK();
582     rt = mfc_find(&req->src, &req->grp);
583     if (rt == NULL) {
584 	    MRW_RUNLOCK();
585 	req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
586 	return EADDRNOTAVAIL;
587     }
588     req->pktcnt = rt->mfc_pkt_cnt;
589     req->bytecnt = rt->mfc_byte_cnt;
590     req->wrong_if = rt->mfc_wrong_if;
591     MRW_RUNLOCK();
592     return 0;
593 }
594 
595 /*
596  * returns the input and output packet and byte counts on the vif provided
597  */
598 static int
599 get_vif_cnt(struct sioc_vif_req *req)
600 {
601     vifi_t vifi = req->vifi;
602 
603     MRW_RLOCK();
604     if (vifi >= V_numvifs) {
605 	MRW_RUNLOCK();
606 	return EINVAL;
607     }
608 
609     mtx_lock_spin(&V_viftable[vifi].v_spin);
610     req->icount = V_viftable[vifi].v_pkt_in;
611     req->ocount = V_viftable[vifi].v_pkt_out;
612     req->ibytes = V_viftable[vifi].v_bytes_in;
613     req->obytes = V_viftable[vifi].v_bytes_out;
614     mtx_unlock_spin(&V_viftable[vifi].v_spin);
615     MRW_RUNLOCK();
616 
617     return 0;
618 }
619 
620 static void
621 if_detached_event(void *arg __unused, struct ifnet *ifp)
622 {
623     vifi_t vifi;
624     u_long i;
625 
626     MRW_WLOCK();
627 
628     if (V_ip_mrouter == NULL) {
629 	MRW_WUNLOCK();
630 	return;
631     }
632 
633     /*
634      * Tear down multicast forwarder state associated with this ifnet.
635      * 1. Walk the vif list, matching vifs against this ifnet.
636      * 2. Walk the multicast forwarding cache (mfc) looking for
637      *    inner matches with this vif's index.
638      * 3. Expire any matching multicast forwarding cache entries.
639      * 4. Free vif state. This should disable ALLMULTI on the interface.
640      */
641     for (vifi = 0; vifi < V_numvifs; vifi++) {
642 	if (V_viftable[vifi].v_ifp != ifp)
643 		continue;
644 	for (i = 0; i < mfchashsize; i++) {
645 		struct mfc *rt, *nrt;
646 
647 		LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
648 			if (rt->mfc_parent == vifi) {
649 				expire_mfc(rt);
650 			}
651 		}
652 	}
653 	del_vif_locked(vifi);
654     }
655 
656     MRW_WUNLOCK();
657 }
658 
659 static void
660 ip_mrouter_upcall_thread(void *arg, int pending __unused)
661 {
662 	CURVNET_SET((struct vnet *) arg);
663 
664 	MRW_WLOCK();
665 	bw_upcalls_send();
666 	MRW_WUNLOCK();
667 
668 	CURVNET_RESTORE();
669 }
670 
671 /*
672  * Enable multicast forwarding.
673  */
674 static int
675 ip_mrouter_init(struct socket *so, int version)
676 {
677 
678     CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
679         so->so_type, so->so_proto->pr_protocol);
680 
681     if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
682 	return EOPNOTSUPP;
683 
684     if (version != 1)
685 	return ENOPROTOOPT;
686 
687     MRW_WLOCK();
688 
689     if (ip_mrouter_unloading) {
690 	MRW_WUNLOCK();
691 	return ENOPROTOOPT;
692     }
693 
694     if (V_ip_mrouter != NULL) {
695 	MRW_WUNLOCK();
696 	return EADDRINUSE;
697     }
698 
699     V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
700 	HASH_NOWAIT);
701 
702     /* Create upcall ring */
703     mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
704     V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
705 	M_NOWAIT, &V_bw_upcalls_ring_mtx);
706     if (!V_bw_upcalls_ring) {
707 	MRW_WUNLOCK();
708 	return (ENOMEM);
709     }
710 
711     TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
712     taskqueue_cancel(V_task_queue, &V_task, NULL);
713     taskqueue_unblock(V_task_queue);
714 
715     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
716 	curvnet);
717     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
718 	curvnet);
719 
720     V_ip_mrouter = so;
721     atomic_add_int(&ip_mrouter_cnt, 1);
722 
723     /* This is a mutex required by buf_ring init, but not used internally */
724     mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
725 
726     MRW_WUNLOCK();
727 
728     CTR1(KTR_IPMF, "%s: done", __func__);
729 
730     return 0;
731 }
732 
733 /*
734  * Disable multicast forwarding.
735  */
736 static int
737 X_ip_mrouter_done(void *locked)
738 {
739     struct ifnet *ifp;
740     u_long i;
741     vifi_t vifi;
742     struct bw_upcall *bu;
743 
744     if (V_ip_mrouter == NULL) {
745 	if (locked) {
746 		struct epoch_tracker *mrouter_et = locked;
747 		MROUTER_RUNLOCK_PARAM(mrouter_et);
748 	}
749 	return EINVAL;
750     }
751 
752     /*
753      * Detach/disable hooks to the reset of the system.
754      */
755     V_ip_mrouter = NULL;
756     atomic_subtract_int(&ip_mrouter_cnt, 1);
757     V_mrt_api_config = 0;
758 
759     if (locked) {
760 	struct epoch_tracker *mrouter_et = locked;
761 	MROUTER_RUNLOCK_PARAM(mrouter_et);
762     }
763 
764     MROUTER_WAIT();
765 
766     /* Stop and drain task queue */
767     taskqueue_block(V_task_queue);
768     while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
769     	taskqueue_drain(V_task_queue, &V_task);
770     }
771 
772     MRW_WLOCK();
773     taskqueue_cancel(V_task_queue, &V_task, NULL);
774 
775     /* Destroy upcall ring */
776     while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
777 	free(bu, M_MRTABLE);
778     }
779     buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
780     mtx_destroy(&V_bw_upcalls_ring_mtx);
781 
782     /*
783      * For each phyint in use, disable promiscuous reception of all IP
784      * multicasts.
785      */
786     for (vifi = 0; vifi < V_numvifs; vifi++) {
787 	if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
788 		!(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
789 	    ifp = V_viftable[vifi].v_ifp;
790 	    if_allmulti(ifp, 0);
791 	}
792     }
793     bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
794     V_numvifs = 0;
795     V_pim_assert_enabled = 0;
796 
797     callout_stop(&V_expire_upcalls_ch);
798     callout_stop(&V_bw_upcalls_ch);
799 
800     /*
801      * Free all multicast forwarding cache entries.
802      * Do not use hashdestroy(), as we must perform other cleanup.
803      */
804     for (i = 0; i < mfchashsize; i++) {
805 	struct mfc *rt, *nrt;
806 
807 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
808 		expire_mfc(rt);
809 	}
810     }
811     free(V_mfchashtbl, M_MRTABLE);
812     V_mfchashtbl = NULL;
813 
814     bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
815 
816     V_reg_vif_num = VIFI_INVALID;
817 
818     mtx_destroy(&V_buf_ring_mtx);
819 
820     MRW_WUNLOCK();
821 
822     CTR1(KTR_IPMF, "%s: done", __func__);
823 
824     return 0;
825 }
826 
827 /*
828  * Set PIM assert processing global
829  */
830 static int
831 set_assert(int i)
832 {
833     if ((i != 1) && (i != 0))
834 	return EINVAL;
835 
836     V_pim_assert_enabled = i;
837 
838     return 0;
839 }
840 
841 /*
842  * Configure API capabilities
843  */
844 int
845 set_api_config(uint32_t *apival)
846 {
847     u_long i;
848 
849     /*
850      * We can set the API capabilities only if it is the first operation
851      * after MRT_INIT. I.e.:
852      *  - there are no vifs installed
853      *  - pim_assert is not enabled
854      *  - the MFC table is empty
855      */
856     if (V_numvifs > 0) {
857 	*apival = 0;
858 	return EPERM;
859     }
860     if (V_pim_assert_enabled) {
861 	*apival = 0;
862 	return EPERM;
863     }
864 
865     MRW_RLOCK();
866 
867     for (i = 0; i < mfchashsize; i++) {
868 	if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
869 	    MRW_RUNLOCK();
870 	    *apival = 0;
871 	    return EPERM;
872 	}
873     }
874 
875     MRW_RUNLOCK();
876 
877     V_mrt_api_config = *apival & mrt_api_support;
878     *apival = V_mrt_api_config;
879 
880     return 0;
881 }
882 
883 /*
884  * Add a vif to the vif table
885  */
886 static int
887 add_vif(struct vifctl *vifcp)
888 {
889     struct vif *vifp = V_viftable + vifcp->vifc_vifi;
890     struct sockaddr_in sin = {sizeof sin, AF_INET};
891     struct ifaddr *ifa;
892     struct ifnet *ifp;
893     int error;
894 
895 
896     if (vifcp->vifc_vifi >= MAXVIFS)
897 	return EINVAL;
898     /* rate limiting is no longer supported by this code */
899     if (vifcp->vifc_rate_limit != 0) {
900 	log(LOG_ERR, "rate limiting is no longer supported\n");
901 	return EINVAL;
902     }
903 
904     if (in_nullhost(vifcp->vifc_lcl_addr))
905 	return EADDRNOTAVAIL;
906 
907     /* Find the interface with an address in AF_INET family */
908     if (vifcp->vifc_flags & VIFF_REGISTER) {
909 	/*
910 	 * XXX: Because VIFF_REGISTER does not really need a valid
911 	 * local interface (e.g. it could be 127.0.0.2), we don't
912 	 * check its address.
913 	 */
914 	ifp = NULL;
915     } else {
916 	struct epoch_tracker et;
917 
918 	sin.sin_addr = vifcp->vifc_lcl_addr;
919 	NET_EPOCH_ENTER(et);
920 	ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
921 	if (ifa == NULL) {
922 	    NET_EPOCH_EXIT(et);
923 	    return EADDRNOTAVAIL;
924 	}
925 	ifp = ifa->ifa_ifp;
926 	/* XXX FIXME we need to take a ref on ifp and cleanup properly! */
927 	NET_EPOCH_EXIT(et);
928     }
929 
930     if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
931 	CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
932 	return EOPNOTSUPP;
933     } else if (vifcp->vifc_flags & VIFF_REGISTER) {
934 	ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
935 	CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
936 	if (V_reg_vif_num == VIFI_INVALID) {
937 	    if_initname(V_multicast_register_if, "register_vif", 0);
938 	    V_reg_vif_num = vifcp->vifc_vifi;
939 	}
940     } else {		/* Make sure the interface supports multicast */
941 	if ((ifp->if_flags & IFF_MULTICAST) == 0)
942 	    return EOPNOTSUPP;
943 
944 	/* Enable promiscuous reception of all IP multicasts from the if */
945 	error = if_allmulti(ifp, 1);
946 	if (error)
947 	    return error;
948     }
949 
950     MRW_WLOCK();
951 
952     if (!in_nullhost(vifp->v_lcl_addr)) {
953 	if (ifp)
954 		V_multicast_register_if = NULL;
955 	MRW_WUNLOCK();
956 	if (ifp)
957 		if_free(ifp);
958 	return EADDRINUSE;
959     }
960 
961     vifp->v_flags     = vifcp->vifc_flags;
962     vifp->v_threshold = vifcp->vifc_threshold;
963     vifp->v_lcl_addr  = vifcp->vifc_lcl_addr;
964     vifp->v_rmt_addr  = vifcp->vifc_rmt_addr;
965     vifp->v_ifp       = ifp;
966     /* initialize per vif pkt counters */
967     vifp->v_pkt_in    = 0;
968     vifp->v_pkt_out   = 0;
969     vifp->v_bytes_in  = 0;
970     vifp->v_bytes_out = 0;
971     sprintf(vifp->v_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
972     mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
973 
974     /* Adjust numvifs up if the vifi is higher than numvifs */
975     if (V_numvifs <= vifcp->vifc_vifi)
976 	V_numvifs = vifcp->vifc_vifi + 1;
977 
978     MRW_WUNLOCK();
979 
980     CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
981 	(int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
982 	(int)vifcp->vifc_threshold);
983 
984     return 0;
985 }
986 
987 /*
988  * Delete a vif from the vif table
989  */
990 static int
991 del_vif_locked(vifi_t vifi)
992 {
993     struct vif *vifp;
994 
995     MRW_WLOCK_ASSERT();
996 
997     if (vifi >= V_numvifs) {
998 	return EINVAL;
999     }
1000     vifp = &V_viftable[vifi];
1001     if (in_nullhost(vifp->v_lcl_addr)) {
1002 	return EADDRNOTAVAIL;
1003     }
1004 
1005     if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1006 	if_allmulti(vifp->v_ifp, 0);
1007 
1008     if (vifp->v_flags & VIFF_REGISTER) {
1009 	V_reg_vif_num = VIFI_INVALID;
1010 	if (vifp->v_ifp) {
1011 	    if (vifp->v_ifp == V_multicast_register_if)
1012 	        V_multicast_register_if = NULL;
1013 	    if_free(vifp->v_ifp);
1014 	}
1015     }
1016 
1017     mtx_destroy(&vifp->v_spin);
1018 
1019     bzero((caddr_t)vifp, sizeof (*vifp));
1020 
1021     CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1022 
1023     /* Adjust numvifs down */
1024     for (vifi = V_numvifs; vifi > 0; vifi--)
1025 	if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1026 	    break;
1027     V_numvifs = vifi;
1028 
1029     return 0;
1030 }
1031 
1032 static int
1033 del_vif(vifi_t vifi)
1034 {
1035     int cc;
1036 
1037     MRW_WLOCK();
1038     cc = del_vif_locked(vifi);
1039     MRW_WUNLOCK();
1040 
1041     return cc;
1042 }
1043 
1044 /*
1045  * update an mfc entry without resetting counters and S,G addresses.
1046  */
1047 static void
1048 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1049 {
1050     int i;
1051 
1052     rt->mfc_parent = mfccp->mfcc_parent;
1053     for (i = 0; i < V_numvifs; i++) {
1054 	rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1055 	rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1056 	    MRT_MFC_FLAGS_ALL;
1057     }
1058     /* set the RP address */
1059     if (V_mrt_api_config & MRT_MFC_RP)
1060 	rt->mfc_rp = mfccp->mfcc_rp;
1061     else
1062 	rt->mfc_rp.s_addr = INADDR_ANY;
1063 }
1064 
1065 /*
1066  * fully initialize an mfc entry from the parameter.
1067  */
1068 static void
1069 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1070 {
1071     rt->mfc_origin     = mfccp->mfcc_origin;
1072     rt->mfc_mcastgrp   = mfccp->mfcc_mcastgrp;
1073 
1074     update_mfc_params(rt, mfccp);
1075 
1076     /* initialize pkt counters per src-grp */
1077     rt->mfc_pkt_cnt    = 0;
1078     rt->mfc_byte_cnt   = 0;
1079     rt->mfc_wrong_if   = 0;
1080     timevalclear(&rt->mfc_last_assert);
1081 }
1082 
1083 static void
1084 expire_mfc(struct mfc *rt)
1085 {
1086 	struct rtdetq *rte;
1087 
1088 	MRW_WLOCK_ASSERT();
1089 
1090 	free_bw_list(rt->mfc_bw_meter_leq);
1091 	free_bw_list(rt->mfc_bw_meter_geq);
1092 
1093 	while (!buf_ring_empty(rt->mfc_stall_ring)) {
1094 		rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1095 		if (rte) {
1096 			m_freem(rte->m);
1097 			free(rte, M_MRTABLE);
1098 		}
1099 	}
1100 	buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1101 
1102 	LIST_REMOVE(rt, mfc_hash);
1103 	free(rt, M_MRTABLE);
1104 }
1105 
1106 /*
1107  * Add an mfc entry
1108  */
1109 static int
1110 add_mfc(struct mfcctl2 *mfccp)
1111 {
1112     struct mfc *rt;
1113     struct rtdetq *rte;
1114     u_long hash = 0;
1115     u_short nstl;
1116 
1117     MRW_WLOCK();
1118     rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1119 
1120     /* If an entry already exists, just update the fields */
1121     if (rt) {
1122 	CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1123 	    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1124 	    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1125 	    mfccp->mfcc_parent);
1126 	update_mfc_params(rt, mfccp);
1127 	MRW_WUNLOCK();
1128 	return (0);
1129     }
1130 
1131     /*
1132      * Find the entry for which the upcall was made and update
1133      */
1134     nstl = 0;
1135     hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1136     LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1137 	if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1138 	    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1139 	    !buf_ring_empty(rt->mfc_stall_ring)) {
1140 		CTR5(KTR_IPMF,
1141 		    "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1142 		    __func__, ntohl(mfccp->mfcc_origin.s_addr),
1143 		    (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1144 		    mfccp->mfcc_parent,
1145 		    rt->mfc_stall_ring);
1146 		if (nstl++)
1147 			CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1148 
1149 		init_mfc_params(rt, mfccp);
1150 		rt->mfc_expire = 0;	/* Don't clean this guy up */
1151 		V_nexpire[hash]--;
1152 
1153 		/* Free queued packets, but attempt to forward them first. */
1154 		while (!buf_ring_empty(rt->mfc_stall_ring)) {
1155 			rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1156 			if (rte->ifp != NULL)
1157 				ip_mdq(rte->m, rte->ifp, rt, -1);
1158 			m_freem(rte->m);
1159 			free(rte, M_MRTABLE);
1160 		}
1161 	}
1162     }
1163 
1164     /*
1165      * It is possible that an entry is being inserted without an upcall
1166      */
1167     if (nstl == 0) {
1168 	CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1169 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1170 		if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1171 		    in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1172 			init_mfc_params(rt, mfccp);
1173 			if (rt->mfc_expire)
1174 			    V_nexpire[hash]--;
1175 			rt->mfc_expire = 0;
1176 			break; /* XXX */
1177 		}
1178 	}
1179 
1180 	if (rt == NULL) {		/* no upcall, so make a new entry */
1181 	    rt = mfc_alloc();
1182 	    if (rt == NULL) {
1183 		MRW_WUNLOCK();
1184 		return (ENOBUFS);
1185 	    }
1186 
1187 	    init_mfc_params(rt, mfccp);
1188 
1189 	    rt->mfc_expire     = 0;
1190 	    rt->mfc_bw_meter_leq = NULL;
1191 	    rt->mfc_bw_meter_geq = NULL;
1192 
1193 	    /* insert new entry at head of hash chain */
1194 	    LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1195 	}
1196     }
1197 
1198     MRW_WUNLOCK();
1199 
1200     return (0);
1201 }
1202 
1203 /*
1204  * Delete an mfc entry
1205  */
1206 static int
1207 del_mfc(struct mfcctl2 *mfccp)
1208 {
1209     struct in_addr	origin;
1210     struct in_addr	mcastgrp;
1211     struct mfc		*rt;
1212 
1213     origin = mfccp->mfcc_origin;
1214     mcastgrp = mfccp->mfcc_mcastgrp;
1215 
1216     CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1217 	ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1218 
1219     MRW_WLOCK();
1220 
1221     rt = mfc_find(&origin, &mcastgrp);
1222     if (rt == NULL) {
1223 	MRW_WUNLOCK();
1224 	return EADDRNOTAVAIL;
1225     }
1226 
1227     /*
1228      * free the bw_meter entries
1229      */
1230     free_bw_list(rt->mfc_bw_meter_leq);
1231     rt->mfc_bw_meter_leq = NULL;
1232     free_bw_list(rt->mfc_bw_meter_geq);
1233     rt->mfc_bw_meter_geq = NULL;
1234 
1235     LIST_REMOVE(rt, mfc_hash);
1236     free(rt, M_MRTABLE);
1237 
1238     MRW_WUNLOCK();
1239 
1240     return (0);
1241 }
1242 
1243 /*
1244  * Send a message to the routing daemon on the multicast routing socket.
1245  */
1246 static int
1247 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1248 {
1249     if (s) {
1250 	SOCKBUF_LOCK(&s->so_rcv);
1251 	if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1252 	    NULL) != 0) {
1253 	    sorwakeup_locked(s);
1254 	    return 0;
1255 	}
1256 	soroverflow_locked(s);
1257     }
1258     m_freem(mm);
1259     return -1;
1260 }
1261 
1262 /*
1263  * IP multicast forwarding function. This function assumes that the packet
1264  * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1265  * pointed to by "ifp", and the packet is to be relayed to other networks
1266  * that have members of the packet's destination IP multicast group.
1267  *
1268  * The packet is returned unscathed to the caller, unless it is
1269  * erroneous, in which case a non-zero return value tells the caller to
1270  * discard it.
1271  */
1272 
1273 #define TUNNEL_LEN  12  /* # bytes of IP option for tunnel encapsulation  */
1274 
1275 static int
1276 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1277     struct ip_moptions *imo)
1278 {
1279 	struct mfc *rt;
1280 	int error;
1281 	vifi_t vifi;
1282 	struct mbuf *mb0;
1283 	struct rtdetq *rte;
1284 	u_long hash;
1285 	int hlen;
1286 
1287 	CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1288 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1289 
1290 	if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1291 	    ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1292 		/*
1293 		 * Packet arrived via a physical interface or
1294 		 * an encapsulated tunnel or a register_vif.
1295 		 */
1296 	} else {
1297 		/*
1298 		 * Packet arrived through a source-route tunnel.
1299 		 * Source-route tunnels are no longer supported.
1300 		 */
1301 		return (1);
1302 	}
1303 
1304 	/*
1305 	 * BEGIN: MCAST ROUTING HOT PATH
1306 	 */
1307 	MRW_RLOCK();
1308 	if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1309 		if (ip->ip_ttl < MAXTTL)
1310 			ip->ip_ttl++; /* compensate for -1 in *_send routines */
1311 		error = ip_mdq(m, ifp, NULL, vifi);
1312 		MRW_RUNLOCK();
1313 		return error;
1314 	}
1315 
1316 	/*
1317 	 * Don't forward a packet with time-to-live of zero or one,
1318 	 * or a packet destined to a local-only group.
1319 	 */
1320 	if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1321 		MRW_RUNLOCK();
1322 		return 0;
1323 	}
1324 
1325 	mfc_find_retry:
1326 	/*
1327 	 * Determine forwarding vifs from the forwarding cache table
1328 	 */
1329 	MRTSTAT_INC(mrts_mfc_lookups);
1330 	rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1331 
1332 	/* Entry exists, so forward if necessary */
1333 	if (rt != NULL) {
1334 		error = ip_mdq(m, ifp, rt, -1);
1335 		/* Generic unlock here as we might release R or W lock */
1336 		MRW_UNLOCK();
1337 		return error;
1338 	}
1339 
1340 	/*
1341 	 * END: MCAST ROUTING HOT PATH
1342 	 */
1343 
1344 	/* Further processing must be done with WLOCK taken */
1345 	if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1346 		MRW_RUNLOCK();
1347 		MRW_WLOCK();
1348 		goto mfc_find_retry;
1349 	}
1350 
1351 	/*
1352 	 * If we don't have a route for packet's origin,
1353 	 * Make a copy of the packet & send message to routing daemon
1354 	 */
1355 	hlen = ip->ip_hl << 2;
1356 
1357 	MRTSTAT_INC(mrts_mfc_misses);
1358 	MRTSTAT_INC(mrts_no_route);
1359 	CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1360 	    ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1361 
1362 	/*
1363 	 * Allocate mbufs early so that we don't do extra work if we are
1364 	 * just going to fail anyway.  Make sure to pullup the header so
1365 	 * that other people can't step on it.
1366 	 */
1367 	rte = (struct rtdetq*) malloc((sizeof *rte), M_MRTABLE,
1368 	    M_NOWAIT|M_ZERO);
1369 	if (rte == NULL) {
1370 		MRW_WUNLOCK();
1371 		return ENOBUFS;
1372 	}
1373 
1374 	mb0 = m_copypacket(m, M_NOWAIT);
1375 	if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1376 		mb0 = m_pullup(mb0, hlen);
1377 	if (mb0 == NULL) {
1378 		free(rte, M_MRTABLE);
1379 		MRW_WUNLOCK();
1380 		return ENOBUFS;
1381 	}
1382 
1383 	/* is there an upcall waiting for this flow ? */
1384 	hash = MFCHASH(ip->ip_src, ip->ip_dst);
1385 	LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1386 	{
1387 		if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1388 		    in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1389 		    !buf_ring_empty(rt->mfc_stall_ring))
1390 			break;
1391 	}
1392 
1393 	if (rt == NULL) {
1394 		int i;
1395 		struct igmpmsg *im;
1396 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1397 		struct mbuf *mm;
1398 
1399 		/*
1400 		 * Locate the vifi for the incoming interface for this packet.
1401 		 * If none found, drop packet.
1402 		 */
1403 		for (vifi = 0; vifi < V_numvifs &&
1404 		    V_viftable[vifi].v_ifp != ifp; vifi++)
1405 			;
1406 		if (vifi >= V_numvifs) /* vif not found, drop packet */
1407 			goto non_fatal;
1408 
1409 		/* no upcall, so make a new entry */
1410 		rt = mfc_alloc();
1411 		if (rt == NULL)
1412 			goto fail;
1413 
1414 		/* Make a copy of the header to send to the user level process */
1415 		mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1416 		if (mm == NULL)
1417 			goto fail1;
1418 
1419 		/*
1420 		 * Send message to routing daemon to install
1421 		 * a route into the kernel table
1422 		 */
1423 
1424 		im = mtod(mm, struct igmpmsg*);
1425 		im->im_msgtype = IGMPMSG_NOCACHE;
1426 		im->im_mbz = 0;
1427 		im->im_vif = vifi;
1428 
1429 		MRTSTAT_INC(mrts_upcalls);
1430 
1431 		k_igmpsrc.sin_addr = ip->ip_src;
1432 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1433 			CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1434 			MRTSTAT_INC(mrts_upq_sockfull);
1435 			fail1: free(rt, M_MRTABLE);
1436 			fail: free(rte, M_MRTABLE);
1437 			m_freem(mb0);
1438 			MRW_WUNLOCK();
1439 			return ENOBUFS;
1440 		}
1441 
1442 		/* insert new entry at head of hash chain */
1443 		rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1444 		rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1445 		rt->mfc_expire = UPCALL_EXPIRE;
1446 		V_nexpire[hash]++;
1447 		for (i = 0; i < V_numvifs; i++) {
1448 			rt->mfc_ttls[i] = 0;
1449 			rt->mfc_flags[i] = 0;
1450 		}
1451 		rt->mfc_parent = -1;
1452 
1453 		/* clear the RP address */
1454 		rt->mfc_rp.s_addr = INADDR_ANY;
1455 		rt->mfc_bw_meter_leq = NULL;
1456 		rt->mfc_bw_meter_geq = NULL;
1457 
1458 		/* initialize pkt counters per src-grp */
1459 		rt->mfc_pkt_cnt = 0;
1460 		rt->mfc_byte_cnt = 0;
1461 		rt->mfc_wrong_if = 0;
1462 		timevalclear(&rt->mfc_last_assert);
1463 
1464 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1465 
1466 		/* Add RT to hashtable as it didn't exist before */
1467 		LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1468 	} else {
1469 		/* determine if queue has overflowed */
1470 		if (buf_ring_full(rt->mfc_stall_ring)) {
1471 			MRTSTAT_INC(mrts_upq_ovflw);
1472 			non_fatal: free(rte, M_MRTABLE);
1473 			m_freem(mb0);
1474 			MRW_WUNLOCK();
1475 			return (0);
1476 		}
1477 
1478 		buf_ring_enqueue(rt->mfc_stall_ring, rte);
1479 	}
1480 
1481 	rte->m = mb0;
1482 	rte->ifp = ifp;
1483 
1484 	MRW_WUNLOCK();
1485 
1486 	return 0;
1487 }
1488 
1489 /*
1490  * Clean up the cache entry if upcall is not serviced
1491  */
1492 static void
1493 expire_upcalls(void *arg)
1494 {
1495     u_long i;
1496 
1497     CURVNET_SET((struct vnet *) arg);
1498 
1499     /*This callout is always run with MRW_WLOCK taken. */
1500 
1501     for (i = 0; i < mfchashsize; i++) {
1502 	struct mfc *rt, *nrt;
1503 
1504 	if (V_nexpire[i] == 0)
1505 	    continue;
1506 
1507 	LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1508 		if (buf_ring_empty(rt->mfc_stall_ring))
1509 			continue;
1510 
1511 		if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1512 			continue;
1513 
1514 		MRTSTAT_INC(mrts_cache_cleanups);
1515 		CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1516 		    (u_long)ntohl(rt->mfc_origin.s_addr),
1517 		    (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1518 
1519 		expire_mfc(rt);
1520 	    }
1521     }
1522 
1523     callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1524 	curvnet);
1525 
1526     CURVNET_RESTORE();
1527 }
1528 
1529 /*
1530  * Packet forwarding routine once entry in the cache is made
1531  */
1532 static int
1533 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1534 {
1535     struct ip  *ip = mtod(m, struct ip *);
1536     vifi_t vifi;
1537     int plen = ntohs(ip->ip_len);
1538 
1539     MRW_LOCK_ASSERT();
1540 
1541     /*
1542      * If xmt_vif is not -1, send on only the requested vif.
1543      *
1544      * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1545      */
1546     if (xmt_vif < V_numvifs) {
1547 	if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1548 		pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1549 	else
1550 		phyint_send(ip, V_viftable + xmt_vif, m);
1551 	return 1;
1552     }
1553 
1554     /*
1555      * Don't forward if it didn't arrive from the parent vif for its origin.
1556      */
1557     vifi = rt->mfc_parent;
1558     if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1559 	CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1560 	    __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1561 	MRTSTAT_INC(mrts_wrong_if);
1562 	++rt->mfc_wrong_if;
1563 	/*
1564 	 * If we are doing PIM assert processing, send a message
1565 	 * to the routing daemon.
1566 	 *
1567 	 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1568 	 * can complete the SPT switch, regardless of the type
1569 	 * of the iif (broadcast media, GRE tunnel, etc).
1570 	 */
1571 	if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1572 	    V_viftable[vifi].v_ifp) {
1573 	    if (ifp == V_multicast_register_if)
1574 		PIMSTAT_INC(pims_rcv_registers_wrongiif);
1575 
1576 	    /* Get vifi for the incoming packet */
1577 	    for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1578 		vifi++)
1579 		;
1580 	    if (vifi >= V_numvifs)
1581 		return 0;	/* The iif is not found: ignore the packet. */
1582 
1583 	    if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1584 		return 0;	/* WRONGVIF disabled: ignore the packet */
1585 
1586 	    if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1587 		struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1588 		struct igmpmsg *im;
1589 		int hlen = ip->ip_hl << 2;
1590 		struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1591 
1592 		if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1593 		    mm = m_pullup(mm, hlen);
1594 		if (mm == NULL)
1595 		    return ENOBUFS;
1596 
1597 		im = mtod(mm, struct igmpmsg *);
1598 		im->im_msgtype	= IGMPMSG_WRONGVIF;
1599 		im->im_mbz		= 0;
1600 		im->im_vif		= vifi;
1601 
1602 		MRTSTAT_INC(mrts_upcalls);
1603 
1604 		k_igmpsrc.sin_addr = im->im_src;
1605 		if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1606 		    CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1607 		    MRTSTAT_INC(mrts_upq_sockfull);
1608 		    return ENOBUFS;
1609 		}
1610 	    }
1611 	}
1612 	return 0;
1613     }
1614 
1615     /* If I sourced this packet, it counts as output, else it was input. */
1616     mtx_lock_spin(&V_viftable[vifi].v_spin);
1617     if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1618 	V_viftable[vifi].v_pkt_out++;
1619 	V_viftable[vifi].v_bytes_out += plen;
1620     } else {
1621 	V_viftable[vifi].v_pkt_in++;
1622 	V_viftable[vifi].v_bytes_in += plen;
1623     }
1624     mtx_unlock_spin(&V_viftable[vifi].v_spin);
1625 
1626     rt->mfc_pkt_cnt++;
1627     rt->mfc_byte_cnt += plen;
1628 
1629     /*
1630      * For each vif, decide if a copy of the packet should be forwarded.
1631      * Forward if:
1632      *		- the ttl exceeds the vif's threshold
1633      *		- there are group members downstream on interface
1634      */
1635     for (vifi = 0; vifi < V_numvifs; vifi++)
1636 	if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1637 	    V_viftable[vifi].v_pkt_out++;
1638 	    V_viftable[vifi].v_bytes_out += plen;
1639 	    if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1640 		pim_register_send(ip, V_viftable + vifi, m, rt);
1641 	    else
1642 		phyint_send(ip, V_viftable + vifi, m);
1643 	}
1644 
1645     /*
1646      * Perform upcall-related bw measuring.
1647      */
1648     if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1649 	struct bw_meter *x;
1650 	struct timeval now;
1651 
1652 	microtime(&now);
1653 	/* Process meters for Greater-or-EQual case */
1654 	for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1655 		bw_meter_geq_receive_packet(x, plen, &now);
1656 
1657 	/* Process meters for Lower-or-EQual case */
1658 	for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1659 		/*
1660 		 * Record that a packet is received.
1661 		 * Spin lock has to be taken as callout context
1662 		 * (expire_bw_meter_leq) might modify these fields
1663 		 * as well
1664 		 */
1665 		mtx_lock_spin(&x->bm_spin);
1666 		x->bm_measured.b_packets++;
1667 		x->bm_measured.b_bytes += plen;
1668 		mtx_unlock_spin(&x->bm_spin);
1669 	}
1670     }
1671 
1672     return 0;
1673 }
1674 
1675 /*
1676  * Check if a vif number is legal/ok. This is used by in_mcast.c.
1677  */
1678 static int
1679 X_legal_vif_num(int vif)
1680 {
1681 	int ret;
1682 
1683 	ret = 0;
1684 	if (vif < 0)
1685 		return (ret);
1686 
1687 	MRW_RLOCK();
1688 	if (vif < V_numvifs)
1689 		ret = 1;
1690 	MRW_RUNLOCK();
1691 
1692 	return (ret);
1693 }
1694 
1695 /*
1696  * Return the local address used by this vif
1697  */
1698 static u_long
1699 X_ip_mcast_src(int vifi)
1700 {
1701 	in_addr_t addr;
1702 
1703 	addr = INADDR_ANY;
1704 	if (vifi < 0)
1705 		return (addr);
1706 
1707 	MRW_RLOCK();
1708 	if (vifi < V_numvifs)
1709 		addr = V_viftable[vifi].v_lcl_addr.s_addr;
1710 	MRW_RUNLOCK();
1711 
1712 	return (addr);
1713 }
1714 
1715 static void
1716 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1717 {
1718     struct mbuf *mb_copy;
1719     int hlen = ip->ip_hl << 2;
1720 
1721     MRW_LOCK_ASSERT();
1722 
1723     /*
1724      * Make a new reference to the packet; make sure that
1725      * the IP header is actually copied, not just referenced,
1726      * so that ip_output() only scribbles on the copy.
1727      */
1728     mb_copy = m_copypacket(m, M_NOWAIT);
1729     if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1730 	mb_copy = m_pullup(mb_copy, hlen);
1731     if (mb_copy == NULL)
1732 	return;
1733 
1734     send_packet(vifp, mb_copy);
1735 }
1736 
1737 static void
1738 send_packet(struct vif *vifp, struct mbuf *m)
1739 {
1740 	struct ip_moptions imo;
1741 	int error __unused;
1742 
1743 	MRW_LOCK_ASSERT();
1744 
1745 	imo.imo_multicast_ifp  = vifp->v_ifp;
1746 	imo.imo_multicast_ttl  = mtod(m, struct ip *)->ip_ttl - 1;
1747 	imo.imo_multicast_loop = !!in_mcast_loop;
1748 	imo.imo_multicast_vif  = -1;
1749 	STAILQ_INIT(&imo.imo_head);
1750 
1751 	/*
1752 	 * Re-entrancy should not be a problem here, because
1753 	 * the packets that we send out and are looped back at us
1754 	 * should get rejected because they appear to come from
1755 	 * the loopback interface, thus preventing looping.
1756 	 */
1757 	error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1758 	CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1759 	    (ptrdiff_t)(vifp - V_viftable), error);
1760 }
1761 
1762 /*
1763  * Stubs for old RSVP socket shim implementation.
1764  */
1765 
1766 static int
1767 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1768 {
1769 
1770 	return (EOPNOTSUPP);
1771 }
1772 
1773 static void
1774 X_ip_rsvp_force_done(struct socket *so __unused)
1775 {
1776 
1777 }
1778 
1779 static int
1780 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1781 {
1782 	struct mbuf *m;
1783 
1784 	m = *mp;
1785 	*mp = NULL;
1786 	if (!V_rsvp_on)
1787 		m_freem(m);
1788 	return (IPPROTO_DONE);
1789 }
1790 
1791 /*
1792  * Code for bandwidth monitors
1793  */
1794 
1795 /*
1796  * Define common interface for timeval-related methods
1797  */
1798 #define	BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1799 #define	BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1800 #define	BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1801 
1802 static uint32_t
1803 compute_bw_meter_flags(struct bw_upcall *req)
1804 {
1805     uint32_t flags = 0;
1806 
1807     if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1808 	flags |= BW_METER_UNIT_PACKETS;
1809     if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1810 	flags |= BW_METER_UNIT_BYTES;
1811     if (req->bu_flags & BW_UPCALL_GEQ)
1812 	flags |= BW_METER_GEQ;
1813     if (req->bu_flags & BW_UPCALL_LEQ)
1814 	flags |= BW_METER_LEQ;
1815 
1816     return flags;
1817 }
1818 
1819 static void
1820 expire_bw_meter_leq(void *arg)
1821 {
1822 	struct bw_meter *x = arg;
1823 	struct timeval now;
1824 	/*
1825 	 * INFO:
1826 	 * callout is always executed with MRW_WLOCK taken
1827 	 */
1828 
1829 	CURVNET_SET((struct vnet *)x->arg);
1830 
1831 	microtime(&now);
1832 
1833 	/*
1834 	 * Test if we should deliver an upcall
1835 	 */
1836 	if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1837 	    (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1838 	    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1839 	    (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1840 		/* Prepare an upcall for delivery */
1841 		bw_meter_prepare_upcall(x, &now);
1842 	}
1843 
1844 	/* Send all upcalls that are pending delivery */
1845 	taskqueue_enqueue(V_task_queue, &V_task);
1846 
1847 	/* Reset counters */
1848 	x->bm_start_time = now;
1849 	/* Spin lock has to be taken as ip_forward context
1850 	 * might modify these fields as well
1851 	 */
1852 	mtx_lock_spin(&x->bm_spin);
1853 	x->bm_measured.b_bytes = 0;
1854 	x->bm_measured.b_packets = 0;
1855 	mtx_unlock_spin(&x->bm_spin);
1856 
1857 	callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1858 
1859 	CURVNET_RESTORE();
1860 }
1861 
1862 /*
1863  * Add a bw_meter entry
1864  */
1865 static int
1866 add_bw_upcall(struct bw_upcall *req)
1867 {
1868 	struct mfc *mfc;
1869 	struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1870 	BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1871 	struct timeval now;
1872 	struct bw_meter *x, **bwm_ptr;
1873 	uint32_t flags;
1874 
1875 	if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1876 		return EOPNOTSUPP;
1877 
1878 	/* Test if the flags are valid */
1879 	if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1880 		return EINVAL;
1881 	if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1882 		return EINVAL;
1883 	if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1884 			== (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1885 		return EINVAL;
1886 
1887 	/* Test if the threshold time interval is valid */
1888 	if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1889 		return EINVAL;
1890 
1891 	flags = compute_bw_meter_flags(req);
1892 
1893 	/*
1894 	 * Find if we have already same bw_meter entry
1895 	 */
1896 	MRW_WLOCK();
1897 	mfc = mfc_find(&req->bu_src, &req->bu_dst);
1898 	if (mfc == NULL) {
1899 		MRW_WUNLOCK();
1900 		return EADDRNOTAVAIL;
1901 	}
1902 
1903 	/* Choose an appropriate bw_meter list */
1904 	if (req->bu_flags & BW_UPCALL_GEQ)
1905 		bwm_ptr = &mfc->mfc_bw_meter_geq;
1906 	else
1907 		bwm_ptr = &mfc->mfc_bw_meter_leq;
1908 
1909 	for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1910 		if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1911 		    &req->bu_threshold.b_time, ==))
1912 		    && (x->bm_threshold.b_packets
1913 		    == req->bu_threshold.b_packets)
1914 		    && (x->bm_threshold.b_bytes
1915 		    == req->bu_threshold.b_bytes)
1916 		    && (x->bm_flags & BW_METER_USER_FLAGS)
1917 		    == flags) {
1918 			MRW_WUNLOCK();
1919 			return 0; /* XXX Already installed */
1920 		}
1921 	}
1922 
1923 	/* Allocate the new bw_meter entry */
1924 	x = (struct bw_meter*) malloc(sizeof(*x), M_BWMETER,
1925 	    M_ZERO | M_NOWAIT);
1926 	if (x == NULL) {
1927 		MRW_WUNLOCK();
1928 		return ENOBUFS;
1929 	}
1930 
1931 	/* Set the new bw_meter entry */
1932 	x->bm_threshold.b_time = req->bu_threshold.b_time;
1933 	microtime(&now);
1934 	x->bm_start_time = now;
1935 	x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1936 	x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1937 	x->bm_measured.b_packets = 0;
1938 	x->bm_measured.b_bytes = 0;
1939 	x->bm_flags = flags;
1940 	x->bm_time_next = NULL;
1941 	x->bm_mfc = mfc;
1942 	x->arg = curvnet;
1943 	sprintf(x->bm_spin_name, "BM spin %p", x);
1944 	mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1945 
1946 	/* For LEQ case create periodic callout */
1947 	if (req->bu_flags & BW_UPCALL_LEQ) {
1948 		callout_init_rw(&x->bm_meter_callout, &mrouter_mtx, CALLOUT_SHAREDLOCK);
1949 		callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1950 		    expire_bw_meter_leq, x);
1951 	}
1952 
1953 	/* Add the new bw_meter entry to the front of entries for this MFC */
1954 	x->bm_mfc_next = *bwm_ptr;
1955 	*bwm_ptr = x;
1956 
1957 	MRW_WUNLOCK();
1958 
1959 	return 0;
1960 }
1961 
1962 static void
1963 free_bw_list(struct bw_meter *list)
1964 {
1965     while (list != NULL) {
1966 	struct bw_meter *x = list;
1967 
1968 	/* MRW_WLOCK must be held here */
1969 	if (x->bm_flags & BW_METER_LEQ) {
1970 		callout_drain(&x->bm_meter_callout);
1971 		mtx_destroy(&x->bm_spin);
1972 	}
1973 
1974 	list = list->bm_mfc_next;
1975 	free(x, M_BWMETER);
1976     }
1977 }
1978 
1979 /*
1980  * Delete one or multiple bw_meter entries
1981  */
1982 static int
1983 del_bw_upcall(struct bw_upcall *req)
1984 {
1985     struct mfc *mfc;
1986     struct bw_meter *x, **bwm_ptr;
1987 
1988     if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1989 	return EOPNOTSUPP;
1990 
1991     MRW_WLOCK();
1992 
1993     /* Find the corresponding MFC entry */
1994     mfc = mfc_find(&req->bu_src, &req->bu_dst);
1995     if (mfc == NULL) {
1996 	MRW_WUNLOCK();
1997 	return EADDRNOTAVAIL;
1998     } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1999 	/*
2000 	 * Delete all bw_meter entries for this mfc
2001 	 */
2002 	struct bw_meter *list;
2003 
2004 	/* Free LEQ list */
2005 	list = mfc->mfc_bw_meter_leq;
2006 	mfc->mfc_bw_meter_leq = NULL;
2007 	free_bw_list(list);
2008 
2009 	/* Free GEQ list */
2010 	list = mfc->mfc_bw_meter_geq;
2011 	mfc->mfc_bw_meter_geq = NULL;
2012 	free_bw_list(list);
2013 	MRW_WUNLOCK();
2014 	return 0;
2015     } else {			/* Delete a single bw_meter entry */
2016 	struct bw_meter *prev;
2017 	uint32_t flags = 0;
2018 
2019 	flags = compute_bw_meter_flags(req);
2020 
2021 	/* Choose an appropriate bw_meter list */
2022 	if (req->bu_flags & BW_UPCALL_GEQ)
2023 		bwm_ptr = &mfc->mfc_bw_meter_geq;
2024 	else
2025 		bwm_ptr = &mfc->mfc_bw_meter_leq;
2026 
2027 	/* Find the bw_meter entry to delete */
2028 	for (prev = NULL, x = *bwm_ptr; x != NULL;
2029 	     prev = x, x = x->bm_mfc_next) {
2030 	    if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2031 			       &req->bu_threshold.b_time, ==)) &&
2032 		(x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2033 		(x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2034 		(x->bm_flags & BW_METER_USER_FLAGS) == flags)
2035 		break;
2036 	}
2037 	if (x != NULL) { /* Delete entry from the list for this MFC */
2038 	    if (prev != NULL)
2039 		prev->bm_mfc_next = x->bm_mfc_next;	/* remove from middle*/
2040 	    else
2041 		*bwm_ptr = x->bm_mfc_next;/* new head of list */
2042 
2043 	    if (req->bu_flags & BW_UPCALL_LEQ)
2044 		    callout_stop(&x->bm_meter_callout);
2045 
2046 	    MRW_WUNLOCK();
2047 	    /* Free the bw_meter entry */
2048 	    free(x, M_BWMETER);
2049 	    return 0;
2050 	} else {
2051 	    MRW_WUNLOCK();
2052 	    return EINVAL;
2053 	}
2054     }
2055     /* NOTREACHED */
2056 }
2057 
2058 /*
2059  * Perform bandwidth measurement processing that may result in an upcall
2060  */
2061 static void
2062 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2063 {
2064 	struct timeval delta;
2065 
2066 	MRW_LOCK_ASSERT();
2067 
2068 	delta = *nowp;
2069 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2070 
2071 	/*
2072 	 * Processing for ">=" type of bw_meter entry.
2073 	 * bm_spin does not have to be hold here as in GEQ
2074 	 * case this is the only context accessing bm_measured.
2075 	 */
2076 	if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2077 	    /* Reset the bw_meter entry */
2078 	    x->bm_start_time = *nowp;
2079 	    x->bm_measured.b_packets = 0;
2080 	    x->bm_measured.b_bytes = 0;
2081 	    x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2082 	}
2083 
2084 	/* Record that a packet is received */
2085 	x->bm_measured.b_packets++;
2086 	x->bm_measured.b_bytes += plen;
2087 
2088 	/*
2089 	 * Test if we should deliver an upcall
2090 	 */
2091 	if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2092 		if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2093 		    (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2094 		    ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2095 		    (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2096 			/* Prepare an upcall for delivery */
2097 			bw_meter_prepare_upcall(x, nowp);
2098 			x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2099 		}
2100 	}
2101 }
2102 
2103 /*
2104  * Prepare a bandwidth-related upcall
2105  */
2106 static void
2107 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2108 {
2109 	struct timeval delta;
2110 	struct bw_upcall *u;
2111 
2112 	MRW_LOCK_ASSERT();
2113 
2114 	/*
2115 	 * Compute the measured time interval
2116 	 */
2117 	delta = *nowp;
2118 	BW_TIMEVALDECR(&delta, &x->bm_start_time);
2119 
2120 	/*
2121 	 * Set the bw_upcall entry
2122 	 */
2123 	u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2124 	if (!u) {
2125 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2126 		return;
2127 	}
2128 	u->bu_src = x->bm_mfc->mfc_origin;
2129 	u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2130 	u->bu_threshold.b_time = x->bm_threshold.b_time;
2131 	u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2132 	u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2133 	u->bu_measured.b_time = delta;
2134 	u->bu_measured.b_packets = x->bm_measured.b_packets;
2135 	u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2136 	u->bu_flags = 0;
2137 	if (x->bm_flags & BW_METER_UNIT_PACKETS)
2138 		u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2139 	if (x->bm_flags & BW_METER_UNIT_BYTES)
2140 		u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2141 	if (x->bm_flags & BW_METER_GEQ)
2142 		u->bu_flags |= BW_UPCALL_GEQ;
2143 	if (x->bm_flags & BW_METER_LEQ)
2144 		u->bu_flags |= BW_UPCALL_LEQ;
2145 
2146 	if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2147 		log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2148 	if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2149 		taskqueue_enqueue(V_task_queue, &V_task);
2150 	}
2151 }
2152 /*
2153  * Send the pending bandwidth-related upcalls
2154  */
2155 static void
2156 bw_upcalls_send(void)
2157 {
2158     struct mbuf *m;
2159     int len = 0;
2160     struct bw_upcall *bu;
2161     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2162     static struct igmpmsg igmpmsg = { 0,		/* unused1 */
2163 				      0,		/* unused2 */
2164 				      IGMPMSG_BW_UPCALL,/* im_msgtype */
2165 				      0,		/* im_mbz  */
2166 				      0,		/* im_vif  */
2167 				      0,		/* unused3 */
2168 				      { 0 },		/* im_src  */
2169 				      { 0 } };		/* im_dst  */
2170 
2171     MRW_LOCK_ASSERT();
2172 
2173     if (buf_ring_empty(V_bw_upcalls_ring))
2174 	return;
2175 
2176     /*
2177      * Allocate a new mbuf, initialize it with the header and
2178      * the payload for the pending calls.
2179      */
2180     m = m_gethdr(M_NOWAIT, MT_DATA);
2181     if (m == NULL) {
2182 	log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2183 	return;
2184     }
2185 
2186     m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2187     len += sizeof(struct igmpmsg);
2188     while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2189 	m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2190 	len += sizeof(struct bw_upcall);
2191 	free(bu, M_MRTABLE);
2192     }
2193 
2194     /*
2195      * Send the upcalls
2196      * XXX do we need to set the address in k_igmpsrc ?
2197      */
2198     MRTSTAT_INC(mrts_upcalls);
2199     if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2200 	log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2201 	MRTSTAT_INC(mrts_upq_sockfull);
2202     }
2203 }
2204 
2205 /*
2206  * A periodic function for sending all upcalls that are pending delivery
2207  */
2208 static void
2209 expire_bw_upcalls_send(void *arg)
2210 {
2211     CURVNET_SET((struct vnet *) arg);
2212 
2213     /* This callout is run with MRW_RLOCK taken */
2214 
2215     bw_upcalls_send();
2216 
2217     callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2218 	curvnet);
2219     CURVNET_RESTORE();
2220 }
2221 
2222 /*
2223  * End of bandwidth monitoring code
2224  */
2225 
2226 /*
2227  * Send the packet up to the user daemon, or eventually do kernel encapsulation
2228  *
2229  */
2230 static int
2231 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2232     struct mfc *rt)
2233 {
2234     struct mbuf *mb_copy, *mm;
2235 
2236     /*
2237      * Do not send IGMP_WHOLEPKT notifications to userland, if the
2238      * rendezvous point was unspecified, and we were told not to.
2239      */
2240     if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2241 	in_nullhost(rt->mfc_rp))
2242 	return 0;
2243 
2244     mb_copy = pim_register_prepare(ip, m);
2245     if (mb_copy == NULL)
2246 	return ENOBUFS;
2247 
2248     /*
2249      * Send all the fragments. Note that the mbuf for each fragment
2250      * is freed by the sending machinery.
2251      */
2252     for (mm = mb_copy; mm; mm = mb_copy) {
2253 	mb_copy = mm->m_nextpkt;
2254 	mm->m_nextpkt = 0;
2255 	mm = m_pullup(mm, sizeof(struct ip));
2256 	if (mm != NULL) {
2257 	    ip = mtod(mm, struct ip *);
2258 	    if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2259 		pim_register_send_rp(ip, vifp, mm, rt);
2260 	    } else {
2261 		pim_register_send_upcall(ip, vifp, mm, rt);
2262 	    }
2263 	}
2264     }
2265 
2266     return 0;
2267 }
2268 
2269 /*
2270  * Return a copy of the data packet that is ready for PIM Register
2271  * encapsulation.
2272  * XXX: Note that in the returned copy the IP header is a valid one.
2273  */
2274 static struct mbuf *
2275 pim_register_prepare(struct ip *ip, struct mbuf *m)
2276 {
2277     struct mbuf *mb_copy = NULL;
2278     int mtu;
2279 
2280     /* Take care of delayed checksums */
2281     if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2282 	in_delayed_cksum(m);
2283 	m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2284     }
2285 
2286     /*
2287      * Copy the old packet & pullup its IP header into the
2288      * new mbuf so we can modify it.
2289      */
2290     mb_copy = m_copypacket(m, M_NOWAIT);
2291     if (mb_copy == NULL)
2292 	return NULL;
2293     mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2294     if (mb_copy == NULL)
2295 	return NULL;
2296 
2297     /* take care of the TTL */
2298     ip = mtod(mb_copy, struct ip *);
2299     --ip->ip_ttl;
2300 
2301     /* Compute the MTU after the PIM Register encapsulation */
2302     mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2303 
2304     if (ntohs(ip->ip_len) <= mtu) {
2305 	/* Turn the IP header into a valid one */
2306 	ip->ip_sum = 0;
2307 	ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2308     } else {
2309 	/* Fragment the packet */
2310 	mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2311 	if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2312 	    m_freem(mb_copy);
2313 	    return NULL;
2314 	}
2315     }
2316     return mb_copy;
2317 }
2318 
2319 /*
2320  * Send an upcall with the data packet to the user-level process.
2321  */
2322 static int
2323 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2324     struct mbuf *mb_copy, struct mfc *rt)
2325 {
2326     struct mbuf *mb_first;
2327     int len = ntohs(ip->ip_len);
2328     struct igmpmsg *im;
2329     struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2330 
2331     MRW_LOCK_ASSERT();
2332 
2333     /*
2334      * Add a new mbuf with an upcall header
2335      */
2336     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2337     if (mb_first == NULL) {
2338 	m_freem(mb_copy);
2339 	return ENOBUFS;
2340     }
2341     mb_first->m_data += max_linkhdr;
2342     mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2343     mb_first->m_len = sizeof(struct igmpmsg);
2344     mb_first->m_next = mb_copy;
2345 
2346     /* Send message to routing daemon */
2347     im = mtod(mb_first, struct igmpmsg *);
2348     im->im_msgtype	= IGMPMSG_WHOLEPKT;
2349     im->im_mbz		= 0;
2350     im->im_vif		= vifp - V_viftable;
2351     im->im_src		= ip->ip_src;
2352     im->im_dst		= ip->ip_dst;
2353 
2354     k_igmpsrc.sin_addr	= ip->ip_src;
2355 
2356     MRTSTAT_INC(mrts_upcalls);
2357 
2358     if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2359 	CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2360 	MRTSTAT_INC(mrts_upq_sockfull);
2361 	return ENOBUFS;
2362     }
2363 
2364     /* Keep statistics */
2365     PIMSTAT_INC(pims_snd_registers_msgs);
2366     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2367 
2368     return 0;
2369 }
2370 
2371 /*
2372  * Encapsulate the data packet in PIM Register message and send it to the RP.
2373  */
2374 static int
2375 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2376     struct mfc *rt)
2377 {
2378     struct mbuf *mb_first;
2379     struct ip *ip_outer;
2380     struct pim_encap_pimhdr *pimhdr;
2381     int len = ntohs(ip->ip_len);
2382     vifi_t vifi = rt->mfc_parent;
2383 
2384     MRW_LOCK_ASSERT();
2385 
2386     if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2387 	m_freem(mb_copy);
2388 	return EADDRNOTAVAIL;		/* The iif vif is invalid */
2389     }
2390 
2391     /*
2392      * Add a new mbuf with the encapsulating header
2393      */
2394     mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2395     if (mb_first == NULL) {
2396 	m_freem(mb_copy);
2397 	return ENOBUFS;
2398     }
2399     mb_first->m_data += max_linkhdr;
2400     mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2401     mb_first->m_next = mb_copy;
2402 
2403     mb_first->m_pkthdr.len = len + mb_first->m_len;
2404 
2405     /*
2406      * Fill in the encapsulating IP and PIM header
2407      */
2408     ip_outer = mtod(mb_first, struct ip *);
2409     *ip_outer = pim_encap_iphdr;
2410     ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2411 	sizeof(pim_encap_pimhdr));
2412     ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2413     ip_outer->ip_dst = rt->mfc_rp;
2414     /*
2415      * Copy the inner header TOS to the outer header, and take care of the
2416      * IP_DF bit.
2417      */
2418     ip_outer->ip_tos = ip->ip_tos;
2419     if (ip->ip_off & htons(IP_DF))
2420 	ip_outer->ip_off |= htons(IP_DF);
2421     ip_fillid(ip_outer);
2422     pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2423 					 + sizeof(pim_encap_iphdr));
2424     *pimhdr = pim_encap_pimhdr;
2425     /* If the iif crosses a border, set the Border-bit */
2426     if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2427 	pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2428 
2429     mb_first->m_data += sizeof(pim_encap_iphdr);
2430     pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2431     mb_first->m_data -= sizeof(pim_encap_iphdr);
2432 
2433     send_packet(vifp, mb_first);
2434 
2435     /* Keep statistics */
2436     PIMSTAT_INC(pims_snd_registers_msgs);
2437     PIMSTAT_ADD(pims_snd_registers_bytes, len);
2438 
2439     return 0;
2440 }
2441 
2442 /*
2443  * pim_encapcheck() is called by the encap4_input() path at runtime to
2444  * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2445  * into the kernel.
2446  */
2447 static int
2448 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2449     int proto __unused, void *arg __unused)
2450 {
2451 
2452     KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2453     return (8);		/* claim the datagram. */
2454 }
2455 
2456 /*
2457  * PIM-SMv2 and PIM-DM messages processing.
2458  * Receives and verifies the PIM control messages, and passes them
2459  * up to the listening socket, using rip_input().
2460  * The only message with special processing is the PIM_REGISTER message
2461  * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2462  * is passed to if_simloop().
2463  */
2464 static int
2465 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2466 {
2467     struct ip *ip = mtod(m, struct ip *);
2468     struct pim *pim;
2469     int iphlen = off;
2470     int minlen;
2471     int datalen = ntohs(ip->ip_len) - iphlen;
2472     int ip_tos;
2473 
2474     /* Keep statistics */
2475     PIMSTAT_INC(pims_rcv_total_msgs);
2476     PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2477 
2478     /*
2479      * Validate lengths
2480      */
2481     if (datalen < PIM_MINLEN) {
2482 	PIMSTAT_INC(pims_rcv_tooshort);
2483 	CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2484 	    __func__, datalen, ntohl(ip->ip_src.s_addr));
2485 	m_freem(m);
2486 	return (IPPROTO_DONE);
2487     }
2488 
2489     /*
2490      * If the packet is at least as big as a REGISTER, go agead
2491      * and grab the PIM REGISTER header size, to avoid another
2492      * possible m_pullup() later.
2493      *
2494      * PIM_MINLEN       == pimhdr + u_int32_t == 4 + 4 = 8
2495      * PIM_REG_MINLEN   == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2496      */
2497     minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2498     /*
2499      * Get the IP and PIM headers in contiguous memory, and
2500      * possibly the PIM REGISTER header.
2501      */
2502     if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2503 	CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2504 	return (IPPROTO_DONE);
2505     }
2506 
2507     /* m_pullup() may have given us a new mbuf so reset ip. */
2508     ip = mtod(m, struct ip *);
2509     ip_tos = ip->ip_tos;
2510 
2511     /* adjust mbuf to point to the PIM header */
2512     m->m_data += iphlen;
2513     m->m_len  -= iphlen;
2514     pim = mtod(m, struct pim *);
2515 
2516     /*
2517      * Validate checksum. If PIM REGISTER, exclude the data packet.
2518      *
2519      * XXX: some older PIMv2 implementations don't make this distinction,
2520      * so for compatibility reason perform the checksum over part of the
2521      * message, and if error, then over the whole message.
2522      */
2523     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2524 	/* do nothing, checksum okay */
2525     } else if (in_cksum(m, datalen)) {
2526 	PIMSTAT_INC(pims_rcv_badsum);
2527 	CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2528 	m_freem(m);
2529 	return (IPPROTO_DONE);
2530     }
2531 
2532     /* PIM version check */
2533     if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2534 	PIMSTAT_INC(pims_rcv_badversion);
2535 	CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2536 	    (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2537 	m_freem(m);
2538 	return (IPPROTO_DONE);
2539     }
2540 
2541     /* restore mbuf back to the outer IP */
2542     m->m_data -= iphlen;
2543     m->m_len  += iphlen;
2544 
2545     if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2546 	/*
2547 	 * Since this is a REGISTER, we'll make a copy of the register
2548 	 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2549 	 * routing daemon.
2550 	 */
2551 	struct sockaddr_in dst = { sizeof(dst), AF_INET };
2552 	struct mbuf *mcp;
2553 	struct ip *encap_ip;
2554 	u_int32_t *reghdr;
2555 	struct ifnet *vifp;
2556 
2557 	MRW_RLOCK();
2558 	if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2559 	    MRW_RUNLOCK();
2560 	    CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2561 		(int)V_reg_vif_num);
2562 	    m_freem(m);
2563 	    return (IPPROTO_DONE);
2564 	}
2565 	/* XXX need refcnt? */
2566 	vifp = V_viftable[V_reg_vif_num].v_ifp;
2567 	MRW_RUNLOCK();
2568 
2569 	/*
2570 	 * Validate length
2571 	 */
2572 	if (datalen < PIM_REG_MINLEN) {
2573 	    PIMSTAT_INC(pims_rcv_tooshort);
2574 	    PIMSTAT_INC(pims_rcv_badregisters);
2575 	    CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2576 	    m_freem(m);
2577 	    return (IPPROTO_DONE);
2578 	}
2579 
2580 	reghdr = (u_int32_t *)(pim + 1);
2581 	encap_ip = (struct ip *)(reghdr + 1);
2582 
2583 	CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2584 	    __func__, ntohl(encap_ip->ip_src.s_addr),
2585 	    ntohs(encap_ip->ip_len));
2586 
2587 	/* verify the version number of the inner packet */
2588 	if (encap_ip->ip_v != IPVERSION) {
2589 	    PIMSTAT_INC(pims_rcv_badregisters);
2590 	    CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2591 	    m_freem(m);
2592 	    return (IPPROTO_DONE);
2593 	}
2594 
2595 	/* verify the inner packet is destined to a mcast group */
2596 	if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2597 	    PIMSTAT_INC(pims_rcv_badregisters);
2598 	    CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2599 		ntohl(encap_ip->ip_dst.s_addr));
2600 	    m_freem(m);
2601 	    return (IPPROTO_DONE);
2602 	}
2603 
2604 	/* If a NULL_REGISTER, pass it to the daemon */
2605 	if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2606 	    goto pim_input_to_daemon;
2607 
2608 	/*
2609 	 * Copy the TOS from the outer IP header to the inner IP header.
2610 	 */
2611 	if (encap_ip->ip_tos != ip_tos) {
2612 	    /* Outer TOS -> inner TOS */
2613 	    encap_ip->ip_tos = ip_tos;
2614 	    /* Recompute the inner header checksum. Sigh... */
2615 
2616 	    /* adjust mbuf to point to the inner IP header */
2617 	    m->m_data += (iphlen + PIM_MINLEN);
2618 	    m->m_len  -= (iphlen + PIM_MINLEN);
2619 
2620 	    encap_ip->ip_sum = 0;
2621 	    encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2622 
2623 	    /* restore mbuf to point back to the outer IP header */
2624 	    m->m_data -= (iphlen + PIM_MINLEN);
2625 	    m->m_len  += (iphlen + PIM_MINLEN);
2626 	}
2627 
2628 	/*
2629 	 * Decapsulate the inner IP packet and loopback to forward it
2630 	 * as a normal multicast packet. Also, make a copy of the
2631 	 *     outer_iphdr + pimhdr + reghdr + encap_iphdr
2632 	 * to pass to the daemon later, so it can take the appropriate
2633 	 * actions (e.g., send back PIM_REGISTER_STOP).
2634 	 * XXX: here m->m_data points to the outer IP header.
2635 	 */
2636 	mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2637 	if (mcp == NULL) {
2638 	    CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2639 	    m_freem(m);
2640 	    return (IPPROTO_DONE);
2641 	}
2642 
2643 	/* Keep statistics */
2644 	/* XXX: registers_bytes include only the encap. mcast pkt */
2645 	PIMSTAT_INC(pims_rcv_registers_msgs);
2646 	PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2647 
2648 	/*
2649 	 * forward the inner ip packet; point m_data at the inner ip.
2650 	 */
2651 	m_adj(m, iphlen + PIM_MINLEN);
2652 
2653 	CTR4(KTR_IPMF,
2654 	    "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2655 	    __func__,
2656 	    (u_long)ntohl(encap_ip->ip_src.s_addr),
2657 	    (u_long)ntohl(encap_ip->ip_dst.s_addr),
2658 	    (int)V_reg_vif_num);
2659 
2660 	/* NB: vifp was collected above; can it change on us? */
2661 	if_simloop(vifp, m, dst.sin_family, 0);
2662 
2663 	/* prepare the register head to send to the mrouting daemon */
2664 	m = mcp;
2665     }
2666 
2667 pim_input_to_daemon:
2668     /*
2669      * Pass the PIM message up to the daemon; if it is a Register message,
2670      * pass the 'head' only up to the daemon. This includes the
2671      * outer IP header, PIM header, PIM-Register header and the
2672      * inner IP header.
2673      * XXX: the outer IP header pkt size of a Register is not adjust to
2674      * reflect the fact that the inner multicast data is truncated.
2675      */
2676     return (rip_input(&m, &off, proto));
2677 }
2678 
2679 static int
2680 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2681 {
2682 	struct mfc	*rt;
2683 	int		 error, i;
2684 
2685 	if (req->newptr)
2686 		return (EPERM);
2687 	if (V_mfchashtbl == NULL)	/* XXX unlocked */
2688 		return (0);
2689 	error = sysctl_wire_old_buffer(req, 0);
2690 	if (error)
2691 		return (error);
2692 
2693 	MRW_RLOCK();
2694 	for (i = 0; i < mfchashsize; i++) {
2695 		LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2696 			error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2697 			if (error)
2698 				goto out_locked;
2699 		}
2700 	}
2701 out_locked:
2702 	MRW_RUNLOCK();
2703 	return (error);
2704 }
2705 
2706 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2707     CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2708     "IPv4 Multicast Forwarding Table "
2709     "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2710 
2711 static int
2712 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2713 {
2714 	int error;
2715 
2716 	if (req->newptr)
2717 		return (EPERM);
2718 	if (V_viftable == NULL)		/* XXX unlocked */
2719 		return (0);
2720 	error = sysctl_wire_old_buffer(req, sizeof(*V_viftable) * MAXVIFS);
2721 	if (error)
2722 		return (error);
2723 
2724 	MRW_RLOCK();
2725 	error = SYSCTL_OUT(req, V_viftable, sizeof(*V_viftable) * MAXVIFS);
2726 	MRW_RUNLOCK();
2727 	return (error);
2728 }
2729 
2730 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2731     CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2732     sysctl_viflist, "S,vif[MAXVIFS]",
2733     "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2734 
2735 static void
2736 vnet_mroute_init(const void *unused __unused)
2737 {
2738 
2739 	V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2740 
2741 	V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2742 	    M_MRTABLE, M_WAITOK|M_ZERO);
2743 
2744 	callout_init_rw(&V_expire_upcalls_ch, &mrouter_mtx, 0);
2745 	callout_init_rw(&V_bw_upcalls_ch, &mrouter_mtx, 0);
2746 
2747 	/* Prepare taskqueue */
2748 	V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2749 		    taskqueue_thread_enqueue, &V_task_queue);
2750 	taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2751 }
2752 
2753 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2754 	NULL);
2755 
2756 static void
2757 vnet_mroute_uninit(const void *unused __unused)
2758 {
2759 
2760 	/* Taskqueue should be cancelled and drained before freeing */
2761 	taskqueue_free(V_task_queue);
2762 
2763 	free(V_viftable, M_MRTABLE);
2764 	free(V_nexpire, M_MRTABLE);
2765 	V_nexpire = NULL;
2766 }
2767 
2768 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2769 	vnet_mroute_uninit, NULL);
2770 
2771 static int
2772 ip_mroute_modevent(module_t mod, int type, void *unused)
2773 {
2774 
2775     switch (type) {
2776     case MOD_LOAD:
2777 	MRW_LOCK_INIT();
2778 
2779 	if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2780 	    if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2781 	if (if_detach_event_tag == NULL) {
2782 		printf("ip_mroute: unable to register "
2783 		    "ifnet_departure_event handler\n");
2784 		MRW_LOCK_DESTROY();
2785 		return (EINVAL);
2786 	}
2787 
2788 	mfchashsize = MFCHASHSIZE;
2789 	if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2790 	    !powerof2(mfchashsize)) {
2791 		printf("WARNING: %s not a power of 2; using default\n",
2792 		    "net.inet.ip.mfchashsize");
2793 		mfchashsize = MFCHASHSIZE;
2794 	}
2795 
2796 	pim_squelch_wholepkt = 0;
2797 	TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2798 	    &pim_squelch_wholepkt);
2799 
2800 	pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2801 	if (pim_encap_cookie == NULL) {
2802 		printf("ip_mroute: unable to attach pim encap\n");
2803 		MRW_LOCK_DESTROY();
2804 		return (EINVAL);
2805 	}
2806 
2807 	ip_mcast_src = X_ip_mcast_src;
2808 	ip_mforward = X_ip_mforward;
2809 	ip_mrouter_done = X_ip_mrouter_done;
2810 	ip_mrouter_get = X_ip_mrouter_get;
2811 	ip_mrouter_set = X_ip_mrouter_set;
2812 
2813 	ip_rsvp_force_done = X_ip_rsvp_force_done;
2814 	ip_rsvp_vif = X_ip_rsvp_vif;
2815 
2816 	legal_vif_num = X_legal_vif_num;
2817 	mrt_ioctl = X_mrt_ioctl;
2818 	rsvp_input_p = X_rsvp_input;
2819 	break;
2820 
2821     case MOD_UNLOAD:
2822 	/*
2823 	 * Typically module unload happens after the user-level
2824 	 * process has shutdown the kernel services (the check
2825 	 * below insures someone can't just yank the module out
2826 	 * from under a running process).  But if the module is
2827 	 * just loaded and then unloaded w/o starting up a user
2828 	 * process we still need to cleanup.
2829 	 */
2830 	MRW_WLOCK();
2831 	if (ip_mrouter_cnt != 0) {
2832 	    MRW_WUNLOCK();
2833 	    return (EINVAL);
2834 	}
2835 	ip_mrouter_unloading = 1;
2836 	MRW_WUNLOCK();
2837 
2838 	EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2839 
2840 	if (pim_encap_cookie) {
2841 	    ip_encap_detach(pim_encap_cookie);
2842 	    pim_encap_cookie = NULL;
2843 	}
2844 
2845 	ip_mcast_src = NULL;
2846 	ip_mforward = NULL;
2847 	ip_mrouter_done = NULL;
2848 	ip_mrouter_get = NULL;
2849 	ip_mrouter_set = NULL;
2850 
2851 	ip_rsvp_force_done = NULL;
2852 	ip_rsvp_vif = NULL;
2853 
2854 	legal_vif_num = NULL;
2855 	mrt_ioctl = NULL;
2856 	rsvp_input_p = NULL;
2857 
2858 	MRW_LOCK_DESTROY();
2859 	break;
2860 
2861     default:
2862 	return EOPNOTSUPP;
2863     }
2864     return 0;
2865 }
2866 
2867 static moduledata_t ip_mroutemod = {
2868     "ip_mroute",
2869     ip_mroute_modevent,
2870     0
2871 };
2872 
2873 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);
2874