2 * SPDX-License-Identifier: BSD-3-Clause
4 * Copyright (c) 1989 Stephen Deering
5 * Copyright (c) 1992, 1993
6 * The Regents of the University of California. All rights reserved.
8 * This code is derived from software contributed to Berkeley by
9 * Stephen Deering of Stanford University.
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
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.
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
35 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
39 * IP multicast forwarding procedures
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
54 * MROUTING Revision: 3.5
55 * and PIM-SMv2 and PIM-DM support, advanced API support,
56 * bandwidth metering and signaling
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.
72 #include <sys/cdefs.h>
73 __FBSDID("$FreeBSD$");
76 #include "opt_mrouting.h"
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>
87 #include <sys/kthread.h>
89 #include <sys/malloc.h>
91 #include <sys/module.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>
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>
108 #include <net/if_var.h>
109 #include <net/if_private.h>
110 #include <net/if_types.h>
111 #include <net/netisr.h>
112 #include <net/route.h>
113 #include <net/vnet.h>
115 #include <netinet/in.h>
116 #include <netinet/igmp.h>
117 #include <netinet/in_systm.h>
118 #include <netinet/in_var.h>
119 #include <netinet/ip.h>
120 #include <netinet/ip_encap.h>
121 #include <netinet/ip_mroute.h>
122 #include <netinet/ip_var.h>
123 #include <netinet/ip_options.h>
124 #include <netinet/pim.h>
125 #include <netinet/pim_var.h>
126 #include <netinet/udp.h>
128 #include <machine/in_cksum.h>
131 #define KTR_IPMF KTR_INET
134 #define VIFI_INVALID ((vifi_t) -1)
136 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
139 * Locking. We use two locks: one for the virtual interface table and
140 * one for the forwarding table. These locks may be nested in which case
141 * the VIF lock must always be taken first. Note that each lock is used
142 * to cover not only the specific data structure but also related data
146 static struct rwlock mrouter_mtx;
147 #define MRW_RLOCK() rw_rlock(&mrouter_mtx)
148 #define MRW_WLOCK() rw_wlock(&mrouter_mtx)
149 #define MRW_RUNLOCK() rw_runlock(&mrouter_mtx)
150 #define MRW_WUNLOCK() rw_wunlock(&mrouter_mtx)
151 #define MRW_UNLOCK() rw_unlock(&mrouter_mtx)
152 #define MRW_LOCK_ASSERT() rw_assert(&mrouter_mtx, RA_LOCKED)
153 #define MRW_WLOCK_ASSERT() rw_assert(&mrouter_mtx, RA_WLOCKED)
154 #define MRW_LOCK_TRY_UPGRADE() rw_try_upgrade(&mrouter_mtx)
155 #define MRW_WOWNED() rw_wowned(&mrouter_mtx)
156 #define MRW_LOCK_INIT() \
157 rw_init(&mrouter_mtx, "IPv4 multicast forwarding")
158 #define MRW_LOCK_DESTROY() rw_destroy(&mrouter_mtx)
160 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
161 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
163 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
164 VNET_PCPUSTAT_SYSINIT(mrtstat);
165 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
166 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
167 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
168 "netinet/ip_mroute.h)");
170 VNET_DEFINE_STATIC(u_long, mfchash);
171 #define V_mfchash VNET(mfchash)
172 #define MFCHASH(a, g) \
173 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
174 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
175 #define MFCHASHSIZE 256
177 static u_long mfchashsize; /* Hash size */
178 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */
179 #define V_nexpire VNET(nexpire)
180 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
181 #define V_mfchashtbl VNET(mfchashtbl)
182 VNET_DEFINE_STATIC(struct taskqueue *, task_queue);
183 #define V_task_queue VNET(task_queue)
184 VNET_DEFINE_STATIC(struct task, task);
185 #define V_task VNET(task)
187 VNET_DEFINE_STATIC(vifi_t, numvifs);
188 #define V_numvifs VNET(numvifs)
189 VNET_DEFINE_STATIC(struct vif *, viftable);
190 #define V_viftable VNET(viftable)
192 static eventhandler_tag if_detach_event_tag = NULL;
194 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
195 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
197 VNET_DEFINE_STATIC(struct mtx, buf_ring_mtx);
198 #define V_buf_ring_mtx VNET(buf_ring_mtx)
200 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
201 #define UPCALL_EXPIRE 6 /* number of timeouts */
204 * Bandwidth meter variables and constants
206 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
209 * Pending upcalls are stored in a ring which is flushed when
210 * full, or periodically
212 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
213 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
214 VNET_DEFINE_STATIC(struct buf_ring *, bw_upcalls_ring);
215 #define V_bw_upcalls_ring VNET(bw_upcalls_ring)
216 VNET_DEFINE_STATIC(struct mtx, bw_upcalls_ring_mtx);
217 #define V_bw_upcalls_ring_mtx VNET(bw_upcalls_ring_mtx)
219 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
221 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
222 VNET_PCPUSTAT_SYSINIT(pimstat);
223 VNET_PCPUSTAT_SYSUNINIT(pimstat);
225 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
227 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
228 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
230 static u_long pim_squelch_wholepkt = 0;
231 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
232 &pim_squelch_wholepkt, 0,
233 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
235 static const struct encaptab *pim_encap_cookie;
236 static int pim_encapcheck(const struct mbuf *, int, int, void *);
237 static int pim_input(struct mbuf *, int, int, void *);
239 extern int in_mcast_loop;
241 static const struct encap_config ipv4_encap_cfg = {
242 .proto = IPPROTO_PIM,
243 .min_length = sizeof(struct ip) + PIM_MINLEN,
245 .check = pim_encapcheck,
250 * Note: the PIM Register encapsulation adds the following in front of a
253 * struct pim_encap_hdr {
255 * struct pim_encap_pimhdr pim;
260 struct pim_encap_pimhdr {
264 #define PIM_ENCAP_TTL 64
266 static struct ip pim_encap_iphdr = {
267 #if BYTE_ORDER == LITTLE_ENDIAN
268 sizeof(struct ip) >> 2,
272 sizeof(struct ip) >> 2,
275 sizeof(struct ip), /* total length */
283 static struct pim_encap_pimhdr pim_encap_pimhdr = {
285 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
292 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
293 #define V_reg_vif_num VNET(reg_vif_num)
294 VNET_DEFINE_STATIC(struct ifnet *, multicast_register_if);
295 #define V_multicast_register_if VNET(multicast_register_if)
301 static u_long X_ip_mcast_src(int);
302 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
303 struct ip_moptions *);
304 static int X_ip_mrouter_done(void);
305 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
306 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
307 static int X_legal_vif_num(int);
308 static int X_mrt_ioctl(u_long, caddr_t, int);
310 static int add_bw_upcall(struct bw_upcall *);
311 static int add_mfc(struct mfcctl2 *);
312 static int add_vif(struct vifctl *);
313 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
314 static void bw_meter_geq_receive_packet(struct bw_meter *, int,
316 static void bw_upcalls_send(void);
317 static int del_bw_upcall(struct bw_upcall *);
318 static int del_mfc(struct mfcctl2 *);
319 static int del_vif(vifi_t);
320 static int del_vif_locked(vifi_t, struct ifnet **);
321 static void expire_bw_upcalls_send(void *);
322 static void expire_mfc(struct mfc *);
323 static void expire_upcalls(void *);
324 static void free_bw_list(struct bw_meter *);
325 static int get_sg_cnt(struct sioc_sg_req *);
326 static int get_vif_cnt(struct sioc_vif_req *);
327 static void if_detached_event(void *, struct ifnet *);
328 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
329 static int ip_mrouter_init(struct socket *, int);
330 static __inline struct mfc *
331 mfc_find(struct in_addr *, struct in_addr *);
332 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
334 pim_register_prepare(struct ip *, struct mbuf *);
335 static int pim_register_send(struct ip *, struct vif *,
336 struct mbuf *, struct mfc *);
337 static int pim_register_send_rp(struct ip *, struct vif *,
338 struct mbuf *, struct mfc *);
339 static int pim_register_send_upcall(struct ip *, struct vif *,
340 struct mbuf *, struct mfc *);
341 static void send_packet(struct vif *, struct mbuf *);
342 static int set_api_config(uint32_t *);
343 static int set_assert(int);
344 static int socket_send(struct socket *, struct mbuf *,
345 struct sockaddr_in *);
348 * Kernel multicast forwarding API capabilities and setup.
349 * If more API capabilities are added to the kernel, they should be
350 * recorded in `mrt_api_support'.
352 #define MRT_API_VERSION 0x0305
354 static const int mrt_api_version = MRT_API_VERSION;
355 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
356 MRT_MFC_FLAGS_BORDER_VIF |
359 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
360 #define V_mrt_api_config VNET(mrt_api_config)
361 VNET_DEFINE_STATIC(int, pim_assert_enabled);
362 #define V_pim_assert_enabled VNET(pim_assert_enabled)
363 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
366 * Find a route for a given origin IP address and multicast group address.
367 * Statistics must be updated by the caller.
369 static __inline struct mfc *
370 mfc_find(struct in_addr *o, struct in_addr *g)
375 * Might be called both RLOCK and WLOCK.
376 * Check if any, it's caller responsibility
377 * to choose correct option.
381 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
382 if (in_hosteq(rt->mfc_origin, *o) &&
383 in_hosteq(rt->mfc_mcastgrp, *g) &&
384 buf_ring_empty(rt->mfc_stall_ring))
391 static __inline struct mfc *
395 rt = (struct mfc*) malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT | M_ZERO);
399 rt->mfc_stall_ring = buf_ring_alloc(MAX_UPQ, M_MRTABLE,
400 M_NOWAIT, &V_buf_ring_mtx);
401 if (rt->mfc_stall_ring == NULL) {
410 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
413 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
419 struct bw_upcall bw_upcall;
422 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
426 switch (sopt->sopt_name) {
428 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
431 error = ip_mrouter_init(so, optval);
435 error = ip_mrouter_done();
439 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
442 error = add_vif(&vifc);
446 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
449 error = del_vif(vifi);
455 * select data size depending on API version.
457 if (sopt->sopt_name == MRT_ADD_MFC &&
458 V_mrt_api_config & MRT_API_FLAGS_ALL) {
459 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
460 sizeof(struct mfcctl2));
462 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
463 sizeof(struct mfcctl));
464 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
465 sizeof(mfc) - sizeof(struct mfcctl));
469 if (sopt->sopt_name == MRT_ADD_MFC)
470 error = add_mfc(&mfc);
472 error = del_mfc(&mfc);
476 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
483 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
485 error = set_api_config(&i);
487 error = sooptcopyout(sopt, &i, sizeof i);
490 case MRT_ADD_BW_UPCALL:
491 case MRT_DEL_BW_UPCALL:
492 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
496 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
497 error = add_bw_upcall(&bw_upcall);
499 error = del_bw_upcall(&bw_upcall);
510 * Handle MRT getsockopt commands
513 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
517 switch (sopt->sopt_name) {
519 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
523 error = sooptcopyout(sopt, &V_pim_assert_enabled,
524 sizeof V_pim_assert_enabled);
527 case MRT_API_SUPPORT:
528 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
532 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
543 * Handle ioctl commands to obtain information from the cache
546 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
551 * Currently the only function calling this ioctl routine is rtioctl_fib().
552 * Typically, only root can create the raw socket in order to execute
553 * this ioctl method, however the request might be coming from a prison
555 error = priv_check(curthread, PRIV_NETINET_MROUTE);
559 case (SIOCGETVIFCNT):
560 error = get_vif_cnt((struct sioc_vif_req *)data);
564 error = get_sg_cnt((struct sioc_sg_req *)data);
575 * returns the packet, byte, rpf-failure count for the source group provided
578 get_sg_cnt(struct sioc_sg_req *req)
583 rt = mfc_find(&req->src, &req->grp);
586 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
587 return EADDRNOTAVAIL;
589 req->pktcnt = rt->mfc_pkt_cnt;
590 req->bytecnt = rt->mfc_byte_cnt;
591 req->wrong_if = rt->mfc_wrong_if;
597 * returns the input and output packet and byte counts on the vif provided
600 get_vif_cnt(struct sioc_vif_req *req)
602 vifi_t vifi = req->vifi;
605 if (vifi >= V_numvifs) {
610 mtx_lock_spin(&V_viftable[vifi].v_spin);
611 req->icount = V_viftable[vifi].v_pkt_in;
612 req->ocount = V_viftable[vifi].v_pkt_out;
613 req->ibytes = V_viftable[vifi].v_bytes_in;
614 req->obytes = V_viftable[vifi].v_bytes_out;
615 mtx_unlock_spin(&V_viftable[vifi].v_spin);
622 if_detached_event(void *arg __unused, struct ifnet *ifp)
625 u_long i, vifi_cnt = 0;
626 struct ifnet *free_ptr;
630 if (V_ip_mrouter == NULL) {
636 * Tear down multicast forwarder state associated with this ifnet.
637 * 1. Walk the vif list, matching vifs against this ifnet.
638 * 2. Walk the multicast forwarding cache (mfc) looking for
639 * inner matches with this vif's index.
640 * 3. Expire any matching multicast forwarding cache entries.
641 * 4. Free vif state. This should disable ALLMULTI on the interface.
643 for (vifi = 0; vifi < V_numvifs; vifi++) {
644 if (V_viftable[vifi].v_ifp != ifp)
646 for (i = 0; i < mfchashsize; i++) {
647 struct mfc *rt, *nrt;
649 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
650 if (rt->mfc_parent == vifi) {
655 del_vif_locked(vifi, &free_ptr);
656 if (free_ptr != NULL)
663 * Free IFP. We don't have to use free_ptr here as it is the same
664 * that ifp. Perform free as many times as required in case
665 * refcount is greater than 1.
667 for (i = 0; i < vifi_cnt; i++)
672 ip_mrouter_upcall_thread(void *arg, int pending __unused)
674 CURVNET_SET((struct vnet *) arg);
684 * Enable multicast forwarding.
687 ip_mrouter_init(struct socket *so, int version)
690 CTR2(KTR_IPMF, "%s: so %p", __func__, so);
697 if (ip_mrouter_unloading) {
702 if (V_ip_mrouter != NULL) {
707 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
710 /* Create upcall ring */
711 mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
712 V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
713 M_NOWAIT, &V_bw_upcalls_ring_mtx);
714 if (!V_bw_upcalls_ring) {
719 TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
720 taskqueue_cancel(V_task_queue, &V_task, NULL);
721 taskqueue_unblock(V_task_queue);
723 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
725 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
729 atomic_add_int(&ip_mrouter_cnt, 1);
731 /* This is a mutex required by buf_ring init, but not used internally */
732 mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
736 CTR1(KTR_IPMF, "%s: done", __func__);
742 * Disable multicast forwarding.
745 X_ip_mrouter_done(void)
751 struct bw_upcall *bu;
753 if (V_ip_mrouter == NULL)
757 * Detach/disable hooks to the reset of the system.
760 atomic_subtract_int(&ip_mrouter_cnt, 1);
761 V_mrt_api_config = 0;
764 * Wait for all epoch sections to complete to ensure
765 * V_ip_mrouter = NULL is visible to others.
767 epoch_wait_preempt(net_epoch_preempt);
769 /* Stop and drain task queue */
770 taskqueue_block(V_task_queue);
771 while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
772 taskqueue_drain(V_task_queue, &V_task);
775 ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
778 taskqueue_cancel(V_task_queue, &V_task, NULL);
780 /* Destroy upcall ring */
781 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
784 buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
785 mtx_destroy(&V_bw_upcalls_ring_mtx);
788 * For each phyint in use, prepare to disable promiscuous reception
789 * of all IP multicasts. Defer the actual call until the lock is released;
790 * just record the list of interfaces while locked. Some interfaces use
791 * sx locks in their ioctl routines, which is not allowed while holding
792 * a non-sleepable lock.
794 KASSERT(V_numvifs <= MAXVIFS, ("More vifs than possible"));
795 for (vifi = 0, nifp = 0; vifi < V_numvifs; vifi++) {
796 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
797 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
798 ifps[nifp++] = V_viftable[vifi].v_ifp;
801 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
803 V_pim_assert_enabled = 0;
805 callout_stop(&V_expire_upcalls_ch);
806 callout_stop(&V_bw_upcalls_ch);
809 * Free all multicast forwarding cache entries.
810 * Do not use hashdestroy(), as we must perform other cleanup.
812 for (i = 0; i < mfchashsize; i++) {
813 struct mfc *rt, *nrt;
815 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
819 free(V_mfchashtbl, M_MRTABLE);
822 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
824 V_reg_vif_num = VIFI_INVALID;
826 mtx_destroy(&V_buf_ring_mtx);
831 * Now drop our claim on promiscuous multicast on the interfaces recorded
832 * above. This is safe to do now because ALLMULTI is reference counted.
834 for (vifi = 0; vifi < nifp; vifi++)
835 if_allmulti(ifps[vifi], 0);
838 CTR1(KTR_IPMF, "%s: done", __func__);
844 * Set PIM assert processing global
849 if ((i != 1) && (i != 0))
852 V_pim_assert_enabled = i;
858 * Configure API capabilities
861 set_api_config(uint32_t *apival)
866 * We can set the API capabilities only if it is the first operation
867 * after MRT_INIT. I.e.:
868 * - there are no vifs installed
869 * - pim_assert is not enabled
870 * - the MFC table is empty
876 if (V_pim_assert_enabled) {
883 for (i = 0; i < mfchashsize; i++) {
884 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
893 V_mrt_api_config = *apival & mrt_api_support;
894 *apival = V_mrt_api_config;
900 * Add a vif to the vif table
903 add_vif(struct vifctl *vifcp)
905 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
906 struct sockaddr_in sin = {sizeof sin, AF_INET};
912 if (vifcp->vifc_vifi >= MAXVIFS)
914 /* rate limiting is no longer supported by this code */
915 if (vifcp->vifc_rate_limit != 0) {
916 log(LOG_ERR, "rate limiting is no longer supported\n");
920 if (in_nullhost(vifcp->vifc_lcl_addr))
921 return EADDRNOTAVAIL;
923 /* Find the interface with an address in AF_INET family */
924 if (vifcp->vifc_flags & VIFF_REGISTER) {
926 * XXX: Because VIFF_REGISTER does not really need a valid
927 * local interface (e.g. it could be 127.0.0.2), we don't
932 struct epoch_tracker et;
934 sin.sin_addr = vifcp->vifc_lcl_addr;
936 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
939 return EADDRNOTAVAIL;
942 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
946 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
947 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
949 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
950 ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
951 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
952 if (V_reg_vif_num == VIFI_INVALID) {
953 if_initname(V_multicast_register_if, "register_vif", 0);
954 V_reg_vif_num = vifcp->vifc_vifi;
956 } else { /* Make sure the interface supports multicast */
957 if ((ifp->if_flags & IFF_MULTICAST) == 0)
960 /* Enable promiscuous reception of all IP multicasts from the if */
961 error = if_allmulti(ifp, 1);
968 if (!in_nullhost(vifp->v_lcl_addr)) {
970 V_multicast_register_if = NULL;
977 vifp->v_flags = vifcp->vifc_flags;
978 vifp->v_threshold = vifcp->vifc_threshold;
979 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
980 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
982 /* initialize per vif pkt counters */
985 vifp->v_bytes_in = 0;
986 vifp->v_bytes_out = 0;
987 sprintf(vifp->v_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
988 mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
990 /* Adjust numvifs up if the vifi is higher than numvifs */
991 if (V_numvifs <= vifcp->vifc_vifi)
992 V_numvifs = vifcp->vifc_vifi + 1;
996 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
997 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
998 (int)vifcp->vifc_threshold);
1004 * Delete a vif from the vif table
1007 del_vif_locked(vifi_t vifi, struct ifnet **ifp_free)
1015 if (vifi >= V_numvifs) {
1018 vifp = &V_viftable[vifi];
1019 if (in_nullhost(vifp->v_lcl_addr)) {
1020 return EADDRNOTAVAIL;
1023 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1024 if_allmulti(vifp->v_ifp, 0);
1026 if (vifp->v_flags & VIFF_REGISTER) {
1027 V_reg_vif_num = VIFI_INVALID;
1029 if (vifp->v_ifp == V_multicast_register_if)
1030 V_multicast_register_if = NULL;
1031 *ifp_free = vifp->v_ifp;
1035 mtx_destroy(&vifp->v_spin);
1037 bzero((caddr_t)vifp, sizeof (*vifp));
1039 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1041 /* Adjust numvifs down */
1042 for (vifi = V_numvifs; vifi > 0; vifi--)
1043 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1051 del_vif(vifi_t vifi)
1054 struct ifnet *free_ptr;
1057 cc = del_vif_locked(vifi, &free_ptr);
1067 * update an mfc entry without resetting counters and S,G addresses.
1070 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1074 rt->mfc_parent = mfccp->mfcc_parent;
1075 for (i = 0; i < V_numvifs; i++) {
1076 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1077 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1080 /* set the RP address */
1081 if (V_mrt_api_config & MRT_MFC_RP)
1082 rt->mfc_rp = mfccp->mfcc_rp;
1084 rt->mfc_rp.s_addr = INADDR_ANY;
1088 * fully initialize an mfc entry from the parameter.
1091 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1093 rt->mfc_origin = mfccp->mfcc_origin;
1094 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1096 update_mfc_params(rt, mfccp);
1098 /* initialize pkt counters per src-grp */
1099 rt->mfc_pkt_cnt = 0;
1100 rt->mfc_byte_cnt = 0;
1101 rt->mfc_wrong_if = 0;
1102 timevalclear(&rt->mfc_last_assert);
1106 expire_mfc(struct mfc *rt)
1112 free_bw_list(rt->mfc_bw_meter_leq);
1113 free_bw_list(rt->mfc_bw_meter_geq);
1115 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1116 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1119 free(rte, M_MRTABLE);
1122 buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1124 LIST_REMOVE(rt, mfc_hash);
1125 free(rt, M_MRTABLE);
1132 add_mfc(struct mfcctl2 *mfccp)
1138 struct epoch_tracker et;
1141 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1143 /* If an entry already exists, just update the fields */
1145 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1146 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1147 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1148 mfccp->mfcc_parent);
1149 update_mfc_params(rt, mfccp);
1155 * Find the entry for which the upcall was made and update
1158 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1159 NET_EPOCH_ENTER(et);
1160 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1161 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1162 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1163 !buf_ring_empty(rt->mfc_stall_ring)) {
1165 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1166 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1167 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1169 rt->mfc_stall_ring);
1171 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1173 init_mfc_params(rt, mfccp);
1174 rt->mfc_expire = 0; /* Don't clean this guy up */
1177 /* Free queued packets, but attempt to forward them first. */
1178 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1179 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1180 if (rte->ifp != NULL)
1181 ip_mdq(rte->m, rte->ifp, rt, -1);
1183 free(rte, M_MRTABLE);
1190 * It is possible that an entry is being inserted without an upcall
1193 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1194 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1195 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1196 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1197 init_mfc_params(rt, mfccp);
1205 if (rt == NULL) { /* no upcall, so make a new entry */
1212 init_mfc_params(rt, mfccp);
1215 rt->mfc_bw_meter_leq = NULL;
1216 rt->mfc_bw_meter_geq = NULL;
1218 /* insert new entry at head of hash chain */
1219 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1229 * Delete an mfc entry
1232 del_mfc(struct mfcctl2 *mfccp)
1234 struct in_addr origin;
1235 struct in_addr mcastgrp;
1238 origin = mfccp->mfcc_origin;
1239 mcastgrp = mfccp->mfcc_mcastgrp;
1241 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1242 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1246 rt = mfc_find(&origin, &mcastgrp);
1249 return EADDRNOTAVAIL;
1253 * free the bw_meter entries
1255 free_bw_list(rt->mfc_bw_meter_leq);
1256 rt->mfc_bw_meter_leq = NULL;
1257 free_bw_list(rt->mfc_bw_meter_geq);
1258 rt->mfc_bw_meter_geq = NULL;
1260 LIST_REMOVE(rt, mfc_hash);
1261 free(rt, M_MRTABLE);
1269 * Send a message to the routing daemon on the multicast routing socket.
1272 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1275 SOCKBUF_LOCK(&s->so_rcv);
1276 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1278 sorwakeup_locked(s);
1281 soroverflow_locked(s);
1288 * IP multicast forwarding function. This function assumes that the packet
1289 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1290 * pointed to by "ifp", and the packet is to be relayed to other networks
1291 * that have members of the packet's destination IP multicast group.
1293 * The packet is returned unscathed to the caller, unless it is
1294 * erroneous, in which case a non-zero return value tells the caller to
1298 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1301 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1302 struct ip_moptions *imo)
1312 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1313 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1315 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1316 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1318 * Packet arrived via a physical interface or
1319 * an encapsulated tunnel or a register_vif.
1323 * Packet arrived through a source-route tunnel.
1324 * Source-route tunnels are no longer supported.
1330 * BEGIN: MCAST ROUTING HOT PATH
1333 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1334 if (ip->ip_ttl < MAXTTL)
1335 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1336 error = ip_mdq(m, ifp, NULL, vifi);
1342 * Don't forward a packet with time-to-live of zero or one,
1343 * or a packet destined to a local-only group.
1345 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1352 * Determine forwarding vifs from the forwarding cache table
1354 MRTSTAT_INC(mrts_mfc_lookups);
1355 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1357 /* Entry exists, so forward if necessary */
1359 error = ip_mdq(m, ifp, rt, -1);
1360 /* Generic unlock here as we might release R or W lock */
1366 * END: MCAST ROUTING HOT PATH
1369 /* Further processing must be done with WLOCK taken */
1370 if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1373 goto mfc_find_retry;
1377 * If we don't have a route for packet's origin,
1378 * Make a copy of the packet & send message to routing daemon
1380 hlen = ip->ip_hl << 2;
1382 MRTSTAT_INC(mrts_mfc_misses);
1383 MRTSTAT_INC(mrts_no_route);
1384 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1385 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1388 * Allocate mbufs early so that we don't do extra work if we are
1389 * just going to fail anyway. Make sure to pullup the header so
1390 * that other people can't step on it.
1392 rte = (struct rtdetq*) malloc((sizeof *rte), M_MRTABLE,
1399 mb0 = m_copypacket(m, M_NOWAIT);
1400 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1401 mb0 = m_pullup(mb0, hlen);
1403 free(rte, M_MRTABLE);
1408 /* is there an upcall waiting for this flow ? */
1409 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1410 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1412 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1413 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1414 !buf_ring_empty(rt->mfc_stall_ring))
1421 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1425 * Locate the vifi for the incoming interface for this packet.
1426 * If none found, drop packet.
1428 for (vifi = 0; vifi < V_numvifs &&
1429 V_viftable[vifi].v_ifp != ifp; vifi++)
1431 if (vifi >= V_numvifs) /* vif not found, drop packet */
1434 /* no upcall, so make a new entry */
1439 /* Make a copy of the header to send to the user level process */
1440 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1445 * Send message to routing daemon to install
1446 * a route into the kernel table
1449 im = mtod(mm, struct igmpmsg*);
1450 im->im_msgtype = IGMPMSG_NOCACHE;
1454 MRTSTAT_INC(mrts_upcalls);
1456 k_igmpsrc.sin_addr = ip->ip_src;
1457 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1458 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1459 MRTSTAT_INC(mrts_upq_sockfull);
1460 fail1: free(rt, M_MRTABLE);
1461 fail: free(rte, M_MRTABLE);
1467 /* insert new entry at head of hash chain */
1468 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1469 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1470 rt->mfc_expire = UPCALL_EXPIRE;
1472 for (i = 0; i < V_numvifs; i++) {
1473 rt->mfc_ttls[i] = 0;
1474 rt->mfc_flags[i] = 0;
1476 rt->mfc_parent = -1;
1478 /* clear the RP address */
1479 rt->mfc_rp.s_addr = INADDR_ANY;
1480 rt->mfc_bw_meter_leq = NULL;
1481 rt->mfc_bw_meter_geq = NULL;
1483 /* initialize pkt counters per src-grp */
1484 rt->mfc_pkt_cnt = 0;
1485 rt->mfc_byte_cnt = 0;
1486 rt->mfc_wrong_if = 0;
1487 timevalclear(&rt->mfc_last_assert);
1489 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1491 /* Add RT to hashtable as it didn't exist before */
1492 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1494 /* determine if queue has overflowed */
1495 if (buf_ring_full(rt->mfc_stall_ring)) {
1496 MRTSTAT_INC(mrts_upq_ovflw);
1497 non_fatal: free(rte, M_MRTABLE);
1503 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1515 * Clean up the cache entry if upcall is not serviced
1518 expire_upcalls(void *arg)
1522 CURVNET_SET((struct vnet *) arg);
1524 /*This callout is always run with MRW_WLOCK taken. */
1526 for (i = 0; i < mfchashsize; i++) {
1527 struct mfc *rt, *nrt;
1529 if (V_nexpire[i] == 0)
1532 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1533 if (buf_ring_empty(rt->mfc_stall_ring))
1536 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1539 MRTSTAT_INC(mrts_cache_cleanups);
1540 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1541 (u_long)ntohl(rt->mfc_origin.s_addr),
1542 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1548 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1555 * Packet forwarding routine once entry in the cache is made
1558 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1560 struct ip *ip = mtod(m, struct ip *);
1562 int plen = ntohs(ip->ip_len);
1568 * If xmt_vif is not -1, send on only the requested vif.
1570 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1572 if (xmt_vif < V_numvifs) {
1573 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1574 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1576 phyint_send(ip, V_viftable + xmt_vif, m);
1581 * Don't forward if it didn't arrive from the parent vif for its origin.
1583 vifi = rt->mfc_parent;
1584 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1585 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1586 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1587 MRTSTAT_INC(mrts_wrong_if);
1590 * If we are doing PIM assert processing, send a message
1591 * to the routing daemon.
1593 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1594 * can complete the SPT switch, regardless of the type
1595 * of the iif (broadcast media, GRE tunnel, etc).
1597 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1598 V_viftable[vifi].v_ifp) {
1599 if (ifp == V_multicast_register_if)
1600 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1602 /* Get vifi for the incoming packet */
1603 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1606 if (vifi >= V_numvifs)
1607 return 0; /* The iif is not found: ignore the packet. */
1609 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1610 return 0; /* WRONGVIF disabled: ignore the packet */
1612 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1613 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1615 int hlen = ip->ip_hl << 2;
1616 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1618 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1619 mm = m_pullup(mm, hlen);
1623 im = mtod(mm, struct igmpmsg *);
1624 im->im_msgtype = IGMPMSG_WRONGVIF;
1628 MRTSTAT_INC(mrts_upcalls);
1630 k_igmpsrc.sin_addr = im->im_src;
1631 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1632 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1633 MRTSTAT_INC(mrts_upq_sockfull);
1641 /* If I sourced this packet, it counts as output, else it was input. */
1642 mtx_lock_spin(&V_viftable[vifi].v_spin);
1643 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1644 V_viftable[vifi].v_pkt_out++;
1645 V_viftable[vifi].v_bytes_out += plen;
1647 V_viftable[vifi].v_pkt_in++;
1648 V_viftable[vifi].v_bytes_in += plen;
1650 mtx_unlock_spin(&V_viftable[vifi].v_spin);
1653 rt->mfc_byte_cnt += plen;
1656 * For each vif, decide if a copy of the packet should be forwarded.
1658 * - the ttl exceeds the vif's threshold
1659 * - there are group members downstream on interface
1661 for (vifi = 0; vifi < V_numvifs; vifi++)
1662 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1663 V_viftable[vifi].v_pkt_out++;
1664 V_viftable[vifi].v_bytes_out += plen;
1665 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1666 pim_register_send(ip, V_viftable + vifi, m, rt);
1668 phyint_send(ip, V_viftable + vifi, m);
1672 * Perform upcall-related bw measuring.
1674 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1679 /* Process meters for Greater-or-EQual case */
1680 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1681 bw_meter_geq_receive_packet(x, plen, &now);
1683 /* Process meters for Lower-or-EQual case */
1684 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1686 * Record that a packet is received.
1687 * Spin lock has to be taken as callout context
1688 * (expire_bw_meter_leq) might modify these fields
1691 mtx_lock_spin(&x->bm_spin);
1692 x->bm_measured.b_packets++;
1693 x->bm_measured.b_bytes += plen;
1694 mtx_unlock_spin(&x->bm_spin);
1702 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1705 X_legal_vif_num(int vif)
1714 if (vif < V_numvifs)
1722 * Return the local address used by this vif
1725 X_ip_mcast_src(int vifi)
1734 if (vifi < V_numvifs)
1735 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1742 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1744 struct mbuf *mb_copy;
1745 int hlen = ip->ip_hl << 2;
1750 * Make a new reference to the packet; make sure that
1751 * the IP header is actually copied, not just referenced,
1752 * so that ip_output() only scribbles on the copy.
1754 mb_copy = m_copypacket(m, M_NOWAIT);
1755 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1756 mb_copy = m_pullup(mb_copy, hlen);
1757 if (mb_copy == NULL)
1760 send_packet(vifp, mb_copy);
1764 send_packet(struct vif *vifp, struct mbuf *m)
1766 struct ip_moptions imo;
1772 imo.imo_multicast_ifp = vifp->v_ifp;
1773 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1774 imo.imo_multicast_loop = !!in_mcast_loop;
1775 imo.imo_multicast_vif = -1;
1776 STAILQ_INIT(&imo.imo_head);
1779 * Re-entrancy should not be a problem here, because
1780 * the packets that we send out and are looped back at us
1781 * should get rejected because they appear to come from
1782 * the loopback interface, thus preventing looping.
1784 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1785 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1786 (ptrdiff_t)(vifp - V_viftable), error);
1790 * Stubs for old RSVP socket shim implementation.
1794 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1797 return (EOPNOTSUPP);
1801 X_ip_rsvp_force_done(struct socket *so __unused)
1807 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1815 return (IPPROTO_DONE);
1819 * Code for bandwidth monitors
1823 * Define common interface for timeval-related methods
1825 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1826 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1827 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1830 compute_bw_meter_flags(struct bw_upcall *req)
1834 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1835 flags |= BW_METER_UNIT_PACKETS;
1836 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1837 flags |= BW_METER_UNIT_BYTES;
1838 if (req->bu_flags & BW_UPCALL_GEQ)
1839 flags |= BW_METER_GEQ;
1840 if (req->bu_flags & BW_UPCALL_LEQ)
1841 flags |= BW_METER_LEQ;
1847 expire_bw_meter_leq(void *arg)
1849 struct bw_meter *x = arg;
1853 * callout is always executed with MRW_WLOCK taken
1856 CURVNET_SET((struct vnet *)x->arg);
1861 * Test if we should deliver an upcall
1863 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1864 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1865 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1866 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1867 /* Prepare an upcall for delivery */
1868 bw_meter_prepare_upcall(x, &now);
1871 /* Send all upcalls that are pending delivery */
1872 taskqueue_enqueue(V_task_queue, &V_task);
1874 /* Reset counters */
1875 x->bm_start_time = now;
1876 /* Spin lock has to be taken as ip_forward context
1877 * might modify these fields as well
1879 mtx_lock_spin(&x->bm_spin);
1880 x->bm_measured.b_bytes = 0;
1881 x->bm_measured.b_packets = 0;
1882 mtx_unlock_spin(&x->bm_spin);
1884 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1890 * Add a bw_meter entry
1893 add_bw_upcall(struct bw_upcall *req)
1896 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1897 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1899 struct bw_meter *x, **bwm_ptr;
1902 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1905 /* Test if the flags are valid */
1906 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1908 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1910 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1911 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1914 /* Test if the threshold time interval is valid */
1915 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1918 flags = compute_bw_meter_flags(req);
1921 * Find if we have already same bw_meter entry
1924 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1927 return EADDRNOTAVAIL;
1930 /* Choose an appropriate bw_meter list */
1931 if (req->bu_flags & BW_UPCALL_GEQ)
1932 bwm_ptr = &mfc->mfc_bw_meter_geq;
1934 bwm_ptr = &mfc->mfc_bw_meter_leq;
1936 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1937 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1938 &req->bu_threshold.b_time, ==))
1939 && (x->bm_threshold.b_packets
1940 == req->bu_threshold.b_packets)
1941 && (x->bm_threshold.b_bytes
1942 == req->bu_threshold.b_bytes)
1943 && (x->bm_flags & BW_METER_USER_FLAGS)
1946 return 0; /* XXX Already installed */
1950 /* Allocate the new bw_meter entry */
1951 x = (struct bw_meter*) malloc(sizeof(*x), M_BWMETER,
1958 /* Set the new bw_meter entry */
1959 x->bm_threshold.b_time = req->bu_threshold.b_time;
1961 x->bm_start_time = now;
1962 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1963 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1964 x->bm_measured.b_packets = 0;
1965 x->bm_measured.b_bytes = 0;
1966 x->bm_flags = flags;
1967 x->bm_time_next = NULL;
1970 sprintf(x->bm_spin_name, "BM spin %p", x);
1971 mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1973 /* For LEQ case create periodic callout */
1974 if (req->bu_flags & BW_UPCALL_LEQ) {
1975 callout_init_rw(&x->bm_meter_callout, &mrouter_mtx, CALLOUT_SHAREDLOCK);
1976 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1977 expire_bw_meter_leq, x);
1980 /* Add the new bw_meter entry to the front of entries for this MFC */
1981 x->bm_mfc_next = *bwm_ptr;
1990 free_bw_list(struct bw_meter *list)
1992 while (list != NULL) {
1993 struct bw_meter *x = list;
1995 /* MRW_WLOCK must be held here */
1996 if (x->bm_flags & BW_METER_LEQ) {
1997 callout_drain(&x->bm_meter_callout);
1998 mtx_destroy(&x->bm_spin);
2001 list = list->bm_mfc_next;
2007 * Delete one or multiple bw_meter entries
2010 del_bw_upcall(struct bw_upcall *req)
2013 struct bw_meter *x, **bwm_ptr;
2015 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2020 /* Find the corresponding MFC entry */
2021 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2024 return EADDRNOTAVAIL;
2025 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2027 * Delete all bw_meter entries for this mfc
2029 struct bw_meter *list;
2032 list = mfc->mfc_bw_meter_leq;
2033 mfc->mfc_bw_meter_leq = NULL;
2037 list = mfc->mfc_bw_meter_geq;
2038 mfc->mfc_bw_meter_geq = NULL;
2042 } else { /* Delete a single bw_meter entry */
2043 struct bw_meter *prev;
2046 flags = compute_bw_meter_flags(req);
2048 /* Choose an appropriate bw_meter list */
2049 if (req->bu_flags & BW_UPCALL_GEQ)
2050 bwm_ptr = &mfc->mfc_bw_meter_geq;
2052 bwm_ptr = &mfc->mfc_bw_meter_leq;
2054 /* Find the bw_meter entry to delete */
2055 for (prev = NULL, x = *bwm_ptr; x != NULL;
2056 prev = x, x = x->bm_mfc_next) {
2057 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2058 &req->bu_threshold.b_time, ==)) &&
2059 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2060 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2061 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2064 if (x != NULL) { /* Delete entry from the list for this MFC */
2066 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2068 *bwm_ptr = x->bm_mfc_next;/* new head of list */
2070 if (req->bu_flags & BW_UPCALL_LEQ)
2071 callout_stop(&x->bm_meter_callout);
2074 /* Free the bw_meter entry */
2086 * Perform bandwidth measurement processing that may result in an upcall
2089 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2091 struct timeval delta;
2096 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2099 * Processing for ">=" type of bw_meter entry.
2100 * bm_spin does not have to be hold here as in GEQ
2101 * case this is the only context accessing bm_measured.
2103 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2104 /* Reset the bw_meter entry */
2105 x->bm_start_time = *nowp;
2106 x->bm_measured.b_packets = 0;
2107 x->bm_measured.b_bytes = 0;
2108 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2111 /* Record that a packet is received */
2112 x->bm_measured.b_packets++;
2113 x->bm_measured.b_bytes += plen;
2116 * Test if we should deliver an upcall
2118 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2119 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2120 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2121 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2122 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2123 /* Prepare an upcall for delivery */
2124 bw_meter_prepare_upcall(x, nowp);
2125 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2131 * Prepare a bandwidth-related upcall
2134 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2136 struct timeval delta;
2137 struct bw_upcall *u;
2142 * Compute the measured time interval
2145 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2148 * Set the bw_upcall entry
2150 u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2152 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2155 u->bu_src = x->bm_mfc->mfc_origin;
2156 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2157 u->bu_threshold.b_time = x->bm_threshold.b_time;
2158 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2159 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2160 u->bu_measured.b_time = delta;
2161 u->bu_measured.b_packets = x->bm_measured.b_packets;
2162 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2164 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2165 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2166 if (x->bm_flags & BW_METER_UNIT_BYTES)
2167 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2168 if (x->bm_flags & BW_METER_GEQ)
2169 u->bu_flags |= BW_UPCALL_GEQ;
2170 if (x->bm_flags & BW_METER_LEQ)
2171 u->bu_flags |= BW_UPCALL_LEQ;
2173 if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2174 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2175 if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2176 taskqueue_enqueue(V_task_queue, &V_task);
2180 * Send the pending bandwidth-related upcalls
2183 bw_upcalls_send(void)
2187 struct bw_upcall *bu;
2188 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2189 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2191 IGMPMSG_BW_UPCALL,/* im_msgtype */
2196 { 0 } }; /* im_dst */
2200 if (buf_ring_empty(V_bw_upcalls_ring))
2204 * Allocate a new mbuf, initialize it with the header and
2205 * the payload for the pending calls.
2207 m = m_gethdr(M_NOWAIT, MT_DATA);
2209 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2213 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2214 len += sizeof(struct igmpmsg);
2215 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2216 m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2217 len += sizeof(struct bw_upcall);
2218 free(bu, M_MRTABLE);
2223 * XXX do we need to set the address in k_igmpsrc ?
2225 MRTSTAT_INC(mrts_upcalls);
2226 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2227 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2228 MRTSTAT_INC(mrts_upq_sockfull);
2233 * A periodic function for sending all upcalls that are pending delivery
2236 expire_bw_upcalls_send(void *arg)
2238 CURVNET_SET((struct vnet *) arg);
2240 /* This callout is run with MRW_RLOCK taken */
2244 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2250 * End of bandwidth monitoring code
2254 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2258 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2261 struct mbuf *mb_copy, *mm;
2264 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2265 * rendezvous point was unspecified, and we were told not to.
2267 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2268 in_nullhost(rt->mfc_rp))
2271 mb_copy = pim_register_prepare(ip, m);
2272 if (mb_copy == NULL)
2276 * Send all the fragments. Note that the mbuf for each fragment
2277 * is freed by the sending machinery.
2279 for (mm = mb_copy; mm; mm = mb_copy) {
2280 mb_copy = mm->m_nextpkt;
2282 mm = m_pullup(mm, sizeof(struct ip));
2284 ip = mtod(mm, struct ip *);
2285 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2286 pim_register_send_rp(ip, vifp, mm, rt);
2288 pim_register_send_upcall(ip, vifp, mm, rt);
2297 * Return a copy of the data packet that is ready for PIM Register
2299 * XXX: Note that in the returned copy the IP header is a valid one.
2301 static struct mbuf *
2302 pim_register_prepare(struct ip *ip, struct mbuf *m)
2304 struct mbuf *mb_copy = NULL;
2307 /* Take care of delayed checksums */
2308 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2309 in_delayed_cksum(m);
2310 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2314 * Copy the old packet & pullup its IP header into the
2315 * new mbuf so we can modify it.
2317 mb_copy = m_copypacket(m, M_NOWAIT);
2318 if (mb_copy == NULL)
2320 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2321 if (mb_copy == NULL)
2324 /* take care of the TTL */
2325 ip = mtod(mb_copy, struct ip *);
2328 /* Compute the MTU after the PIM Register encapsulation */
2329 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2331 if (ntohs(ip->ip_len) <= mtu) {
2332 /* Turn the IP header into a valid one */
2334 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2336 /* Fragment the packet */
2337 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2338 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2347 * Send an upcall with the data packet to the user-level process.
2350 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2351 struct mbuf *mb_copy, struct mfc *rt)
2353 struct mbuf *mb_first;
2354 int len = ntohs(ip->ip_len);
2356 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2361 * Add a new mbuf with an upcall header
2363 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2364 if (mb_first == NULL) {
2368 mb_first->m_data += max_linkhdr;
2369 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2370 mb_first->m_len = sizeof(struct igmpmsg);
2371 mb_first->m_next = mb_copy;
2373 /* Send message to routing daemon */
2374 im = mtod(mb_first, struct igmpmsg *);
2375 im->im_msgtype = IGMPMSG_WHOLEPKT;
2377 im->im_vif = vifp - V_viftable;
2378 im->im_src = ip->ip_src;
2379 im->im_dst = ip->ip_dst;
2381 k_igmpsrc.sin_addr = ip->ip_src;
2383 MRTSTAT_INC(mrts_upcalls);
2385 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2386 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2387 MRTSTAT_INC(mrts_upq_sockfull);
2391 /* Keep statistics */
2392 PIMSTAT_INC(pims_snd_registers_msgs);
2393 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2399 * Encapsulate the data packet in PIM Register message and send it to the RP.
2402 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2405 struct mbuf *mb_first;
2406 struct ip *ip_outer;
2407 struct pim_encap_pimhdr *pimhdr;
2408 int len = ntohs(ip->ip_len);
2409 vifi_t vifi = rt->mfc_parent;
2413 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2415 return EADDRNOTAVAIL; /* The iif vif is invalid */
2419 * Add a new mbuf with the encapsulating header
2421 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2422 if (mb_first == NULL) {
2426 mb_first->m_data += max_linkhdr;
2427 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2428 mb_first->m_next = mb_copy;
2430 mb_first->m_pkthdr.len = len + mb_first->m_len;
2433 * Fill in the encapsulating IP and PIM header
2435 ip_outer = mtod(mb_first, struct ip *);
2436 *ip_outer = pim_encap_iphdr;
2437 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2438 sizeof(pim_encap_pimhdr));
2439 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2440 ip_outer->ip_dst = rt->mfc_rp;
2442 * Copy the inner header TOS to the outer header, and take care of the
2445 ip_outer->ip_tos = ip->ip_tos;
2446 if (ip->ip_off & htons(IP_DF))
2447 ip_outer->ip_off |= htons(IP_DF);
2448 ip_fillid(ip_outer);
2449 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2450 + sizeof(pim_encap_iphdr));
2451 *pimhdr = pim_encap_pimhdr;
2452 /* If the iif crosses a border, set the Border-bit */
2453 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2454 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2456 mb_first->m_data += sizeof(pim_encap_iphdr);
2457 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2458 mb_first->m_data -= sizeof(pim_encap_iphdr);
2460 send_packet(vifp, mb_first);
2462 /* Keep statistics */
2463 PIMSTAT_INC(pims_snd_registers_msgs);
2464 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2470 * pim_encapcheck() is called by the encap4_input() path at runtime to
2471 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2475 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2476 int proto __unused, void *arg __unused)
2479 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2480 return (8); /* claim the datagram. */
2484 * PIM-SMv2 and PIM-DM messages processing.
2485 * Receives and verifies the PIM control messages, and passes them
2486 * up to the listening socket, using rip_input().
2487 * The only message with special processing is the PIM_REGISTER message
2488 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2489 * is passed to if_simloop().
2492 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2494 struct ip *ip = mtod(m, struct ip *);
2498 int datalen = ntohs(ip->ip_len) - iphlen;
2501 /* Keep statistics */
2502 PIMSTAT_INC(pims_rcv_total_msgs);
2503 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2508 if (datalen < PIM_MINLEN) {
2509 PIMSTAT_INC(pims_rcv_tooshort);
2510 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2511 __func__, datalen, ntohl(ip->ip_src.s_addr));
2513 return (IPPROTO_DONE);
2517 * If the packet is at least as big as a REGISTER, go agead
2518 * and grab the PIM REGISTER header size, to avoid another
2519 * possible m_pullup() later.
2521 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2522 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2524 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2526 * Get the IP and PIM headers in contiguous memory, and
2527 * possibly the PIM REGISTER header.
2529 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2530 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2531 return (IPPROTO_DONE);
2534 /* m_pullup() may have given us a new mbuf so reset ip. */
2535 ip = mtod(m, struct ip *);
2536 ip_tos = ip->ip_tos;
2538 /* adjust mbuf to point to the PIM header */
2539 m->m_data += iphlen;
2541 pim = mtod(m, struct pim *);
2544 * Validate checksum. If PIM REGISTER, exclude the data packet.
2546 * XXX: some older PIMv2 implementations don't make this distinction,
2547 * so for compatibility reason perform the checksum over part of the
2548 * message, and if error, then over the whole message.
2550 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2551 /* do nothing, checksum okay */
2552 } else if (in_cksum(m, datalen)) {
2553 PIMSTAT_INC(pims_rcv_badsum);
2554 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2556 return (IPPROTO_DONE);
2559 /* PIM version check */
2560 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2561 PIMSTAT_INC(pims_rcv_badversion);
2562 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2563 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2565 return (IPPROTO_DONE);
2568 /* restore mbuf back to the outer IP */
2569 m->m_data -= iphlen;
2572 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2574 * Since this is a REGISTER, we'll make a copy of the register
2575 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2578 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2580 struct ip *encap_ip;
2585 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2587 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2588 (int)V_reg_vif_num);
2590 return (IPPROTO_DONE);
2592 /* XXX need refcnt? */
2593 vifp = V_viftable[V_reg_vif_num].v_ifp;
2599 if (datalen < PIM_REG_MINLEN) {
2600 PIMSTAT_INC(pims_rcv_tooshort);
2601 PIMSTAT_INC(pims_rcv_badregisters);
2602 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2604 return (IPPROTO_DONE);
2607 reghdr = (u_int32_t *)(pim + 1);
2608 encap_ip = (struct ip *)(reghdr + 1);
2610 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2611 __func__, ntohl(encap_ip->ip_src.s_addr),
2612 ntohs(encap_ip->ip_len));
2614 /* verify the version number of the inner packet */
2615 if (encap_ip->ip_v != IPVERSION) {
2616 PIMSTAT_INC(pims_rcv_badregisters);
2617 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2619 return (IPPROTO_DONE);
2622 /* verify the inner packet is destined to a mcast group */
2623 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2624 PIMSTAT_INC(pims_rcv_badregisters);
2625 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2626 ntohl(encap_ip->ip_dst.s_addr));
2628 return (IPPROTO_DONE);
2631 /* If a NULL_REGISTER, pass it to the daemon */
2632 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2633 goto pim_input_to_daemon;
2636 * Copy the TOS from the outer IP header to the inner IP header.
2638 if (encap_ip->ip_tos != ip_tos) {
2639 /* Outer TOS -> inner TOS */
2640 encap_ip->ip_tos = ip_tos;
2641 /* Recompute the inner header checksum. Sigh... */
2643 /* adjust mbuf to point to the inner IP header */
2644 m->m_data += (iphlen + PIM_MINLEN);
2645 m->m_len -= (iphlen + PIM_MINLEN);
2647 encap_ip->ip_sum = 0;
2648 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2650 /* restore mbuf to point back to the outer IP header */
2651 m->m_data -= (iphlen + PIM_MINLEN);
2652 m->m_len += (iphlen + PIM_MINLEN);
2656 * Decapsulate the inner IP packet and loopback to forward it
2657 * as a normal multicast packet. Also, make a copy of the
2658 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2659 * to pass to the daemon later, so it can take the appropriate
2660 * actions (e.g., send back PIM_REGISTER_STOP).
2661 * XXX: here m->m_data points to the outer IP header.
2663 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2665 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2667 return (IPPROTO_DONE);
2670 /* Keep statistics */
2671 /* XXX: registers_bytes include only the encap. mcast pkt */
2672 PIMSTAT_INC(pims_rcv_registers_msgs);
2673 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2676 * forward the inner ip packet; point m_data at the inner ip.
2678 m_adj(m, iphlen + PIM_MINLEN);
2681 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2683 (u_long)ntohl(encap_ip->ip_src.s_addr),
2684 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2685 (int)V_reg_vif_num);
2687 /* NB: vifp was collected above; can it change on us? */
2688 if_simloop(vifp, m, dst.sin_family, 0);
2690 /* prepare the register head to send to the mrouting daemon */
2694 pim_input_to_daemon:
2696 * Pass the PIM message up to the daemon; if it is a Register message,
2697 * pass the 'head' only up to the daemon. This includes the
2698 * outer IP header, PIM header, PIM-Register header and the
2700 * XXX: the outer IP header pkt size of a Register is not adjust to
2701 * reflect the fact that the inner multicast data is truncated.
2703 return (rip_input(&m, &off, proto));
2707 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2714 if (V_mfchashtbl == NULL) /* XXX unlocked */
2716 error = sysctl_wire_old_buffer(req, 0);
2721 for (i = 0; i < mfchashsize; i++) {
2722 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2723 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2733 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2734 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2735 "IPv4 Multicast Forwarding Table "
2736 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2739 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2745 if (V_viftable == NULL) /* XXX unlocked */
2747 error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2752 /* Copy out user-visible portion of vif entry. */
2753 for (i = 0; i < MAXVIFS; i++) {
2754 error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2762 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2763 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2764 sysctl_viflist, "S,vif[MAXVIFS]",
2765 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2768 vnet_mroute_init(const void *unused __unused)
2771 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2773 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2774 M_MRTABLE, M_WAITOK|M_ZERO);
2776 callout_init_rw(&V_expire_upcalls_ch, &mrouter_mtx, 0);
2777 callout_init_rw(&V_bw_upcalls_ch, &mrouter_mtx, 0);
2779 /* Prepare taskqueue */
2780 V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2781 taskqueue_thread_enqueue, &V_task_queue);
2782 taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2785 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2789 vnet_mroute_uninit(const void *unused __unused)
2792 /* Taskqueue should be cancelled and drained before freeing */
2793 taskqueue_free(V_task_queue);
2795 free(V_viftable, M_MRTABLE);
2796 free(V_nexpire, M_MRTABLE);
2800 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2801 vnet_mroute_uninit, NULL);
2804 ip_mroute_modevent(module_t mod, int type, void *unused)
2811 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2812 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2813 if (if_detach_event_tag == NULL) {
2814 printf("ip_mroute: unable to register "
2815 "ifnet_departure_event handler\n");
2820 mfchashsize = MFCHASHSIZE;
2821 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2822 !powerof2(mfchashsize)) {
2823 printf("WARNING: %s not a power of 2; using default\n",
2824 "net.inet.ip.mfchashsize");
2825 mfchashsize = MFCHASHSIZE;
2828 pim_squelch_wholepkt = 0;
2829 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2830 &pim_squelch_wholepkt);
2832 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2833 if (pim_encap_cookie == NULL) {
2834 printf("ip_mroute: unable to attach pim encap\n");
2839 ip_mcast_src = X_ip_mcast_src;
2840 ip_mforward = X_ip_mforward;
2841 ip_mrouter_done = X_ip_mrouter_done;
2842 ip_mrouter_get = X_ip_mrouter_get;
2843 ip_mrouter_set = X_ip_mrouter_set;
2845 ip_rsvp_force_done = X_ip_rsvp_force_done;
2846 ip_rsvp_vif = X_ip_rsvp_vif;
2848 legal_vif_num = X_legal_vif_num;
2849 mrt_ioctl = X_mrt_ioctl;
2850 rsvp_input_p = X_rsvp_input;
2855 * Typically module unload happens after the user-level
2856 * process has shutdown the kernel services (the check
2857 * below insures someone can't just yank the module out
2858 * from under a running process). But if the module is
2859 * just loaded and then unloaded w/o starting up a user
2860 * process we still need to cleanup.
2863 if (ip_mrouter_cnt != 0) {
2867 ip_mrouter_unloading = 1;
2870 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2872 if (pim_encap_cookie) {
2873 ip_encap_detach(pim_encap_cookie);
2874 pim_encap_cookie = NULL;
2877 ip_mcast_src = NULL;
2879 ip_mrouter_done = NULL;
2880 ip_mrouter_get = NULL;
2881 ip_mrouter_set = NULL;
2883 ip_rsvp_force_done = NULL;
2886 legal_vif_num = NULL;
2888 rsvp_input_p = NULL;
2899 static moduledata_t ip_mroutemod = {
2905 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);