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_lock;
147 #define MRW_RLOCK() rw_rlock(&mrouter_lock)
148 #define MRW_WLOCK() rw_wlock(&mrouter_lock)
149 #define MRW_RUNLOCK() rw_runlock(&mrouter_lock)
150 #define MRW_WUNLOCK() rw_wunlock(&mrouter_lock)
151 #define MRW_UNLOCK() rw_unlock(&mrouter_lock)
152 #define MRW_LOCK_ASSERT() rw_assert(&mrouter_lock, RA_LOCKED)
153 #define MRW_WLOCK_ASSERT() rw_assert(&mrouter_lock, RA_WLOCKED)
154 #define MRW_LOCK_TRY_UPGRADE() rw_try_upgrade(&mrouter_lock)
155 #define MRW_WOWNED() rw_wowned(&mrouter_lock)
156 #define MRW_LOCK_INIT() \
157 rw_init(&mrouter_lock, "IPv4 multicast forwarding")
158 #define MRW_LOCK_DESTROY() rw_destroy(&mrouter_lock)
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 = 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);
434 error = ip_mrouter_done();
437 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
440 error = add_vif(&vifc);
443 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
446 error = del_vif(vifi);
451 * select data size depending on API version.
453 if (sopt->sopt_name == MRT_ADD_MFC &&
454 V_mrt_api_config & MRT_API_FLAGS_ALL) {
455 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
456 sizeof(struct mfcctl2));
458 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
459 sizeof(struct mfcctl));
460 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
461 sizeof(mfc) - sizeof(struct mfcctl));
465 if (sopt->sopt_name == MRT_ADD_MFC)
466 error = add_mfc(&mfc);
468 error = del_mfc(&mfc);
472 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
479 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
481 error = set_api_config(&i);
483 error = sooptcopyout(sopt, &i, sizeof i);
486 case MRT_ADD_BW_UPCALL:
487 case MRT_DEL_BW_UPCALL:
488 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
492 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
493 error = add_bw_upcall(&bw_upcall);
495 error = del_bw_upcall(&bw_upcall);
506 * Handle MRT getsockopt commands
509 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
513 switch (sopt->sopt_name) {
515 error = sooptcopyout(sopt, &mrt_api_version,
516 sizeof mrt_api_version);
519 error = sooptcopyout(sopt, &V_pim_assert_enabled,
520 sizeof V_pim_assert_enabled);
522 case MRT_API_SUPPORT:
523 error = sooptcopyout(sopt, &mrt_api_support,
524 sizeof mrt_api_support);
527 error = sooptcopyout(sopt, &V_mrt_api_config,
528 sizeof V_mrt_api_config);
538 * Handle ioctl commands to obtain information from the cache
541 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
546 * Currently the only function calling this ioctl routine is rtioctl_fib().
547 * Typically, only root can create the raw socket in order to execute
548 * this ioctl method, however the request might be coming from a prison
550 error = priv_check(curthread, PRIV_NETINET_MROUTE);
554 case (SIOCGETVIFCNT):
555 error = get_vif_cnt((struct sioc_vif_req *)data);
559 error = get_sg_cnt((struct sioc_sg_req *)data);
570 * returns the packet, byte, rpf-failure count for the source group provided
573 get_sg_cnt(struct sioc_sg_req *req)
578 rt = mfc_find(&req->src, &req->grp);
581 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
582 return EADDRNOTAVAIL;
584 req->pktcnt = rt->mfc_pkt_cnt;
585 req->bytecnt = rt->mfc_byte_cnt;
586 req->wrong_if = rt->mfc_wrong_if;
592 * returns the input and output packet and byte counts on the vif provided
595 get_vif_cnt(struct sioc_vif_req *req)
597 vifi_t vifi = req->vifi;
600 if (vifi >= V_numvifs) {
605 mtx_lock_spin(&V_viftable[vifi].v_spin);
606 req->icount = V_viftable[vifi].v_pkt_in;
607 req->ocount = V_viftable[vifi].v_pkt_out;
608 req->ibytes = V_viftable[vifi].v_bytes_in;
609 req->obytes = V_viftable[vifi].v_bytes_out;
610 mtx_unlock_spin(&V_viftable[vifi].v_spin);
617 if_detached_event(void *arg __unused, struct ifnet *ifp)
620 u_long i, vifi_cnt = 0;
621 struct ifnet *free_ptr;
625 if (V_ip_mrouter == NULL) {
631 * Tear down multicast forwarder state associated with this ifnet.
632 * 1. Walk the vif list, matching vifs against this ifnet.
633 * 2. Walk the multicast forwarding cache (mfc) looking for
634 * inner matches with this vif's index.
635 * 3. Expire any matching multicast forwarding cache entries.
636 * 4. Free vif state. This should disable ALLMULTI on the interface.
638 for (vifi = 0; vifi < V_numvifs; vifi++) {
639 if (V_viftable[vifi].v_ifp != ifp)
641 for (i = 0; i < mfchashsize; i++) {
642 struct mfc *rt, *nrt;
644 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
645 if (rt->mfc_parent == vifi) {
650 del_vif_locked(vifi, &free_ptr);
651 if (free_ptr != NULL)
658 * Free IFP. We don't have to use free_ptr here as it is the same
659 * that ifp. Perform free as many times as required in case
660 * refcount is greater than 1.
662 for (i = 0; i < vifi_cnt; i++)
667 ip_mrouter_upcall_thread(void *arg, int pending __unused)
669 CURVNET_SET((struct vnet *) arg);
679 * Enable multicast forwarding.
682 ip_mrouter_init(struct socket *so, int version)
685 CTR2(KTR_IPMF, "%s: so %p", __func__, so);
692 if (ip_mrouter_unloading) {
697 if (V_ip_mrouter != NULL) {
702 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
705 /* Create upcall ring */
706 mtx_init(&V_bw_upcalls_ring_mtx, "mroute upcall buf_ring mtx", NULL, MTX_DEF);
707 V_bw_upcalls_ring = buf_ring_alloc(BW_UPCALLS_MAX, M_MRTABLE,
708 M_NOWAIT, &V_bw_upcalls_ring_mtx);
709 if (!V_bw_upcalls_ring) {
714 TASK_INIT(&V_task, 0, ip_mrouter_upcall_thread, curvnet);
715 taskqueue_cancel(V_task_queue, &V_task, NULL);
716 taskqueue_unblock(V_task_queue);
718 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
720 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
724 atomic_add_int(&ip_mrouter_cnt, 1);
726 /* This is a mutex required by buf_ring init, but not used internally */
727 mtx_init(&V_buf_ring_mtx, "mroute buf_ring mtx", NULL, MTX_DEF);
731 CTR1(KTR_IPMF, "%s: done", __func__);
737 * Disable multicast forwarding.
740 X_ip_mrouter_done(void)
746 struct bw_upcall *bu;
748 if (V_ip_mrouter == NULL)
752 * Detach/disable hooks to the reset of the system.
755 atomic_subtract_int(&ip_mrouter_cnt, 1);
756 V_mrt_api_config = 0;
759 * Wait for all epoch sections to complete to ensure
760 * V_ip_mrouter = NULL is visible to others.
762 epoch_wait_preempt(net_epoch_preempt);
764 /* Stop and drain task queue */
765 taskqueue_block(V_task_queue);
766 while (taskqueue_cancel(V_task_queue, &V_task, NULL)) {
767 taskqueue_drain(V_task_queue, &V_task);
770 ifps = malloc(MAXVIFS * sizeof(*ifps), M_TEMP, M_WAITOK);
773 taskqueue_cancel(V_task_queue, &V_task, NULL);
775 /* Destroy upcall ring */
776 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
779 buf_ring_free(V_bw_upcalls_ring, M_MRTABLE);
780 mtx_destroy(&V_bw_upcalls_ring_mtx);
783 * For each phyint in use, prepare to disable promiscuous reception
784 * of all IP multicasts. Defer the actual call until the lock is released;
785 * just record the list of interfaces while locked. Some interfaces use
786 * sx locks in their ioctl routines, which is not allowed while holding
787 * a non-sleepable lock.
789 KASSERT(V_numvifs <= MAXVIFS, ("More vifs than possible"));
790 for (vifi = 0, nifp = 0; vifi < V_numvifs; vifi++) {
791 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
792 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
793 ifps[nifp++] = V_viftable[vifi].v_ifp;
796 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
798 V_pim_assert_enabled = 0;
800 callout_stop(&V_expire_upcalls_ch);
801 callout_stop(&V_bw_upcalls_ch);
804 * Free all multicast forwarding cache entries.
805 * Do not use hashdestroy(), as we must perform other cleanup.
807 for (i = 0; i < mfchashsize; i++) {
808 struct mfc *rt, *nrt;
810 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
814 free(V_mfchashtbl, M_MRTABLE);
817 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
819 V_reg_vif_num = VIFI_INVALID;
821 mtx_destroy(&V_buf_ring_mtx);
826 * Now drop our claim on promiscuous multicast on the interfaces recorded
827 * above. This is safe to do now because ALLMULTI is reference counted.
829 for (vifi = 0; vifi < nifp; vifi++)
830 if_allmulti(ifps[vifi], 0);
833 CTR1(KTR_IPMF, "%s: done", __func__);
839 * Set PIM assert processing global
844 if ((i != 1) && (i != 0))
847 V_pim_assert_enabled = i;
853 * Configure API capabilities
856 set_api_config(uint32_t *apival)
861 * We can set the API capabilities only if it is the first operation
862 * after MRT_INIT. I.e.:
863 * - there are no vifs installed
864 * - pim_assert is not enabled
865 * - the MFC table is empty
871 if (V_pim_assert_enabled) {
878 for (i = 0; i < mfchashsize; i++) {
879 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
888 V_mrt_api_config = *apival & mrt_api_support;
889 *apival = V_mrt_api_config;
895 * Add a vif to the vif table
898 add_vif(struct vifctl *vifcp)
900 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
901 struct sockaddr_in sin = {sizeof sin, AF_INET};
906 if (vifcp->vifc_vifi >= MAXVIFS)
908 /* rate limiting is no longer supported by this code */
909 if (vifcp->vifc_rate_limit != 0) {
910 log(LOG_ERR, "rate limiting is no longer supported\n");
914 if (in_nullhost(vifcp->vifc_lcl_addr))
915 return EADDRNOTAVAIL;
917 /* Find the interface with an address in AF_INET family */
918 if (vifcp->vifc_flags & VIFF_REGISTER) {
920 * XXX: Because VIFF_REGISTER does not really need a valid
921 * local interface (e.g. it could be 127.0.0.2), we don't
926 struct epoch_tracker et;
928 sin.sin_addr = vifcp->vifc_lcl_addr;
930 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
933 return EADDRNOTAVAIL;
936 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
940 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
941 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
943 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
944 ifp = V_multicast_register_if = if_alloc(IFT_LOOP);
945 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
946 if (V_reg_vif_num == VIFI_INVALID) {
947 if_initname(V_multicast_register_if, "register_vif", 0);
948 V_reg_vif_num = vifcp->vifc_vifi;
950 } else { /* Make sure the interface supports multicast */
951 if ((ifp->if_flags & IFF_MULTICAST) == 0)
954 /* Enable promiscuous reception of all IP multicasts from the if */
955 error = if_allmulti(ifp, 1);
962 if (!in_nullhost(vifp->v_lcl_addr)) {
964 V_multicast_register_if = NULL;
971 vifp->v_flags = vifcp->vifc_flags;
972 vifp->v_threshold = vifcp->vifc_threshold;
973 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
974 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
976 /* initialize per vif pkt counters */
979 vifp->v_bytes_in = 0;
980 vifp->v_bytes_out = 0;
981 sprintf(vifp->v_spin_name, "BM[%d] spin", vifcp->vifc_vifi);
982 mtx_init(&vifp->v_spin, vifp->v_spin_name, NULL, MTX_SPIN);
984 /* Adjust numvifs up if the vifi is higher than numvifs */
985 if (V_numvifs <= vifcp->vifc_vifi)
986 V_numvifs = vifcp->vifc_vifi + 1;
990 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
991 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
992 (int)vifcp->vifc_threshold);
998 * Delete a vif from the vif table
1001 del_vif_locked(vifi_t vifi, struct ifnet **ifp_free)
1009 if (vifi >= V_numvifs) {
1012 vifp = &V_viftable[vifi];
1013 if (in_nullhost(vifp->v_lcl_addr)) {
1014 return EADDRNOTAVAIL;
1017 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
1018 if_allmulti(vifp->v_ifp, 0);
1020 if (vifp->v_flags & VIFF_REGISTER) {
1021 V_reg_vif_num = VIFI_INVALID;
1023 if (vifp->v_ifp == V_multicast_register_if)
1024 V_multicast_register_if = NULL;
1025 *ifp_free = vifp->v_ifp;
1029 mtx_destroy(&vifp->v_spin);
1031 bzero((caddr_t)vifp, sizeof (*vifp));
1033 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
1035 /* Adjust numvifs down */
1036 for (vifi = V_numvifs; vifi > 0; vifi--)
1037 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
1045 del_vif(vifi_t vifi)
1048 struct ifnet *free_ptr;
1051 cc = del_vif_locked(vifi, &free_ptr);
1061 * update an mfc entry without resetting counters and S,G addresses.
1064 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1068 rt->mfc_parent = mfccp->mfcc_parent;
1069 for (i = 0; i < V_numvifs; i++) {
1070 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1071 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1074 /* set the RP address */
1075 if (V_mrt_api_config & MRT_MFC_RP)
1076 rt->mfc_rp = mfccp->mfcc_rp;
1078 rt->mfc_rp.s_addr = INADDR_ANY;
1082 * fully initialize an mfc entry from the parameter.
1085 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1087 rt->mfc_origin = mfccp->mfcc_origin;
1088 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1090 update_mfc_params(rt, mfccp);
1092 /* initialize pkt counters per src-grp */
1093 rt->mfc_pkt_cnt = 0;
1094 rt->mfc_byte_cnt = 0;
1095 rt->mfc_wrong_if = 0;
1096 timevalclear(&rt->mfc_last_assert);
1100 expire_mfc(struct mfc *rt)
1106 free_bw_list(rt->mfc_bw_meter_leq);
1107 free_bw_list(rt->mfc_bw_meter_geq);
1109 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1110 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1113 free(rte, M_MRTABLE);
1116 buf_ring_free(rt->mfc_stall_ring, M_MRTABLE);
1118 LIST_REMOVE(rt, mfc_hash);
1119 free(rt, M_MRTABLE);
1126 add_mfc(struct mfcctl2 *mfccp)
1132 struct epoch_tracker et;
1135 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1137 /* If an entry already exists, just update the fields */
1139 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1140 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1141 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1142 mfccp->mfcc_parent);
1143 update_mfc_params(rt, mfccp);
1149 * Find the entry for which the upcall was made and update
1152 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1153 NET_EPOCH_ENTER(et);
1154 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1155 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1156 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1157 !buf_ring_empty(rt->mfc_stall_ring)) {
1159 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1160 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1161 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1163 rt->mfc_stall_ring);
1165 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1167 init_mfc_params(rt, mfccp);
1168 rt->mfc_expire = 0; /* Don't clean this guy up */
1171 /* Free queued packets, but attempt to forward them first. */
1172 while (!buf_ring_empty(rt->mfc_stall_ring)) {
1173 rte = buf_ring_dequeue_mc(rt->mfc_stall_ring);
1174 if (rte->ifp != NULL)
1175 ip_mdq(rte->m, rte->ifp, rt, -1);
1177 free(rte, M_MRTABLE);
1184 * It is possible that an entry is being inserted without an upcall
1187 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1188 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1189 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1190 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1191 init_mfc_params(rt, mfccp);
1199 if (rt == NULL) { /* no upcall, so make a new entry */
1206 init_mfc_params(rt, mfccp);
1209 rt->mfc_bw_meter_leq = NULL;
1210 rt->mfc_bw_meter_geq = NULL;
1212 /* insert new entry at head of hash chain */
1213 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1223 * Delete an mfc entry
1226 del_mfc(struct mfcctl2 *mfccp)
1228 struct in_addr origin;
1229 struct in_addr mcastgrp;
1232 origin = mfccp->mfcc_origin;
1233 mcastgrp = mfccp->mfcc_mcastgrp;
1235 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1236 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1240 rt = mfc_find(&origin, &mcastgrp);
1243 return EADDRNOTAVAIL;
1247 * free the bw_meter entries
1249 free_bw_list(rt->mfc_bw_meter_leq);
1250 rt->mfc_bw_meter_leq = NULL;
1251 free_bw_list(rt->mfc_bw_meter_geq);
1252 rt->mfc_bw_meter_geq = NULL;
1254 LIST_REMOVE(rt, mfc_hash);
1255 free(rt, M_MRTABLE);
1263 * Send a message to the routing daemon on the multicast routing socket.
1266 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1269 SOCKBUF_LOCK(&s->so_rcv);
1270 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1272 sorwakeup_locked(s);
1275 soroverflow_locked(s);
1282 * IP multicast forwarding function. This function assumes that the packet
1283 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1284 * pointed to by "ifp", and the packet is to be relayed to other networks
1285 * that have members of the packet's destination IP multicast group.
1287 * The packet is returned unscathed to the caller, unless it is
1288 * erroneous, in which case a non-zero return value tells the caller to
1292 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1295 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1296 struct ip_moptions *imo)
1306 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1307 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1309 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1310 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) {
1312 * Packet arrived via a physical interface or
1313 * an encapsulated tunnel or a register_vif.
1317 * Packet arrived through a source-route tunnel.
1318 * Source-route tunnels are no longer supported.
1324 * BEGIN: MCAST ROUTING HOT PATH
1327 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1328 if (ip->ip_ttl < MAXTTL)
1329 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1330 error = ip_mdq(m, ifp, NULL, vifi);
1336 * Don't forward a packet with time-to-live of zero or one,
1337 * or a packet destined to a local-only group.
1339 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1346 * Determine forwarding vifs from the forwarding cache table
1348 MRTSTAT_INC(mrts_mfc_lookups);
1349 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1351 /* Entry exists, so forward if necessary */
1353 error = ip_mdq(m, ifp, rt, -1);
1354 /* Generic unlock here as we might release R or W lock */
1360 * END: MCAST ROUTING HOT PATH
1363 /* Further processing must be done with WLOCK taken */
1364 if ((MRW_WOWNED() == 0) && (MRW_LOCK_TRY_UPGRADE() == 0)) {
1367 goto mfc_find_retry;
1371 * If we don't have a route for packet's origin,
1372 * Make a copy of the packet & send message to routing daemon
1374 hlen = ip->ip_hl << 2;
1376 MRTSTAT_INC(mrts_mfc_misses);
1377 MRTSTAT_INC(mrts_no_route);
1378 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1379 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1382 * Allocate mbufs early so that we don't do extra work if we are
1383 * just going to fail anyway. Make sure to pullup the header so
1384 * that other people can't step on it.
1386 rte = malloc((sizeof *rte), M_MRTABLE, M_NOWAIT|M_ZERO);
1392 mb0 = m_copypacket(m, M_NOWAIT);
1393 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1394 mb0 = m_pullup(mb0, hlen);
1396 free(rte, M_MRTABLE);
1401 /* is there an upcall waiting for this flow ? */
1402 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1403 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash)
1405 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1406 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1407 !buf_ring_empty(rt->mfc_stall_ring))
1414 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1418 * Locate the vifi for the incoming interface for this packet.
1419 * If none found, drop packet.
1421 for (vifi = 0; vifi < V_numvifs &&
1422 V_viftable[vifi].v_ifp != ifp; vifi++)
1424 if (vifi >= V_numvifs) /* vif not found, drop packet */
1427 /* no upcall, so make a new entry */
1432 /* Make a copy of the header to send to the user level process */
1433 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1438 * Send message to routing daemon to install
1439 * a route into the kernel table
1442 im = mtod(mm, struct igmpmsg*);
1443 im->im_msgtype = IGMPMSG_NOCACHE;
1447 MRTSTAT_INC(mrts_upcalls);
1449 k_igmpsrc.sin_addr = ip->ip_src;
1450 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1451 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1452 MRTSTAT_INC(mrts_upq_sockfull);
1453 fail1: free(rt, M_MRTABLE);
1454 fail: free(rte, M_MRTABLE);
1460 /* insert new entry at head of hash chain */
1461 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1462 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1463 rt->mfc_expire = UPCALL_EXPIRE;
1465 for (i = 0; i < V_numvifs; i++) {
1466 rt->mfc_ttls[i] = 0;
1467 rt->mfc_flags[i] = 0;
1469 rt->mfc_parent = -1;
1471 /* clear the RP address */
1472 rt->mfc_rp.s_addr = INADDR_ANY;
1473 rt->mfc_bw_meter_leq = NULL;
1474 rt->mfc_bw_meter_geq = NULL;
1476 /* initialize pkt counters per src-grp */
1477 rt->mfc_pkt_cnt = 0;
1478 rt->mfc_byte_cnt = 0;
1479 rt->mfc_wrong_if = 0;
1480 timevalclear(&rt->mfc_last_assert);
1482 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1484 /* Add RT to hashtable as it didn't exist before */
1485 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1487 /* determine if queue has overflowed */
1488 if (buf_ring_full(rt->mfc_stall_ring)) {
1489 MRTSTAT_INC(mrts_upq_ovflw);
1490 non_fatal: free(rte, M_MRTABLE);
1496 buf_ring_enqueue(rt->mfc_stall_ring, rte);
1508 * Clean up the cache entry if upcall is not serviced
1511 expire_upcalls(void *arg)
1515 CURVNET_SET((struct vnet *) arg);
1517 /*This callout is always run with MRW_WLOCK taken. */
1519 for (i = 0; i < mfchashsize; i++) {
1520 struct mfc *rt, *nrt;
1522 if (V_nexpire[i] == 0)
1525 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1526 if (buf_ring_empty(rt->mfc_stall_ring))
1529 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1532 MRTSTAT_INC(mrts_cache_cleanups);
1533 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1534 (u_long)ntohl(rt->mfc_origin.s_addr),
1535 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1541 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1548 * Packet forwarding routine once entry in the cache is made
1551 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1553 struct ip *ip = mtod(m, struct ip *);
1555 int plen = ntohs(ip->ip_len);
1561 * If xmt_vif is not -1, send on only the requested vif.
1563 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1565 if (xmt_vif < V_numvifs) {
1566 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1567 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1569 phyint_send(ip, V_viftable + xmt_vif, m);
1574 * Don't forward if it didn't arrive from the parent vif for its origin.
1576 vifi = rt->mfc_parent;
1577 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1578 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1579 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1580 MRTSTAT_INC(mrts_wrong_if);
1583 * If we are doing PIM assert processing, send a message
1584 * to the routing daemon.
1586 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1587 * can complete the SPT switch, regardless of the type
1588 * of the iif (broadcast media, GRE tunnel, etc).
1590 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1591 V_viftable[vifi].v_ifp) {
1592 if (ifp == V_multicast_register_if)
1593 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1595 /* Get vifi for the incoming packet */
1596 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; vifi++)
1598 if (vifi >= V_numvifs)
1599 return 0; /* The iif is not found: ignore the packet. */
1601 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1602 return 0; /* WRONGVIF disabled: ignore the packet */
1604 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1605 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1607 int hlen = ip->ip_hl << 2;
1608 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1610 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1611 mm = m_pullup(mm, hlen);
1615 im = mtod(mm, struct igmpmsg *);
1616 im->im_msgtype = IGMPMSG_WRONGVIF;
1620 MRTSTAT_INC(mrts_upcalls);
1622 k_igmpsrc.sin_addr = im->im_src;
1623 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1624 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1625 MRTSTAT_INC(mrts_upq_sockfull);
1633 /* If I sourced this packet, it counts as output, else it was input. */
1634 mtx_lock_spin(&V_viftable[vifi].v_spin);
1635 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1636 V_viftable[vifi].v_pkt_out++;
1637 V_viftable[vifi].v_bytes_out += plen;
1639 V_viftable[vifi].v_pkt_in++;
1640 V_viftable[vifi].v_bytes_in += plen;
1642 mtx_unlock_spin(&V_viftable[vifi].v_spin);
1645 rt->mfc_byte_cnt += plen;
1648 * For each vif, decide if a copy of the packet should be forwarded.
1650 * - the ttl exceeds the vif's threshold
1651 * - there are group members downstream on interface
1653 for (vifi = 0; vifi < V_numvifs; vifi++)
1654 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1655 V_viftable[vifi].v_pkt_out++;
1656 V_viftable[vifi].v_bytes_out += plen;
1657 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1658 pim_register_send(ip, V_viftable + vifi, m, rt);
1660 phyint_send(ip, V_viftable + vifi, m);
1664 * Perform upcall-related bw measuring.
1666 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1671 /* Process meters for Greater-or-EQual case */
1672 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1673 bw_meter_geq_receive_packet(x, plen, &now);
1675 /* Process meters for Lower-or-EQual case */
1676 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1678 * Record that a packet is received.
1679 * Spin lock has to be taken as callout context
1680 * (expire_bw_meter_leq) might modify these fields
1683 mtx_lock_spin(&x->bm_spin);
1684 x->bm_measured.b_packets++;
1685 x->bm_measured.b_bytes += plen;
1686 mtx_unlock_spin(&x->bm_spin);
1694 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1697 X_legal_vif_num(int vif)
1706 if (vif < V_numvifs)
1714 * Return the local address used by this vif
1717 X_ip_mcast_src(int vifi)
1726 if (vifi < V_numvifs)
1727 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1734 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1736 struct mbuf *mb_copy;
1737 int hlen = ip->ip_hl << 2;
1742 * Make a new reference to the packet; make sure that
1743 * the IP header is actually copied, not just referenced,
1744 * so that ip_output() only scribbles on the copy.
1746 mb_copy = m_copypacket(m, M_NOWAIT);
1747 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1748 mb_copy = m_pullup(mb_copy, hlen);
1749 if (mb_copy == NULL)
1752 send_packet(vifp, mb_copy);
1756 send_packet(struct vif *vifp, struct mbuf *m)
1758 struct ip_moptions imo;
1764 imo.imo_multicast_ifp = vifp->v_ifp;
1765 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1766 imo.imo_multicast_loop = !!in_mcast_loop;
1767 imo.imo_multicast_vif = -1;
1768 STAILQ_INIT(&imo.imo_head);
1771 * Re-entrancy should not be a problem here, because
1772 * the packets that we send out and are looped back at us
1773 * should get rejected because they appear to come from
1774 * the loopback interface, thus preventing looping.
1776 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1777 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1778 (ptrdiff_t)(vifp - V_viftable), error);
1782 * Stubs for old RSVP socket shim implementation.
1786 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1789 return (EOPNOTSUPP);
1793 X_ip_rsvp_force_done(struct socket *so __unused)
1799 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1807 return (IPPROTO_DONE);
1811 * Code for bandwidth monitors
1815 * Define common interface for timeval-related methods
1817 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1818 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1819 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1822 compute_bw_meter_flags(struct bw_upcall *req)
1826 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1827 flags |= BW_METER_UNIT_PACKETS;
1828 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1829 flags |= BW_METER_UNIT_BYTES;
1830 if (req->bu_flags & BW_UPCALL_GEQ)
1831 flags |= BW_METER_GEQ;
1832 if (req->bu_flags & BW_UPCALL_LEQ)
1833 flags |= BW_METER_LEQ;
1839 expire_bw_meter_leq(void *arg)
1841 struct bw_meter *x = arg;
1845 * callout is always executed with MRW_WLOCK taken
1848 CURVNET_SET((struct vnet *)x->arg);
1853 * Test if we should deliver an upcall
1855 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1856 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1857 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1858 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1859 /* Prepare an upcall for delivery */
1860 bw_meter_prepare_upcall(x, &now);
1863 /* Send all upcalls that are pending delivery */
1864 taskqueue_enqueue(V_task_queue, &V_task);
1866 /* Reset counters */
1867 x->bm_start_time = now;
1868 /* Spin lock has to be taken as ip_forward context
1869 * might modify these fields as well
1871 mtx_lock_spin(&x->bm_spin);
1872 x->bm_measured.b_bytes = 0;
1873 x->bm_measured.b_packets = 0;
1874 mtx_unlock_spin(&x->bm_spin);
1876 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1882 * Add a bw_meter entry
1885 add_bw_upcall(struct bw_upcall *req)
1888 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1889 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1891 struct bw_meter *x, **bwm_ptr;
1894 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1897 /* Test if the flags are valid */
1898 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1900 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1902 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1905 /* Test if the threshold time interval is valid */
1906 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1909 flags = compute_bw_meter_flags(req);
1912 * Find if we have already same bw_meter entry
1915 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1918 return EADDRNOTAVAIL;
1921 /* Choose an appropriate bw_meter list */
1922 if (req->bu_flags & BW_UPCALL_GEQ)
1923 bwm_ptr = &mfc->mfc_bw_meter_geq;
1925 bwm_ptr = &mfc->mfc_bw_meter_leq;
1927 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1928 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1929 &req->bu_threshold.b_time, ==))
1930 && (x->bm_threshold.b_packets
1931 == req->bu_threshold.b_packets)
1932 && (x->bm_threshold.b_bytes
1933 == req->bu_threshold.b_bytes)
1934 && (x->bm_flags & BW_METER_USER_FLAGS)
1937 return 0; /* XXX Already installed */
1941 /* Allocate the new bw_meter entry */
1942 x = malloc(sizeof(*x), M_BWMETER, M_ZERO | M_NOWAIT);
1948 /* Set the new bw_meter entry */
1949 x->bm_threshold.b_time = req->bu_threshold.b_time;
1951 x->bm_start_time = now;
1952 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1953 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1954 x->bm_measured.b_packets = 0;
1955 x->bm_measured.b_bytes = 0;
1956 x->bm_flags = flags;
1957 x->bm_time_next = NULL;
1960 sprintf(x->bm_spin_name, "BM spin %p", x);
1961 mtx_init(&x->bm_spin, x->bm_spin_name, NULL, MTX_SPIN);
1963 /* For LEQ case create periodic callout */
1964 if (req->bu_flags & BW_UPCALL_LEQ) {
1965 callout_init_rw(&x->bm_meter_callout, &mrouter_lock, CALLOUT_SHAREDLOCK);
1966 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1967 expire_bw_meter_leq, x);
1970 /* Add the new bw_meter entry to the front of entries for this MFC */
1971 x->bm_mfc_next = *bwm_ptr;
1980 free_bw_list(struct bw_meter *list)
1982 while (list != NULL) {
1983 struct bw_meter *x = list;
1985 /* MRW_WLOCK must be held here */
1986 if (x->bm_flags & BW_METER_LEQ) {
1987 callout_drain(&x->bm_meter_callout);
1988 mtx_destroy(&x->bm_spin);
1991 list = list->bm_mfc_next;
1997 * Delete one or multiple bw_meter entries
2000 del_bw_upcall(struct bw_upcall *req)
2003 struct bw_meter *x, **bwm_ptr;
2005 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
2010 /* Find the corresponding MFC entry */
2011 mfc = mfc_find(&req->bu_src, &req->bu_dst);
2014 return EADDRNOTAVAIL;
2015 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2017 * Delete all bw_meter entries for this mfc
2019 struct bw_meter *list;
2022 list = mfc->mfc_bw_meter_leq;
2023 mfc->mfc_bw_meter_leq = NULL;
2027 list = mfc->mfc_bw_meter_geq;
2028 mfc->mfc_bw_meter_geq = NULL;
2032 } else { /* Delete a single bw_meter entry */
2033 struct bw_meter *prev;
2036 flags = compute_bw_meter_flags(req);
2038 /* Choose an appropriate bw_meter list */
2039 if (req->bu_flags & BW_UPCALL_GEQ)
2040 bwm_ptr = &mfc->mfc_bw_meter_geq;
2042 bwm_ptr = &mfc->mfc_bw_meter_leq;
2044 /* Find the bw_meter entry to delete */
2045 for (prev = NULL, x = *bwm_ptr; x != NULL;
2046 prev = x, x = x->bm_mfc_next) {
2047 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, &req->bu_threshold.b_time, ==)) &&
2048 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2049 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2050 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2053 if (x != NULL) { /* Delete entry from the list for this MFC */
2055 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2057 *bwm_ptr = x->bm_mfc_next;/* new head of list */
2059 if (req->bu_flags & BW_UPCALL_LEQ)
2060 callout_stop(&x->bm_meter_callout);
2063 /* Free the bw_meter entry */
2071 __assert_unreachable();
2075 * Perform bandwidth measurement processing that may result in an upcall
2078 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2080 struct timeval delta;
2085 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2088 * Processing for ">=" type of bw_meter entry.
2089 * bm_spin does not have to be hold here as in GEQ
2090 * case this is the only context accessing bm_measured.
2092 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2093 /* Reset the bw_meter entry */
2094 x->bm_start_time = *nowp;
2095 x->bm_measured.b_packets = 0;
2096 x->bm_measured.b_bytes = 0;
2097 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2100 /* Record that a packet is received */
2101 x->bm_measured.b_packets++;
2102 x->bm_measured.b_bytes += plen;
2105 * Test if we should deliver an upcall
2107 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2108 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2109 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2110 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2111 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2112 /* Prepare an upcall for delivery */
2113 bw_meter_prepare_upcall(x, nowp);
2114 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2120 * Prepare a bandwidth-related upcall
2123 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2125 struct timeval delta;
2126 struct bw_upcall *u;
2131 * Compute the measured time interval
2134 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2137 * Set the bw_upcall entry
2139 u = malloc(sizeof(struct bw_upcall), M_MRTABLE, M_NOWAIT | M_ZERO);
2141 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot allocate entry\n");
2144 u->bu_src = x->bm_mfc->mfc_origin;
2145 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2146 u->bu_threshold.b_time = x->bm_threshold.b_time;
2147 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2148 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2149 u->bu_measured.b_time = delta;
2150 u->bu_measured.b_packets = x->bm_measured.b_packets;
2151 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2153 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2154 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2155 if (x->bm_flags & BW_METER_UNIT_BYTES)
2156 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2157 if (x->bm_flags & BW_METER_GEQ)
2158 u->bu_flags |= BW_UPCALL_GEQ;
2159 if (x->bm_flags & BW_METER_LEQ)
2160 u->bu_flags |= BW_UPCALL_LEQ;
2162 if (buf_ring_enqueue(V_bw_upcalls_ring, u))
2163 log(LOG_WARNING, "bw_meter_prepare_upcall: cannot enqueue upcall\n");
2164 if (buf_ring_count(V_bw_upcalls_ring) > (BW_UPCALLS_MAX / 2)) {
2165 taskqueue_enqueue(V_task_queue, &V_task);
2169 * Send the pending bandwidth-related upcalls
2172 bw_upcalls_send(void)
2176 struct bw_upcall *bu;
2177 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2178 static struct igmpmsg igmpmsg = {
2181 IGMPMSG_BW_UPCALL,/* im_msgtype */
2191 if (buf_ring_empty(V_bw_upcalls_ring))
2195 * Allocate a new mbuf, initialize it with the header and
2196 * the payload for the pending calls.
2198 m = m_gethdr(M_NOWAIT, MT_DATA);
2200 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2204 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2205 len += sizeof(struct igmpmsg);
2206 while ((bu = buf_ring_dequeue_mc(V_bw_upcalls_ring)) != NULL) {
2207 m_copyback(m, len, sizeof(struct bw_upcall), (caddr_t)bu);
2208 len += sizeof(struct bw_upcall);
2209 free(bu, M_MRTABLE);
2214 * XXX do we need to set the address in k_igmpsrc ?
2216 MRTSTAT_INC(mrts_upcalls);
2217 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2218 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2219 MRTSTAT_INC(mrts_upq_sockfull);
2224 * A periodic function for sending all upcalls that are pending delivery
2227 expire_bw_upcalls_send(void *arg)
2229 CURVNET_SET((struct vnet *) arg);
2231 /* This callout is run with MRW_RLOCK taken */
2235 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2241 * End of bandwidth monitoring code
2245 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2249 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2252 struct mbuf *mb_copy, *mm;
2255 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2256 * rendezvous point was unspecified, and we were told not to.
2258 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2259 in_nullhost(rt->mfc_rp))
2262 mb_copy = pim_register_prepare(ip, m);
2263 if (mb_copy == NULL)
2267 * Send all the fragments. Note that the mbuf for each fragment
2268 * is freed by the sending machinery.
2270 for (mm = mb_copy; mm; mm = mb_copy) {
2271 mb_copy = mm->m_nextpkt;
2273 mm = m_pullup(mm, sizeof(struct ip));
2275 ip = mtod(mm, struct ip *);
2276 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2277 pim_register_send_rp(ip, vifp, mm, rt);
2279 pim_register_send_upcall(ip, vifp, mm, rt);
2288 * Return a copy of the data packet that is ready for PIM Register
2290 * XXX: Note that in the returned copy the IP header is a valid one.
2292 static struct mbuf *
2293 pim_register_prepare(struct ip *ip, struct mbuf *m)
2295 struct mbuf *mb_copy = NULL;
2298 /* Take care of delayed checksums */
2299 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2300 in_delayed_cksum(m);
2301 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2305 * Copy the old packet & pullup its IP header into the
2306 * new mbuf so we can modify it.
2308 mb_copy = m_copypacket(m, M_NOWAIT);
2309 if (mb_copy == NULL)
2311 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2312 if (mb_copy == NULL)
2315 /* take care of the TTL */
2316 ip = mtod(mb_copy, struct ip *);
2319 /* Compute the MTU after the PIM Register encapsulation */
2320 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2322 if (ntohs(ip->ip_len) <= mtu) {
2323 /* Turn the IP header into a valid one */
2325 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2327 /* Fragment the packet */
2328 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2329 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2338 * Send an upcall with the data packet to the user-level process.
2341 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2342 struct mbuf *mb_copy, struct mfc *rt)
2344 struct mbuf *mb_first;
2345 int len = ntohs(ip->ip_len);
2347 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2352 * Add a new mbuf with an upcall header
2354 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2355 if (mb_first == NULL) {
2359 mb_first->m_data += max_linkhdr;
2360 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2361 mb_first->m_len = sizeof(struct igmpmsg);
2362 mb_first->m_next = mb_copy;
2364 /* Send message to routing daemon */
2365 im = mtod(mb_first, struct igmpmsg *);
2366 im->im_msgtype = IGMPMSG_WHOLEPKT;
2368 im->im_vif = vifp - V_viftable;
2369 im->im_src = ip->ip_src;
2370 im->im_dst = ip->ip_dst;
2372 k_igmpsrc.sin_addr = ip->ip_src;
2374 MRTSTAT_INC(mrts_upcalls);
2376 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2377 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2378 MRTSTAT_INC(mrts_upq_sockfull);
2382 /* Keep statistics */
2383 PIMSTAT_INC(pims_snd_registers_msgs);
2384 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2390 * Encapsulate the data packet in PIM Register message and send it to the RP.
2393 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2396 struct mbuf *mb_first;
2397 struct ip *ip_outer;
2398 struct pim_encap_pimhdr *pimhdr;
2399 int len = ntohs(ip->ip_len);
2400 vifi_t vifi = rt->mfc_parent;
2404 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2406 return EADDRNOTAVAIL; /* The iif vif is invalid */
2410 * Add a new mbuf with the encapsulating header
2412 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2413 if (mb_first == NULL) {
2417 mb_first->m_data += max_linkhdr;
2418 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2419 mb_first->m_next = mb_copy;
2421 mb_first->m_pkthdr.len = len + mb_first->m_len;
2424 * Fill in the encapsulating IP and PIM header
2426 ip_outer = mtod(mb_first, struct ip *);
2427 *ip_outer = pim_encap_iphdr;
2428 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2429 sizeof(pim_encap_pimhdr));
2430 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2431 ip_outer->ip_dst = rt->mfc_rp;
2433 * Copy the inner header TOS to the outer header, and take care of the
2436 ip_outer->ip_tos = ip->ip_tos;
2437 if (ip->ip_off & htons(IP_DF))
2438 ip_outer->ip_off |= htons(IP_DF);
2439 ip_fillid(ip_outer);
2440 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2441 + sizeof(pim_encap_iphdr));
2442 *pimhdr = pim_encap_pimhdr;
2443 /* If the iif crosses a border, set the Border-bit */
2444 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2445 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2447 mb_first->m_data += sizeof(pim_encap_iphdr);
2448 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2449 mb_first->m_data -= sizeof(pim_encap_iphdr);
2451 send_packet(vifp, mb_first);
2453 /* Keep statistics */
2454 PIMSTAT_INC(pims_snd_registers_msgs);
2455 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2461 * pim_encapcheck() is called by the encap4_input() path at runtime to
2462 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2466 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2467 int proto __unused, void *arg __unused)
2470 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2471 return (8); /* claim the datagram. */
2475 * PIM-SMv2 and PIM-DM messages processing.
2476 * Receives and verifies the PIM control messages, and passes them
2477 * up to the listening socket, using rip_input().
2478 * The only message with special processing is the PIM_REGISTER message
2479 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2480 * is passed to if_simloop().
2483 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2485 struct ip *ip = mtod(m, struct ip *);
2489 int datalen = ntohs(ip->ip_len) - iphlen;
2492 /* Keep statistics */
2493 PIMSTAT_INC(pims_rcv_total_msgs);
2494 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2499 if (datalen < PIM_MINLEN) {
2500 PIMSTAT_INC(pims_rcv_tooshort);
2501 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2502 __func__, datalen, ntohl(ip->ip_src.s_addr));
2504 return (IPPROTO_DONE);
2508 * If the packet is at least as big as a REGISTER, go agead
2509 * and grab the PIM REGISTER header size, to avoid another
2510 * possible m_pullup() later.
2512 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2513 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2515 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2517 * Get the IP and PIM headers in contiguous memory, and
2518 * possibly the PIM REGISTER header.
2520 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2521 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2522 return (IPPROTO_DONE);
2525 /* m_pullup() may have given us a new mbuf so reset ip. */
2526 ip = mtod(m, struct ip *);
2527 ip_tos = ip->ip_tos;
2529 /* adjust mbuf to point to the PIM header */
2530 m->m_data += iphlen;
2532 pim = mtod(m, struct pim *);
2535 * Validate checksum. If PIM REGISTER, exclude the data packet.
2537 * XXX: some older PIMv2 implementations don't make this distinction,
2538 * so for compatibility reason perform the checksum over part of the
2539 * message, and if error, then over the whole message.
2541 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2542 /* do nothing, checksum okay */
2543 } else if (in_cksum(m, datalen)) {
2544 PIMSTAT_INC(pims_rcv_badsum);
2545 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2547 return (IPPROTO_DONE);
2550 /* PIM version check */
2551 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2552 PIMSTAT_INC(pims_rcv_badversion);
2553 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2554 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2556 return (IPPROTO_DONE);
2559 /* restore mbuf back to the outer IP */
2560 m->m_data -= iphlen;
2563 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2565 * Since this is a REGISTER, we'll make a copy of the register
2566 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2569 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2571 struct ip *encap_ip;
2576 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2578 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2579 (int)V_reg_vif_num);
2581 return (IPPROTO_DONE);
2583 /* XXX need refcnt? */
2584 vifp = V_viftable[V_reg_vif_num].v_ifp;
2590 if (datalen < PIM_REG_MINLEN) {
2591 PIMSTAT_INC(pims_rcv_tooshort);
2592 PIMSTAT_INC(pims_rcv_badregisters);
2593 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2595 return (IPPROTO_DONE);
2598 reghdr = (u_int32_t *)(pim + 1);
2599 encap_ip = (struct ip *)(reghdr + 1);
2601 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2602 __func__, ntohl(encap_ip->ip_src.s_addr),
2603 ntohs(encap_ip->ip_len));
2605 /* verify the version number of the inner packet */
2606 if (encap_ip->ip_v != IPVERSION) {
2607 PIMSTAT_INC(pims_rcv_badregisters);
2608 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2610 return (IPPROTO_DONE);
2613 /* verify the inner packet is destined to a mcast group */
2614 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2615 PIMSTAT_INC(pims_rcv_badregisters);
2616 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2617 ntohl(encap_ip->ip_dst.s_addr));
2619 return (IPPROTO_DONE);
2622 /* If a NULL_REGISTER, pass it to the daemon */
2623 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2624 goto pim_input_to_daemon;
2627 * Copy the TOS from the outer IP header to the inner IP header.
2629 if (encap_ip->ip_tos != ip_tos) {
2630 /* Outer TOS -> inner TOS */
2631 encap_ip->ip_tos = ip_tos;
2632 /* Recompute the inner header checksum. Sigh... */
2634 /* adjust mbuf to point to the inner IP header */
2635 m->m_data += (iphlen + PIM_MINLEN);
2636 m->m_len -= (iphlen + PIM_MINLEN);
2638 encap_ip->ip_sum = 0;
2639 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2641 /* restore mbuf to point back to the outer IP header */
2642 m->m_data -= (iphlen + PIM_MINLEN);
2643 m->m_len += (iphlen + PIM_MINLEN);
2647 * Decapsulate the inner IP packet and loopback to forward it
2648 * as a normal multicast packet. Also, make a copy of the
2649 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2650 * to pass to the daemon later, so it can take the appropriate
2651 * actions (e.g., send back PIM_REGISTER_STOP).
2652 * XXX: here m->m_data points to the outer IP header.
2654 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2656 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2658 return (IPPROTO_DONE);
2661 /* Keep statistics */
2662 /* XXX: registers_bytes include only the encap. mcast pkt */
2663 PIMSTAT_INC(pims_rcv_registers_msgs);
2664 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2667 * forward the inner ip packet; point m_data at the inner ip.
2669 m_adj(m, iphlen + PIM_MINLEN);
2672 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2674 (u_long)ntohl(encap_ip->ip_src.s_addr),
2675 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2676 (int)V_reg_vif_num);
2678 /* NB: vifp was collected above; can it change on us? */
2679 if_simloop(vifp, m, dst.sin_family, 0);
2681 /* prepare the register head to send to the mrouting daemon */
2685 pim_input_to_daemon:
2687 * Pass the PIM message up to the daemon; if it is a Register message,
2688 * pass the 'head' only up to the daemon. This includes the
2689 * outer IP header, PIM header, PIM-Register header and the
2691 * XXX: the outer IP header pkt size of a Register is not adjust to
2692 * reflect the fact that the inner multicast data is truncated.
2694 return (rip_input(&m, &off, proto));
2698 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2705 if (V_mfchashtbl == NULL) /* XXX unlocked */
2707 error = sysctl_wire_old_buffer(req, 0);
2712 for (i = 0; i < mfchashsize; i++) {
2713 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2714 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2724 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2725 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2726 "IPv4 Multicast Forwarding Table "
2727 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2730 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2736 if (V_viftable == NULL) /* XXX unlocked */
2738 error = sysctl_wire_old_buffer(req, MROUTE_VIF_SYSCTL_LEN * MAXVIFS);
2743 /* Copy out user-visible portion of vif entry. */
2744 for (i = 0; i < MAXVIFS; i++) {
2745 error = SYSCTL_OUT(req, &V_viftable[i], MROUTE_VIF_SYSCTL_LEN);
2753 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2754 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2755 sysctl_viflist, "S,vif[MAXVIFS]",
2756 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2759 vnet_mroute_init(const void *unused __unused)
2762 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2764 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2765 M_MRTABLE, M_WAITOK|M_ZERO);
2767 callout_init_rw(&V_expire_upcalls_ch, &mrouter_lock, 0);
2768 callout_init_rw(&V_bw_upcalls_ch, &mrouter_lock, 0);
2770 /* Prepare taskqueue */
2771 V_task_queue = taskqueue_create_fast("ip_mroute_tskq", M_NOWAIT,
2772 taskqueue_thread_enqueue, &V_task_queue);
2773 taskqueue_start_threads(&V_task_queue, 1, PI_NET, "ip_mroute_tskq task");
2776 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2780 vnet_mroute_uninit(const void *unused __unused)
2783 /* Taskqueue should be cancelled and drained before freeing */
2784 taskqueue_free(V_task_queue);
2786 free(V_viftable, M_MRTABLE);
2787 free(V_nexpire, M_MRTABLE);
2791 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2792 vnet_mroute_uninit, NULL);
2795 ip_mroute_modevent(module_t mod, int type, void *unused)
2802 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2803 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2804 if (if_detach_event_tag == NULL) {
2805 printf("ip_mroute: unable to register "
2806 "ifnet_departure_event handler\n");
2811 mfchashsize = MFCHASHSIZE;
2812 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2813 !powerof2(mfchashsize)) {
2814 printf("WARNING: %s not a power of 2; using default\n",
2815 "net.inet.ip.mfchashsize");
2816 mfchashsize = MFCHASHSIZE;
2819 pim_squelch_wholepkt = 0;
2820 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2821 &pim_squelch_wholepkt);
2823 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2825 ip_mcast_src = X_ip_mcast_src;
2826 ip_mforward = X_ip_mforward;
2827 ip_mrouter_done = X_ip_mrouter_done;
2828 ip_mrouter_get = X_ip_mrouter_get;
2829 ip_mrouter_set = X_ip_mrouter_set;
2831 ip_rsvp_force_done = X_ip_rsvp_force_done;
2832 ip_rsvp_vif = X_ip_rsvp_vif;
2834 legal_vif_num = X_legal_vif_num;
2835 mrt_ioctl = X_mrt_ioctl;
2836 rsvp_input_p = X_rsvp_input;
2841 * Typically module unload happens after the user-level
2842 * process has shutdown the kernel services (the check
2843 * below insures someone can't just yank the module out
2844 * from under a running process). But if the module is
2845 * just loaded and then unloaded w/o starting up a user
2846 * process we still need to cleanup.
2849 if (ip_mrouter_cnt != 0) {
2853 ip_mrouter_unloading = 1;
2856 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2858 if (pim_encap_cookie) {
2859 ip_encap_detach(pim_encap_cookie);
2860 pim_encap_cookie = NULL;
2863 ip_mcast_src = NULL;
2865 ip_mrouter_done = NULL;
2866 ip_mrouter_get = NULL;
2867 ip_mrouter_set = NULL;
2869 ip_rsvp_force_done = NULL;
2872 legal_vif_num = NULL;
2874 rsvp_input_p = NULL;
2885 static moduledata_t ip_mroutemod = {
2891 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);