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/param.h>
81 #include <sys/kernel.h>
82 #include <sys/stddef.h>
83 #include <sys/eventhandler.h>
86 #include <sys/malloc.h>
88 #include <sys/module.h>
90 #include <sys/protosw.h>
91 #include <sys/signalvar.h>
92 #include <sys/socket.h>
93 #include <sys/socketvar.h>
94 #include <sys/sockio.h>
96 #include <sys/sysctl.h>
97 #include <sys/syslog.h>
98 #include <sys/systm.h>
100 #include <sys/counter.h>
103 #include <net/if_var.h>
104 #include <net/netisr.h>
105 #include <net/route.h>
106 #include <net/vnet.h>
108 #include <netinet/in.h>
109 #include <netinet/igmp.h>
110 #include <netinet/in_systm.h>
111 #include <netinet/in_var.h>
112 #include <netinet/ip.h>
113 #include <netinet/ip_encap.h>
114 #include <netinet/ip_mroute.h>
115 #include <netinet/ip_var.h>
116 #include <netinet/ip_options.h>
117 #include <netinet/pim.h>
118 #include <netinet/pim_var.h>
119 #include <netinet/udp.h>
121 #include <machine/in_cksum.h>
124 #define KTR_IPMF KTR_INET
127 #define VIFI_INVALID ((vifi_t) -1)
129 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
132 * Locking. We use two locks: one for the virtual interface table and
133 * one for the forwarding table. These locks may be nested in which case
134 * the VIF lock must always be taken first. Note that each lock is used
135 * to cover not only the specific data structure but also related data
139 static struct mtx mrouter_mtx;
140 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
141 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
142 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
143 #define MROUTER_LOCK_INIT() \
144 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
145 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
147 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
148 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
150 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
151 VNET_PCPUSTAT_SYSINIT(mrtstat);
152 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
153 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
154 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
155 "netinet/ip_mroute.h)");
157 VNET_DEFINE_STATIC(u_long, mfchash);
158 #define V_mfchash VNET(mfchash)
159 #define MFCHASH(a, g) \
160 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
161 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
162 #define MFCHASHSIZE 256
164 static u_long mfchashsize; /* Hash size */
165 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */
166 #define V_nexpire VNET(nexpire)
167 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
168 #define V_mfchashtbl VNET(mfchashtbl)
170 static struct mtx mfc_mtx;
171 #define MFC_LOCK() mtx_lock(&mfc_mtx)
172 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
173 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
174 #define MFC_LOCK_INIT() \
175 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
176 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
178 VNET_DEFINE_STATIC(vifi_t, numvifs);
179 #define V_numvifs VNET(numvifs)
180 VNET_DEFINE_STATIC(struct vif *, viftable);
181 #define V_viftable VNET(viftable)
183 static struct mtx vif_mtx;
184 #define VIF_LOCK() mtx_lock(&vif_mtx)
185 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
186 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
187 #define VIF_LOCK_INIT() \
188 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
189 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
191 static eventhandler_tag if_detach_event_tag = NULL;
193 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
194 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
196 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
197 #define UPCALL_EXPIRE 6 /* number of timeouts */
200 * Bandwidth meter variables and constants
202 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
205 * Pending upcalls are stored in a vector which is flushed when
206 * full, or periodically
208 VNET_DEFINE_STATIC(struct bw_upcall *, bw_upcalls);
209 #define V_bw_upcalls VNET(bw_upcalls)
210 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
211 #define V_bw_upcalls_n VNET(bw_upcalls_n)
212 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
213 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
215 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
217 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
218 VNET_PCPUSTAT_SYSINIT(pimstat);
219 VNET_PCPUSTAT_SYSUNINIT(pimstat);
221 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW | CTLFLAG_MPSAFE, 0,
223 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
224 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
226 static u_long pim_squelch_wholepkt = 0;
227 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
228 &pim_squelch_wholepkt, 0,
229 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
231 static const struct encaptab *pim_encap_cookie;
232 static int pim_encapcheck(const struct mbuf *, int, int, void *);
233 static int pim_input(struct mbuf *, int, int, void *);
235 extern int in_mcast_loop;
237 static const struct encap_config ipv4_encap_cfg = {
238 .proto = IPPROTO_PIM,
239 .min_length = sizeof(struct ip) + PIM_MINLEN,
241 .check = pim_encapcheck,
246 * Note: the PIM Register encapsulation adds the following in front of a
249 * struct pim_encap_hdr {
251 * struct pim_encap_pimhdr pim;
256 struct pim_encap_pimhdr {
260 #define PIM_ENCAP_TTL 64
262 static struct ip pim_encap_iphdr = {
263 #if BYTE_ORDER == LITTLE_ENDIAN
264 sizeof(struct ip) >> 2,
268 sizeof(struct ip) >> 2,
271 sizeof(struct ip), /* total length */
279 static struct pim_encap_pimhdr pim_encap_pimhdr = {
281 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
288 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
289 #define V_reg_vif_num VNET(reg_vif_num)
290 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
291 #define V_multicast_register_if VNET(multicast_register_if)
297 static u_long X_ip_mcast_src(int);
298 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
299 struct ip_moptions *);
300 static int X_ip_mrouter_done(void);
301 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
302 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
303 static int X_legal_vif_num(int);
304 static int X_mrt_ioctl(u_long, caddr_t, int);
306 static int add_bw_upcall(struct bw_upcall *);
307 static int add_mfc(struct mfcctl2 *);
308 static int add_vif(struct vifctl *);
309 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
310 static void bw_meter_geq_receive_packet(struct bw_meter *, int,
312 static void bw_upcalls_send(void);
313 static int del_bw_upcall(struct bw_upcall *);
314 static int del_mfc(struct mfcctl2 *);
315 static int del_vif(vifi_t);
316 static int del_vif_locked(vifi_t);
317 static void expire_bw_upcalls_send(void *);
318 static void expire_mfc(struct mfc *);
319 static void expire_upcalls(void *);
320 static void free_bw_list(struct bw_meter *);
321 static int get_sg_cnt(struct sioc_sg_req *);
322 static int get_vif_cnt(struct sioc_vif_req *);
323 static void if_detached_event(void *, struct ifnet *);
324 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
325 static int ip_mrouter_init(struct socket *, int);
326 static __inline struct mfc *
327 mfc_find(struct in_addr *, struct in_addr *);
328 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
330 pim_register_prepare(struct ip *, struct mbuf *);
331 static int pim_register_send(struct ip *, struct vif *,
332 struct mbuf *, struct mfc *);
333 static int pim_register_send_rp(struct ip *, struct vif *,
334 struct mbuf *, struct mfc *);
335 static int pim_register_send_upcall(struct ip *, struct vif *,
336 struct mbuf *, struct mfc *);
337 static void send_packet(struct vif *, struct mbuf *);
338 static int set_api_config(uint32_t *);
339 static int set_assert(int);
340 static int socket_send(struct socket *, struct mbuf *,
341 struct sockaddr_in *);
344 * Kernel multicast forwarding API capabilities and setup.
345 * If more API capabilities are added to the kernel, they should be
346 * recorded in `mrt_api_support'.
348 #define MRT_API_VERSION 0x0305
350 static const int mrt_api_version = MRT_API_VERSION;
351 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
352 MRT_MFC_FLAGS_BORDER_VIF |
355 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
356 #define V_mrt_api_config VNET(mrt_api_config)
357 VNET_DEFINE_STATIC(int, pim_assert_enabled);
358 #define V_pim_assert_enabled VNET(pim_assert_enabled)
359 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
362 * Find a route for a given origin IP address and multicast group address.
363 * Statistics must be updated by the caller.
365 static __inline struct mfc *
366 mfc_find(struct in_addr *o, struct in_addr *g)
372 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
373 if (in_hosteq(rt->mfc_origin, *o) &&
374 in_hosteq(rt->mfc_mcastgrp, *g) &&
375 TAILQ_EMPTY(&rt->mfc_stall))
383 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
386 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
392 struct bw_upcall bw_upcall;
395 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
399 switch (sopt->sopt_name) {
401 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
404 error = ip_mrouter_init(so, optval);
408 error = ip_mrouter_done();
412 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
415 error = add_vif(&vifc);
419 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
422 error = del_vif(vifi);
428 * select data size depending on API version.
430 if (sopt->sopt_name == MRT_ADD_MFC &&
431 V_mrt_api_config & MRT_API_FLAGS_ALL) {
432 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
433 sizeof(struct mfcctl2));
435 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
436 sizeof(struct mfcctl));
437 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
438 sizeof(mfc) - sizeof(struct mfcctl));
442 if (sopt->sopt_name == MRT_ADD_MFC)
443 error = add_mfc(&mfc);
445 error = del_mfc(&mfc);
449 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
456 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
458 error = set_api_config(&i);
460 error = sooptcopyout(sopt, &i, sizeof i);
463 case MRT_ADD_BW_UPCALL:
464 case MRT_DEL_BW_UPCALL:
465 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
469 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
470 error = add_bw_upcall(&bw_upcall);
472 error = del_bw_upcall(&bw_upcall);
483 * Handle MRT getsockopt commands
486 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
490 switch (sopt->sopt_name) {
492 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
496 error = sooptcopyout(sopt, &V_pim_assert_enabled,
497 sizeof V_pim_assert_enabled);
500 case MRT_API_SUPPORT:
501 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
505 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
516 * Handle ioctl commands to obtain information from the cache
519 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
524 * Currently the only function calling this ioctl routine is rtioctl_fib().
525 * Typically, only root can create the raw socket in order to execute
526 * this ioctl method, however the request might be coming from a prison
528 error = priv_check(curthread, PRIV_NETINET_MROUTE);
532 case (SIOCGETVIFCNT):
533 error = get_vif_cnt((struct sioc_vif_req *)data);
537 error = get_sg_cnt((struct sioc_sg_req *)data);
548 * returns the packet, byte, rpf-failure count for the source group provided
551 get_sg_cnt(struct sioc_sg_req *req)
556 rt = mfc_find(&req->src, &req->grp);
559 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
560 return EADDRNOTAVAIL;
562 req->pktcnt = rt->mfc_pkt_cnt;
563 req->bytecnt = rt->mfc_byte_cnt;
564 req->wrong_if = rt->mfc_wrong_if;
570 * returns the input and output packet and byte counts on the vif provided
573 get_vif_cnt(struct sioc_vif_req *req)
575 vifi_t vifi = req->vifi;
578 if (vifi >= V_numvifs) {
583 req->icount = V_viftable[vifi].v_pkt_in;
584 req->ocount = V_viftable[vifi].v_pkt_out;
585 req->ibytes = V_viftable[vifi].v_bytes_in;
586 req->obytes = V_viftable[vifi].v_bytes_out;
593 if_detached_event(void *arg __unused, struct ifnet *ifp)
600 if (V_ip_mrouter == NULL) {
609 * Tear down multicast forwarder state associated with this ifnet.
610 * 1. Walk the vif list, matching vifs against this ifnet.
611 * 2. Walk the multicast forwarding cache (mfc) looking for
612 * inner matches with this vif's index.
613 * 3. Expire any matching multicast forwarding cache entries.
614 * 4. Free vif state. This should disable ALLMULTI on the interface.
616 for (vifi = 0; vifi < V_numvifs; vifi++) {
617 if (V_viftable[vifi].v_ifp != ifp)
619 for (i = 0; i < mfchashsize; i++) {
620 struct mfc *rt, *nrt;
622 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
623 if (rt->mfc_parent == vifi) {
628 del_vif_locked(vifi);
638 * Enable multicast forwarding.
641 ip_mrouter_init(struct socket *so, int version)
644 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
645 so->so_type, so->so_proto->pr_protocol);
647 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
655 if (ip_mrouter_unloading) {
660 if (V_ip_mrouter != NULL) {
665 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
668 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
670 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
678 CTR1(KTR_IPMF, "%s: done", __func__);
684 * Disable multicast forwarding.
687 X_ip_mrouter_done(void)
695 if (V_ip_mrouter == NULL) {
701 * Detach/disable hooks to the reset of the system.
705 V_mrt_api_config = 0;
712 * For each phyint in use, disable promiscuous reception of all IP
715 for (vifi = 0; vifi < V_numvifs; vifi++) {
716 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
717 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
718 ifp = V_viftable[vifi].v_ifp;
722 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
724 V_pim_assert_enabled = 0;
728 callout_stop(&V_expire_upcalls_ch);
729 callout_stop(&V_bw_upcalls_ch);
734 * Free all multicast forwarding cache entries.
735 * Do not use hashdestroy(), as we must perform other cleanup.
737 for (i = 0; i < mfchashsize; i++) {
738 struct mfc *rt, *nrt;
740 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
744 free(V_mfchashtbl, M_MRTABLE);
747 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
753 V_reg_vif_num = VIFI_INVALID;
757 CTR1(KTR_IPMF, "%s: done", __func__);
763 * Set PIM assert processing global
768 if ((i != 1) && (i != 0))
771 V_pim_assert_enabled = i;
777 * Configure API capabilities
780 set_api_config(uint32_t *apival)
785 * We can set the API capabilities only if it is the first operation
786 * after MRT_INIT. I.e.:
787 * - there are no vifs installed
788 * - pim_assert is not enabled
789 * - the MFC table is empty
795 if (V_pim_assert_enabled) {
802 for (i = 0; i < mfchashsize; i++) {
803 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
812 V_mrt_api_config = *apival & mrt_api_support;
813 *apival = V_mrt_api_config;
819 * Add a vif to the vif table
822 add_vif(struct vifctl *vifcp)
824 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
825 struct sockaddr_in sin = {sizeof sin, AF_INET};
831 if (vifcp->vifc_vifi >= MAXVIFS) {
835 /* rate limiting is no longer supported by this code */
836 if (vifcp->vifc_rate_limit != 0) {
837 log(LOG_ERR, "rate limiting is no longer supported\n");
841 if (!in_nullhost(vifp->v_lcl_addr)) {
845 if (in_nullhost(vifcp->vifc_lcl_addr)) {
847 return EADDRNOTAVAIL;
850 /* Find the interface with an address in AF_INET family */
851 if (vifcp->vifc_flags & VIFF_REGISTER) {
853 * XXX: Because VIFF_REGISTER does not really need a valid
854 * local interface (e.g. it could be 127.0.0.2), we don't
859 struct epoch_tracker et;
861 sin.sin_addr = vifcp->vifc_lcl_addr;
863 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
867 return EADDRNOTAVAIL;
870 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
874 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
875 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
878 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
879 ifp = &V_multicast_register_if;
880 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
881 if (V_reg_vif_num == VIFI_INVALID) {
882 if_initname(&V_multicast_register_if, "register_vif", 0);
883 V_multicast_register_if.if_flags = IFF_LOOPBACK;
884 V_reg_vif_num = vifcp->vifc_vifi;
886 } else { /* Make sure the interface supports multicast */
887 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
892 /* Enable promiscuous reception of all IP multicasts from the if */
893 error = if_allmulti(ifp, 1);
900 vifp->v_flags = vifcp->vifc_flags;
901 vifp->v_threshold = vifcp->vifc_threshold;
902 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
903 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
905 /* initialize per vif pkt counters */
908 vifp->v_bytes_in = 0;
909 vifp->v_bytes_out = 0;
911 /* Adjust numvifs up if the vifi is higher than numvifs */
912 if (V_numvifs <= vifcp->vifc_vifi)
913 V_numvifs = vifcp->vifc_vifi + 1;
917 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
918 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
919 (int)vifcp->vifc_threshold);
925 * Delete a vif from the vif table
928 del_vif_locked(vifi_t vifi)
934 if (vifi >= V_numvifs) {
937 vifp = &V_viftable[vifi];
938 if (in_nullhost(vifp->v_lcl_addr)) {
939 return EADDRNOTAVAIL;
942 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
943 if_allmulti(vifp->v_ifp, 0);
945 if (vifp->v_flags & VIFF_REGISTER)
946 V_reg_vif_num = VIFI_INVALID;
948 bzero((caddr_t)vifp, sizeof (*vifp));
950 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
952 /* Adjust numvifs down */
953 for (vifi = V_numvifs; vifi > 0; vifi--)
954 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
967 cc = del_vif_locked(vifi);
974 * update an mfc entry without resetting counters and S,G addresses.
977 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
981 rt->mfc_parent = mfccp->mfcc_parent;
982 for (i = 0; i < V_numvifs; i++) {
983 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
984 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
987 /* set the RP address */
988 if (V_mrt_api_config & MRT_MFC_RP)
989 rt->mfc_rp = mfccp->mfcc_rp;
991 rt->mfc_rp.s_addr = INADDR_ANY;
995 * fully initialize an mfc entry from the parameter.
998 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1000 rt->mfc_origin = mfccp->mfcc_origin;
1001 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1003 update_mfc_params(rt, mfccp);
1005 /* initialize pkt counters per src-grp */
1006 rt->mfc_pkt_cnt = 0;
1007 rt->mfc_byte_cnt = 0;
1008 rt->mfc_wrong_if = 0;
1009 timevalclear(&rt->mfc_last_assert);
1013 expire_mfc(struct mfc *rt)
1015 struct rtdetq *rte, *nrte;
1019 free_bw_list(rt->mfc_bw_meter_leq);
1020 free_bw_list(rt->mfc_bw_meter_geq);
1022 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1024 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1025 free(rte, M_MRTABLE);
1028 LIST_REMOVE(rt, mfc_hash);
1029 free(rt, M_MRTABLE);
1036 add_mfc(struct mfcctl2 *mfccp)
1039 struct rtdetq *rte, *nrte;
1046 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1048 /* If an entry already exists, just update the fields */
1050 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1051 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1052 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1053 mfccp->mfcc_parent);
1054 update_mfc_params(rt, mfccp);
1061 * Find the entry for which the upcall was made and update
1064 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1065 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1066 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1067 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1068 !TAILQ_EMPTY(&rt->mfc_stall)) {
1070 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1071 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1072 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1074 TAILQ_FIRST(&rt->mfc_stall));
1076 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1078 init_mfc_params(rt, mfccp);
1079 rt->mfc_expire = 0; /* Don't clean this guy up */
1082 /* Free queued packets, but attempt to forward them first. */
1083 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1084 if (rte->ifp != NULL)
1085 ip_mdq(rte->m, rte->ifp, rt, -1);
1087 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1089 free(rte, M_MRTABLE);
1095 * It is possible that an entry is being inserted without an upcall
1098 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1099 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1100 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1101 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1102 init_mfc_params(rt, mfccp);
1110 if (rt == NULL) { /* no upcall, so make a new entry */
1111 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1118 init_mfc_params(rt, mfccp);
1119 TAILQ_INIT(&rt->mfc_stall);
1123 rt->mfc_bw_meter_leq = NULL;
1124 rt->mfc_bw_meter_geq = NULL;
1126 /* insert new entry at head of hash chain */
1127 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1138 * Delete an mfc entry
1141 del_mfc(struct mfcctl2 *mfccp)
1143 struct in_addr origin;
1144 struct in_addr mcastgrp;
1147 origin = mfccp->mfcc_origin;
1148 mcastgrp = mfccp->mfcc_mcastgrp;
1150 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1151 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1155 rt = mfc_find(&origin, &mcastgrp);
1158 return EADDRNOTAVAIL;
1162 * free the bw_meter entries
1164 free_bw_list(rt->mfc_bw_meter_leq);
1165 rt->mfc_bw_meter_leq = NULL;
1166 free_bw_list(rt->mfc_bw_meter_geq);
1167 rt->mfc_bw_meter_geq = NULL;
1169 LIST_REMOVE(rt, mfc_hash);
1170 free(rt, M_MRTABLE);
1178 * Send a message to the routing daemon on the multicast routing socket.
1181 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1184 SOCKBUF_LOCK(&s->so_rcv);
1185 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1187 sorwakeup_locked(s);
1190 SOCKBUF_UNLOCK(&s->so_rcv);
1197 * IP multicast forwarding function. This function assumes that the packet
1198 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1199 * pointed to by "ifp", and the packet is to be relayed to other networks
1200 * that have members of the packet's destination IP multicast group.
1202 * The packet is returned unscathed to the caller, unless it is
1203 * erroneous, in which case a non-zero return value tells the caller to
1207 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1210 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1211 struct ip_moptions *imo)
1217 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1218 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1220 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1221 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1223 * Packet arrived via a physical interface or
1224 * an encapsulated tunnel or a register_vif.
1228 * Packet arrived through a source-route tunnel.
1229 * Source-route tunnels are no longer supported.
1236 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1237 if (ip->ip_ttl < MAXTTL)
1238 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1239 error = ip_mdq(m, ifp, NULL, vifi);
1246 * Don't forward a packet with time-to-live of zero or one,
1247 * or a packet destined to a local-only group.
1249 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1256 * Determine forwarding vifs from the forwarding cache table
1258 MRTSTAT_INC(mrts_mfc_lookups);
1259 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1261 /* Entry exists, so forward if necessary */
1263 error = ip_mdq(m, ifp, rt, -1);
1269 * If we don't have a route for packet's origin,
1270 * Make a copy of the packet & send message to routing daemon
1276 int hlen = ip->ip_hl << 2;
1278 MRTSTAT_INC(mrts_mfc_misses);
1279 MRTSTAT_INC(mrts_no_route);
1280 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1281 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1284 * Allocate mbufs early so that we don't do extra work if we are
1285 * just going to fail anyway. Make sure to pullup the header so
1286 * that other people can't step on it.
1288 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1296 mb0 = m_copypacket(m, M_NOWAIT);
1297 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1298 mb0 = m_pullup(mb0, hlen);
1300 free(rte, M_MRTABLE);
1306 /* is there an upcall waiting for this flow ? */
1307 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1308 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1309 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1310 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1311 !TAILQ_EMPTY(&rt->mfc_stall))
1318 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1322 * Locate the vifi for the incoming interface for this packet.
1323 * If none found, drop packet.
1325 for (vifi = 0; vifi < V_numvifs &&
1326 V_viftable[vifi].v_ifp != ifp; vifi++)
1328 if (vifi >= V_numvifs) /* vif not found, drop packet */
1331 /* no upcall, so make a new entry */
1332 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1336 /* Make a copy of the header to send to the user level process */
1337 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1342 * Send message to routing daemon to install
1343 * a route into the kernel table
1346 im = mtod(mm, struct igmpmsg *);
1347 im->im_msgtype = IGMPMSG_NOCACHE;
1351 MRTSTAT_INC(mrts_upcalls);
1353 k_igmpsrc.sin_addr = ip->ip_src;
1354 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1355 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1356 MRTSTAT_INC(mrts_upq_sockfull);
1358 free(rt, M_MRTABLE);
1360 free(rte, M_MRTABLE);
1367 /* insert new entry at head of hash chain */
1368 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1369 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1370 rt->mfc_expire = UPCALL_EXPIRE;
1372 for (i = 0; i < V_numvifs; i++) {
1373 rt->mfc_ttls[i] = 0;
1374 rt->mfc_flags[i] = 0;
1376 rt->mfc_parent = -1;
1378 /* clear the RP address */
1379 rt->mfc_rp.s_addr = INADDR_ANY;
1380 rt->mfc_bw_meter_leq = NULL;
1381 rt->mfc_bw_meter_geq = NULL;
1383 /* initialize pkt counters per src-grp */
1384 rt->mfc_pkt_cnt = 0;
1385 rt->mfc_byte_cnt = 0;
1386 rt->mfc_wrong_if = 0;
1387 timevalclear(&rt->mfc_last_assert);
1389 TAILQ_INIT(&rt->mfc_stall);
1392 /* link into table */
1393 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1394 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1398 /* determine if queue has overflowed */
1399 if (rt->mfc_nstall > MAX_UPQ) {
1400 MRTSTAT_INC(mrts_upq_ovflw);
1402 free(rte, M_MRTABLE);
1408 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1423 * Clean up the cache entry if upcall is not serviced
1426 expire_upcalls(void *arg)
1430 CURVNET_SET((struct vnet *) arg);
1434 for (i = 0; i < mfchashsize; i++) {
1435 struct mfc *rt, *nrt;
1437 if (V_nexpire[i] == 0)
1440 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1441 if (TAILQ_EMPTY(&rt->mfc_stall))
1444 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1447 MRTSTAT_INC(mrts_cache_cleanups);
1448 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1449 (u_long)ntohl(rt->mfc_origin.s_addr),
1450 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1458 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1465 * Packet forwarding routine once entry in the cache is made
1468 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1470 struct ip *ip = mtod(m, struct ip *);
1472 int plen = ntohs(ip->ip_len);
1477 * If xmt_vif is not -1, send on only the requested vif.
1479 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1481 if (xmt_vif < V_numvifs) {
1482 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1483 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1485 phyint_send(ip, V_viftable + xmt_vif, m);
1490 * Don't forward if it didn't arrive from the parent vif for its origin.
1492 vifi = rt->mfc_parent;
1493 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1494 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1495 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1496 MRTSTAT_INC(mrts_wrong_if);
1499 * If we are doing PIM assert processing, send a message
1500 * to the routing daemon.
1502 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1503 * can complete the SPT switch, regardless of the type
1504 * of the iif (broadcast media, GRE tunnel, etc).
1506 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1507 V_viftable[vifi].v_ifp) {
1508 if (ifp == &V_multicast_register_if)
1509 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1511 /* Get vifi for the incoming packet */
1512 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1515 if (vifi >= V_numvifs)
1516 return 0; /* The iif is not found: ignore the packet. */
1518 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1519 return 0; /* WRONGVIF disabled: ignore the packet */
1521 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1522 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1524 int hlen = ip->ip_hl << 2;
1525 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1527 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1528 mm = m_pullup(mm, hlen);
1532 im = mtod(mm, struct igmpmsg *);
1533 im->im_msgtype = IGMPMSG_WRONGVIF;
1537 MRTSTAT_INC(mrts_upcalls);
1539 k_igmpsrc.sin_addr = im->im_src;
1540 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1541 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1542 MRTSTAT_INC(mrts_upq_sockfull);
1550 /* If I sourced this packet, it counts as output, else it was input. */
1551 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1552 V_viftable[vifi].v_pkt_out++;
1553 V_viftable[vifi].v_bytes_out += plen;
1555 V_viftable[vifi].v_pkt_in++;
1556 V_viftable[vifi].v_bytes_in += plen;
1559 rt->mfc_byte_cnt += plen;
1562 * For each vif, decide if a copy of the packet should be forwarded.
1564 * - the ttl exceeds the vif's threshold
1565 * - there are group members downstream on interface
1567 for (vifi = 0; vifi < V_numvifs; vifi++)
1568 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1569 V_viftable[vifi].v_pkt_out++;
1570 V_viftable[vifi].v_bytes_out += plen;
1571 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1572 pim_register_send(ip, V_viftable + vifi, m, rt);
1574 phyint_send(ip, V_viftable + vifi, m);
1578 * Perform upcall-related bw measuring.
1580 if ((rt->mfc_bw_meter_geq != NULL) || (rt->mfc_bw_meter_leq != NULL)) {
1586 /* Process meters for Greater-or-EQual case */
1587 for (x = rt->mfc_bw_meter_geq; x != NULL; x = x->bm_mfc_next)
1588 bw_meter_geq_receive_packet(x, plen, &now);
1590 /* Process meters for Lower-or-EQual case */
1591 for (x = rt->mfc_bw_meter_leq; x != NULL; x = x->bm_mfc_next) {
1592 /* Record that a packet is received */
1593 x->bm_measured.b_packets++;
1594 x->bm_measured.b_bytes += plen;
1602 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1605 X_legal_vif_num(int vif)
1614 if (vif < V_numvifs)
1622 * Return the local address used by this vif
1625 X_ip_mcast_src(int vifi)
1634 if (vifi < V_numvifs)
1635 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1642 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1644 struct mbuf *mb_copy;
1645 int hlen = ip->ip_hl << 2;
1650 * Make a new reference to the packet; make sure that
1651 * the IP header is actually copied, not just referenced,
1652 * so that ip_output() only scribbles on the copy.
1654 mb_copy = m_copypacket(m, M_NOWAIT);
1655 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1656 mb_copy = m_pullup(mb_copy, hlen);
1657 if (mb_copy == NULL)
1660 send_packet(vifp, mb_copy);
1664 send_packet(struct vif *vifp, struct mbuf *m)
1666 struct ip_moptions imo;
1671 imo.imo_multicast_ifp = vifp->v_ifp;
1672 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1673 imo.imo_multicast_loop = !!in_mcast_loop;
1674 imo.imo_multicast_vif = -1;
1675 STAILQ_INIT(&imo.imo_head);
1678 * Re-entrancy should not be a problem here, because
1679 * the packets that we send out and are looped back at us
1680 * should get rejected because they appear to come from
1681 * the loopback interface, thus preventing looping.
1683 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1684 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1685 (ptrdiff_t)(vifp - V_viftable), error);
1689 * Stubs for old RSVP socket shim implementation.
1693 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1696 return (EOPNOTSUPP);
1700 X_ip_rsvp_force_done(struct socket *so __unused)
1706 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1714 return (IPPROTO_DONE);
1718 * Code for bandwidth monitors
1722 * Define common interface for timeval-related methods
1724 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1725 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1726 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1729 compute_bw_meter_flags(struct bw_upcall *req)
1733 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1734 flags |= BW_METER_UNIT_PACKETS;
1735 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1736 flags |= BW_METER_UNIT_BYTES;
1737 if (req->bu_flags & BW_UPCALL_GEQ)
1738 flags |= BW_METER_GEQ;
1739 if (req->bu_flags & BW_UPCALL_LEQ)
1740 flags |= BW_METER_LEQ;
1746 expire_bw_meter_leq(void *arg)
1748 struct bw_meter *x = arg;
1752 * callout is always executed with MFC_LOCK taken
1755 CURVNET_SET((struct vnet *)x->arg);
1760 * Test if we should deliver an upcall
1762 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1763 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1764 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1765 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1766 /* Prepare an upcall for delivery */
1767 bw_meter_prepare_upcall(x, &now);
1770 /* Send all upcalls that are pending delivery */
1773 /* Reset counters */
1774 x->bm_start_time = now;
1775 x->bm_measured.b_bytes = 0;
1776 x->bm_measured.b_packets = 0;
1778 callout_schedule(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time));
1784 * Add a bw_meter entry
1787 add_bw_upcall(struct bw_upcall *req)
1790 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1791 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1793 struct bw_meter *x, **bwm_ptr;
1796 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1799 /* Test if the flags are valid */
1800 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1802 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1804 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1805 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1808 /* Test if the threshold time interval is valid */
1809 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1812 flags = compute_bw_meter_flags(req);
1815 * Find if we have already same bw_meter entry
1818 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1821 return EADDRNOTAVAIL;
1824 /* Choose an appropriate bw_meter list */
1825 if (req->bu_flags & BW_UPCALL_GEQ)
1826 bwm_ptr = &mfc->mfc_bw_meter_geq;
1828 bwm_ptr = &mfc->mfc_bw_meter_leq;
1830 for (x = *bwm_ptr; x != NULL; x = x->bm_mfc_next) {
1831 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1832 &req->bu_threshold.b_time, ==))
1833 && (x->bm_threshold.b_packets
1834 == req->bu_threshold.b_packets)
1835 && (x->bm_threshold.b_bytes
1836 == req->bu_threshold.b_bytes)
1837 && (x->bm_flags & BW_METER_USER_FLAGS)
1840 return 0; /* XXX Already installed */
1844 /* Allocate the new bw_meter entry */
1845 x = (struct bw_meter*) malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1851 /* Set the new bw_meter entry */
1852 x->bm_threshold.b_time = req->bu_threshold.b_time;
1854 x->bm_start_time = now;
1855 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1856 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1857 x->bm_measured.b_packets = 0;
1858 x->bm_measured.b_bytes = 0;
1859 x->bm_flags = flags;
1860 x->bm_time_next = NULL;
1864 /* For LEQ case create periodic callout */
1865 if (req->bu_flags & BW_UPCALL_LEQ) {
1866 callout_init_mtx(&x->bm_meter_callout, &mfc_mtx,0);
1867 callout_reset(&x->bm_meter_callout, tvtohz(&x->bm_threshold.b_time),
1868 expire_bw_meter_leq, x);
1871 /* Add the new bw_meter entry to the front of entries for this MFC */
1872 x->bm_mfc_next = *bwm_ptr;
1881 free_bw_list(struct bw_meter *list)
1883 while (list != NULL) {
1884 struct bw_meter *x = list;
1886 /* MFC_LOCK must be held here */
1887 if (x->bm_flags & BW_METER_LEQ)
1888 callout_drain(&x->bm_meter_callout);
1890 list = list->bm_mfc_next;
1896 * Delete one or multiple bw_meter entries
1899 del_bw_upcall(struct bw_upcall *req)
1902 struct bw_meter *x, **bwm_ptr;
1904 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1909 /* Find the corresponding MFC entry */
1910 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1913 return EADDRNOTAVAIL;
1914 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1916 * Delete all bw_meter entries for this mfc
1918 struct bw_meter *list;
1921 list = mfc->mfc_bw_meter_leq;
1922 mfc->mfc_bw_meter_leq = NULL;
1926 list = mfc->mfc_bw_meter_geq;
1927 mfc->mfc_bw_meter_geq = NULL;
1931 } else { /* Delete a single bw_meter entry */
1932 struct bw_meter *prev;
1935 flags = compute_bw_meter_flags(req);
1937 /* Choose an appropriate bw_meter list */
1938 if (req->bu_flags & BW_UPCALL_GEQ)
1939 bwm_ptr = &mfc->mfc_bw_meter_geq;
1941 bwm_ptr = &mfc->mfc_bw_meter_leq;
1943 /* Find the bw_meter entry to delete */
1944 for (prev = NULL, x = *bwm_ptr; x != NULL;
1945 prev = x, x = x->bm_mfc_next) {
1946 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1947 &req->bu_threshold.b_time, ==)) &&
1948 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1949 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1950 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1953 if (x != NULL) { /* Delete entry from the list for this MFC */
1955 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1957 *bwm_ptr = x->bm_mfc_next;/* new head of list */
1959 if (req->bu_flags & BW_UPCALL_LEQ)
1960 callout_stop(&x->bm_meter_callout);
1963 /* Free the bw_meter entry */
1975 * Perform bandwidth measurement processing that may result in an upcall
1978 bw_meter_geq_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1980 struct timeval delta;
1985 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1988 * Processing for ">=" type of bw_meter entry
1990 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1991 /* Reset the bw_meter entry */
1992 x->bm_start_time = *nowp;
1993 x->bm_measured.b_packets = 0;
1994 x->bm_measured.b_bytes = 0;
1995 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1998 /* Record that a packet is received */
1999 x->bm_measured.b_packets++;
2000 x->bm_measured.b_bytes += plen;
2003 * Test if we should deliver an upcall
2005 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2006 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2007 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2008 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2009 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2010 /* Prepare an upcall for delivery */
2011 bw_meter_prepare_upcall(x, nowp);
2012 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2018 * Prepare a bandwidth-related upcall
2021 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2023 struct timeval delta;
2024 struct bw_upcall *u;
2029 * Compute the measured time interval
2032 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2035 * If there are too many pending upcalls, deliver them now
2037 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2041 * Set the bw_upcall entry
2043 u = &V_bw_upcalls[V_bw_upcalls_n++];
2044 u->bu_src = x->bm_mfc->mfc_origin;
2045 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2046 u->bu_threshold.b_time = x->bm_threshold.b_time;
2047 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2048 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2049 u->bu_measured.b_time = delta;
2050 u->bu_measured.b_packets = x->bm_measured.b_packets;
2051 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2053 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2054 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2055 if (x->bm_flags & BW_METER_UNIT_BYTES)
2056 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2057 if (x->bm_flags & BW_METER_GEQ)
2058 u->bu_flags |= BW_UPCALL_GEQ;
2059 if (x->bm_flags & BW_METER_LEQ)
2060 u->bu_flags |= BW_UPCALL_LEQ;
2064 * Send the pending bandwidth-related upcalls
2067 bw_upcalls_send(void)
2070 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2071 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2072 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2074 IGMPMSG_BW_UPCALL,/* im_msgtype */
2079 { 0 } }; /* im_dst */
2083 if (V_bw_upcalls_n == 0)
2084 return; /* No pending upcalls */
2089 * Allocate a new mbuf, initialize it with the header and
2090 * the payload for the pending calls.
2092 m = m_gethdr(M_NOWAIT, MT_DATA);
2094 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2098 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2099 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2103 * XXX do we need to set the address in k_igmpsrc ?
2105 MRTSTAT_INC(mrts_upcalls);
2106 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2107 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2108 MRTSTAT_INC(mrts_upq_sockfull);
2113 * A periodic function for sending all upcalls that are pending delivery
2116 expire_bw_upcalls_send(void *arg)
2118 CURVNET_SET((struct vnet *) arg);
2124 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2130 * End of bandwidth monitoring code
2134 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2138 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2141 struct mbuf *mb_copy, *mm;
2144 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2145 * rendezvous point was unspecified, and we were told not to.
2147 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2148 in_nullhost(rt->mfc_rp))
2151 mb_copy = pim_register_prepare(ip, m);
2152 if (mb_copy == NULL)
2156 * Send all the fragments. Note that the mbuf for each fragment
2157 * is freed by the sending machinery.
2159 for (mm = mb_copy; mm; mm = mb_copy) {
2160 mb_copy = mm->m_nextpkt;
2162 mm = m_pullup(mm, sizeof(struct ip));
2164 ip = mtod(mm, struct ip *);
2165 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2166 pim_register_send_rp(ip, vifp, mm, rt);
2168 pim_register_send_upcall(ip, vifp, mm, rt);
2177 * Return a copy of the data packet that is ready for PIM Register
2179 * XXX: Note that in the returned copy the IP header is a valid one.
2181 static struct mbuf *
2182 pim_register_prepare(struct ip *ip, struct mbuf *m)
2184 struct mbuf *mb_copy = NULL;
2187 /* Take care of delayed checksums */
2188 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2189 in_delayed_cksum(m);
2190 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2194 * Copy the old packet & pullup its IP header into the
2195 * new mbuf so we can modify it.
2197 mb_copy = m_copypacket(m, M_NOWAIT);
2198 if (mb_copy == NULL)
2200 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2201 if (mb_copy == NULL)
2204 /* take care of the TTL */
2205 ip = mtod(mb_copy, struct ip *);
2208 /* Compute the MTU after the PIM Register encapsulation */
2209 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2211 if (ntohs(ip->ip_len) <= mtu) {
2212 /* Turn the IP header into a valid one */
2214 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2216 /* Fragment the packet */
2217 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2218 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2227 * Send an upcall with the data packet to the user-level process.
2230 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2231 struct mbuf *mb_copy, struct mfc *rt)
2233 struct mbuf *mb_first;
2234 int len = ntohs(ip->ip_len);
2236 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2241 * Add a new mbuf with an upcall header
2243 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2244 if (mb_first == NULL) {
2248 mb_first->m_data += max_linkhdr;
2249 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2250 mb_first->m_len = sizeof(struct igmpmsg);
2251 mb_first->m_next = mb_copy;
2253 /* Send message to routing daemon */
2254 im = mtod(mb_first, struct igmpmsg *);
2255 im->im_msgtype = IGMPMSG_WHOLEPKT;
2257 im->im_vif = vifp - V_viftable;
2258 im->im_src = ip->ip_src;
2259 im->im_dst = ip->ip_dst;
2261 k_igmpsrc.sin_addr = ip->ip_src;
2263 MRTSTAT_INC(mrts_upcalls);
2265 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2266 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2267 MRTSTAT_INC(mrts_upq_sockfull);
2271 /* Keep statistics */
2272 PIMSTAT_INC(pims_snd_registers_msgs);
2273 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2279 * Encapsulate the data packet in PIM Register message and send it to the RP.
2282 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2285 struct mbuf *mb_first;
2286 struct ip *ip_outer;
2287 struct pim_encap_pimhdr *pimhdr;
2288 int len = ntohs(ip->ip_len);
2289 vifi_t vifi = rt->mfc_parent;
2293 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2295 return EADDRNOTAVAIL; /* The iif vif is invalid */
2299 * Add a new mbuf with the encapsulating header
2301 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2302 if (mb_first == NULL) {
2306 mb_first->m_data += max_linkhdr;
2307 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2308 mb_first->m_next = mb_copy;
2310 mb_first->m_pkthdr.len = len + mb_first->m_len;
2313 * Fill in the encapsulating IP and PIM header
2315 ip_outer = mtod(mb_first, struct ip *);
2316 *ip_outer = pim_encap_iphdr;
2317 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2318 sizeof(pim_encap_pimhdr));
2319 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2320 ip_outer->ip_dst = rt->mfc_rp;
2322 * Copy the inner header TOS to the outer header, and take care of the
2325 ip_outer->ip_tos = ip->ip_tos;
2326 if (ip->ip_off & htons(IP_DF))
2327 ip_outer->ip_off |= htons(IP_DF);
2328 ip_fillid(ip_outer);
2329 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2330 + sizeof(pim_encap_iphdr));
2331 *pimhdr = pim_encap_pimhdr;
2332 /* If the iif crosses a border, set the Border-bit */
2333 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2334 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2336 mb_first->m_data += sizeof(pim_encap_iphdr);
2337 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2338 mb_first->m_data -= sizeof(pim_encap_iphdr);
2340 send_packet(vifp, mb_first);
2342 /* Keep statistics */
2343 PIMSTAT_INC(pims_snd_registers_msgs);
2344 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2350 * pim_encapcheck() is called by the encap4_input() path at runtime to
2351 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2355 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2356 int proto __unused, void *arg __unused)
2359 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2360 return (8); /* claim the datagram. */
2364 * PIM-SMv2 and PIM-DM messages processing.
2365 * Receives and verifies the PIM control messages, and passes them
2366 * up to the listening socket, using rip_input().
2367 * The only message with special processing is the PIM_REGISTER message
2368 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2369 * is passed to if_simloop().
2372 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2374 struct ip *ip = mtod(m, struct ip *);
2378 int datalen = ntohs(ip->ip_len) - iphlen;
2381 /* Keep statistics */
2382 PIMSTAT_INC(pims_rcv_total_msgs);
2383 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2388 if (datalen < PIM_MINLEN) {
2389 PIMSTAT_INC(pims_rcv_tooshort);
2390 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2391 __func__, datalen, ntohl(ip->ip_src.s_addr));
2393 return (IPPROTO_DONE);
2397 * If the packet is at least as big as a REGISTER, go agead
2398 * and grab the PIM REGISTER header size, to avoid another
2399 * possible m_pullup() later.
2401 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2402 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2404 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2406 * Get the IP and PIM headers in contiguous memory, and
2407 * possibly the PIM REGISTER header.
2409 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2410 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2411 return (IPPROTO_DONE);
2414 /* m_pullup() may have given us a new mbuf so reset ip. */
2415 ip = mtod(m, struct ip *);
2416 ip_tos = ip->ip_tos;
2418 /* adjust mbuf to point to the PIM header */
2419 m->m_data += iphlen;
2421 pim = mtod(m, struct pim *);
2424 * Validate checksum. If PIM REGISTER, exclude the data packet.
2426 * XXX: some older PIMv2 implementations don't make this distinction,
2427 * so for compatibility reason perform the checksum over part of the
2428 * message, and if error, then over the whole message.
2430 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2431 /* do nothing, checksum okay */
2432 } else if (in_cksum(m, datalen)) {
2433 PIMSTAT_INC(pims_rcv_badsum);
2434 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2436 return (IPPROTO_DONE);
2439 /* PIM version check */
2440 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2441 PIMSTAT_INC(pims_rcv_badversion);
2442 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2443 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2445 return (IPPROTO_DONE);
2448 /* restore mbuf back to the outer IP */
2449 m->m_data -= iphlen;
2452 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2454 * Since this is a REGISTER, we'll make a copy of the register
2455 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2458 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2460 struct ip *encap_ip;
2465 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2467 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2468 (int)V_reg_vif_num);
2470 return (IPPROTO_DONE);
2472 /* XXX need refcnt? */
2473 vifp = V_viftable[V_reg_vif_num].v_ifp;
2479 if (datalen < PIM_REG_MINLEN) {
2480 PIMSTAT_INC(pims_rcv_tooshort);
2481 PIMSTAT_INC(pims_rcv_badregisters);
2482 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2484 return (IPPROTO_DONE);
2487 reghdr = (u_int32_t *)(pim + 1);
2488 encap_ip = (struct ip *)(reghdr + 1);
2490 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2491 __func__, ntohl(encap_ip->ip_src.s_addr),
2492 ntohs(encap_ip->ip_len));
2494 /* verify the version number of the inner packet */
2495 if (encap_ip->ip_v != IPVERSION) {
2496 PIMSTAT_INC(pims_rcv_badregisters);
2497 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2499 return (IPPROTO_DONE);
2502 /* verify the inner packet is destined to a mcast group */
2503 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2504 PIMSTAT_INC(pims_rcv_badregisters);
2505 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2506 ntohl(encap_ip->ip_dst.s_addr));
2508 return (IPPROTO_DONE);
2511 /* If a NULL_REGISTER, pass it to the daemon */
2512 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2513 goto pim_input_to_daemon;
2516 * Copy the TOS from the outer IP header to the inner IP header.
2518 if (encap_ip->ip_tos != ip_tos) {
2519 /* Outer TOS -> inner TOS */
2520 encap_ip->ip_tos = ip_tos;
2521 /* Recompute the inner header checksum. Sigh... */
2523 /* adjust mbuf to point to the inner IP header */
2524 m->m_data += (iphlen + PIM_MINLEN);
2525 m->m_len -= (iphlen + PIM_MINLEN);
2527 encap_ip->ip_sum = 0;
2528 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2530 /* restore mbuf to point back to the outer IP header */
2531 m->m_data -= (iphlen + PIM_MINLEN);
2532 m->m_len += (iphlen + PIM_MINLEN);
2536 * Decapsulate the inner IP packet and loopback to forward it
2537 * as a normal multicast packet. Also, make a copy of the
2538 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2539 * to pass to the daemon later, so it can take the appropriate
2540 * actions (e.g., send back PIM_REGISTER_STOP).
2541 * XXX: here m->m_data points to the outer IP header.
2543 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2545 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2547 return (IPPROTO_DONE);
2550 /* Keep statistics */
2551 /* XXX: registers_bytes include only the encap. mcast pkt */
2552 PIMSTAT_INC(pims_rcv_registers_msgs);
2553 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2556 * forward the inner ip packet; point m_data at the inner ip.
2558 m_adj(m, iphlen + PIM_MINLEN);
2561 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2563 (u_long)ntohl(encap_ip->ip_src.s_addr),
2564 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2565 (int)V_reg_vif_num);
2567 /* NB: vifp was collected above; can it change on us? */
2568 if_simloop(vifp, m, dst.sin_family, 0);
2570 /* prepare the register head to send to the mrouting daemon */
2574 pim_input_to_daemon:
2576 * Pass the PIM message up to the daemon; if it is a Register message,
2577 * pass the 'head' only up to the daemon. This includes the
2578 * outer IP header, PIM header, PIM-Register header and the
2580 * XXX: the outer IP header pkt size of a Register is not adjust to
2581 * reflect the fact that the inner multicast data is truncated.
2583 return (rip_input(&m, &off, proto));
2587 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2594 if (V_mfchashtbl == NULL) /* XXX unlocked */
2596 error = sysctl_wire_old_buffer(req, 0);
2601 for (i = 0; i < mfchashsize; i++) {
2602 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2603 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2613 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2614 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2615 "IPv4 Multicast Forwarding Table "
2616 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2619 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2625 if (V_viftable == NULL) /* XXX unlocked */
2627 error = sysctl_wire_old_buffer(req, sizeof(*V_viftable) * MAXVIFS);
2632 error = SYSCTL_OUT(req, V_viftable, sizeof(*V_viftable) * MAXVIFS);
2637 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2638 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2639 sysctl_viflist, "S,vif[MAXVIFS]",
2640 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2643 vnet_mroute_init(const void *unused __unused)
2646 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2648 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2649 M_MRTABLE, M_WAITOK|M_ZERO);
2650 V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2651 M_MRTABLE, M_WAITOK|M_ZERO);
2653 callout_init(&V_expire_upcalls_ch, 1);
2654 callout_init(&V_bw_upcalls_ch, 1);
2657 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2661 vnet_mroute_uninit(const void *unused __unused)
2664 free(V_bw_upcalls, M_MRTABLE);
2665 free(V_viftable, M_MRTABLE);
2666 free(V_nexpire, M_MRTABLE);
2670 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2671 vnet_mroute_uninit, NULL);
2674 ip_mroute_modevent(module_t mod, int type, void *unused)
2679 MROUTER_LOCK_INIT();
2681 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2682 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2683 if (if_detach_event_tag == NULL) {
2684 printf("ip_mroute: unable to register "
2685 "ifnet_departure_event handler\n");
2686 MROUTER_LOCK_DESTROY();
2693 mfchashsize = MFCHASHSIZE;
2694 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2695 !powerof2(mfchashsize)) {
2696 printf("WARNING: %s not a power of 2; using default\n",
2697 "net.inet.ip.mfchashsize");
2698 mfchashsize = MFCHASHSIZE;
2701 pim_squelch_wholepkt = 0;
2702 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2703 &pim_squelch_wholepkt);
2705 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2706 if (pim_encap_cookie == NULL) {
2707 printf("ip_mroute: unable to attach pim encap\n");
2710 MROUTER_LOCK_DESTROY();
2714 ip_mcast_src = X_ip_mcast_src;
2715 ip_mforward = X_ip_mforward;
2716 ip_mrouter_done = X_ip_mrouter_done;
2717 ip_mrouter_get = X_ip_mrouter_get;
2718 ip_mrouter_set = X_ip_mrouter_set;
2720 ip_rsvp_force_done = X_ip_rsvp_force_done;
2721 ip_rsvp_vif = X_ip_rsvp_vif;
2723 legal_vif_num = X_legal_vif_num;
2724 mrt_ioctl = X_mrt_ioctl;
2725 rsvp_input_p = X_rsvp_input;
2730 * Typically module unload happens after the user-level
2731 * process has shutdown the kernel services (the check
2732 * below insures someone can't just yank the module out
2733 * from under a running process). But if the module is
2734 * just loaded and then unloaded w/o starting up a user
2735 * process we still need to cleanup.
2738 if (ip_mrouter_cnt != 0) {
2742 ip_mrouter_unloading = 1;
2745 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2747 if (pim_encap_cookie) {
2748 ip_encap_detach(pim_encap_cookie);
2749 pim_encap_cookie = NULL;
2752 ip_mcast_src = NULL;
2754 ip_mrouter_done = NULL;
2755 ip_mrouter_get = NULL;
2756 ip_mrouter_set = NULL;
2758 ip_rsvp_force_done = NULL;
2761 legal_vif_num = NULL;
2763 rsvp_input_p = NULL;
2767 MROUTER_LOCK_DESTROY();
2776 static moduledata_t ip_mroutemod = {
2782 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);