2 * Copyright (c) 1989 Stephen Deering
3 * Copyright (c) 1992, 1993
4 * The Regents of the University of California. All rights reserved.
6 * This code is derived from software contributed to Berkeley by
7 * Stephen Deering of Stanford University.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93
37 * IP multicast forwarding procedures
39 * Written by David Waitzman, BBN Labs, August 1988.
40 * Modified by Steve Deering, Stanford, February 1989.
41 * Modified by Mark J. Steiglitz, Stanford, May, 1991
42 * Modified by Van Jacobson, LBL, January 1993
43 * Modified by Ajit Thyagarajan, PARC, August 1993
44 * Modified by Bill Fenner, PARC, April 1995
45 * Modified by Ahmed Helmy, SGI, June 1996
46 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
47 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
48 * Modified by Hitoshi Asaeda, WIDE, August 2000
49 * Modified by Pavlin Radoslavov, ICSI, October 2002
51 * MROUTING Revision: 3.5
52 * and PIM-SMv2 and PIM-DM support, advanced API support,
53 * bandwidth metering and signaling
57 * TODO: Prefix functions with ipmf_.
58 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
59 * domain attachment (if_afdata) so we can track consumers of that service.
60 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
61 * move it to socket options.
62 * TODO: Cleanup LSRR removal further.
63 * TODO: Push RSVP stubs into raw_ip.c.
64 * TODO: Use bitstring.h for vif set.
65 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
66 * TODO: Sync ip6_mroute.c with this file.
69 #include <sys/cdefs.h>
70 __FBSDID("$FreeBSD$");
73 #include "opt_mrouting.h"
77 #include <sys/param.h>
78 #include <sys/kernel.h>
79 #include <sys/stddef.h>
82 #include <sys/malloc.h>
84 #include <sys/module.h>
86 #include <sys/protosw.h>
87 #include <sys/signalvar.h>
88 #include <sys/socket.h>
89 #include <sys/socketvar.h>
90 #include <sys/sockio.h>
92 #include <sys/sysctl.h>
93 #include <sys/syslog.h>
94 #include <sys/systm.h>
98 #include <net/netisr.h>
99 #include <net/route.h>
100 #include <net/vnet.h>
102 #include <netinet/in.h>
103 #include <netinet/igmp.h>
104 #include <netinet/in_systm.h>
105 #include <netinet/in_var.h>
106 #include <netinet/ip.h>
107 #include <netinet/ip_encap.h>
108 #include <netinet/ip_mroute.h>
109 #include <netinet/ip_var.h>
110 #include <netinet/ip_options.h>
111 #include <netinet/pim.h>
112 #include <netinet/pim_var.h>
113 #include <netinet/udp.h>
115 #include <machine/in_cksum.h>
118 #define KTR_IPMF KTR_INET
121 #define VIFI_INVALID ((vifi_t) -1)
122 #define M_HASCL(m) ((m)->m_flags & M_EXT)
124 static VNET_DEFINE(uint32_t, last_tv_sec); /* last time we processed this */
125 #define V_last_tv_sec VNET(last_tv_sec)
127 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
130 * Locking. We use two locks: one for the virtual interface table and
131 * one for the forwarding table. These locks may be nested in which case
132 * the VIF lock must always be taken first. Note that each lock is used
133 * to cover not only the specific data structure but also related data
137 static struct mtx mrouter_mtx;
138 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
139 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
140 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
141 #define MROUTER_LOCK_INIT() \
142 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
143 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
145 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
146 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
148 static VNET_DEFINE(struct mrtstat, mrtstat);
149 #define V_mrtstat VNET(mrtstat)
150 SYSCTL_VNET_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
151 &VNET_NAME(mrtstat), mrtstat,
152 "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
153 "netinet/ip_mroute.h)");
155 static VNET_DEFINE(u_long, mfchash);
156 #define V_mfchash VNET(mfchash)
157 #define MFCHASH(a, g) \
158 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
159 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
160 #define MFCHASHSIZE 256
162 static u_long mfchashsize; /* Hash size */
163 static VNET_DEFINE(u_char *, nexpire); /* 0..mfchashsize-1 */
164 #define V_nexpire VNET(nexpire)
165 static VNET_DEFINE(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
166 #define V_mfchashtbl VNET(mfchashtbl)
168 static struct mtx mfc_mtx;
169 #define MFC_LOCK() mtx_lock(&mfc_mtx)
170 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
171 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
172 #define MFC_LOCK_INIT() \
173 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
174 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
176 static VNET_DEFINE(vifi_t, numvifs);
177 #define V_numvifs VNET(numvifs)
178 static VNET_DEFINE(struct vif, viftable[MAXVIFS]);
179 #define V_viftable VNET(viftable)
180 SYSCTL_VNET_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
181 &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]",
182 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
184 static struct mtx vif_mtx;
185 #define VIF_LOCK() mtx_lock(&vif_mtx)
186 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
187 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
188 #define VIF_LOCK_INIT() \
189 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
190 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
192 static eventhandler_tag if_detach_event_tag = NULL;
194 static VNET_DEFINE(struct callout, expire_upcalls_ch);
195 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
197 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
198 #define UPCALL_EXPIRE 6 /* number of timeouts */
201 * Bandwidth meter variables and constants
203 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
205 * Pending timeouts are stored in a hash table, the key being the
206 * expiration time. Periodically, the entries are analysed and processed.
208 #define BW_METER_BUCKETS 1024
209 static VNET_DEFINE(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]);
210 #define V_bw_meter_timers VNET(bw_meter_timers)
211 static VNET_DEFINE(struct callout, bw_meter_ch);
212 #define V_bw_meter_ch VNET(bw_meter_ch)
213 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
216 * Pending upcalls are stored in a vector which is flushed when
217 * full, or periodically
219 static VNET_DEFINE(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]);
220 #define V_bw_upcalls VNET(bw_upcalls)
221 static VNET_DEFINE(u_int, bw_upcalls_n); /* # of pending upcalls */
222 #define V_bw_upcalls_n VNET(bw_upcalls_n)
223 static VNET_DEFINE(struct callout, bw_upcalls_ch);
224 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
226 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
228 static VNET_DEFINE(struct pimstat, pimstat);
229 #define V_pimstat VNET(pimstat)
231 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
232 SYSCTL_VNET_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
233 &VNET_NAME(pimstat), pimstat,
234 "PIM Statistics (struct pimstat, netinet/pim_var.h)");
236 static u_long pim_squelch_wholepkt = 0;
237 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
238 &pim_squelch_wholepkt, 0,
239 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
241 extern struct domain inetdomain;
242 static const struct protosw in_pim_protosw = {
244 .pr_domain = &inetdomain,
245 .pr_protocol = IPPROTO_PIM,
246 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR,
247 .pr_input = pim_input,
248 .pr_output = (pr_output_t*)rip_output,
249 .pr_ctloutput = rip_ctloutput,
250 .pr_usrreqs = &rip_usrreqs
252 static const struct encaptab *pim_encap_cookie;
254 static int pim_encapcheck(const struct mbuf *, int, int, void *);
257 * Note: the PIM Register encapsulation adds the following in front of a
260 * struct pim_encap_hdr {
262 * struct pim_encap_pimhdr pim;
267 struct pim_encap_pimhdr {
271 #define PIM_ENCAP_TTL 64
273 static struct ip pim_encap_iphdr = {
274 #if BYTE_ORDER == LITTLE_ENDIAN
275 sizeof(struct ip) >> 2,
279 sizeof(struct ip) >> 2,
282 sizeof(struct ip), /* total length */
290 static struct pim_encap_pimhdr pim_encap_pimhdr = {
292 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
299 static VNET_DEFINE(vifi_t, reg_vif_num) = VIFI_INVALID;
300 #define V_reg_vif_num VNET(reg_vif_num)
301 static VNET_DEFINE(struct ifnet, multicast_register_if);
302 #define V_multicast_register_if VNET(multicast_register_if)
308 static u_long X_ip_mcast_src(int);
309 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
310 struct ip_moptions *);
311 static int X_ip_mrouter_done(void);
312 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
313 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
314 static int X_legal_vif_num(int);
315 static int X_mrt_ioctl(u_long, caddr_t, int);
317 static int add_bw_upcall(struct bw_upcall *);
318 static int add_mfc(struct mfcctl2 *);
319 static int add_vif(struct vifctl *);
320 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
321 static void bw_meter_process(void);
322 static void bw_meter_receive_packet(struct bw_meter *, int,
324 static void bw_upcalls_send(void);
325 static int del_bw_upcall(struct bw_upcall *);
326 static int del_mfc(struct mfcctl2 *);
327 static int del_vif(vifi_t);
328 static int del_vif_locked(vifi_t);
329 static void expire_bw_meter_process(void *);
330 static void expire_bw_upcalls_send(void *);
331 static void expire_mfc(struct mfc *);
332 static void expire_upcalls(void *);
333 static void free_bw_list(struct bw_meter *);
334 static int get_sg_cnt(struct sioc_sg_req *);
335 static int get_vif_cnt(struct sioc_vif_req *);
336 static void if_detached_event(void *, struct ifnet *);
337 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
338 static int ip_mrouter_init(struct socket *, int);
339 static __inline struct mfc *
340 mfc_find(struct in_addr *, struct in_addr *);
341 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
343 pim_register_prepare(struct ip *, struct mbuf *);
344 static int pim_register_send(struct ip *, struct vif *,
345 struct mbuf *, struct mfc *);
346 static int pim_register_send_rp(struct ip *, struct vif *,
347 struct mbuf *, struct mfc *);
348 static int pim_register_send_upcall(struct ip *, struct vif *,
349 struct mbuf *, struct mfc *);
350 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
351 static void send_packet(struct vif *, struct mbuf *);
352 static int set_api_config(uint32_t *);
353 static int set_assert(int);
354 static int socket_send(struct socket *, struct mbuf *,
355 struct sockaddr_in *);
356 static void unschedule_bw_meter(struct bw_meter *);
359 * Kernel multicast forwarding API capabilities and setup.
360 * If more API capabilities are added to the kernel, they should be
361 * recorded in `mrt_api_support'.
363 #define MRT_API_VERSION 0x0305
365 static const int mrt_api_version = MRT_API_VERSION;
366 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
367 MRT_MFC_FLAGS_BORDER_VIF |
370 static VNET_DEFINE(uint32_t, mrt_api_config);
371 #define V_mrt_api_config VNET(mrt_api_config)
372 static VNET_DEFINE(int, pim_assert_enabled);
373 #define V_pim_assert_enabled VNET(pim_assert_enabled)
374 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
377 * Find a route for a given origin IP address and multicast group address.
378 * Statistics must be updated by the caller.
380 static __inline struct mfc *
381 mfc_find(struct in_addr *o, struct in_addr *g)
387 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
388 if (in_hosteq(rt->mfc_origin, *o) &&
389 in_hosteq(rt->mfc_mcastgrp, *g) &&
390 TAILQ_EMPTY(&rt->mfc_stall))
398 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
401 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
407 struct bw_upcall bw_upcall;
410 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
414 switch (sopt->sopt_name) {
416 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
419 error = ip_mrouter_init(so, optval);
423 error = ip_mrouter_done();
427 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
430 error = add_vif(&vifc);
434 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
437 error = del_vif(vifi);
443 * select data size depending on API version.
445 if (sopt->sopt_name == MRT_ADD_MFC &&
446 V_mrt_api_config & MRT_API_FLAGS_ALL) {
447 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
448 sizeof(struct mfcctl2));
450 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
451 sizeof(struct mfcctl));
452 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
453 sizeof(mfc) - sizeof(struct mfcctl));
457 if (sopt->sopt_name == MRT_ADD_MFC)
458 error = add_mfc(&mfc);
460 error = del_mfc(&mfc);
464 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
471 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
473 error = set_api_config(&i);
475 error = sooptcopyout(sopt, &i, sizeof i);
478 case MRT_ADD_BW_UPCALL:
479 case MRT_DEL_BW_UPCALL:
480 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
484 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
485 error = add_bw_upcall(&bw_upcall);
487 error = del_bw_upcall(&bw_upcall);
498 * Handle MRT getsockopt commands
501 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
505 switch (sopt->sopt_name) {
507 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
511 error = sooptcopyout(sopt, &V_pim_assert_enabled,
512 sizeof V_pim_assert_enabled);
515 case MRT_API_SUPPORT:
516 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
520 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
531 * Handle ioctl commands to obtain information from the cache
534 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
539 * Currently the only function calling this ioctl routine is rtioctl().
540 * Typically, only root can create the raw socket in order to execute
541 * this ioctl method, however the request might be coming from a prison
543 error = priv_check(curthread, PRIV_NETINET_MROUTE);
547 case (SIOCGETVIFCNT):
548 error = get_vif_cnt((struct sioc_vif_req *)data);
552 error = get_sg_cnt((struct sioc_sg_req *)data);
563 * returns the packet, byte, rpf-failure count for the source group provided
566 get_sg_cnt(struct sioc_sg_req *req)
571 rt = mfc_find(&req->src, &req->grp);
574 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
575 return EADDRNOTAVAIL;
577 req->pktcnt = rt->mfc_pkt_cnt;
578 req->bytecnt = rt->mfc_byte_cnt;
579 req->wrong_if = rt->mfc_wrong_if;
585 * returns the input and output packet and byte counts on the vif provided
588 get_vif_cnt(struct sioc_vif_req *req)
590 vifi_t vifi = req->vifi;
593 if (vifi >= V_numvifs) {
598 req->icount = V_viftable[vifi].v_pkt_in;
599 req->ocount = V_viftable[vifi].v_pkt_out;
600 req->ibytes = V_viftable[vifi].v_bytes_in;
601 req->obytes = V_viftable[vifi].v_bytes_out;
608 if_detached_event(void *arg __unused, struct ifnet *ifp)
615 if (V_ip_mrouter == NULL) {
624 * Tear down multicast forwarder state associated with this ifnet.
625 * 1. Walk the vif list, matching vifs against this ifnet.
626 * 2. Walk the multicast forwarding cache (mfc) looking for
627 * inner matches with this vif's index.
628 * 3. Expire any matching multicast forwarding cache entries.
629 * 4. Free vif state. This should disable ALLMULTI on the interface.
631 for (vifi = 0; vifi < V_numvifs; vifi++) {
632 if (V_viftable[vifi].v_ifp != ifp)
634 for (i = 0; i < mfchashsize; i++) {
635 struct mfc *rt, *nrt;
636 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
637 nrt = LIST_NEXT(rt, mfc_hash);
638 if (rt->mfc_parent == vifi) {
643 del_vif_locked(vifi);
653 * Enable multicast forwarding.
656 ip_mrouter_init(struct socket *so, int version)
659 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
660 so->so_type, so->so_proto->pr_protocol);
662 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
670 if (ip_mrouter_unloading) {
675 if (V_ip_mrouter != NULL) {
680 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
683 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
685 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
687 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
695 CTR1(KTR_IPMF, "%s: done", __func__);
701 * Disable multicast forwarding.
704 X_ip_mrouter_done(void)
713 if (V_ip_mrouter == NULL) {
719 * Detach/disable hooks to the reset of the system.
723 V_mrt_api_config = 0;
728 * For each phyint in use, disable promiscuous reception of all IP
731 for (vifi = 0; vifi < V_numvifs; vifi++) {
732 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
733 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
734 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
736 so->sin_len = sizeof(struct sockaddr_in);
737 so->sin_family = AF_INET;
738 so->sin_addr.s_addr = INADDR_ANY;
739 ifp = V_viftable[vifi].v_ifp;
743 bzero((caddr_t)V_viftable, sizeof(V_viftable));
745 V_pim_assert_enabled = 0;
749 callout_stop(&V_expire_upcalls_ch);
750 callout_stop(&V_bw_upcalls_ch);
751 callout_stop(&V_bw_meter_ch);
756 * Free all multicast forwarding cache entries.
757 * Do not use hashdestroy(), as we must perform other cleanup.
759 for (i = 0; i < mfchashsize; i++) {
760 struct mfc *rt, *nrt;
761 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
762 nrt = LIST_NEXT(rt, mfc_hash);
766 free(V_mfchashtbl, M_MRTABLE);
769 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
772 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
776 V_reg_vif_num = VIFI_INVALID;
780 CTR1(KTR_IPMF, "%s: done", __func__);
786 * Set PIM assert processing global
791 if ((i != 1) && (i != 0))
794 V_pim_assert_enabled = i;
800 * Configure API capabilities
803 set_api_config(uint32_t *apival)
808 * We can set the API capabilities only if it is the first operation
809 * after MRT_INIT. I.e.:
810 * - there are no vifs installed
811 * - pim_assert is not enabled
812 * - the MFC table is empty
818 if (V_pim_assert_enabled) {
825 for (i = 0; i < mfchashsize; i++) {
826 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
834 V_mrt_api_config = *apival & mrt_api_support;
835 *apival = V_mrt_api_config;
841 * Add a vif to the vif table
844 add_vif(struct vifctl *vifcp)
846 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
847 struct sockaddr_in sin = {sizeof sin, AF_INET};
853 if (vifcp->vifc_vifi >= MAXVIFS) {
857 /* rate limiting is no longer supported by this code */
858 if (vifcp->vifc_rate_limit != 0) {
859 log(LOG_ERR, "rate limiting is no longer supported\n");
863 if (!in_nullhost(vifp->v_lcl_addr)) {
867 if (in_nullhost(vifcp->vifc_lcl_addr)) {
869 return EADDRNOTAVAIL;
872 /* Find the interface with an address in AF_INET family */
873 if (vifcp->vifc_flags & VIFF_REGISTER) {
875 * XXX: Because VIFF_REGISTER does not really need a valid
876 * local interface (e.g. it could be 127.0.0.2), we don't
881 sin.sin_addr = vifcp->vifc_lcl_addr;
882 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
885 return EADDRNOTAVAIL;
891 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
892 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
895 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
896 ifp = &V_multicast_register_if;
897 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
898 if (V_reg_vif_num == VIFI_INVALID) {
899 if_initname(&V_multicast_register_if, "register_vif", 0);
900 V_multicast_register_if.if_flags = IFF_LOOPBACK;
901 V_reg_vif_num = vifcp->vifc_vifi;
903 } else { /* Make sure the interface supports multicast */
904 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
909 /* Enable promiscuous reception of all IP multicasts from the if */
910 error = if_allmulti(ifp, 1);
917 vifp->v_flags = vifcp->vifc_flags;
918 vifp->v_threshold = vifcp->vifc_threshold;
919 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
920 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
922 /* initialize per vif pkt counters */
925 vifp->v_bytes_in = 0;
926 vifp->v_bytes_out = 0;
927 bzero(&vifp->v_route, sizeof(vifp->v_route));
929 /* Adjust numvifs up if the vifi is higher than numvifs */
930 if (V_numvifs <= vifcp->vifc_vifi)
931 V_numvifs = vifcp->vifc_vifi + 1;
935 CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__,
936 (int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr),
937 (int)vifcp->vifc_threshold);
943 * Delete a vif from the vif table
946 del_vif_locked(vifi_t vifi)
952 if (vifi >= V_numvifs) {
955 vifp = &V_viftable[vifi];
956 if (in_nullhost(vifp->v_lcl_addr)) {
957 return EADDRNOTAVAIL;
960 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
961 if_allmulti(vifp->v_ifp, 0);
963 if (vifp->v_flags & VIFF_REGISTER)
964 V_reg_vif_num = VIFI_INVALID;
966 bzero((caddr_t)vifp, sizeof (*vifp));
968 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
970 /* Adjust numvifs down */
971 for (vifi = V_numvifs; vifi > 0; vifi--)
972 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
985 cc = del_vif_locked(vifi);
992 * update an mfc entry without resetting counters and S,G addresses.
995 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
999 rt->mfc_parent = mfccp->mfcc_parent;
1000 for (i = 0; i < V_numvifs; i++) {
1001 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
1002 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
1005 /* set the RP address */
1006 if (V_mrt_api_config & MRT_MFC_RP)
1007 rt->mfc_rp = mfccp->mfcc_rp;
1009 rt->mfc_rp.s_addr = INADDR_ANY;
1013 * fully initialize an mfc entry from the parameter.
1016 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1018 rt->mfc_origin = mfccp->mfcc_origin;
1019 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1021 update_mfc_params(rt, mfccp);
1023 /* initialize pkt counters per src-grp */
1024 rt->mfc_pkt_cnt = 0;
1025 rt->mfc_byte_cnt = 0;
1026 rt->mfc_wrong_if = 0;
1027 timevalclear(&rt->mfc_last_assert);
1031 expire_mfc(struct mfc *rt)
1033 struct rtdetq *rte, *nrte;
1037 free_bw_list(rt->mfc_bw_meter);
1039 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1041 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1042 free(rte, M_MRTABLE);
1045 LIST_REMOVE(rt, mfc_hash);
1046 free(rt, M_MRTABLE);
1053 add_mfc(struct mfcctl2 *mfccp)
1056 struct rtdetq *rte, *nrte;
1063 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1065 /* If an entry already exists, just update the fields */
1067 CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x",
1068 __func__, inet_ntoa(mfccp->mfcc_origin),
1069 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1070 mfccp->mfcc_parent);
1071 update_mfc_params(rt, mfccp);
1078 * Find the entry for which the upcall was made and update
1081 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1082 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1083 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1084 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1085 !TAILQ_EMPTY(&rt->mfc_stall)) {
1087 "%s: add mfc orig %s group %lx parent %x qh %p",
1088 __func__, inet_ntoa(mfccp->mfcc_origin),
1089 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1091 TAILQ_FIRST(&rt->mfc_stall));
1093 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1095 init_mfc_params(rt, mfccp);
1096 rt->mfc_expire = 0; /* Don't clean this guy up */
1099 /* Free queued packets, but attempt to forward them first. */
1100 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1101 if (rte->ifp != NULL)
1102 ip_mdq(rte->m, rte->ifp, rt, -1);
1104 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1106 free(rte, M_MRTABLE);
1112 * It is possible that an entry is being inserted without an upcall
1115 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1116 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1117 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1118 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1119 init_mfc_params(rt, mfccp);
1127 if (rt == NULL) { /* no upcall, so make a new entry */
1128 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1135 init_mfc_params(rt, mfccp);
1136 TAILQ_INIT(&rt->mfc_stall);
1140 rt->mfc_bw_meter = NULL;
1142 /* insert new entry at head of hash chain */
1143 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1154 * Delete an mfc entry
1157 del_mfc(struct mfcctl2 *mfccp)
1159 struct in_addr origin;
1160 struct in_addr mcastgrp;
1163 origin = mfccp->mfcc_origin;
1164 mcastgrp = mfccp->mfcc_mcastgrp;
1166 CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__,
1167 inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr));
1171 rt = mfc_find(&origin, &mcastgrp);
1174 return EADDRNOTAVAIL;
1178 * free the bw_meter entries
1180 free_bw_list(rt->mfc_bw_meter);
1181 rt->mfc_bw_meter = NULL;
1183 LIST_REMOVE(rt, mfc_hash);
1184 free(rt, M_MRTABLE);
1192 * Send a message to the routing daemon on the multicast routing socket.
1195 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1198 SOCKBUF_LOCK(&s->so_rcv);
1199 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1201 sorwakeup_locked(s);
1204 SOCKBUF_UNLOCK(&s->so_rcv);
1211 * IP multicast forwarding function. This function assumes that the packet
1212 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1213 * pointed to by "ifp", and the packet is to be relayed to other networks
1214 * that have members of the packet's destination IP multicast group.
1216 * The packet is returned unscathed to the caller, unless it is
1217 * erroneous, in which case a non-zero return value tells the caller to
1221 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1224 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1225 struct ip_moptions *imo)
1231 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p",
1232 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1234 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1235 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1237 * Packet arrived via a physical interface or
1238 * an encapsulated tunnel or a register_vif.
1242 * Packet arrived through a source-route tunnel.
1243 * Source-route tunnels are no longer supported.
1250 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1251 if (ip->ip_ttl < MAXTTL)
1252 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1253 error = ip_mdq(m, ifp, NULL, vifi);
1260 * Don't forward a packet with time-to-live of zero or one,
1261 * or a packet destined to a local-only group.
1263 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1270 * Determine forwarding vifs from the forwarding cache table
1272 MRTSTAT_INC(mrts_mfc_lookups);
1273 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1275 /* Entry exists, so forward if necessary */
1277 error = ip_mdq(m, ifp, rt, -1);
1283 * If we don't have a route for packet's origin,
1284 * Make a copy of the packet & send message to routing daemon
1290 int hlen = ip->ip_hl << 2;
1292 MRTSTAT_INC(mrts_mfc_misses);
1293 MRTSTAT_INC(mrts_no_route);
1294 CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)",
1295 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr));
1298 * Allocate mbufs early so that we don't do extra work if we are
1299 * just going to fail anyway. Make sure to pullup the header so
1300 * that other people can't step on it.
1302 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1310 mb0 = m_copypacket(m, M_DONTWAIT);
1311 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1312 mb0 = m_pullup(mb0, hlen);
1314 free(rte, M_MRTABLE);
1320 /* is there an upcall waiting for this flow ? */
1321 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1322 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1323 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1324 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1325 !TAILQ_EMPTY(&rt->mfc_stall))
1332 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1336 * Locate the vifi for the incoming interface for this packet.
1337 * If none found, drop packet.
1339 for (vifi = 0; vifi < V_numvifs &&
1340 V_viftable[vifi].v_ifp != ifp; vifi++)
1342 if (vifi >= V_numvifs) /* vif not found, drop packet */
1345 /* no upcall, so make a new entry */
1346 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1350 /* Make a copy of the header to send to the user level process */
1351 mm = m_copy(mb0, 0, hlen);
1356 * Send message to routing daemon to install
1357 * a route into the kernel table
1360 im = mtod(mm, struct igmpmsg *);
1361 im->im_msgtype = IGMPMSG_NOCACHE;
1365 MRTSTAT_INC(mrts_upcalls);
1367 k_igmpsrc.sin_addr = ip->ip_src;
1368 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1369 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1370 MRTSTAT_INC(mrts_upq_sockfull);
1372 free(rt, M_MRTABLE);
1374 free(rte, M_MRTABLE);
1381 /* insert new entry at head of hash chain */
1382 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1383 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1384 rt->mfc_expire = UPCALL_EXPIRE;
1386 for (i = 0; i < V_numvifs; i++) {
1387 rt->mfc_ttls[i] = 0;
1388 rt->mfc_flags[i] = 0;
1390 rt->mfc_parent = -1;
1392 /* clear the RP address */
1393 rt->mfc_rp.s_addr = INADDR_ANY;
1394 rt->mfc_bw_meter = NULL;
1396 /* initialize pkt counters per src-grp */
1397 rt->mfc_pkt_cnt = 0;
1398 rt->mfc_byte_cnt = 0;
1399 rt->mfc_wrong_if = 0;
1400 timevalclear(&rt->mfc_last_assert);
1402 TAILQ_INIT(&rt->mfc_stall);
1405 /* link into table */
1406 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1407 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1411 /* determine if queue has overflowed */
1412 if (rt->mfc_nstall > MAX_UPQ) {
1413 MRTSTAT_INC(mrts_upq_ovflw);
1415 free(rte, M_MRTABLE);
1421 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1436 * Clean up the cache entry if upcall is not serviced
1439 expire_upcalls(void *arg)
1443 CURVNET_SET((struct vnet *) arg);
1447 for (i = 0; i < mfchashsize; i++) {
1448 struct mfc *rt, *nrt;
1450 if (V_nexpire[i] == 0)
1453 for (rt = LIST_FIRST(&V_mfchashtbl[i]); rt; rt = nrt) {
1454 nrt = LIST_NEXT(rt, mfc_hash);
1456 if (TAILQ_EMPTY(&rt->mfc_stall))
1459 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1463 * free the bw_meter entries
1465 while (rt->mfc_bw_meter != NULL) {
1466 struct bw_meter *x = rt->mfc_bw_meter;
1468 rt->mfc_bw_meter = x->bm_mfc_next;
1472 MRTSTAT_INC(mrts_cache_cleanups);
1473 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1474 (u_long)ntohl(rt->mfc_origin.s_addr),
1475 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1483 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1490 * Packet forwarding routine once entry in the cache is made
1493 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1495 struct ip *ip = mtod(m, struct ip *);
1497 int plen = ip->ip_len;
1502 * If xmt_vif is not -1, send on only the requested vif.
1504 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1506 if (xmt_vif < V_numvifs) {
1507 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1508 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1510 phyint_send(ip, V_viftable + xmt_vif, m);
1515 * Don't forward if it didn't arrive from the parent vif for its origin.
1517 vifi = rt->mfc_parent;
1518 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1519 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1520 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1521 MRTSTAT_INC(mrts_wrong_if);
1524 * If we are doing PIM assert processing, send a message
1525 * to the routing daemon.
1527 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1528 * can complete the SPT switch, regardless of the type
1529 * of the iif (broadcast media, GRE tunnel, etc).
1531 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1532 V_viftable[vifi].v_ifp) {
1534 if (ifp == &V_multicast_register_if)
1535 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1537 /* Get vifi for the incoming packet */
1538 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1541 if (vifi >= V_numvifs)
1542 return 0; /* The iif is not found: ignore the packet. */
1544 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1545 return 0; /* WRONGVIF disabled: ignore the packet */
1547 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1548 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1550 int hlen = ip->ip_hl << 2;
1551 struct mbuf *mm = m_copy(m, 0, hlen);
1553 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1554 mm = m_pullup(mm, hlen);
1558 im = mtod(mm, struct igmpmsg *);
1559 im->im_msgtype = IGMPMSG_WRONGVIF;
1563 MRTSTAT_INC(mrts_upcalls);
1565 k_igmpsrc.sin_addr = im->im_src;
1566 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1567 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1568 MRTSTAT_INC(mrts_upq_sockfull);
1577 /* If I sourced this packet, it counts as output, else it was input. */
1578 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1579 V_viftable[vifi].v_pkt_out++;
1580 V_viftable[vifi].v_bytes_out += plen;
1582 V_viftable[vifi].v_pkt_in++;
1583 V_viftable[vifi].v_bytes_in += plen;
1586 rt->mfc_byte_cnt += plen;
1589 * For each vif, decide if a copy of the packet should be forwarded.
1591 * - the ttl exceeds the vif's threshold
1592 * - there are group members downstream on interface
1594 for (vifi = 0; vifi < V_numvifs; vifi++)
1595 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1596 V_viftable[vifi].v_pkt_out++;
1597 V_viftable[vifi].v_bytes_out += plen;
1598 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1599 pim_register_send(ip, V_viftable + vifi, m, rt);
1601 phyint_send(ip, V_viftable + vifi, m);
1605 * Perform upcall-related bw measuring.
1607 if (rt->mfc_bw_meter != NULL) {
1613 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1614 bw_meter_receive_packet(x, plen, &now);
1621 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1624 X_legal_vif_num(int vif)
1633 if (vif < V_numvifs)
1641 * Return the local address used by this vif
1644 X_ip_mcast_src(int vifi)
1653 if (vifi < V_numvifs)
1654 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1661 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1663 struct mbuf *mb_copy;
1664 int hlen = ip->ip_hl << 2;
1669 * Make a new reference to the packet; make sure that
1670 * the IP header is actually copied, not just referenced,
1671 * so that ip_output() only scribbles on the copy.
1673 mb_copy = m_copypacket(m, M_DONTWAIT);
1674 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1675 mb_copy = m_pullup(mb_copy, hlen);
1676 if (mb_copy == NULL)
1679 send_packet(vifp, mb_copy);
1683 send_packet(struct vif *vifp, struct mbuf *m)
1685 struct ip_moptions imo;
1686 struct in_multi *imm[2];
1691 imo.imo_multicast_ifp = vifp->v_ifp;
1692 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1693 imo.imo_multicast_loop = 1;
1694 imo.imo_multicast_vif = -1;
1695 imo.imo_num_memberships = 0;
1696 imo.imo_max_memberships = 2;
1697 imo.imo_membership = &imm[0];
1700 * Re-entrancy should not be a problem here, because
1701 * the packets that we send out and are looped back at us
1702 * should get rejected because they appear to come from
1703 * the loopback interface, thus preventing looping.
1705 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL);
1706 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1707 (ptrdiff_t)(vifp - V_viftable), error);
1711 * Stubs for old RSVP socket shim implementation.
1715 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1718 return (EOPNOTSUPP);
1722 X_ip_rsvp_force_done(struct socket *so __unused)
1728 X_rsvp_input(struct mbuf *m, int off __unused)
1736 * Code for bandwidth monitors
1740 * Define common interface for timeval-related methods
1742 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1743 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1744 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1747 compute_bw_meter_flags(struct bw_upcall *req)
1751 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1752 flags |= BW_METER_UNIT_PACKETS;
1753 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1754 flags |= BW_METER_UNIT_BYTES;
1755 if (req->bu_flags & BW_UPCALL_GEQ)
1756 flags |= BW_METER_GEQ;
1757 if (req->bu_flags & BW_UPCALL_LEQ)
1758 flags |= BW_METER_LEQ;
1764 * Add a bw_meter entry
1767 add_bw_upcall(struct bw_upcall *req)
1770 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1771 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1776 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1779 /* Test if the flags are valid */
1780 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1782 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1784 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1785 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1788 /* Test if the threshold time interval is valid */
1789 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1792 flags = compute_bw_meter_flags(req);
1795 * Find if we have already same bw_meter entry
1798 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1801 return EADDRNOTAVAIL;
1803 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1804 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1805 &req->bu_threshold.b_time, ==)) &&
1806 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1807 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1808 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1810 return 0; /* XXX Already installed */
1814 /* Allocate the new bw_meter entry */
1815 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1821 /* Set the new bw_meter entry */
1822 x->bm_threshold.b_time = req->bu_threshold.b_time;
1824 x->bm_start_time = now;
1825 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1826 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1827 x->bm_measured.b_packets = 0;
1828 x->bm_measured.b_bytes = 0;
1829 x->bm_flags = flags;
1830 x->bm_time_next = NULL;
1831 x->bm_time_hash = BW_METER_BUCKETS;
1833 /* Add the new bw_meter entry to the front of entries for this MFC */
1835 x->bm_mfc_next = mfc->mfc_bw_meter;
1836 mfc->mfc_bw_meter = x;
1837 schedule_bw_meter(x, &now);
1844 free_bw_list(struct bw_meter *list)
1846 while (list != NULL) {
1847 struct bw_meter *x = list;
1849 list = list->bm_mfc_next;
1850 unschedule_bw_meter(x);
1856 * Delete one or multiple bw_meter entries
1859 del_bw_upcall(struct bw_upcall *req)
1864 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1869 /* Find the corresponding MFC entry */
1870 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1873 return EADDRNOTAVAIL;
1874 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1876 * Delete all bw_meter entries for this mfc
1878 struct bw_meter *list;
1880 list = mfc->mfc_bw_meter;
1881 mfc->mfc_bw_meter = NULL;
1885 } else { /* Delete a single bw_meter entry */
1886 struct bw_meter *prev;
1889 flags = compute_bw_meter_flags(req);
1891 /* Find the bw_meter entry to delete */
1892 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1893 prev = x, x = x->bm_mfc_next) {
1894 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1895 &req->bu_threshold.b_time, ==)) &&
1896 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1897 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1898 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1901 if (x != NULL) { /* Delete entry from the list for this MFC */
1903 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1905 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1907 unschedule_bw_meter(x);
1909 /* Free the bw_meter entry */
1921 * Perform bandwidth measurement processing that may result in an upcall
1924 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1926 struct timeval delta;
1931 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1933 if (x->bm_flags & BW_METER_GEQ) {
1935 * Processing for ">=" type of bw_meter entry
1937 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1938 /* Reset the bw_meter entry */
1939 x->bm_start_time = *nowp;
1940 x->bm_measured.b_packets = 0;
1941 x->bm_measured.b_bytes = 0;
1942 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1945 /* Record that a packet is received */
1946 x->bm_measured.b_packets++;
1947 x->bm_measured.b_bytes += plen;
1950 * Test if we should deliver an upcall
1952 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1953 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1954 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1955 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1956 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1957 /* Prepare an upcall for delivery */
1958 bw_meter_prepare_upcall(x, nowp);
1959 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1962 } else if (x->bm_flags & BW_METER_LEQ) {
1964 * Processing for "<=" type of bw_meter entry
1966 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1968 * We are behind time with the multicast forwarding table
1969 * scanning for "<=" type of bw_meter entries, so test now
1970 * if we should deliver an upcall.
1972 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1973 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1974 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1975 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1976 /* Prepare an upcall for delivery */
1977 bw_meter_prepare_upcall(x, nowp);
1979 /* Reschedule the bw_meter entry */
1980 unschedule_bw_meter(x);
1981 schedule_bw_meter(x, nowp);
1984 /* Record that a packet is received */
1985 x->bm_measured.b_packets++;
1986 x->bm_measured.b_bytes += plen;
1989 * Test if we should restart the measuring interval
1991 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1992 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1993 (x->bm_flags & BW_METER_UNIT_BYTES &&
1994 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1995 /* Don't restart the measuring interval */
1997 /* Do restart the measuring interval */
1999 * XXX: note that we don't unschedule and schedule, because this
2000 * might be too much overhead per packet. Instead, when we process
2001 * all entries for a given timer hash bin, we check whether it is
2002 * really a timeout. If not, we reschedule at that time.
2004 x->bm_start_time = *nowp;
2005 x->bm_measured.b_packets = 0;
2006 x->bm_measured.b_bytes = 0;
2007 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2013 * Prepare a bandwidth-related upcall
2016 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2018 struct timeval delta;
2019 struct bw_upcall *u;
2024 * Compute the measured time interval
2027 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2030 * If there are too many pending upcalls, deliver them now
2032 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2036 * Set the bw_upcall entry
2038 u = &V_bw_upcalls[V_bw_upcalls_n++];
2039 u->bu_src = x->bm_mfc->mfc_origin;
2040 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2041 u->bu_threshold.b_time = x->bm_threshold.b_time;
2042 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2043 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2044 u->bu_measured.b_time = delta;
2045 u->bu_measured.b_packets = x->bm_measured.b_packets;
2046 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2048 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2049 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2050 if (x->bm_flags & BW_METER_UNIT_BYTES)
2051 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2052 if (x->bm_flags & BW_METER_GEQ)
2053 u->bu_flags |= BW_UPCALL_GEQ;
2054 if (x->bm_flags & BW_METER_LEQ)
2055 u->bu_flags |= BW_UPCALL_LEQ;
2059 * Send the pending bandwidth-related upcalls
2062 bw_upcalls_send(void)
2065 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2066 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2067 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2069 IGMPMSG_BW_UPCALL,/* im_msgtype */
2074 { 0 } }; /* im_dst */
2078 if (V_bw_upcalls_n == 0)
2079 return; /* No pending upcalls */
2084 * Allocate a new mbuf, initialize it with the header and
2085 * the payload for the pending calls.
2087 MGETHDR(m, M_DONTWAIT, MT_DATA);
2089 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2093 m->m_len = m->m_pkthdr.len = 0;
2094 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2095 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2099 * XXX do we need to set the address in k_igmpsrc ?
2101 MRTSTAT_INC(mrts_upcalls);
2102 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2103 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2104 MRTSTAT_INC(mrts_upq_sockfull);
2109 * Compute the timeout hash value for the bw_meter entries
2111 #define BW_METER_TIMEHASH(bw_meter, hash) \
2113 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2115 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2116 (hash) = next_timeval.tv_sec; \
2117 if (next_timeval.tv_usec) \
2118 (hash)++; /* XXX: make sure we don't timeout early */ \
2119 (hash) %= BW_METER_BUCKETS; \
2123 * Schedule a timer to process periodically bw_meter entry of type "<="
2124 * by linking the entry in the proper hash bucket.
2127 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2133 if (!(x->bm_flags & BW_METER_LEQ))
2134 return; /* XXX: we schedule timers only for "<=" entries */
2137 * Reset the bw_meter entry
2139 x->bm_start_time = *nowp;
2140 x->bm_measured.b_packets = 0;
2141 x->bm_measured.b_bytes = 0;
2142 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2145 * Compute the timeout hash value and insert the entry
2147 BW_METER_TIMEHASH(x, time_hash);
2148 x->bm_time_next = V_bw_meter_timers[time_hash];
2149 V_bw_meter_timers[time_hash] = x;
2150 x->bm_time_hash = time_hash;
2154 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2155 * by removing the entry from the proper hash bucket.
2158 unschedule_bw_meter(struct bw_meter *x)
2161 struct bw_meter *prev, *tmp;
2165 if (!(x->bm_flags & BW_METER_LEQ))
2166 return; /* XXX: we schedule timers only for "<=" entries */
2169 * Compute the timeout hash value and delete the entry
2171 time_hash = x->bm_time_hash;
2172 if (time_hash >= BW_METER_BUCKETS)
2173 return; /* Entry was not scheduled */
2175 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2176 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2181 panic("unschedule_bw_meter: bw_meter entry not found");
2184 prev->bm_time_next = x->bm_time_next;
2186 V_bw_meter_timers[time_hash] = x->bm_time_next;
2188 x->bm_time_next = NULL;
2189 x->bm_time_hash = BW_METER_BUCKETS;
2194 * Process all "<=" type of bw_meter that should be processed now,
2195 * and for each entry prepare an upcall if necessary. Each processed
2196 * entry is rescheduled again for the (periodic) processing.
2198 * This is run periodically (once per second normally). On each round,
2199 * all the potentially matching entries are in the hash slot that we are
2207 struct timeval now, process_endtime;
2210 if (V_last_tv_sec == now.tv_sec)
2211 return; /* nothing to do */
2213 loops = now.tv_sec - V_last_tv_sec;
2214 V_last_tv_sec = now.tv_sec;
2215 if (loops > BW_METER_BUCKETS)
2216 loops = BW_METER_BUCKETS;
2220 * Process all bins of bw_meter entries from the one after the last
2221 * processed to the current one. On entry, i points to the last bucket
2222 * visited, so we need to increment i at the beginning of the loop.
2224 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2225 struct bw_meter *x, *tmp_list;
2227 if (++i >= BW_METER_BUCKETS)
2230 /* Disconnect the list of bw_meter entries from the bin */
2231 tmp_list = V_bw_meter_timers[i];
2232 V_bw_meter_timers[i] = NULL;
2234 /* Process the list of bw_meter entries */
2235 while (tmp_list != NULL) {
2237 tmp_list = tmp_list->bm_time_next;
2239 /* Test if the time interval is over */
2240 process_endtime = x->bm_start_time;
2241 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2242 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2243 /* Not yet: reschedule, but don't reset */
2246 BW_METER_TIMEHASH(x, time_hash);
2247 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2249 * XXX: somehow the bin processing is a bit ahead of time.
2250 * Put the entry in the next bin.
2252 if (++time_hash >= BW_METER_BUCKETS)
2255 x->bm_time_next = V_bw_meter_timers[time_hash];
2256 V_bw_meter_timers[time_hash] = x;
2257 x->bm_time_hash = time_hash;
2263 * Test if we should deliver an upcall
2265 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2266 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2267 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2268 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2269 /* Prepare an upcall for delivery */
2270 bw_meter_prepare_upcall(x, &now);
2274 * Reschedule for next processing
2276 schedule_bw_meter(x, &now);
2280 /* Send all upcalls that are pending delivery */
2287 * A periodic function for sending all upcalls that are pending delivery
2290 expire_bw_upcalls_send(void *arg)
2292 CURVNET_SET((struct vnet *) arg);
2298 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2304 * A periodic function for periodic scanning of the multicast forwarding
2305 * table for processing all "<=" bw_meter entries.
2308 expire_bw_meter_process(void *arg)
2310 CURVNET_SET((struct vnet *) arg);
2312 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2315 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2321 * End of bandwidth monitoring code
2325 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2329 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2332 struct mbuf *mb_copy, *mm;
2335 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2336 * rendezvous point was unspecified, and we were told not to.
2338 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2339 in_nullhost(rt->mfc_rp))
2342 mb_copy = pim_register_prepare(ip, m);
2343 if (mb_copy == NULL)
2347 * Send all the fragments. Note that the mbuf for each fragment
2348 * is freed by the sending machinery.
2350 for (mm = mb_copy; mm; mm = mb_copy) {
2351 mb_copy = mm->m_nextpkt;
2353 mm = m_pullup(mm, sizeof(struct ip));
2355 ip = mtod(mm, struct ip *);
2356 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2357 pim_register_send_rp(ip, vifp, mm, rt);
2359 pim_register_send_upcall(ip, vifp, mm, rt);
2368 * Return a copy of the data packet that is ready for PIM Register
2370 * XXX: Note that in the returned copy the IP header is a valid one.
2372 static struct mbuf *
2373 pim_register_prepare(struct ip *ip, struct mbuf *m)
2375 struct mbuf *mb_copy = NULL;
2378 /* Take care of delayed checksums */
2379 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2380 in_delayed_cksum(m);
2381 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2385 * Copy the old packet & pullup its IP header into the
2386 * new mbuf so we can modify it.
2388 mb_copy = m_copypacket(m, M_DONTWAIT);
2389 if (mb_copy == NULL)
2391 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2392 if (mb_copy == NULL)
2395 /* take care of the TTL */
2396 ip = mtod(mb_copy, struct ip *);
2399 /* Compute the MTU after the PIM Register encapsulation */
2400 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2402 if (ip->ip_len <= mtu) {
2403 /* Turn the IP header into a valid one */
2404 ip->ip_len = htons(ip->ip_len);
2405 ip->ip_off = htons(ip->ip_off);
2407 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2409 /* Fragment the packet */
2410 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2419 * Send an upcall with the data packet to the user-level process.
2422 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2423 struct mbuf *mb_copy, struct mfc *rt)
2425 struct mbuf *mb_first;
2426 int len = ntohs(ip->ip_len);
2428 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2433 * Add a new mbuf with an upcall header
2435 MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2436 if (mb_first == NULL) {
2440 mb_first->m_data += max_linkhdr;
2441 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2442 mb_first->m_len = sizeof(struct igmpmsg);
2443 mb_first->m_next = mb_copy;
2445 /* Send message to routing daemon */
2446 im = mtod(mb_first, struct igmpmsg *);
2447 im->im_msgtype = IGMPMSG_WHOLEPKT;
2449 im->im_vif = vifp - V_viftable;
2450 im->im_src = ip->ip_src;
2451 im->im_dst = ip->ip_dst;
2453 k_igmpsrc.sin_addr = ip->ip_src;
2455 MRTSTAT_INC(mrts_upcalls);
2457 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2458 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2459 MRTSTAT_INC(mrts_upq_sockfull);
2463 /* Keep statistics */
2464 PIMSTAT_INC(pims_snd_registers_msgs);
2465 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2471 * Encapsulate the data packet in PIM Register message and send it to the RP.
2474 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2477 struct mbuf *mb_first;
2478 struct ip *ip_outer;
2479 struct pim_encap_pimhdr *pimhdr;
2480 int len = ntohs(ip->ip_len);
2481 vifi_t vifi = rt->mfc_parent;
2485 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2487 return EADDRNOTAVAIL; /* The iif vif is invalid */
2491 * Add a new mbuf with the encapsulating header
2493 MGETHDR(mb_first, M_DONTWAIT, MT_DATA);
2494 if (mb_first == NULL) {
2498 mb_first->m_data += max_linkhdr;
2499 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2500 mb_first->m_next = mb_copy;
2502 mb_first->m_pkthdr.len = len + mb_first->m_len;
2505 * Fill in the encapsulating IP and PIM header
2507 ip_outer = mtod(mb_first, struct ip *);
2508 *ip_outer = pim_encap_iphdr;
2509 ip_outer->ip_id = ip_newid();
2510 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2511 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2512 ip_outer->ip_dst = rt->mfc_rp;
2514 * Copy the inner header TOS to the outer header, and take care of the
2517 ip_outer->ip_tos = ip->ip_tos;
2518 if (ntohs(ip->ip_off) & IP_DF)
2519 ip_outer->ip_off |= IP_DF;
2520 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2521 + sizeof(pim_encap_iphdr));
2522 *pimhdr = pim_encap_pimhdr;
2523 /* If the iif crosses a border, set the Border-bit */
2524 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2525 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2527 mb_first->m_data += sizeof(pim_encap_iphdr);
2528 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2529 mb_first->m_data -= sizeof(pim_encap_iphdr);
2531 send_packet(vifp, mb_first);
2533 /* Keep statistics */
2534 PIMSTAT_INC(pims_snd_registers_msgs);
2535 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2541 * pim_encapcheck() is called by the encap4_input() path at runtime to
2542 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2546 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg)
2550 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2552 if (proto != IPPROTO_PIM)
2553 return 0; /* not for us; reject the datagram. */
2555 return 64; /* claim the datagram. */
2559 * PIM-SMv2 and PIM-DM messages processing.
2560 * Receives and verifies the PIM control messages, and passes them
2561 * up to the listening socket, using rip_input().
2562 * The only message with special processing is the PIM_REGISTER message
2563 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2564 * is passed to if_simloop().
2567 pim_input(struct mbuf *m, int off)
2569 struct ip *ip = mtod(m, struct ip *);
2572 int datalen = ip->ip_len;
2576 /* Keep statistics */
2577 PIMSTAT_INC(pims_rcv_total_msgs);
2578 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2583 if (datalen < PIM_MINLEN) {
2584 PIMSTAT_INC(pims_rcv_tooshort);
2585 CTR3(KTR_IPMF, "%s: short packet (%d) from %s",
2586 __func__, datalen, inet_ntoa(ip->ip_src));
2592 * If the packet is at least as big as a REGISTER, go agead
2593 * and grab the PIM REGISTER header size, to avoid another
2594 * possible m_pullup() later.
2596 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2597 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2599 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2601 * Get the IP and PIM headers in contiguous memory, and
2602 * possibly the PIM REGISTER header.
2604 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
2605 (m = m_pullup(m, minlen)) == 0) {
2606 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2610 /* m_pullup() may have given us a new mbuf so reset ip. */
2611 ip = mtod(m, struct ip *);
2612 ip_tos = ip->ip_tos;
2614 /* adjust mbuf to point to the PIM header */
2615 m->m_data += iphlen;
2617 pim = mtod(m, struct pim *);
2620 * Validate checksum. If PIM REGISTER, exclude the data packet.
2622 * XXX: some older PIMv2 implementations don't make this distinction,
2623 * so for compatibility reason perform the checksum over part of the
2624 * message, and if error, then over the whole message.
2626 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2627 /* do nothing, checksum okay */
2628 } else if (in_cksum(m, datalen)) {
2629 PIMSTAT_INC(pims_rcv_badsum);
2630 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2635 /* PIM version check */
2636 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2637 PIMSTAT_INC(pims_rcv_badversion);
2638 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2639 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2644 /* restore mbuf back to the outer IP */
2645 m->m_data -= iphlen;
2648 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2650 * Since this is a REGISTER, we'll make a copy of the register
2651 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2654 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2656 struct ip *encap_ip;
2661 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2663 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2664 (int)V_reg_vif_num);
2668 /* XXX need refcnt? */
2669 vifp = V_viftable[V_reg_vif_num].v_ifp;
2675 if (datalen < PIM_REG_MINLEN) {
2676 PIMSTAT_INC(pims_rcv_tooshort);
2677 PIMSTAT_INC(pims_rcv_badregisters);
2678 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2683 reghdr = (u_int32_t *)(pim + 1);
2684 encap_ip = (struct ip *)(reghdr + 1);
2686 CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d",
2687 __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len));
2689 /* verify the version number of the inner packet */
2690 if (encap_ip->ip_v != IPVERSION) {
2691 PIMSTAT_INC(pims_rcv_badregisters);
2692 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2697 /* verify the inner packet is destined to a mcast group */
2698 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2699 PIMSTAT_INC(pims_rcv_badregisters);
2700 CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__,
2701 inet_ntoa(encap_ip->ip_dst));
2706 /* If a NULL_REGISTER, pass it to the daemon */
2707 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2708 goto pim_input_to_daemon;
2711 * Copy the TOS from the outer IP header to the inner IP header.
2713 if (encap_ip->ip_tos != ip_tos) {
2714 /* Outer TOS -> inner TOS */
2715 encap_ip->ip_tos = ip_tos;
2716 /* Recompute the inner header checksum. Sigh... */
2718 /* adjust mbuf to point to the inner IP header */
2719 m->m_data += (iphlen + PIM_MINLEN);
2720 m->m_len -= (iphlen + PIM_MINLEN);
2722 encap_ip->ip_sum = 0;
2723 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2725 /* restore mbuf to point back to the outer IP header */
2726 m->m_data -= (iphlen + PIM_MINLEN);
2727 m->m_len += (iphlen + PIM_MINLEN);
2731 * Decapsulate the inner IP packet and loopback to forward it
2732 * as a normal multicast packet. Also, make a copy of the
2733 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2734 * to pass to the daemon later, so it can take the appropriate
2735 * actions (e.g., send back PIM_REGISTER_STOP).
2736 * XXX: here m->m_data points to the outer IP header.
2738 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
2740 CTR1(KTR_IPMF, "%s: m_copy() failed", __func__);
2745 /* Keep statistics */
2746 /* XXX: registers_bytes include only the encap. mcast pkt */
2747 PIMSTAT_INC(pims_rcv_registers_msgs);
2748 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2751 * forward the inner ip packet; point m_data at the inner ip.
2753 m_adj(m, iphlen + PIM_MINLEN);
2756 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2758 (u_long)ntohl(encap_ip->ip_src.s_addr),
2759 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2760 (int)V_reg_vif_num);
2762 /* NB: vifp was collected above; can it change on us? */
2763 if_simloop(vifp, m, dst.sin_family, 0);
2765 /* prepare the register head to send to the mrouting daemon */
2769 pim_input_to_daemon:
2771 * Pass the PIM message up to the daemon; if it is a Register message,
2772 * pass the 'head' only up to the daemon. This includes the
2773 * outer IP header, PIM header, PIM-Register header and the
2775 * XXX: the outer IP header pkt size of a Register is not adjust to
2776 * reflect the fact that the inner multicast data is truncated.
2778 rip_input(m, iphlen);
2784 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2791 if (V_mfchashtbl == NULL) /* XXX unlocked */
2793 error = sysctl_wire_old_buffer(req, 0);
2798 for (i = 0; i < mfchashsize; i++) {
2799 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2800 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2810 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2811 sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2812 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2815 vnet_mroute_init(const void *unused __unused)
2818 MALLOC(V_nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2819 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers));
2820 callout_init(&V_expire_upcalls_ch, CALLOUT_MPSAFE);
2821 callout_init(&V_bw_upcalls_ch, CALLOUT_MPSAFE);
2822 callout_init(&V_bw_meter_ch, CALLOUT_MPSAFE);
2825 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PSEUDO, SI_ORDER_MIDDLE, vnet_mroute_init,
2829 vnet_mroute_uninit(const void *unused __unused)
2832 FREE(V_nexpire, M_MRTABLE);
2836 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PSEUDO, SI_ORDER_MIDDLE,
2837 vnet_mroute_uninit, NULL);
2840 ip_mroute_modevent(module_t mod, int type, void *unused)
2845 MROUTER_LOCK_INIT();
2847 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2848 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2849 if (if_detach_event_tag == NULL) {
2850 printf("ip_mroute: unable to ifnet_deperture_even handler\n");
2851 MROUTER_LOCK_DESTROY();
2858 mfchashsize = MFCHASHSIZE;
2859 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2860 !powerof2(mfchashsize)) {
2861 printf("WARNING: %s not a power of 2; using default\n",
2862 "net.inet.ip.mfchashsize");
2863 mfchashsize = MFCHASHSIZE;
2866 pim_squelch_wholepkt = 0;
2867 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2868 &pim_squelch_wholepkt);
2870 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM,
2871 pim_encapcheck, &in_pim_protosw, NULL);
2872 if (pim_encap_cookie == NULL) {
2873 printf("ip_mroute: unable to attach pim encap\n");
2876 MROUTER_LOCK_DESTROY();
2880 ip_mcast_src = X_ip_mcast_src;
2881 ip_mforward = X_ip_mforward;
2882 ip_mrouter_done = X_ip_mrouter_done;
2883 ip_mrouter_get = X_ip_mrouter_get;
2884 ip_mrouter_set = X_ip_mrouter_set;
2886 ip_rsvp_force_done = X_ip_rsvp_force_done;
2887 ip_rsvp_vif = X_ip_rsvp_vif;
2889 legal_vif_num = X_legal_vif_num;
2890 mrt_ioctl = X_mrt_ioctl;
2891 rsvp_input_p = X_rsvp_input;
2896 * Typically module unload happens after the user-level
2897 * process has shutdown the kernel services (the check
2898 * below insures someone can't just yank the module out
2899 * from under a running process). But if the module is
2900 * just loaded and then unloaded w/o starting up a user
2901 * process we still need to cleanup.
2904 if (ip_mrouter_cnt != 0) {
2908 ip_mrouter_unloading = 1;
2911 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2913 if (pim_encap_cookie) {
2914 encap_detach(pim_encap_cookie);
2915 pim_encap_cookie = NULL;
2918 ip_mcast_src = NULL;
2920 ip_mrouter_done = NULL;
2921 ip_mrouter_get = NULL;
2922 ip_mrouter_set = NULL;
2924 ip_rsvp_force_done = NULL;
2927 legal_vif_num = NULL;
2929 rsvp_input_p = NULL;
2933 MROUTER_LOCK_DESTROY();
2942 static moduledata_t ip_mroutemod = {
2948 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);