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
53 * MROUTING Revision: 3.5
54 * and PIM-SMv2 and PIM-DM support, advanced API support,
55 * bandwidth metering and signaling
59 * TODO: Prefix functions with ipmf_.
60 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol
61 * domain attachment (if_afdata) so we can track consumers of that service.
62 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT,
63 * move it to socket options.
64 * TODO: Cleanup LSRR removal further.
65 * TODO: Push RSVP stubs into raw_ip.c.
66 * TODO: Use bitstring.h for vif set.
67 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded.
68 * TODO: Sync ip6_mroute.c with this file.
71 #include <sys/cdefs.h>
72 __FBSDID("$FreeBSD$");
75 #include "opt_mrouting.h"
79 #include <sys/param.h>
80 #include <sys/kernel.h>
81 #include <sys/stddef.h>
82 #include <sys/eventhandler.h>
85 #include <sys/malloc.h>
87 #include <sys/module.h>
89 #include <sys/protosw.h>
90 #include <sys/signalvar.h>
91 #include <sys/socket.h>
92 #include <sys/socketvar.h>
93 #include <sys/sockio.h>
95 #include <sys/sysctl.h>
96 #include <sys/syslog.h>
97 #include <sys/systm.h>
99 #include <sys/counter.h>
102 #include <net/if_var.h>
103 #include <net/netisr.h>
104 #include <net/route.h>
105 #include <net/vnet.h>
107 #include <netinet/in.h>
108 #include <netinet/igmp.h>
109 #include <netinet/in_systm.h>
110 #include <netinet/in_var.h>
111 #include <netinet/ip.h>
112 #include <netinet/ip_encap.h>
113 #include <netinet/ip_mroute.h>
114 #include <netinet/ip_var.h>
115 #include <netinet/ip_options.h>
116 #include <netinet/pim.h>
117 #include <netinet/pim_var.h>
118 #include <netinet/udp.h>
120 #include <machine/in_cksum.h>
123 #define KTR_IPMF KTR_INET
126 #define VIFI_INVALID ((vifi_t) -1)
128 VNET_DEFINE_STATIC(uint32_t, last_tv_sec); /* last time we processed this */
129 #define V_last_tv_sec VNET(last_tv_sec)
131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache");
134 * Locking. We use two locks: one for the virtual interface table and
135 * one for the forwarding table. These locks may be nested in which case
136 * the VIF lock must always be taken first. Note that each lock is used
137 * to cover not only the specific data structure but also related data
141 static struct mtx mrouter_mtx;
142 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx)
143 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx)
144 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED)
145 #define MROUTER_LOCK_INIT() \
146 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF)
147 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx)
149 static int ip_mrouter_cnt; /* # of vnets with active mrouters */
150 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */
152 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat);
153 VNET_PCPUSTAT_SYSINIT(mrtstat);
154 VNET_PCPUSTAT_SYSUNINIT(mrtstat);
155 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat,
156 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, "
157 "netinet/ip_mroute.h)");
159 VNET_DEFINE_STATIC(u_long, mfchash);
160 #define V_mfchash VNET(mfchash)
161 #define MFCHASH(a, g) \
162 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \
163 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash)
164 #define MFCHASHSIZE 256
166 static u_long mfchashsize; /* Hash size */
167 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */
168 #define V_nexpire VNET(nexpire)
169 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl);
170 #define V_mfchashtbl VNET(mfchashtbl)
172 static struct mtx mfc_mtx;
173 #define MFC_LOCK() mtx_lock(&mfc_mtx)
174 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx)
175 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED)
176 #define MFC_LOCK_INIT() \
177 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF)
178 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx)
180 VNET_DEFINE_STATIC(vifi_t, numvifs);
181 #define V_numvifs VNET(numvifs)
182 VNET_DEFINE_STATIC(struct vif *, viftable);
183 #define V_viftable VNET(viftable)
185 static struct mtx vif_mtx;
186 #define VIF_LOCK() mtx_lock(&vif_mtx)
187 #define VIF_UNLOCK() mtx_unlock(&vif_mtx)
188 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED)
189 #define VIF_LOCK_INIT() \
190 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF)
191 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx)
193 static eventhandler_tag if_detach_event_tag = NULL;
195 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch);
196 #define V_expire_upcalls_ch VNET(expire_upcalls_ch)
198 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
199 #define UPCALL_EXPIRE 6 /* number of timeouts */
202 * Bandwidth meter variables and constants
204 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
206 * Pending timeouts are stored in a hash table, the key being the
207 * expiration time. Periodically, the entries are analysed and processed.
209 #define BW_METER_BUCKETS 1024
210 VNET_DEFINE_STATIC(struct bw_meter **, bw_meter_timers);
211 #define V_bw_meter_timers VNET(bw_meter_timers)
212 VNET_DEFINE_STATIC(struct callout, bw_meter_ch);
213 #define V_bw_meter_ch VNET(bw_meter_ch)
214 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
217 * Pending upcalls are stored in a vector which is flushed when
218 * full, or periodically
220 VNET_DEFINE_STATIC(struct bw_upcall *, bw_upcalls);
221 #define V_bw_upcalls VNET(bw_upcalls)
222 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */
223 #define V_bw_upcalls_n VNET(bw_upcalls_n)
224 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch);
225 #define V_bw_upcalls_ch VNET(bw_upcalls_ch)
227 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
229 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat);
230 VNET_PCPUSTAT_SYSINIT(pimstat);
231 VNET_PCPUSTAT_SYSUNINIT(pimstat);
233 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM");
234 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
235 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
237 static u_long pim_squelch_wholepkt = 0;
238 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
239 &pim_squelch_wholepkt, 0,
240 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
242 static const struct encaptab *pim_encap_cookie;
243 static int pim_encapcheck(const struct mbuf *, int, int, void *);
244 static int pim_input(struct mbuf *, int, int, void *);
246 static const struct encap_config ipv4_encap_cfg = {
247 .proto = IPPROTO_PIM,
248 .min_length = sizeof(struct ip) + PIM_MINLEN,
250 .check = pim_encapcheck,
255 * Note: the PIM Register encapsulation adds the following in front of a
258 * struct pim_encap_hdr {
260 * struct pim_encap_pimhdr pim;
265 struct pim_encap_pimhdr {
269 #define PIM_ENCAP_TTL 64
271 static struct ip pim_encap_iphdr = {
272 #if BYTE_ORDER == LITTLE_ENDIAN
273 sizeof(struct ip) >> 2,
277 sizeof(struct ip) >> 2,
280 sizeof(struct ip), /* total length */
288 static struct pim_encap_pimhdr pim_encap_pimhdr = {
290 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
297 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
298 #define V_reg_vif_num VNET(reg_vif_num)
299 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
300 #define V_multicast_register_if VNET(multicast_register_if)
306 static u_long X_ip_mcast_src(int);
307 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
308 struct ip_moptions *);
309 static int X_ip_mrouter_done(void);
310 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
311 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
312 static int X_legal_vif_num(int);
313 static int X_mrt_ioctl(u_long, caddr_t, int);
315 static int add_bw_upcall(struct bw_upcall *);
316 static int add_mfc(struct mfcctl2 *);
317 static int add_vif(struct vifctl *);
318 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
319 static void bw_meter_process(void);
320 static void bw_meter_receive_packet(struct bw_meter *, int,
322 static void bw_upcalls_send(void);
323 static int del_bw_upcall(struct bw_upcall *);
324 static int del_mfc(struct mfcctl2 *);
325 static int del_vif(vifi_t);
326 static int del_vif_locked(vifi_t);
327 static void expire_bw_meter_process(void *);
328 static void expire_bw_upcalls_send(void *);
329 static void expire_mfc(struct mfc *);
330 static void expire_upcalls(void *);
331 static void free_bw_list(struct bw_meter *);
332 static int get_sg_cnt(struct sioc_sg_req *);
333 static int get_vif_cnt(struct sioc_vif_req *);
334 static void if_detached_event(void *, struct ifnet *);
335 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
336 static int ip_mrouter_init(struct socket *, int);
337 static __inline struct mfc *
338 mfc_find(struct in_addr *, struct in_addr *);
339 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
341 pim_register_prepare(struct ip *, struct mbuf *);
342 static int pim_register_send(struct ip *, struct vif *,
343 struct mbuf *, struct mfc *);
344 static int pim_register_send_rp(struct ip *, struct vif *,
345 struct mbuf *, struct mfc *);
346 static int pim_register_send_upcall(struct ip *, struct vif *,
347 struct mbuf *, struct mfc *);
348 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
349 static void send_packet(struct vif *, struct mbuf *);
350 static int set_api_config(uint32_t *);
351 static int set_assert(int);
352 static int socket_send(struct socket *, struct mbuf *,
353 struct sockaddr_in *);
354 static void unschedule_bw_meter(struct bw_meter *);
357 * Kernel multicast forwarding API capabilities and setup.
358 * If more API capabilities are added to the kernel, they should be
359 * recorded in `mrt_api_support'.
361 #define MRT_API_VERSION 0x0305
363 static const int mrt_api_version = MRT_API_VERSION;
364 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
365 MRT_MFC_FLAGS_BORDER_VIF |
368 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
369 #define V_mrt_api_config VNET(mrt_api_config)
370 VNET_DEFINE_STATIC(int, pim_assert_enabled);
371 #define V_pim_assert_enabled VNET(pim_assert_enabled)
372 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
375 * Find a route for a given origin IP address and multicast group address.
376 * Statistics must be updated by the caller.
378 static __inline struct mfc *
379 mfc_find(struct in_addr *o, struct in_addr *g)
385 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
386 if (in_hosteq(rt->mfc_origin, *o) &&
387 in_hosteq(rt->mfc_mcastgrp, *g) &&
388 TAILQ_EMPTY(&rt->mfc_stall))
396 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
399 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
405 struct bw_upcall bw_upcall;
408 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
412 switch (sopt->sopt_name) {
414 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
417 error = ip_mrouter_init(so, optval);
421 error = ip_mrouter_done();
425 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
428 error = add_vif(&vifc);
432 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
435 error = del_vif(vifi);
441 * select data size depending on API version.
443 if (sopt->sopt_name == MRT_ADD_MFC &&
444 V_mrt_api_config & MRT_API_FLAGS_ALL) {
445 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
446 sizeof(struct mfcctl2));
448 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
449 sizeof(struct mfcctl));
450 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
451 sizeof(mfc) - sizeof(struct mfcctl));
455 if (sopt->sopt_name == MRT_ADD_MFC)
456 error = add_mfc(&mfc);
458 error = del_mfc(&mfc);
462 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
469 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
471 error = set_api_config(&i);
473 error = sooptcopyout(sopt, &i, sizeof i);
476 case MRT_ADD_BW_UPCALL:
477 case MRT_DEL_BW_UPCALL:
478 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
482 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
483 error = add_bw_upcall(&bw_upcall);
485 error = del_bw_upcall(&bw_upcall);
496 * Handle MRT getsockopt commands
499 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
503 switch (sopt->sopt_name) {
505 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
509 error = sooptcopyout(sopt, &V_pim_assert_enabled,
510 sizeof V_pim_assert_enabled);
513 case MRT_API_SUPPORT:
514 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
518 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
529 * Handle ioctl commands to obtain information from the cache
532 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
537 * Currently the only function calling this ioctl routine is rtioctl_fib().
538 * Typically, only root can create the raw socket in order to execute
539 * this ioctl method, however the request might be coming from a prison
541 error = priv_check(curthread, PRIV_NETINET_MROUTE);
545 case (SIOCGETVIFCNT):
546 error = get_vif_cnt((struct sioc_vif_req *)data);
550 error = get_sg_cnt((struct sioc_sg_req *)data);
561 * returns the packet, byte, rpf-failure count for the source group provided
564 get_sg_cnt(struct sioc_sg_req *req)
569 rt = mfc_find(&req->src, &req->grp);
572 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
573 return EADDRNOTAVAIL;
575 req->pktcnt = rt->mfc_pkt_cnt;
576 req->bytecnt = rt->mfc_byte_cnt;
577 req->wrong_if = rt->mfc_wrong_if;
583 * returns the input and output packet and byte counts on the vif provided
586 get_vif_cnt(struct sioc_vif_req *req)
588 vifi_t vifi = req->vifi;
591 if (vifi >= V_numvifs) {
596 req->icount = V_viftable[vifi].v_pkt_in;
597 req->ocount = V_viftable[vifi].v_pkt_out;
598 req->ibytes = V_viftable[vifi].v_bytes_in;
599 req->obytes = V_viftable[vifi].v_bytes_out;
606 if_detached_event(void *arg __unused, struct ifnet *ifp)
613 if (V_ip_mrouter == NULL) {
622 * Tear down multicast forwarder state associated with this ifnet.
623 * 1. Walk the vif list, matching vifs against this ifnet.
624 * 2. Walk the multicast forwarding cache (mfc) looking for
625 * inner matches with this vif's index.
626 * 3. Expire any matching multicast forwarding cache entries.
627 * 4. Free vif state. This should disable ALLMULTI on the interface.
629 for (vifi = 0; vifi < V_numvifs; vifi++) {
630 if (V_viftable[vifi].v_ifp != ifp)
632 for (i = 0; i < mfchashsize; i++) {
633 struct mfc *rt, *nrt;
635 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
636 if (rt->mfc_parent == vifi) {
641 del_vif_locked(vifi);
651 * Enable multicast forwarding.
654 ip_mrouter_init(struct socket *so, int version)
657 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
658 so->so_type, so->so_proto->pr_protocol);
660 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
668 if (ip_mrouter_unloading) {
673 if (V_ip_mrouter != NULL) {
678 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
681 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
683 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
685 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
693 CTR1(KTR_IPMF, "%s: done", __func__);
699 * Disable multicast forwarding.
702 X_ip_mrouter_done(void)
710 if (V_ip_mrouter == NULL) {
716 * Detach/disable hooks to the reset of the system.
720 V_mrt_api_config = 0;
725 * For each phyint in use, disable promiscuous reception of all IP
728 for (vifi = 0; vifi < V_numvifs; vifi++) {
729 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) &&
730 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
731 ifp = V_viftable[vifi].v_ifp;
735 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
737 V_pim_assert_enabled = 0;
741 callout_stop(&V_expire_upcalls_ch);
742 callout_stop(&V_bw_upcalls_ch);
743 callout_stop(&V_bw_meter_ch);
748 * Free all multicast forwarding cache entries.
749 * Do not use hashdestroy(), as we must perform other cleanup.
751 for (i = 0; i < mfchashsize; i++) {
752 struct mfc *rt, *nrt;
754 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
758 free(V_mfchashtbl, M_MRTABLE);
761 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
764 bzero(V_bw_meter_timers, BW_METER_BUCKETS * sizeof(*V_bw_meter_timers));
768 V_reg_vif_num = VIFI_INVALID;
772 CTR1(KTR_IPMF, "%s: done", __func__);
778 * Set PIM assert processing global
783 if ((i != 1) && (i != 0))
786 V_pim_assert_enabled = i;
792 * Configure API capabilities
795 set_api_config(uint32_t *apival)
800 * We can set the API capabilities only if it is the first operation
801 * after MRT_INIT. I.e.:
802 * - there are no vifs installed
803 * - pim_assert is not enabled
804 * - the MFC table is empty
810 if (V_pim_assert_enabled) {
817 for (i = 0; i < mfchashsize; i++) {
818 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
827 V_mrt_api_config = *apival & mrt_api_support;
828 *apival = V_mrt_api_config;
834 * Add a vif to the vif table
837 add_vif(struct vifctl *vifcp)
839 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
840 struct sockaddr_in sin = {sizeof sin, AF_INET};
846 if (vifcp->vifc_vifi >= MAXVIFS) {
850 /* rate limiting is no longer supported by this code */
851 if (vifcp->vifc_rate_limit != 0) {
852 log(LOG_ERR, "rate limiting is no longer supported\n");
856 if (!in_nullhost(vifp->v_lcl_addr)) {
860 if (in_nullhost(vifcp->vifc_lcl_addr)) {
862 return EADDRNOTAVAIL;
865 /* Find the interface with an address in AF_INET family */
866 if (vifcp->vifc_flags & VIFF_REGISTER) {
868 * XXX: Because VIFF_REGISTER does not really need a valid
869 * local interface (e.g. it could be 127.0.0.2), we don't
874 sin.sin_addr = vifcp->vifc_lcl_addr;
876 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
880 return EADDRNOTAVAIL;
886 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
887 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
890 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
891 ifp = &V_multicast_register_if;
892 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
893 if (V_reg_vif_num == VIFI_INVALID) {
894 if_initname(&V_multicast_register_if, "register_vif", 0);
895 V_multicast_register_if.if_flags = IFF_LOOPBACK;
896 V_reg_vif_num = vifcp->vifc_vifi;
898 } else { /* Make sure the interface supports multicast */
899 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
904 /* Enable promiscuous reception of all IP multicasts from the if */
905 error = if_allmulti(ifp, 1);
912 vifp->v_flags = vifcp->vifc_flags;
913 vifp->v_threshold = vifcp->vifc_threshold;
914 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
915 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
917 /* initialize per vif pkt counters */
920 vifp->v_bytes_in = 0;
921 vifp->v_bytes_out = 0;
923 /* Adjust numvifs up if the vifi is higher than numvifs */
924 if (V_numvifs <= vifcp->vifc_vifi)
925 V_numvifs = vifcp->vifc_vifi + 1;
929 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__,
930 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr),
931 (int)vifcp->vifc_threshold);
937 * Delete a vif from the vif table
940 del_vif_locked(vifi_t vifi)
946 if (vifi >= V_numvifs) {
949 vifp = &V_viftable[vifi];
950 if (in_nullhost(vifp->v_lcl_addr)) {
951 return EADDRNOTAVAIL;
954 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
955 if_allmulti(vifp->v_ifp, 0);
957 if (vifp->v_flags & VIFF_REGISTER)
958 V_reg_vif_num = VIFI_INVALID;
960 bzero((caddr_t)vifp, sizeof (*vifp));
962 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi);
964 /* Adjust numvifs down */
965 for (vifi = V_numvifs; vifi > 0; vifi--)
966 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr))
979 cc = del_vif_locked(vifi);
986 * update an mfc entry without resetting counters and S,G addresses.
989 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
993 rt->mfc_parent = mfccp->mfcc_parent;
994 for (i = 0; i < V_numvifs; i++) {
995 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
996 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config &
999 /* set the RP address */
1000 if (V_mrt_api_config & MRT_MFC_RP)
1001 rt->mfc_rp = mfccp->mfcc_rp;
1003 rt->mfc_rp.s_addr = INADDR_ANY;
1007 * fully initialize an mfc entry from the parameter.
1010 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
1012 rt->mfc_origin = mfccp->mfcc_origin;
1013 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
1015 update_mfc_params(rt, mfccp);
1017 /* initialize pkt counters per src-grp */
1018 rt->mfc_pkt_cnt = 0;
1019 rt->mfc_byte_cnt = 0;
1020 rt->mfc_wrong_if = 0;
1021 timevalclear(&rt->mfc_last_assert);
1025 expire_mfc(struct mfc *rt)
1027 struct rtdetq *rte, *nrte;
1031 free_bw_list(rt->mfc_bw_meter);
1033 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1035 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1036 free(rte, M_MRTABLE);
1039 LIST_REMOVE(rt, mfc_hash);
1040 free(rt, M_MRTABLE);
1047 add_mfc(struct mfcctl2 *mfccp)
1050 struct rtdetq *rte, *nrte;
1057 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp);
1059 /* If an entry already exists, just update the fields */
1061 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x",
1062 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1063 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1064 mfccp->mfcc_parent);
1065 update_mfc_params(rt, mfccp);
1072 * Find the entry for which the upcall was made and update
1075 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp);
1076 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1077 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1078 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) &&
1079 !TAILQ_EMPTY(&rt->mfc_stall)) {
1081 "%s: add mfc orig 0x%08x group %lx parent %x qh %p",
1082 __func__, ntohl(mfccp->mfcc_origin.s_addr),
1083 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1085 TAILQ_FIRST(&rt->mfc_stall));
1087 CTR1(KTR_IPMF, "%s: multiple matches", __func__);
1089 init_mfc_params(rt, mfccp);
1090 rt->mfc_expire = 0; /* Don't clean this guy up */
1093 /* Free queued packets, but attempt to forward them first. */
1094 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) {
1095 if (rte->ifp != NULL)
1096 ip_mdq(rte->m, rte->ifp, rt, -1);
1098 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link);
1100 free(rte, M_MRTABLE);
1106 * It is possible that an entry is being inserted without an upcall
1109 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__);
1110 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1111 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) &&
1112 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) {
1113 init_mfc_params(rt, mfccp);
1121 if (rt == NULL) { /* no upcall, so make a new entry */
1122 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1129 init_mfc_params(rt, mfccp);
1130 TAILQ_INIT(&rt->mfc_stall);
1134 rt->mfc_bw_meter = NULL;
1136 /* insert new entry at head of hash chain */
1137 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1148 * Delete an mfc entry
1151 del_mfc(struct mfcctl2 *mfccp)
1153 struct in_addr origin;
1154 struct in_addr mcastgrp;
1157 origin = mfccp->mfcc_origin;
1158 mcastgrp = mfccp->mfcc_mcastgrp;
1160 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__,
1161 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1165 rt = mfc_find(&origin, &mcastgrp);
1168 return EADDRNOTAVAIL;
1172 * free the bw_meter entries
1174 free_bw_list(rt->mfc_bw_meter);
1175 rt->mfc_bw_meter = NULL;
1177 LIST_REMOVE(rt, mfc_hash);
1178 free(rt, M_MRTABLE);
1186 * Send a message to the routing daemon on the multicast routing socket.
1189 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1192 SOCKBUF_LOCK(&s->so_rcv);
1193 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm,
1195 sorwakeup_locked(s);
1198 SOCKBUF_UNLOCK(&s->so_rcv);
1205 * IP multicast forwarding function. This function assumes that the packet
1206 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1207 * pointed to by "ifp", and the packet is to be relayed to other networks
1208 * that have members of the packet's destination IP multicast group.
1210 * The packet is returned unscathed to the caller, unless it is
1211 * erroneous, in which case a non-zero return value tells the caller to
1215 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1218 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1219 struct ip_moptions *imo)
1225 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p",
1226 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp);
1228 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1229 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1231 * Packet arrived via a physical interface or
1232 * an encapsulated tunnel or a register_vif.
1236 * Packet arrived through a source-route tunnel.
1237 * Source-route tunnels are no longer supported.
1244 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) {
1245 if (ip->ip_ttl < MAXTTL)
1246 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1247 error = ip_mdq(m, ifp, NULL, vifi);
1254 * Don't forward a packet with time-to-live of zero or one,
1255 * or a packet destined to a local-only group.
1257 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) {
1264 * Determine forwarding vifs from the forwarding cache table
1266 MRTSTAT_INC(mrts_mfc_lookups);
1267 rt = mfc_find(&ip->ip_src, &ip->ip_dst);
1269 /* Entry exists, so forward if necessary */
1271 error = ip_mdq(m, ifp, rt, -1);
1277 * If we don't have a route for packet's origin,
1278 * Make a copy of the packet & send message to routing daemon
1284 int hlen = ip->ip_hl << 2;
1286 MRTSTAT_INC(mrts_mfc_misses);
1287 MRTSTAT_INC(mrts_no_route);
1288 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)",
1289 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr));
1292 * Allocate mbufs early so that we don't do extra work if we are
1293 * just going to fail anyway. Make sure to pullup the header so
1294 * that other people can't step on it.
1296 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE,
1304 mb0 = m_copypacket(m, M_NOWAIT);
1305 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen))
1306 mb0 = m_pullup(mb0, hlen);
1308 free(rte, M_MRTABLE);
1314 /* is there an upcall waiting for this flow ? */
1315 hash = MFCHASH(ip->ip_src, ip->ip_dst);
1316 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) {
1317 if (in_hosteq(ip->ip_src, rt->mfc_origin) &&
1318 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) &&
1319 !TAILQ_EMPTY(&rt->mfc_stall))
1326 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1330 * Locate the vifi for the incoming interface for this packet.
1331 * If none found, drop packet.
1333 for (vifi = 0; vifi < V_numvifs &&
1334 V_viftable[vifi].v_ifp != ifp; vifi++)
1336 if (vifi >= V_numvifs) /* vif not found, drop packet */
1339 /* no upcall, so make a new entry */
1340 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT);
1344 /* Make a copy of the header to send to the user level process */
1345 mm = m_copym(mb0, 0, hlen, M_NOWAIT);
1350 * Send message to routing daemon to install
1351 * a route into the kernel table
1354 im = mtod(mm, struct igmpmsg *);
1355 im->im_msgtype = IGMPMSG_NOCACHE;
1359 MRTSTAT_INC(mrts_upcalls);
1361 k_igmpsrc.sin_addr = ip->ip_src;
1362 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1363 CTR0(KTR_IPMF, "ip_mforward: socket queue full");
1364 MRTSTAT_INC(mrts_upq_sockfull);
1366 free(rt, M_MRTABLE);
1368 free(rte, M_MRTABLE);
1375 /* insert new entry at head of hash chain */
1376 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1377 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1378 rt->mfc_expire = UPCALL_EXPIRE;
1380 for (i = 0; i < V_numvifs; i++) {
1381 rt->mfc_ttls[i] = 0;
1382 rt->mfc_flags[i] = 0;
1384 rt->mfc_parent = -1;
1386 /* clear the RP address */
1387 rt->mfc_rp.s_addr = INADDR_ANY;
1388 rt->mfc_bw_meter = NULL;
1390 /* initialize pkt counters per src-grp */
1391 rt->mfc_pkt_cnt = 0;
1392 rt->mfc_byte_cnt = 0;
1393 rt->mfc_wrong_if = 0;
1394 timevalclear(&rt->mfc_last_assert);
1396 TAILQ_INIT(&rt->mfc_stall);
1399 /* link into table */
1400 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash);
1401 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link);
1405 /* determine if queue has overflowed */
1406 if (rt->mfc_nstall > MAX_UPQ) {
1407 MRTSTAT_INC(mrts_upq_ovflw);
1409 free(rte, M_MRTABLE);
1415 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link);
1430 * Clean up the cache entry if upcall is not serviced
1433 expire_upcalls(void *arg)
1437 CURVNET_SET((struct vnet *) arg);
1441 for (i = 0; i < mfchashsize; i++) {
1442 struct mfc *rt, *nrt;
1444 if (V_nexpire[i] == 0)
1447 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1448 if (TAILQ_EMPTY(&rt->mfc_stall))
1451 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1455 * free the bw_meter entries
1457 while (rt->mfc_bw_meter != NULL) {
1458 struct bw_meter *x = rt->mfc_bw_meter;
1460 rt->mfc_bw_meter = x->bm_mfc_next;
1464 MRTSTAT_INC(mrts_cache_cleanups);
1465 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1466 (u_long)ntohl(rt->mfc_origin.s_addr),
1467 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1475 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1482 * Packet forwarding routine once entry in the cache is made
1485 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1487 struct ip *ip = mtod(m, struct ip *);
1489 int plen = ntohs(ip->ip_len);
1494 * If xmt_vif is not -1, send on only the requested vif.
1496 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1498 if (xmt_vif < V_numvifs) {
1499 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1500 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1502 phyint_send(ip, V_viftable + xmt_vif, m);
1507 * Don't forward if it didn't arrive from the parent vif for its origin.
1509 vifi = rt->mfc_parent;
1510 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1511 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1512 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1513 MRTSTAT_INC(mrts_wrong_if);
1516 * If we are doing PIM assert processing, send a message
1517 * to the routing daemon.
1519 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1520 * can complete the SPT switch, regardless of the type
1521 * of the iif (broadcast media, GRE tunnel, etc).
1523 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1524 V_viftable[vifi].v_ifp) {
1526 if (ifp == &V_multicast_register_if)
1527 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1529 /* Get vifi for the incoming packet */
1530 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1533 if (vifi >= V_numvifs)
1534 return 0; /* The iif is not found: ignore the packet. */
1536 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1537 return 0; /* WRONGVIF disabled: ignore the packet */
1539 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1540 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1542 int hlen = ip->ip_hl << 2;
1543 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1545 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1546 mm = m_pullup(mm, hlen);
1550 im = mtod(mm, struct igmpmsg *);
1551 im->im_msgtype = IGMPMSG_WRONGVIF;
1555 MRTSTAT_INC(mrts_upcalls);
1557 k_igmpsrc.sin_addr = im->im_src;
1558 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1559 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1560 MRTSTAT_INC(mrts_upq_sockfull);
1569 /* If I sourced this packet, it counts as output, else it was input. */
1570 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1571 V_viftable[vifi].v_pkt_out++;
1572 V_viftable[vifi].v_bytes_out += plen;
1574 V_viftable[vifi].v_pkt_in++;
1575 V_viftable[vifi].v_bytes_in += plen;
1578 rt->mfc_byte_cnt += plen;
1581 * For each vif, decide if a copy of the packet should be forwarded.
1583 * - the ttl exceeds the vif's threshold
1584 * - there are group members downstream on interface
1586 for (vifi = 0; vifi < V_numvifs; vifi++)
1587 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1588 V_viftable[vifi].v_pkt_out++;
1589 V_viftable[vifi].v_bytes_out += plen;
1590 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1591 pim_register_send(ip, V_viftable + vifi, m, rt);
1593 phyint_send(ip, V_viftable + vifi, m);
1597 * Perform upcall-related bw measuring.
1599 if (rt->mfc_bw_meter != NULL) {
1605 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1606 bw_meter_receive_packet(x, plen, &now);
1613 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1616 X_legal_vif_num(int vif)
1625 if (vif < V_numvifs)
1633 * Return the local address used by this vif
1636 X_ip_mcast_src(int vifi)
1645 if (vifi < V_numvifs)
1646 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1653 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1655 struct mbuf *mb_copy;
1656 int hlen = ip->ip_hl << 2;
1661 * Make a new reference to the packet; make sure that
1662 * the IP header is actually copied, not just referenced,
1663 * so that ip_output() only scribbles on the copy.
1665 mb_copy = m_copypacket(m, M_NOWAIT);
1666 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1667 mb_copy = m_pullup(mb_copy, hlen);
1668 if (mb_copy == NULL)
1671 send_packet(vifp, mb_copy);
1675 send_packet(struct vif *vifp, struct mbuf *m)
1677 struct ip_moptions imo;
1682 imo.imo_multicast_ifp = vifp->v_ifp;
1683 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1684 imo.imo_multicast_loop = 1;
1685 imo.imo_multicast_vif = -1;
1686 STAILQ_INIT(&imo.imo_head);
1689 * Re-entrancy should not be a problem here, because
1690 * the packets that we send out and are looped back at us
1691 * should get rejected because they appear to come from
1692 * the loopback interface, thus preventing looping.
1694 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1695 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1696 (ptrdiff_t)(vifp - V_viftable), error);
1700 * Stubs for old RSVP socket shim implementation.
1704 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1707 return (EOPNOTSUPP);
1711 X_ip_rsvp_force_done(struct socket *so __unused)
1717 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1725 return (IPPROTO_DONE);
1729 * Code for bandwidth monitors
1733 * Define common interface for timeval-related methods
1735 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1736 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1737 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1740 compute_bw_meter_flags(struct bw_upcall *req)
1744 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1745 flags |= BW_METER_UNIT_PACKETS;
1746 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1747 flags |= BW_METER_UNIT_BYTES;
1748 if (req->bu_flags & BW_UPCALL_GEQ)
1749 flags |= BW_METER_GEQ;
1750 if (req->bu_flags & BW_UPCALL_LEQ)
1751 flags |= BW_METER_LEQ;
1757 * Add a bw_meter entry
1760 add_bw_upcall(struct bw_upcall *req)
1763 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1764 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1769 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1772 /* Test if the flags are valid */
1773 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1775 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1777 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1778 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1781 /* Test if the threshold time interval is valid */
1782 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1785 flags = compute_bw_meter_flags(req);
1788 * Find if we have already same bw_meter entry
1791 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1794 return EADDRNOTAVAIL;
1796 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1797 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1798 &req->bu_threshold.b_time, ==)) &&
1799 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1800 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1801 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1803 return 0; /* XXX Already installed */
1807 /* Allocate the new bw_meter entry */
1808 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1814 /* Set the new bw_meter entry */
1815 x->bm_threshold.b_time = req->bu_threshold.b_time;
1817 x->bm_start_time = now;
1818 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1819 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1820 x->bm_measured.b_packets = 0;
1821 x->bm_measured.b_bytes = 0;
1822 x->bm_flags = flags;
1823 x->bm_time_next = NULL;
1824 x->bm_time_hash = BW_METER_BUCKETS;
1826 /* Add the new bw_meter entry to the front of entries for this MFC */
1828 x->bm_mfc_next = mfc->mfc_bw_meter;
1829 mfc->mfc_bw_meter = x;
1830 schedule_bw_meter(x, &now);
1837 free_bw_list(struct bw_meter *list)
1839 while (list != NULL) {
1840 struct bw_meter *x = list;
1842 list = list->bm_mfc_next;
1843 unschedule_bw_meter(x);
1849 * Delete one or multiple bw_meter entries
1852 del_bw_upcall(struct bw_upcall *req)
1857 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1862 /* Find the corresponding MFC entry */
1863 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1866 return EADDRNOTAVAIL;
1867 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1869 * Delete all bw_meter entries for this mfc
1871 struct bw_meter *list;
1873 list = mfc->mfc_bw_meter;
1874 mfc->mfc_bw_meter = NULL;
1878 } else { /* Delete a single bw_meter entry */
1879 struct bw_meter *prev;
1882 flags = compute_bw_meter_flags(req);
1884 /* Find the bw_meter entry to delete */
1885 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1886 prev = x, x = x->bm_mfc_next) {
1887 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1888 &req->bu_threshold.b_time, ==)) &&
1889 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1890 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1891 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1894 if (x != NULL) { /* Delete entry from the list for this MFC */
1896 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1898 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1900 unschedule_bw_meter(x);
1902 /* Free the bw_meter entry */
1914 * Perform bandwidth measurement processing that may result in an upcall
1917 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1919 struct timeval delta;
1924 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1926 if (x->bm_flags & BW_METER_GEQ) {
1928 * Processing for ">=" type of bw_meter entry
1930 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1931 /* Reset the bw_meter entry */
1932 x->bm_start_time = *nowp;
1933 x->bm_measured.b_packets = 0;
1934 x->bm_measured.b_bytes = 0;
1935 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1938 /* Record that a packet is received */
1939 x->bm_measured.b_packets++;
1940 x->bm_measured.b_bytes += plen;
1943 * Test if we should deliver an upcall
1945 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1946 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1947 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1948 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1949 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1950 /* Prepare an upcall for delivery */
1951 bw_meter_prepare_upcall(x, nowp);
1952 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1955 } else if (x->bm_flags & BW_METER_LEQ) {
1957 * Processing for "<=" type of bw_meter entry
1959 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1961 * We are behind time with the multicast forwarding table
1962 * scanning for "<=" type of bw_meter entries, so test now
1963 * if we should deliver an upcall.
1965 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1966 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1967 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1968 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1969 /* Prepare an upcall for delivery */
1970 bw_meter_prepare_upcall(x, nowp);
1972 /* Reschedule the bw_meter entry */
1973 unschedule_bw_meter(x);
1974 schedule_bw_meter(x, nowp);
1977 /* Record that a packet is received */
1978 x->bm_measured.b_packets++;
1979 x->bm_measured.b_bytes += plen;
1982 * Test if we should restart the measuring interval
1984 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1985 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1986 (x->bm_flags & BW_METER_UNIT_BYTES &&
1987 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1988 /* Don't restart the measuring interval */
1990 /* Do restart the measuring interval */
1992 * XXX: note that we don't unschedule and schedule, because this
1993 * might be too much overhead per packet. Instead, when we process
1994 * all entries for a given timer hash bin, we check whether it is
1995 * really a timeout. If not, we reschedule at that time.
1997 x->bm_start_time = *nowp;
1998 x->bm_measured.b_packets = 0;
1999 x->bm_measured.b_bytes = 0;
2000 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2006 * Prepare a bandwidth-related upcall
2009 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2011 struct timeval delta;
2012 struct bw_upcall *u;
2017 * Compute the measured time interval
2020 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2023 * If there are too many pending upcalls, deliver them now
2025 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2029 * Set the bw_upcall entry
2031 u = &V_bw_upcalls[V_bw_upcalls_n++];
2032 u->bu_src = x->bm_mfc->mfc_origin;
2033 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2034 u->bu_threshold.b_time = x->bm_threshold.b_time;
2035 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2036 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2037 u->bu_measured.b_time = delta;
2038 u->bu_measured.b_packets = x->bm_measured.b_packets;
2039 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2041 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2042 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2043 if (x->bm_flags & BW_METER_UNIT_BYTES)
2044 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2045 if (x->bm_flags & BW_METER_GEQ)
2046 u->bu_flags |= BW_UPCALL_GEQ;
2047 if (x->bm_flags & BW_METER_LEQ)
2048 u->bu_flags |= BW_UPCALL_LEQ;
2052 * Send the pending bandwidth-related upcalls
2055 bw_upcalls_send(void)
2058 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2059 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2060 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2062 IGMPMSG_BW_UPCALL,/* im_msgtype */
2067 { 0 } }; /* im_dst */
2071 if (V_bw_upcalls_n == 0)
2072 return; /* No pending upcalls */
2077 * Allocate a new mbuf, initialize it with the header and
2078 * the payload for the pending calls.
2080 m = m_gethdr(M_NOWAIT, MT_DATA);
2082 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2086 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2087 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2091 * XXX do we need to set the address in k_igmpsrc ?
2093 MRTSTAT_INC(mrts_upcalls);
2094 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2095 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2096 MRTSTAT_INC(mrts_upq_sockfull);
2101 * Compute the timeout hash value for the bw_meter entries
2103 #define BW_METER_TIMEHASH(bw_meter, hash) \
2105 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2107 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2108 (hash) = next_timeval.tv_sec; \
2109 if (next_timeval.tv_usec) \
2110 (hash)++; /* XXX: make sure we don't timeout early */ \
2111 (hash) %= BW_METER_BUCKETS; \
2115 * Schedule a timer to process periodically bw_meter entry of type "<="
2116 * by linking the entry in the proper hash bucket.
2119 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2125 if (!(x->bm_flags & BW_METER_LEQ))
2126 return; /* XXX: we schedule timers only for "<=" entries */
2129 * Reset the bw_meter entry
2131 x->bm_start_time = *nowp;
2132 x->bm_measured.b_packets = 0;
2133 x->bm_measured.b_bytes = 0;
2134 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2137 * Compute the timeout hash value and insert the entry
2139 BW_METER_TIMEHASH(x, time_hash);
2140 x->bm_time_next = V_bw_meter_timers[time_hash];
2141 V_bw_meter_timers[time_hash] = x;
2142 x->bm_time_hash = time_hash;
2146 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2147 * by removing the entry from the proper hash bucket.
2150 unschedule_bw_meter(struct bw_meter *x)
2153 struct bw_meter *prev, *tmp;
2157 if (!(x->bm_flags & BW_METER_LEQ))
2158 return; /* XXX: we schedule timers only for "<=" entries */
2161 * Compute the timeout hash value and delete the entry
2163 time_hash = x->bm_time_hash;
2164 if (time_hash >= BW_METER_BUCKETS)
2165 return; /* Entry was not scheduled */
2167 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2168 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2173 panic("unschedule_bw_meter: bw_meter entry not found");
2176 prev->bm_time_next = x->bm_time_next;
2178 V_bw_meter_timers[time_hash] = x->bm_time_next;
2180 x->bm_time_next = NULL;
2181 x->bm_time_hash = BW_METER_BUCKETS;
2186 * Process all "<=" type of bw_meter that should be processed now,
2187 * and for each entry prepare an upcall if necessary. Each processed
2188 * entry is rescheduled again for the (periodic) processing.
2190 * This is run periodically (once per second normally). On each round,
2191 * all the potentially matching entries are in the hash slot that we are
2199 struct timeval now, process_endtime;
2202 if (V_last_tv_sec == now.tv_sec)
2203 return; /* nothing to do */
2205 loops = now.tv_sec - V_last_tv_sec;
2206 V_last_tv_sec = now.tv_sec;
2207 if (loops > BW_METER_BUCKETS)
2208 loops = BW_METER_BUCKETS;
2212 * Process all bins of bw_meter entries from the one after the last
2213 * processed to the current one. On entry, i points to the last bucket
2214 * visited, so we need to increment i at the beginning of the loop.
2216 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2217 struct bw_meter *x, *tmp_list;
2219 if (++i >= BW_METER_BUCKETS)
2222 /* Disconnect the list of bw_meter entries from the bin */
2223 tmp_list = V_bw_meter_timers[i];
2224 V_bw_meter_timers[i] = NULL;
2226 /* Process the list of bw_meter entries */
2227 while (tmp_list != NULL) {
2229 tmp_list = tmp_list->bm_time_next;
2231 /* Test if the time interval is over */
2232 process_endtime = x->bm_start_time;
2233 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2234 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2235 /* Not yet: reschedule, but don't reset */
2238 BW_METER_TIMEHASH(x, time_hash);
2239 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2241 * XXX: somehow the bin processing is a bit ahead of time.
2242 * Put the entry in the next bin.
2244 if (++time_hash >= BW_METER_BUCKETS)
2247 x->bm_time_next = V_bw_meter_timers[time_hash];
2248 V_bw_meter_timers[time_hash] = x;
2249 x->bm_time_hash = time_hash;
2255 * Test if we should deliver an upcall
2257 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2258 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2259 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2260 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2261 /* Prepare an upcall for delivery */
2262 bw_meter_prepare_upcall(x, &now);
2266 * Reschedule for next processing
2268 schedule_bw_meter(x, &now);
2272 /* Send all upcalls that are pending delivery */
2279 * A periodic function for sending all upcalls that are pending delivery
2282 expire_bw_upcalls_send(void *arg)
2284 CURVNET_SET((struct vnet *) arg);
2290 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2296 * A periodic function for periodic scanning of the multicast forwarding
2297 * table for processing all "<=" bw_meter entries.
2300 expire_bw_meter_process(void *arg)
2302 CURVNET_SET((struct vnet *) arg);
2304 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2307 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2313 * End of bandwidth monitoring code
2317 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2321 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2324 struct mbuf *mb_copy, *mm;
2327 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2328 * rendezvous point was unspecified, and we were told not to.
2330 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2331 in_nullhost(rt->mfc_rp))
2334 mb_copy = pim_register_prepare(ip, m);
2335 if (mb_copy == NULL)
2339 * Send all the fragments. Note that the mbuf for each fragment
2340 * is freed by the sending machinery.
2342 for (mm = mb_copy; mm; mm = mb_copy) {
2343 mb_copy = mm->m_nextpkt;
2345 mm = m_pullup(mm, sizeof(struct ip));
2347 ip = mtod(mm, struct ip *);
2348 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2349 pim_register_send_rp(ip, vifp, mm, rt);
2351 pim_register_send_upcall(ip, vifp, mm, rt);
2360 * Return a copy of the data packet that is ready for PIM Register
2362 * XXX: Note that in the returned copy the IP header is a valid one.
2364 static struct mbuf *
2365 pim_register_prepare(struct ip *ip, struct mbuf *m)
2367 struct mbuf *mb_copy = NULL;
2370 /* Take care of delayed checksums */
2371 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2372 in_delayed_cksum(m);
2373 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2377 * Copy the old packet & pullup its IP header into the
2378 * new mbuf so we can modify it.
2380 mb_copy = m_copypacket(m, M_NOWAIT);
2381 if (mb_copy == NULL)
2383 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2384 if (mb_copy == NULL)
2387 /* take care of the TTL */
2388 ip = mtod(mb_copy, struct ip *);
2391 /* Compute the MTU after the PIM Register encapsulation */
2392 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2394 if (ntohs(ip->ip_len) <= mtu) {
2395 /* Turn the IP header into a valid one */
2397 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2399 /* Fragment the packet */
2400 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2401 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2410 * Send an upcall with the data packet to the user-level process.
2413 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2414 struct mbuf *mb_copy, struct mfc *rt)
2416 struct mbuf *mb_first;
2417 int len = ntohs(ip->ip_len);
2419 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2424 * Add a new mbuf with an upcall header
2426 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2427 if (mb_first == NULL) {
2431 mb_first->m_data += max_linkhdr;
2432 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2433 mb_first->m_len = sizeof(struct igmpmsg);
2434 mb_first->m_next = mb_copy;
2436 /* Send message to routing daemon */
2437 im = mtod(mb_first, struct igmpmsg *);
2438 im->im_msgtype = IGMPMSG_WHOLEPKT;
2440 im->im_vif = vifp - V_viftable;
2441 im->im_src = ip->ip_src;
2442 im->im_dst = ip->ip_dst;
2444 k_igmpsrc.sin_addr = ip->ip_src;
2446 MRTSTAT_INC(mrts_upcalls);
2448 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2449 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2450 MRTSTAT_INC(mrts_upq_sockfull);
2454 /* Keep statistics */
2455 PIMSTAT_INC(pims_snd_registers_msgs);
2456 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2462 * Encapsulate the data packet in PIM Register message and send it to the RP.
2465 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2468 struct mbuf *mb_first;
2469 struct ip *ip_outer;
2470 struct pim_encap_pimhdr *pimhdr;
2471 int len = ntohs(ip->ip_len);
2472 vifi_t vifi = rt->mfc_parent;
2476 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2478 return EADDRNOTAVAIL; /* The iif vif is invalid */
2482 * Add a new mbuf with the encapsulating header
2484 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2485 if (mb_first == NULL) {
2489 mb_first->m_data += max_linkhdr;
2490 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2491 mb_first->m_next = mb_copy;
2493 mb_first->m_pkthdr.len = len + mb_first->m_len;
2496 * Fill in the encapsulating IP and PIM header
2498 ip_outer = mtod(mb_first, struct ip *);
2499 *ip_outer = pim_encap_iphdr;
2500 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2501 sizeof(pim_encap_pimhdr));
2502 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2503 ip_outer->ip_dst = rt->mfc_rp;
2505 * Copy the inner header TOS to the outer header, and take care of the
2508 ip_outer->ip_tos = ip->ip_tos;
2509 if (ip->ip_off & htons(IP_DF))
2510 ip_outer->ip_off |= htons(IP_DF);
2511 ip_fillid(ip_outer);
2512 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2513 + sizeof(pim_encap_iphdr));
2514 *pimhdr = pim_encap_pimhdr;
2515 /* If the iif crosses a border, set the Border-bit */
2516 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2517 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2519 mb_first->m_data += sizeof(pim_encap_iphdr);
2520 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2521 mb_first->m_data -= sizeof(pim_encap_iphdr);
2523 send_packet(vifp, mb_first);
2525 /* Keep statistics */
2526 PIMSTAT_INC(pims_snd_registers_msgs);
2527 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2533 * pim_encapcheck() is called by the encap4_input() path at runtime to
2534 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2538 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2539 int proto __unused, void *arg __unused)
2542 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2543 return (8); /* claim the datagram. */
2547 * PIM-SMv2 and PIM-DM messages processing.
2548 * Receives and verifies the PIM control messages, and passes them
2549 * up to the listening socket, using rip_input().
2550 * The only message with special processing is the PIM_REGISTER message
2551 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2552 * is passed to if_simloop().
2555 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2557 struct ip *ip = mtod(m, struct ip *);
2561 int datalen = ntohs(ip->ip_len) - iphlen;
2564 /* Keep statistics */
2565 PIMSTAT_INC(pims_rcv_total_msgs);
2566 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2571 if (datalen < PIM_MINLEN) {
2572 PIMSTAT_INC(pims_rcv_tooshort);
2573 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2574 __func__, datalen, ntohl(ip->ip_src.s_addr));
2576 return (IPPROTO_DONE);
2580 * If the packet is at least as big as a REGISTER, go agead
2581 * and grab the PIM REGISTER header size, to avoid another
2582 * possible m_pullup() later.
2584 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2585 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2587 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2589 * Get the IP and PIM headers in contiguous memory, and
2590 * possibly the PIM REGISTER header.
2592 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2593 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2594 return (IPPROTO_DONE);
2597 /* m_pullup() may have given us a new mbuf so reset ip. */
2598 ip = mtod(m, struct ip *);
2599 ip_tos = ip->ip_tos;
2601 /* adjust mbuf to point to the PIM header */
2602 m->m_data += iphlen;
2604 pim = mtod(m, struct pim *);
2607 * Validate checksum. If PIM REGISTER, exclude the data packet.
2609 * XXX: some older PIMv2 implementations don't make this distinction,
2610 * so for compatibility reason perform the checksum over part of the
2611 * message, and if error, then over the whole message.
2613 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2614 /* do nothing, checksum okay */
2615 } else if (in_cksum(m, datalen)) {
2616 PIMSTAT_INC(pims_rcv_badsum);
2617 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2619 return (IPPROTO_DONE);
2622 /* PIM version check */
2623 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2624 PIMSTAT_INC(pims_rcv_badversion);
2625 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2626 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2628 return (IPPROTO_DONE);
2631 /* restore mbuf back to the outer IP */
2632 m->m_data -= iphlen;
2635 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2637 * Since this is a REGISTER, we'll make a copy of the register
2638 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2641 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2643 struct ip *encap_ip;
2648 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2650 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2651 (int)V_reg_vif_num);
2653 return (IPPROTO_DONE);
2655 /* XXX need refcnt? */
2656 vifp = V_viftable[V_reg_vif_num].v_ifp;
2662 if (datalen < PIM_REG_MINLEN) {
2663 PIMSTAT_INC(pims_rcv_tooshort);
2664 PIMSTAT_INC(pims_rcv_badregisters);
2665 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2667 return (IPPROTO_DONE);
2670 reghdr = (u_int32_t *)(pim + 1);
2671 encap_ip = (struct ip *)(reghdr + 1);
2673 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2674 __func__, ntohl(encap_ip->ip_src.s_addr),
2675 ntohs(encap_ip->ip_len));
2677 /* verify the version number of the inner packet */
2678 if (encap_ip->ip_v != IPVERSION) {
2679 PIMSTAT_INC(pims_rcv_badregisters);
2680 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2682 return (IPPROTO_DONE);
2685 /* verify the inner packet is destined to a mcast group */
2686 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2687 PIMSTAT_INC(pims_rcv_badregisters);
2688 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2689 ntohl(encap_ip->ip_dst.s_addr));
2691 return (IPPROTO_DONE);
2694 /* If a NULL_REGISTER, pass it to the daemon */
2695 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2696 goto pim_input_to_daemon;
2699 * Copy the TOS from the outer IP header to the inner IP header.
2701 if (encap_ip->ip_tos != ip_tos) {
2702 /* Outer TOS -> inner TOS */
2703 encap_ip->ip_tos = ip_tos;
2704 /* Recompute the inner header checksum. Sigh... */
2706 /* adjust mbuf to point to the inner IP header */
2707 m->m_data += (iphlen + PIM_MINLEN);
2708 m->m_len -= (iphlen + PIM_MINLEN);
2710 encap_ip->ip_sum = 0;
2711 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2713 /* restore mbuf to point back to the outer IP header */
2714 m->m_data -= (iphlen + PIM_MINLEN);
2715 m->m_len += (iphlen + PIM_MINLEN);
2719 * Decapsulate the inner IP packet and loopback to forward it
2720 * as a normal multicast packet. Also, make a copy of the
2721 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2722 * to pass to the daemon later, so it can take the appropriate
2723 * actions (e.g., send back PIM_REGISTER_STOP).
2724 * XXX: here m->m_data points to the outer IP header.
2726 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2728 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2730 return (IPPROTO_DONE);
2733 /* Keep statistics */
2734 /* XXX: registers_bytes include only the encap. mcast pkt */
2735 PIMSTAT_INC(pims_rcv_registers_msgs);
2736 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2739 * forward the inner ip packet; point m_data at the inner ip.
2741 m_adj(m, iphlen + PIM_MINLEN);
2744 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2746 (u_long)ntohl(encap_ip->ip_src.s_addr),
2747 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2748 (int)V_reg_vif_num);
2750 /* NB: vifp was collected above; can it change on us? */
2751 if_simloop(vifp, m, dst.sin_family, 0);
2753 /* prepare the register head to send to the mrouting daemon */
2757 pim_input_to_daemon:
2759 * Pass the PIM message up to the daemon; if it is a Register message,
2760 * pass the 'head' only up to the daemon. This includes the
2761 * outer IP header, PIM header, PIM-Register header and the
2763 * XXX: the outer IP header pkt size of a Register is not adjust to
2764 * reflect the fact that the inner multicast data is truncated.
2766 return (rip_input(&m, &off, proto));
2770 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2777 if (V_mfchashtbl == NULL) /* XXX unlocked */
2779 error = sysctl_wire_old_buffer(req, 0);
2784 for (i = 0; i < mfchashsize; i++) {
2785 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2786 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2796 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
2797 sysctl_mfctable, "IPv4 Multicast Forwarding Table "
2798 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2801 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2807 if (V_viftable == NULL) /* XXX unlocked */
2809 error = sysctl_wire_old_buffer(req, sizeof(*V_viftable) * MAXVIFS);
2814 error = SYSCTL_OUT(req, V_viftable, sizeof(*V_viftable) * MAXVIFS);
2819 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2820 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2821 sysctl_viflist, "S,vif[MAXVIFS]",
2822 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2825 vnet_mroute_init(const void *unused __unused)
2828 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2830 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2831 M_MRTABLE, M_WAITOK|M_ZERO);
2832 V_bw_meter_timers = mallocarray(BW_METER_BUCKETS,
2833 sizeof(*V_bw_meter_timers), M_MRTABLE, M_WAITOK|M_ZERO);
2834 V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2835 M_MRTABLE, M_WAITOK|M_ZERO);
2837 callout_init(&V_expire_upcalls_ch, 1);
2838 callout_init(&V_bw_upcalls_ch, 1);
2839 callout_init(&V_bw_meter_ch, 1);
2842 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2846 vnet_mroute_uninit(const void *unused __unused)
2849 free(V_bw_upcalls, M_MRTABLE);
2850 free(V_bw_meter_timers, M_MRTABLE);
2851 free(V_viftable, M_MRTABLE);
2852 free(V_nexpire, M_MRTABLE);
2856 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2857 vnet_mroute_uninit, NULL);
2860 ip_mroute_modevent(module_t mod, int type, void *unused)
2865 MROUTER_LOCK_INIT();
2867 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2868 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2869 if (if_detach_event_tag == NULL) {
2870 printf("ip_mroute: unable to register "
2871 "ifnet_departure_event handler\n");
2872 MROUTER_LOCK_DESTROY();
2879 mfchashsize = MFCHASHSIZE;
2880 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2881 !powerof2(mfchashsize)) {
2882 printf("WARNING: %s not a power of 2; using default\n",
2883 "net.inet.ip.mfchashsize");
2884 mfchashsize = MFCHASHSIZE;
2887 pim_squelch_wholepkt = 0;
2888 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2889 &pim_squelch_wholepkt);
2891 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2892 if (pim_encap_cookie == NULL) {
2893 printf("ip_mroute: unable to attach pim encap\n");
2896 MROUTER_LOCK_DESTROY();
2900 ip_mcast_src = X_ip_mcast_src;
2901 ip_mforward = X_ip_mforward;
2902 ip_mrouter_done = X_ip_mrouter_done;
2903 ip_mrouter_get = X_ip_mrouter_get;
2904 ip_mrouter_set = X_ip_mrouter_set;
2906 ip_rsvp_force_done = X_ip_rsvp_force_done;
2907 ip_rsvp_vif = X_ip_rsvp_vif;
2909 legal_vif_num = X_legal_vif_num;
2910 mrt_ioctl = X_mrt_ioctl;
2911 rsvp_input_p = X_rsvp_input;
2916 * Typically module unload happens after the user-level
2917 * process has shutdown the kernel services (the check
2918 * below insures someone can't just yank the module out
2919 * from under a running process). But if the module is
2920 * just loaded and then unloaded w/o starting up a user
2921 * process we still need to cleanup.
2924 if (ip_mrouter_cnt != 0) {
2928 ip_mrouter_unloading = 1;
2931 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2933 if (pim_encap_cookie) {
2934 ip_encap_detach(pim_encap_cookie);
2935 pim_encap_cookie = NULL;
2938 ip_mcast_src = NULL;
2940 ip_mrouter_done = NULL;
2941 ip_mrouter_get = NULL;
2942 ip_mrouter_set = NULL;
2944 ip_rsvp_force_done = NULL;
2947 legal_vif_num = NULL;
2949 rsvp_input_p = NULL;
2953 MROUTER_LOCK_DESTROY();
2962 static moduledata_t ip_mroutemod = {
2968 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);