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 | CTLFLAG_MPSAFE, 0,
235 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat,
236 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)");
238 static u_long pim_squelch_wholepkt = 0;
239 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW,
240 &pim_squelch_wholepkt, 0,
241 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified");
243 static const struct encaptab *pim_encap_cookie;
244 static int pim_encapcheck(const struct mbuf *, int, int, void *);
245 static int pim_input(struct mbuf *, int, int, void *);
247 static const struct encap_config ipv4_encap_cfg = {
248 .proto = IPPROTO_PIM,
249 .min_length = sizeof(struct ip) + PIM_MINLEN,
251 .check = pim_encapcheck,
256 * Note: the PIM Register encapsulation adds the following in front of a
259 * struct pim_encap_hdr {
261 * struct pim_encap_pimhdr pim;
266 struct pim_encap_pimhdr {
270 #define PIM_ENCAP_TTL 64
272 static struct ip pim_encap_iphdr = {
273 #if BYTE_ORDER == LITTLE_ENDIAN
274 sizeof(struct ip) >> 2,
278 sizeof(struct ip) >> 2,
281 sizeof(struct ip), /* total length */
289 static struct pim_encap_pimhdr pim_encap_pimhdr = {
291 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
298 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID;
299 #define V_reg_vif_num VNET(reg_vif_num)
300 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if);
301 #define V_multicast_register_if VNET(multicast_register_if)
307 static u_long X_ip_mcast_src(int);
308 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *,
309 struct ip_moptions *);
310 static int X_ip_mrouter_done(void);
311 static int X_ip_mrouter_get(struct socket *, struct sockopt *);
312 static int X_ip_mrouter_set(struct socket *, struct sockopt *);
313 static int X_legal_vif_num(int);
314 static int X_mrt_ioctl(u_long, caddr_t, int);
316 static int add_bw_upcall(struct bw_upcall *);
317 static int add_mfc(struct mfcctl2 *);
318 static int add_vif(struct vifctl *);
319 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *);
320 static void bw_meter_process(void);
321 static void bw_meter_receive_packet(struct bw_meter *, int,
323 static void bw_upcalls_send(void);
324 static int del_bw_upcall(struct bw_upcall *);
325 static int del_mfc(struct mfcctl2 *);
326 static int del_vif(vifi_t);
327 static int del_vif_locked(vifi_t);
328 static void expire_bw_meter_process(void *);
329 static void expire_bw_upcalls_send(void *);
330 static void expire_mfc(struct mfc *);
331 static void expire_upcalls(void *);
332 static void free_bw_list(struct bw_meter *);
333 static int get_sg_cnt(struct sioc_sg_req *);
334 static int get_vif_cnt(struct sioc_vif_req *);
335 static void if_detached_event(void *, struct ifnet *);
336 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
337 static int ip_mrouter_init(struct socket *, int);
338 static __inline struct mfc *
339 mfc_find(struct in_addr *, struct in_addr *);
340 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
342 pim_register_prepare(struct ip *, struct mbuf *);
343 static int pim_register_send(struct ip *, struct vif *,
344 struct mbuf *, struct mfc *);
345 static int pim_register_send_rp(struct ip *, struct vif *,
346 struct mbuf *, struct mfc *);
347 static int pim_register_send_upcall(struct ip *, struct vif *,
348 struct mbuf *, struct mfc *);
349 static void schedule_bw_meter(struct bw_meter *, struct timeval *);
350 static void send_packet(struct vif *, struct mbuf *);
351 static int set_api_config(uint32_t *);
352 static int set_assert(int);
353 static int socket_send(struct socket *, struct mbuf *,
354 struct sockaddr_in *);
355 static void unschedule_bw_meter(struct bw_meter *);
358 * Kernel multicast forwarding API capabilities and setup.
359 * If more API capabilities are added to the kernel, they should be
360 * recorded in `mrt_api_support'.
362 #define MRT_API_VERSION 0x0305
364 static const int mrt_api_version = MRT_API_VERSION;
365 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
366 MRT_MFC_FLAGS_BORDER_VIF |
369 VNET_DEFINE_STATIC(uint32_t, mrt_api_config);
370 #define V_mrt_api_config VNET(mrt_api_config)
371 VNET_DEFINE_STATIC(int, pim_assert_enabled);
372 #define V_pim_assert_enabled VNET(pim_assert_enabled)
373 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */
376 * Find a route for a given origin IP address and multicast group address.
377 * Statistics must be updated by the caller.
379 static __inline struct mfc *
380 mfc_find(struct in_addr *o, struct in_addr *g)
386 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) {
387 if (in_hosteq(rt->mfc_origin, *o) &&
388 in_hosteq(rt->mfc_mcastgrp, *g) &&
389 TAILQ_EMPTY(&rt->mfc_stall))
397 * Handle MRT setsockopt commands to modify the multicast forwarding tables.
400 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
406 struct bw_upcall bw_upcall;
409 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT)
413 switch (sopt->sopt_name) {
415 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
418 error = ip_mrouter_init(so, optval);
422 error = ip_mrouter_done();
426 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
429 error = add_vif(&vifc);
433 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
436 error = del_vif(vifi);
442 * select data size depending on API version.
444 if (sopt->sopt_name == MRT_ADD_MFC &&
445 V_mrt_api_config & MRT_API_FLAGS_ALL) {
446 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
447 sizeof(struct mfcctl2));
449 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
450 sizeof(struct mfcctl));
451 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
452 sizeof(mfc) - sizeof(struct mfcctl));
456 if (sopt->sopt_name == MRT_ADD_MFC)
457 error = add_mfc(&mfc);
459 error = del_mfc(&mfc);
463 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
470 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
472 error = set_api_config(&i);
474 error = sooptcopyout(sopt, &i, sizeof i);
477 case MRT_ADD_BW_UPCALL:
478 case MRT_DEL_BW_UPCALL:
479 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
483 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
484 error = add_bw_upcall(&bw_upcall);
486 error = del_bw_upcall(&bw_upcall);
497 * Handle MRT getsockopt commands
500 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
504 switch (sopt->sopt_name) {
506 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version);
510 error = sooptcopyout(sopt, &V_pim_assert_enabled,
511 sizeof V_pim_assert_enabled);
514 case MRT_API_SUPPORT:
515 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
519 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config);
530 * Handle ioctl commands to obtain information from the cache
533 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused)
538 * Currently the only function calling this ioctl routine is rtioctl_fib().
539 * Typically, only root can create the raw socket in order to execute
540 * this ioctl method, however the request might be coming from a prison
542 error = priv_check(curthread, PRIV_NETINET_MROUTE);
546 case (SIOCGETVIFCNT):
547 error = get_vif_cnt((struct sioc_vif_req *)data);
551 error = get_sg_cnt((struct sioc_sg_req *)data);
562 * returns the packet, byte, rpf-failure count for the source group provided
565 get_sg_cnt(struct sioc_sg_req *req)
570 rt = mfc_find(&req->src, &req->grp);
573 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
574 return EADDRNOTAVAIL;
576 req->pktcnt = rt->mfc_pkt_cnt;
577 req->bytecnt = rt->mfc_byte_cnt;
578 req->wrong_if = rt->mfc_wrong_if;
584 * returns the input and output packet and byte counts on the vif provided
587 get_vif_cnt(struct sioc_vif_req *req)
589 vifi_t vifi = req->vifi;
592 if (vifi >= V_numvifs) {
597 req->icount = V_viftable[vifi].v_pkt_in;
598 req->ocount = V_viftable[vifi].v_pkt_out;
599 req->ibytes = V_viftable[vifi].v_bytes_in;
600 req->obytes = V_viftable[vifi].v_bytes_out;
607 if_detached_event(void *arg __unused, struct ifnet *ifp)
614 if (V_ip_mrouter == NULL) {
623 * Tear down multicast forwarder state associated with this ifnet.
624 * 1. Walk the vif list, matching vifs against this ifnet.
625 * 2. Walk the multicast forwarding cache (mfc) looking for
626 * inner matches with this vif's index.
627 * 3. Expire any matching multicast forwarding cache entries.
628 * 4. Free vif state. This should disable ALLMULTI on the interface.
630 for (vifi = 0; vifi < V_numvifs; vifi++) {
631 if (V_viftable[vifi].v_ifp != ifp)
633 for (i = 0; i < mfchashsize; i++) {
634 struct mfc *rt, *nrt;
636 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
637 if (rt->mfc_parent == vifi) {
642 del_vif_locked(vifi);
652 * Enable multicast forwarding.
655 ip_mrouter_init(struct socket *so, int version)
658 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__,
659 so->so_type, so->so_proto->pr_protocol);
661 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
669 if (ip_mrouter_unloading) {
674 if (V_ip_mrouter != NULL) {
679 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash,
682 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
684 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
686 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
694 CTR1(KTR_IPMF, "%s: done", __func__);
700 * Disable multicast forwarding.
703 X_ip_mrouter_done(void)
711 if (V_ip_mrouter == NULL) {
717 * Detach/disable hooks to the reset of the system.
721 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 ifp = V_viftable[vifi].v_ifp;
738 bzero((caddr_t)V_viftable, sizeof(*V_viftable) * MAXVIFS);
740 V_pim_assert_enabled = 0;
744 callout_stop(&V_expire_upcalls_ch);
745 callout_stop(&V_bw_upcalls_ch);
746 callout_stop(&V_bw_meter_ch);
751 * Free all multicast forwarding cache entries.
752 * Do not use hashdestroy(), as we must perform other cleanup.
754 for (i = 0; i < mfchashsize; i++) {
755 struct mfc *rt, *nrt;
757 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
761 free(V_mfchashtbl, M_MRTABLE);
764 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize);
767 bzero(V_bw_meter_timers, BW_METER_BUCKETS * sizeof(*V_bw_meter_timers));
771 V_reg_vif_num = VIFI_INVALID;
775 CTR1(KTR_IPMF, "%s: done", __func__);
781 * Set PIM assert processing global
786 if ((i != 1) && (i != 0))
789 V_pim_assert_enabled = i;
795 * Configure API capabilities
798 set_api_config(uint32_t *apival)
803 * We can set the API capabilities only if it is the first operation
804 * after MRT_INIT. I.e.:
805 * - there are no vifs installed
806 * - pim_assert is not enabled
807 * - the MFC table is empty
813 if (V_pim_assert_enabled) {
820 for (i = 0; i < mfchashsize; i++) {
821 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) {
830 V_mrt_api_config = *apival & mrt_api_support;
831 *apival = V_mrt_api_config;
837 * Add a vif to the vif table
840 add_vif(struct vifctl *vifcp)
842 struct vif *vifp = V_viftable + vifcp->vifc_vifi;
843 struct sockaddr_in sin = {sizeof sin, AF_INET};
849 if (vifcp->vifc_vifi >= MAXVIFS) {
853 /* rate limiting is no longer supported by this code */
854 if (vifcp->vifc_rate_limit != 0) {
855 log(LOG_ERR, "rate limiting is no longer supported\n");
859 if (!in_nullhost(vifp->v_lcl_addr)) {
863 if (in_nullhost(vifcp->vifc_lcl_addr)) {
865 return EADDRNOTAVAIL;
868 /* Find the interface with an address in AF_INET family */
869 if (vifcp->vifc_flags & VIFF_REGISTER) {
871 * XXX: Because VIFF_REGISTER does not really need a valid
872 * local interface (e.g. it could be 127.0.0.2), we don't
877 struct epoch_tracker et;
879 sin.sin_addr = vifcp->vifc_lcl_addr;
881 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
885 return EADDRNOTAVAIL;
888 /* XXX FIXME we need to take a ref on ifp and cleanup properly! */
892 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) {
893 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__);
896 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
897 ifp = &V_multicast_register_if;
898 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp);
899 if (V_reg_vif_num == VIFI_INVALID) {
900 if_initname(&V_multicast_register_if, "register_vif", 0);
901 V_multicast_register_if.if_flags = IFF_LOOPBACK;
902 V_reg_vif_num = vifcp->vifc_vifi;
904 } else { /* Make sure the interface supports multicast */
905 if ((ifp->if_flags & IFF_MULTICAST) == 0) {
910 /* Enable promiscuous reception of all IP multicasts from the if */
911 error = if_allmulti(ifp, 1);
918 vifp->v_flags = vifcp->vifc_flags;
919 vifp->v_threshold = vifcp->vifc_threshold;
920 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
921 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
923 /* initialize per vif pkt counters */
926 vifp->v_bytes_in = 0;
927 vifp->v_bytes_out = 0;
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 0x%08x thresh %x", __func__,
936 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_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 0x%08x group %lx parent %x",
1068 __func__, ntohl(mfccp->mfcc_origin.s_addr),
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 0x%08x group %lx parent %x qh %p",
1088 __func__, ntohl(mfccp->mfcc_origin.s_addr),
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 0x%08x group %lx", __func__,
1167 ntohl(origin.s_addr), (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 0x%08x group %lx ifp %p",
1232 ntohl(ip->ip_src.s_addr), (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 (0x%08x,%lx)",
1295 ntohl(ip->ip_src.s_addr), (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_NOWAIT);
1311 if (mb0 && (!M_WRITABLE(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_copym(mb0, 0, hlen, M_NOWAIT);
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 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) {
1454 if (TAILQ_EMPTY(&rt->mfc_stall))
1457 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0)
1461 * free the bw_meter entries
1463 while (rt->mfc_bw_meter != NULL) {
1464 struct bw_meter *x = rt->mfc_bw_meter;
1466 rt->mfc_bw_meter = x->bm_mfc_next;
1470 MRTSTAT_INC(mrts_cache_cleanups);
1471 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__,
1472 (u_long)ntohl(rt->mfc_origin.s_addr),
1473 (u_long)ntohl(rt->mfc_mcastgrp.s_addr));
1481 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls,
1488 * Packet forwarding routine once entry in the cache is made
1491 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1493 struct ip *ip = mtod(m, struct ip *);
1495 int plen = ntohs(ip->ip_len);
1500 * If xmt_vif is not -1, send on only the requested vif.
1502 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1504 if (xmt_vif < V_numvifs) {
1505 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER)
1506 pim_register_send(ip, V_viftable + xmt_vif, m, rt);
1508 phyint_send(ip, V_viftable + xmt_vif, m);
1513 * Don't forward if it didn't arrive from the parent vif for its origin.
1515 vifi = rt->mfc_parent;
1516 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) {
1517 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)",
1518 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp);
1519 MRTSTAT_INC(mrts_wrong_if);
1522 * If we are doing PIM assert processing, send a message
1523 * to the routing daemon.
1525 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1526 * can complete the SPT switch, regardless of the type
1527 * of the iif (broadcast media, GRE tunnel, etc).
1529 if (V_pim_assert_enabled && (vifi < V_numvifs) &&
1530 V_viftable[vifi].v_ifp) {
1531 if (ifp == &V_multicast_register_if)
1532 PIMSTAT_INC(pims_rcv_registers_wrongiif);
1534 /* Get vifi for the incoming packet */
1535 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp;
1538 if (vifi >= V_numvifs)
1539 return 0; /* The iif is not found: ignore the packet. */
1541 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1542 return 0; /* WRONGVIF disabled: ignore the packet */
1544 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) {
1545 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1547 int hlen = ip->ip_hl << 2;
1548 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT);
1550 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen))
1551 mm = m_pullup(mm, hlen);
1555 im = mtod(mm, struct igmpmsg *);
1556 im->im_msgtype = IGMPMSG_WRONGVIF;
1560 MRTSTAT_INC(mrts_upcalls);
1562 k_igmpsrc.sin_addr = im->im_src;
1563 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) {
1564 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
1565 MRTSTAT_INC(mrts_upq_sockfull);
1573 /* If I sourced this packet, it counts as output, else it was input. */
1574 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) {
1575 V_viftable[vifi].v_pkt_out++;
1576 V_viftable[vifi].v_bytes_out += plen;
1578 V_viftable[vifi].v_pkt_in++;
1579 V_viftable[vifi].v_bytes_in += plen;
1582 rt->mfc_byte_cnt += plen;
1585 * For each vif, decide if a copy of the packet should be forwarded.
1587 * - the ttl exceeds the vif's threshold
1588 * - there are group members downstream on interface
1590 for (vifi = 0; vifi < V_numvifs; vifi++)
1591 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1592 V_viftable[vifi].v_pkt_out++;
1593 V_viftable[vifi].v_bytes_out += plen;
1594 if (V_viftable[vifi].v_flags & VIFF_REGISTER)
1595 pim_register_send(ip, V_viftable + vifi, m, rt);
1597 phyint_send(ip, V_viftable + vifi, m);
1601 * Perform upcall-related bw measuring.
1603 if (rt->mfc_bw_meter != NULL) {
1609 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1610 bw_meter_receive_packet(x, plen, &now);
1617 * Check if a vif number is legal/ok. This is used by in_mcast.c.
1620 X_legal_vif_num(int vif)
1629 if (vif < V_numvifs)
1637 * Return the local address used by this vif
1640 X_ip_mcast_src(int vifi)
1649 if (vifi < V_numvifs)
1650 addr = V_viftable[vifi].v_lcl_addr.s_addr;
1657 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1659 struct mbuf *mb_copy;
1660 int hlen = ip->ip_hl << 2;
1665 * Make a new reference to the packet; make sure that
1666 * the IP header is actually copied, not just referenced,
1667 * so that ip_output() only scribbles on the copy.
1669 mb_copy = m_copypacket(m, M_NOWAIT);
1670 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen))
1671 mb_copy = m_pullup(mb_copy, hlen);
1672 if (mb_copy == NULL)
1675 send_packet(vifp, mb_copy);
1679 send_packet(struct vif *vifp, struct mbuf *m)
1681 struct ip_moptions imo;
1686 imo.imo_multicast_ifp = vifp->v_ifp;
1687 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1688 imo.imo_multicast_loop = 1;
1689 imo.imo_multicast_vif = -1;
1690 STAILQ_INIT(&imo.imo_head);
1693 * Re-entrancy should not be a problem here, because
1694 * the packets that we send out and are looped back at us
1695 * should get rejected because they appear to come from
1696 * the loopback interface, thus preventing looping.
1698 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL);
1699 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__,
1700 (ptrdiff_t)(vifp - V_viftable), error);
1704 * Stubs for old RSVP socket shim implementation.
1708 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused)
1711 return (EOPNOTSUPP);
1715 X_ip_rsvp_force_done(struct socket *so __unused)
1721 X_rsvp_input(struct mbuf **mp, int *offp, int proto)
1729 return (IPPROTO_DONE);
1733 * Code for bandwidth monitors
1737 * Define common interface for timeval-related methods
1739 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
1740 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
1741 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
1744 compute_bw_meter_flags(struct bw_upcall *req)
1748 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
1749 flags |= BW_METER_UNIT_PACKETS;
1750 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
1751 flags |= BW_METER_UNIT_BYTES;
1752 if (req->bu_flags & BW_UPCALL_GEQ)
1753 flags |= BW_METER_GEQ;
1754 if (req->bu_flags & BW_UPCALL_LEQ)
1755 flags |= BW_METER_LEQ;
1761 * Add a bw_meter entry
1764 add_bw_upcall(struct bw_upcall *req)
1767 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
1768 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
1773 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1776 /* Test if the flags are valid */
1777 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
1779 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
1781 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1782 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
1785 /* Test if the threshold time interval is valid */
1786 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
1789 flags = compute_bw_meter_flags(req);
1792 * Find if we have already same bw_meter entry
1795 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1798 return EADDRNOTAVAIL;
1800 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
1801 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1802 &req->bu_threshold.b_time, ==)) &&
1803 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1804 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1805 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
1807 return 0; /* XXX Already installed */
1811 /* Allocate the new bw_meter entry */
1812 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT);
1818 /* Set the new bw_meter entry */
1819 x->bm_threshold.b_time = req->bu_threshold.b_time;
1821 x->bm_start_time = now;
1822 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
1823 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
1824 x->bm_measured.b_packets = 0;
1825 x->bm_measured.b_bytes = 0;
1826 x->bm_flags = flags;
1827 x->bm_time_next = NULL;
1828 x->bm_time_hash = BW_METER_BUCKETS;
1830 /* Add the new bw_meter entry to the front of entries for this MFC */
1832 x->bm_mfc_next = mfc->mfc_bw_meter;
1833 mfc->mfc_bw_meter = x;
1834 schedule_bw_meter(x, &now);
1841 free_bw_list(struct bw_meter *list)
1843 while (list != NULL) {
1844 struct bw_meter *x = list;
1846 list = list->bm_mfc_next;
1847 unschedule_bw_meter(x);
1853 * Delete one or multiple bw_meter entries
1856 del_bw_upcall(struct bw_upcall *req)
1861 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL))
1866 /* Find the corresponding MFC entry */
1867 mfc = mfc_find(&req->bu_src, &req->bu_dst);
1870 return EADDRNOTAVAIL;
1871 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
1873 * Delete all bw_meter entries for this mfc
1875 struct bw_meter *list;
1877 list = mfc->mfc_bw_meter;
1878 mfc->mfc_bw_meter = NULL;
1882 } else { /* Delete a single bw_meter entry */
1883 struct bw_meter *prev;
1886 flags = compute_bw_meter_flags(req);
1888 /* Find the bw_meter entry to delete */
1889 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
1890 prev = x, x = x->bm_mfc_next) {
1891 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
1892 &req->bu_threshold.b_time, ==)) &&
1893 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
1894 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
1895 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
1898 if (x != NULL) { /* Delete entry from the list for this MFC */
1900 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
1902 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
1904 unschedule_bw_meter(x);
1906 /* Free the bw_meter entry */
1918 * Perform bandwidth measurement processing that may result in an upcall
1921 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
1923 struct timeval delta;
1928 BW_TIMEVALDECR(&delta, &x->bm_start_time);
1930 if (x->bm_flags & BW_METER_GEQ) {
1932 * Processing for ">=" type of bw_meter entry
1934 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1935 /* Reset the bw_meter entry */
1936 x->bm_start_time = *nowp;
1937 x->bm_measured.b_packets = 0;
1938 x->bm_measured.b_bytes = 0;
1939 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
1942 /* Record that a packet is received */
1943 x->bm_measured.b_packets++;
1944 x->bm_measured.b_bytes += plen;
1947 * Test if we should deliver an upcall
1949 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
1950 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1951 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
1952 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1953 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
1954 /* Prepare an upcall for delivery */
1955 bw_meter_prepare_upcall(x, nowp);
1956 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
1959 } else if (x->bm_flags & BW_METER_LEQ) {
1961 * Processing for "<=" type of bw_meter entry
1963 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
1965 * We are behind time with the multicast forwarding table
1966 * scanning for "<=" type of bw_meter entries, so test now
1967 * if we should deliver an upcall.
1969 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
1970 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
1971 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
1972 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
1973 /* Prepare an upcall for delivery */
1974 bw_meter_prepare_upcall(x, nowp);
1976 /* Reschedule the bw_meter entry */
1977 unschedule_bw_meter(x);
1978 schedule_bw_meter(x, nowp);
1981 /* Record that a packet is received */
1982 x->bm_measured.b_packets++;
1983 x->bm_measured.b_bytes += plen;
1986 * Test if we should restart the measuring interval
1988 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
1989 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
1990 (x->bm_flags & BW_METER_UNIT_BYTES &&
1991 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
1992 /* Don't restart the measuring interval */
1994 /* Do restart the measuring interval */
1996 * XXX: note that we don't unschedule and schedule, because this
1997 * might be too much overhead per packet. Instead, when we process
1998 * all entries for a given timer hash bin, we check whether it is
1999 * really a timeout. If not, we reschedule at that time.
2001 x->bm_start_time = *nowp;
2002 x->bm_measured.b_packets = 0;
2003 x->bm_measured.b_bytes = 0;
2004 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2010 * Prepare a bandwidth-related upcall
2013 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2015 struct timeval delta;
2016 struct bw_upcall *u;
2021 * Compute the measured time interval
2024 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2027 * If there are too many pending upcalls, deliver them now
2029 if (V_bw_upcalls_n >= BW_UPCALLS_MAX)
2033 * Set the bw_upcall entry
2035 u = &V_bw_upcalls[V_bw_upcalls_n++];
2036 u->bu_src = x->bm_mfc->mfc_origin;
2037 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2038 u->bu_threshold.b_time = x->bm_threshold.b_time;
2039 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2040 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2041 u->bu_measured.b_time = delta;
2042 u->bu_measured.b_packets = x->bm_measured.b_packets;
2043 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2045 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2046 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2047 if (x->bm_flags & BW_METER_UNIT_BYTES)
2048 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2049 if (x->bm_flags & BW_METER_GEQ)
2050 u->bu_flags |= BW_UPCALL_GEQ;
2051 if (x->bm_flags & BW_METER_LEQ)
2052 u->bu_flags |= BW_UPCALL_LEQ;
2056 * Send the pending bandwidth-related upcalls
2059 bw_upcalls_send(void)
2062 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]);
2063 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2064 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2066 IGMPMSG_BW_UPCALL,/* im_msgtype */
2071 { 0 } }; /* im_dst */
2075 if (V_bw_upcalls_n == 0)
2076 return; /* No pending upcalls */
2081 * Allocate a new mbuf, initialize it with the header and
2082 * the payload for the pending calls.
2084 m = m_gethdr(M_NOWAIT, MT_DATA);
2086 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2090 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2091 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]);
2095 * XXX do we need to set the address in k_igmpsrc ?
2097 MRTSTAT_INC(mrts_upcalls);
2098 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) {
2099 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2100 MRTSTAT_INC(mrts_upq_sockfull);
2105 * Compute the timeout hash value for the bw_meter entries
2107 #define BW_METER_TIMEHASH(bw_meter, hash) \
2109 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2111 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2112 (hash) = next_timeval.tv_sec; \
2113 if (next_timeval.tv_usec) \
2114 (hash)++; /* XXX: make sure we don't timeout early */ \
2115 (hash) %= BW_METER_BUCKETS; \
2119 * Schedule a timer to process periodically bw_meter entry of type "<="
2120 * by linking the entry in the proper hash bucket.
2123 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2129 if (!(x->bm_flags & BW_METER_LEQ))
2130 return; /* XXX: we schedule timers only for "<=" entries */
2133 * Reset the bw_meter entry
2135 x->bm_start_time = *nowp;
2136 x->bm_measured.b_packets = 0;
2137 x->bm_measured.b_bytes = 0;
2138 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2141 * Compute the timeout hash value and insert the entry
2143 BW_METER_TIMEHASH(x, time_hash);
2144 x->bm_time_next = V_bw_meter_timers[time_hash];
2145 V_bw_meter_timers[time_hash] = x;
2146 x->bm_time_hash = time_hash;
2150 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2151 * by removing the entry from the proper hash bucket.
2154 unschedule_bw_meter(struct bw_meter *x)
2157 struct bw_meter *prev, *tmp;
2161 if (!(x->bm_flags & BW_METER_LEQ))
2162 return; /* XXX: we schedule timers only for "<=" entries */
2165 * Compute the timeout hash value and delete the entry
2167 time_hash = x->bm_time_hash;
2168 if (time_hash >= BW_METER_BUCKETS)
2169 return; /* Entry was not scheduled */
2171 for (prev = NULL, tmp = V_bw_meter_timers[time_hash];
2172 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2177 panic("unschedule_bw_meter: bw_meter entry not found");
2180 prev->bm_time_next = x->bm_time_next;
2182 V_bw_meter_timers[time_hash] = x->bm_time_next;
2184 x->bm_time_next = NULL;
2185 x->bm_time_hash = BW_METER_BUCKETS;
2189 * Process all "<=" type of bw_meter that should be processed now,
2190 * and for each entry prepare an upcall if necessary. Each processed
2191 * entry is rescheduled again for the (periodic) processing.
2193 * This is run periodically (once per second normally). On each round,
2194 * all the potentially matching entries are in the hash slot that we are
2202 struct timeval now, process_endtime;
2205 if (V_last_tv_sec == now.tv_sec)
2206 return; /* nothing to do */
2208 loops = now.tv_sec - V_last_tv_sec;
2209 V_last_tv_sec = now.tv_sec;
2210 if (loops > BW_METER_BUCKETS)
2211 loops = BW_METER_BUCKETS;
2215 * Process all bins of bw_meter entries from the one after the last
2216 * processed to the current one. On entry, i points to the last bucket
2217 * visited, so we need to increment i at the beginning of the loop.
2219 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2220 struct bw_meter *x, *tmp_list;
2222 if (++i >= BW_METER_BUCKETS)
2225 /* Disconnect the list of bw_meter entries from the bin */
2226 tmp_list = V_bw_meter_timers[i];
2227 V_bw_meter_timers[i] = NULL;
2229 /* Process the list of bw_meter entries */
2230 while (tmp_list != NULL) {
2232 tmp_list = tmp_list->bm_time_next;
2234 /* Test if the time interval is over */
2235 process_endtime = x->bm_start_time;
2236 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2237 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2238 /* Not yet: reschedule, but don't reset */
2241 BW_METER_TIMEHASH(x, time_hash);
2242 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2244 * XXX: somehow the bin processing is a bit ahead of time.
2245 * Put the entry in the next bin.
2247 if (++time_hash >= BW_METER_BUCKETS)
2250 x->bm_time_next = V_bw_meter_timers[time_hash];
2251 V_bw_meter_timers[time_hash] = x;
2252 x->bm_time_hash = time_hash;
2258 * Test if we should deliver an upcall
2260 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2261 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2262 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2263 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2264 /* Prepare an upcall for delivery */
2265 bw_meter_prepare_upcall(x, &now);
2269 * Reschedule for next processing
2271 schedule_bw_meter(x, &now);
2275 /* Send all upcalls that are pending delivery */
2282 * A periodic function for sending all upcalls that are pending delivery
2285 expire_bw_upcalls_send(void *arg)
2287 CURVNET_SET((struct vnet *) arg);
2293 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send,
2299 * A periodic function for periodic scanning of the multicast forwarding
2300 * table for processing all "<=" bw_meter entries.
2303 expire_bw_meter_process(void *arg)
2305 CURVNET_SET((struct vnet *) arg);
2307 if (V_mrt_api_config & MRT_MFC_BW_UPCALL)
2310 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process,
2316 * End of bandwidth monitoring code
2320 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2324 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m,
2327 struct mbuf *mb_copy, *mm;
2330 * Do not send IGMP_WHOLEPKT notifications to userland, if the
2331 * rendezvous point was unspecified, and we were told not to.
2333 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) &&
2334 in_nullhost(rt->mfc_rp))
2337 mb_copy = pim_register_prepare(ip, m);
2338 if (mb_copy == NULL)
2342 * Send all the fragments. Note that the mbuf for each fragment
2343 * is freed by the sending machinery.
2345 for (mm = mb_copy; mm; mm = mb_copy) {
2346 mb_copy = mm->m_nextpkt;
2348 mm = m_pullup(mm, sizeof(struct ip));
2350 ip = mtod(mm, struct ip *);
2351 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) {
2352 pim_register_send_rp(ip, vifp, mm, rt);
2354 pim_register_send_upcall(ip, vifp, mm, rt);
2363 * Return a copy of the data packet that is ready for PIM Register
2365 * XXX: Note that in the returned copy the IP header is a valid one.
2367 static struct mbuf *
2368 pim_register_prepare(struct ip *ip, struct mbuf *m)
2370 struct mbuf *mb_copy = NULL;
2373 /* Take care of delayed checksums */
2374 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2375 in_delayed_cksum(m);
2376 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2380 * Copy the old packet & pullup its IP header into the
2381 * new mbuf so we can modify it.
2383 mb_copy = m_copypacket(m, M_NOWAIT);
2384 if (mb_copy == NULL)
2386 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2387 if (mb_copy == NULL)
2390 /* take care of the TTL */
2391 ip = mtod(mb_copy, struct ip *);
2394 /* Compute the MTU after the PIM Register encapsulation */
2395 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2397 if (ntohs(ip->ip_len) <= mtu) {
2398 /* Turn the IP header into a valid one */
2400 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2402 /* Fragment the packet */
2403 mb_copy->m_pkthdr.csum_flags |= CSUM_IP;
2404 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) {
2413 * Send an upcall with the data packet to the user-level process.
2416 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2417 struct mbuf *mb_copy, struct mfc *rt)
2419 struct mbuf *mb_first;
2420 int len = ntohs(ip->ip_len);
2422 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2427 * Add a new mbuf with an upcall header
2429 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2430 if (mb_first == NULL) {
2434 mb_first->m_data += max_linkhdr;
2435 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2436 mb_first->m_len = sizeof(struct igmpmsg);
2437 mb_first->m_next = mb_copy;
2439 /* Send message to routing daemon */
2440 im = mtod(mb_first, struct igmpmsg *);
2441 im->im_msgtype = IGMPMSG_WHOLEPKT;
2443 im->im_vif = vifp - V_viftable;
2444 im->im_src = ip->ip_src;
2445 im->im_dst = ip->ip_dst;
2447 k_igmpsrc.sin_addr = ip->ip_src;
2449 MRTSTAT_INC(mrts_upcalls);
2451 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2452 CTR1(KTR_IPMF, "%s: socket queue full", __func__);
2453 MRTSTAT_INC(mrts_upq_sockfull);
2457 /* Keep statistics */
2458 PIMSTAT_INC(pims_snd_registers_msgs);
2459 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2465 * Encapsulate the data packet in PIM Register message and send it to the RP.
2468 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy,
2471 struct mbuf *mb_first;
2472 struct ip *ip_outer;
2473 struct pim_encap_pimhdr *pimhdr;
2474 int len = ntohs(ip->ip_len);
2475 vifi_t vifi = rt->mfc_parent;
2479 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) {
2481 return EADDRNOTAVAIL; /* The iif vif is invalid */
2485 * Add a new mbuf with the encapsulating header
2487 mb_first = m_gethdr(M_NOWAIT, MT_DATA);
2488 if (mb_first == NULL) {
2492 mb_first->m_data += max_linkhdr;
2493 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2494 mb_first->m_next = mb_copy;
2496 mb_first->m_pkthdr.len = len + mb_first->m_len;
2499 * Fill in the encapsulating IP and PIM header
2501 ip_outer = mtod(mb_first, struct ip *);
2502 *ip_outer = pim_encap_iphdr;
2503 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) +
2504 sizeof(pim_encap_pimhdr));
2505 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr;
2506 ip_outer->ip_dst = rt->mfc_rp;
2508 * Copy the inner header TOS to the outer header, and take care of the
2511 ip_outer->ip_tos = ip->ip_tos;
2512 if (ip->ip_off & htons(IP_DF))
2513 ip_outer->ip_off |= htons(IP_DF);
2514 ip_fillid(ip_outer);
2515 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
2516 + sizeof(pim_encap_iphdr));
2517 *pimhdr = pim_encap_pimhdr;
2518 /* If the iif crosses a border, set the Border-bit */
2519 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config)
2520 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
2522 mb_first->m_data += sizeof(pim_encap_iphdr);
2523 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
2524 mb_first->m_data -= sizeof(pim_encap_iphdr);
2526 send_packet(vifp, mb_first);
2528 /* Keep statistics */
2529 PIMSTAT_INC(pims_snd_registers_msgs);
2530 PIMSTAT_ADD(pims_snd_registers_bytes, len);
2536 * pim_encapcheck() is called by the encap4_input() path at runtime to
2537 * determine if a packet is for PIM; allowing PIM to be dynamically loaded
2541 pim_encapcheck(const struct mbuf *m __unused, int off __unused,
2542 int proto __unused, void *arg __unused)
2545 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM"));
2546 return (8); /* claim the datagram. */
2550 * PIM-SMv2 and PIM-DM messages processing.
2551 * Receives and verifies the PIM control messages, and passes them
2552 * up to the listening socket, using rip_input().
2553 * The only message with special processing is the PIM_REGISTER message
2554 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
2555 * is passed to if_simloop().
2558 pim_input(struct mbuf *m, int off, int proto, void *arg __unused)
2560 struct ip *ip = mtod(m, struct ip *);
2564 int datalen = ntohs(ip->ip_len) - iphlen;
2567 /* Keep statistics */
2568 PIMSTAT_INC(pims_rcv_total_msgs);
2569 PIMSTAT_ADD(pims_rcv_total_bytes, datalen);
2574 if (datalen < PIM_MINLEN) {
2575 PIMSTAT_INC(pims_rcv_tooshort);
2576 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x",
2577 __func__, datalen, ntohl(ip->ip_src.s_addr));
2579 return (IPPROTO_DONE);
2583 * If the packet is at least as big as a REGISTER, go agead
2584 * and grab the PIM REGISTER header size, to avoid another
2585 * possible m_pullup() later.
2587 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
2588 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
2590 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
2592 * Get the IP and PIM headers in contiguous memory, and
2593 * possibly the PIM REGISTER header.
2595 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) {
2596 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__);
2597 return (IPPROTO_DONE);
2600 /* m_pullup() may have given us a new mbuf so reset ip. */
2601 ip = mtod(m, struct ip *);
2602 ip_tos = ip->ip_tos;
2604 /* adjust mbuf to point to the PIM header */
2605 m->m_data += iphlen;
2607 pim = mtod(m, struct pim *);
2610 * Validate checksum. If PIM REGISTER, exclude the data packet.
2612 * XXX: some older PIMv2 implementations don't make this distinction,
2613 * so for compatibility reason perform the checksum over part of the
2614 * message, and if error, then over the whole message.
2616 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
2617 /* do nothing, checksum okay */
2618 } else if (in_cksum(m, datalen)) {
2619 PIMSTAT_INC(pims_rcv_badsum);
2620 CTR1(KTR_IPMF, "%s: invalid checksum", __func__);
2622 return (IPPROTO_DONE);
2625 /* PIM version check */
2626 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
2627 PIMSTAT_INC(pims_rcv_badversion);
2628 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__,
2629 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION);
2631 return (IPPROTO_DONE);
2634 /* restore mbuf back to the outer IP */
2635 m->m_data -= iphlen;
2638 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
2640 * Since this is a REGISTER, we'll make a copy of the register
2641 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
2644 struct sockaddr_in dst = { sizeof(dst), AF_INET };
2646 struct ip *encap_ip;
2651 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) {
2653 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__,
2654 (int)V_reg_vif_num);
2656 return (IPPROTO_DONE);
2658 /* XXX need refcnt? */
2659 vifp = V_viftable[V_reg_vif_num].v_ifp;
2665 if (datalen < PIM_REG_MINLEN) {
2666 PIMSTAT_INC(pims_rcv_tooshort);
2667 PIMSTAT_INC(pims_rcv_badregisters);
2668 CTR1(KTR_IPMF, "%s: register packet size too small", __func__);
2670 return (IPPROTO_DONE);
2673 reghdr = (u_int32_t *)(pim + 1);
2674 encap_ip = (struct ip *)(reghdr + 1);
2676 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d",
2677 __func__, ntohl(encap_ip->ip_src.s_addr),
2678 ntohs(encap_ip->ip_len));
2680 /* verify the version number of the inner packet */
2681 if (encap_ip->ip_v != IPVERSION) {
2682 PIMSTAT_INC(pims_rcv_badregisters);
2683 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__);
2685 return (IPPROTO_DONE);
2688 /* verify the inner packet is destined to a mcast group */
2689 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
2690 PIMSTAT_INC(pims_rcv_badregisters);
2691 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__,
2692 ntohl(encap_ip->ip_dst.s_addr));
2694 return (IPPROTO_DONE);
2697 /* If a NULL_REGISTER, pass it to the daemon */
2698 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
2699 goto pim_input_to_daemon;
2702 * Copy the TOS from the outer IP header to the inner IP header.
2704 if (encap_ip->ip_tos != ip_tos) {
2705 /* Outer TOS -> inner TOS */
2706 encap_ip->ip_tos = ip_tos;
2707 /* Recompute the inner header checksum. Sigh... */
2709 /* adjust mbuf to point to the inner IP header */
2710 m->m_data += (iphlen + PIM_MINLEN);
2711 m->m_len -= (iphlen + PIM_MINLEN);
2713 encap_ip->ip_sum = 0;
2714 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
2716 /* restore mbuf to point back to the outer IP header */
2717 m->m_data -= (iphlen + PIM_MINLEN);
2718 m->m_len += (iphlen + PIM_MINLEN);
2722 * Decapsulate the inner IP packet and loopback to forward it
2723 * as a normal multicast packet. Also, make a copy of the
2724 * outer_iphdr + pimhdr + reghdr + encap_iphdr
2725 * to pass to the daemon later, so it can take the appropriate
2726 * actions (e.g., send back PIM_REGISTER_STOP).
2727 * XXX: here m->m_data points to the outer IP header.
2729 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT);
2731 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__);
2733 return (IPPROTO_DONE);
2736 /* Keep statistics */
2737 /* XXX: registers_bytes include only the encap. mcast pkt */
2738 PIMSTAT_INC(pims_rcv_registers_msgs);
2739 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len));
2742 * forward the inner ip packet; point m_data at the inner ip.
2744 m_adj(m, iphlen + PIM_MINLEN);
2747 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d",
2749 (u_long)ntohl(encap_ip->ip_src.s_addr),
2750 (u_long)ntohl(encap_ip->ip_dst.s_addr),
2751 (int)V_reg_vif_num);
2753 /* NB: vifp was collected above; can it change on us? */
2754 if_simloop(vifp, m, dst.sin_family, 0);
2756 /* prepare the register head to send to the mrouting daemon */
2760 pim_input_to_daemon:
2762 * Pass the PIM message up to the daemon; if it is a Register message,
2763 * pass the 'head' only up to the daemon. This includes the
2764 * outer IP header, PIM header, PIM-Register header and the
2766 * XXX: the outer IP header pkt size of a Register is not adjust to
2767 * reflect the fact that the inner multicast data is truncated.
2769 return (rip_input(&m, &off, proto));
2773 sysctl_mfctable(SYSCTL_HANDLER_ARGS)
2780 if (V_mfchashtbl == NULL) /* XXX unlocked */
2782 error = sysctl_wire_old_buffer(req, 0);
2787 for (i = 0; i < mfchashsize; i++) {
2788 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) {
2789 error = SYSCTL_OUT(req, rt, sizeof(struct mfc));
2799 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable,
2800 CTLFLAG_RD | CTLFLAG_MPSAFE, sysctl_mfctable,
2801 "IPv4 Multicast Forwarding Table "
2802 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)");
2805 sysctl_viflist(SYSCTL_HANDLER_ARGS)
2811 if (V_viftable == NULL) /* XXX unlocked */
2813 error = sysctl_wire_old_buffer(req, sizeof(*V_viftable) * MAXVIFS);
2818 error = SYSCTL_OUT(req, V_viftable, sizeof(*V_viftable) * MAXVIFS);
2823 SYSCTL_PROC(_net_inet_ip, OID_AUTO, viftable,
2824 CTLTYPE_OPAQUE | CTLFLAG_VNET | CTLFLAG_RD | CTLFLAG_MPSAFE, NULL, 0,
2825 sysctl_viflist, "S,vif[MAXVIFS]",
2826 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
2829 vnet_mroute_init(const void *unused __unused)
2832 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO);
2834 V_viftable = mallocarray(MAXVIFS, sizeof(*V_viftable),
2835 M_MRTABLE, M_WAITOK|M_ZERO);
2836 V_bw_meter_timers = mallocarray(BW_METER_BUCKETS,
2837 sizeof(*V_bw_meter_timers), M_MRTABLE, M_WAITOK|M_ZERO);
2838 V_bw_upcalls = mallocarray(BW_UPCALLS_MAX, sizeof(*V_bw_upcalls),
2839 M_MRTABLE, M_WAITOK|M_ZERO);
2841 callout_init(&V_expire_upcalls_ch, 1);
2842 callout_init(&V_bw_upcalls_ch, 1);
2843 callout_init(&V_bw_meter_ch, 1);
2846 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init,
2850 vnet_mroute_uninit(const void *unused __unused)
2853 free(V_bw_upcalls, M_MRTABLE);
2854 free(V_bw_meter_timers, M_MRTABLE);
2855 free(V_viftable, M_MRTABLE);
2856 free(V_nexpire, M_MRTABLE);
2860 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE,
2861 vnet_mroute_uninit, NULL);
2864 ip_mroute_modevent(module_t mod, int type, void *unused)
2869 MROUTER_LOCK_INIT();
2871 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event,
2872 if_detached_event, NULL, EVENTHANDLER_PRI_ANY);
2873 if (if_detach_event_tag == NULL) {
2874 printf("ip_mroute: unable to register "
2875 "ifnet_departure_event handler\n");
2876 MROUTER_LOCK_DESTROY();
2883 mfchashsize = MFCHASHSIZE;
2884 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) &&
2885 !powerof2(mfchashsize)) {
2886 printf("WARNING: %s not a power of 2; using default\n",
2887 "net.inet.ip.mfchashsize");
2888 mfchashsize = MFCHASHSIZE;
2891 pim_squelch_wholepkt = 0;
2892 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt",
2893 &pim_squelch_wholepkt);
2895 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK);
2896 if (pim_encap_cookie == NULL) {
2897 printf("ip_mroute: unable to attach pim encap\n");
2900 MROUTER_LOCK_DESTROY();
2904 ip_mcast_src = X_ip_mcast_src;
2905 ip_mforward = X_ip_mforward;
2906 ip_mrouter_done = X_ip_mrouter_done;
2907 ip_mrouter_get = X_ip_mrouter_get;
2908 ip_mrouter_set = X_ip_mrouter_set;
2910 ip_rsvp_force_done = X_ip_rsvp_force_done;
2911 ip_rsvp_vif = X_ip_rsvp_vif;
2913 legal_vif_num = X_legal_vif_num;
2914 mrt_ioctl = X_mrt_ioctl;
2915 rsvp_input_p = X_rsvp_input;
2920 * Typically module unload happens after the user-level
2921 * process has shutdown the kernel services (the check
2922 * below insures someone can't just yank the module out
2923 * from under a running process). But if the module is
2924 * just loaded and then unloaded w/o starting up a user
2925 * process we still need to cleanup.
2928 if (ip_mrouter_cnt != 0) {
2932 ip_mrouter_unloading = 1;
2935 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag);
2937 if (pim_encap_cookie) {
2938 ip_encap_detach(pim_encap_cookie);
2939 pim_encap_cookie = NULL;
2942 ip_mcast_src = NULL;
2944 ip_mrouter_done = NULL;
2945 ip_mrouter_get = NULL;
2946 ip_mrouter_set = NULL;
2948 ip_rsvp_force_done = NULL;
2951 legal_vif_num = NULL;
2953 rsvp_input_p = NULL;
2957 MROUTER_LOCK_DESTROY();
2966 static moduledata_t ip_mroutemod = {
2972 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE);