2 * IP multicast forwarding procedures
4 * Written by David Waitzman, BBN Labs, August 1988.
5 * Modified by Steve Deering, Stanford, February 1989.
6 * Modified by Mark J. Steiglitz, Stanford, May, 1991
7 * Modified by Van Jacobson, LBL, January 1993
8 * Modified by Ajit Thyagarajan, PARC, August 1993
9 * Modified by Bill Fenner, PARC, April 1995
10 * Modified by Ahmed Helmy, SGI, June 1996
11 * Modified by George Edmond Eddy (Rusty), ISI, February 1998
12 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000
13 * Modified by Hitoshi Asaeda, WIDE, August 2000
14 * Modified by Pavlin Radoslavov, ICSI, October 2002
16 * MROUTING Revision: 3.5
17 * and PIM-SMv2 and PIM-DM support, advanced API support,
18 * bandwidth metering and signaling
20 * $FreeBSD: src/sys/netinet/ip_mroute.c,v 1.56.2.10 2003/08/24 21:37:34 hsu Exp $
21 * $DragonFly: src/sys/net/ip_mroute/ip_mroute.c,v 1.16 2005/02/11 22:25:57 joerg Exp $
24 #include "opt_mrouting.h"
25 #include "opt_random_ip_id.h"
31 #include <sys/param.h>
32 #include <sys/kernel.h>
33 #include <sys/malloc.h>
35 #include <sys/protosw.h>
36 #include <sys/socket.h>
37 #include <sys/socketvar.h>
38 #include <sys/sockio.h>
39 #include <sys/sysctl.h>
40 #include <sys/syslog.h>
41 #include <sys/systm.h>
43 #include <sys/in_cksum.h>
45 #include <machine/stdarg.h>
48 #include <net/netisr.h>
49 #include <net/route.h>
50 #include <netinet/in.h>
51 #include <netinet/igmp.h>
52 #include <netinet/in_systm.h>
53 #include <netinet/in_var.h>
54 #include <netinet/ip.h>
55 #include "ip_mroute.h"
56 #include <netinet/ip_var.h>
58 #include <netinet/pim.h>
59 #include <netinet/pim_var.h>
62 #include <netinet/in_pcb.h>
64 #include <netinet/udp.h>
67 * Control debugging code for rsvp and multicast routing code.
68 * Can only set them with the debugger.
70 static u_int rsvpdebug; /* non-zero enables debugging */
72 static u_int mrtdebug; /* any set of the flags below */
74 #define DEBUG_MFC 0x02
75 #define DEBUG_FORWARD 0x04
76 #define DEBUG_EXPIRE 0x08
77 #define DEBUG_XMIT 0x10
78 #define DEBUG_PIM 0x20
80 #define VIFI_INVALID ((vifi_t) -1)
82 #define M_HASCL(m) ((m)->m_flags & M_EXT)
84 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables");
86 static struct mrtstat mrtstat;
87 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW,
89 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)");
91 static struct mfc *mfctable[MFCTBLSIZ];
92 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD,
93 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]",
94 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)");
96 static struct vif viftable[MAXVIFS];
97 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD,
98 &viftable, sizeof(viftable), "S,vif[MAXVIFS]",
99 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)");
101 static u_char nexpire[MFCTBLSIZ];
103 static struct callout expire_upcalls_ch;
104 static struct callout tbf_reprocess_q_ch;
105 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */
106 #define UPCALL_EXPIRE 6 /* number of timeouts */
109 * Define the token bucket filter structures
110 * tbftable -> each vif has one of these for storing info
113 static struct tbf tbftable[MAXVIFS];
114 #define TBF_REPROCESS (hz / 100) /* 100x / second */
117 * 'Interfaces' associated with decapsulator (so we can tell
118 * packets that went through it from ones that get reflected
119 * by a broken gateway). These interfaces are never linked into
120 * the system ifnet list & no routes point to them. I.e., packets
121 * can't be sent this way. They only exist as a placeholder for
122 * multicast source verification.
124 static struct ifnet multicast_decap_if[MAXVIFS];
127 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */
129 /* prototype IP hdr for encapsulated packets */
130 static struct ip multicast_encap_iphdr = {
131 #if BYTE_ORDER == LITTLE_ENDIAN
132 sizeof(struct ip) >> 2, IPVERSION,
134 IPVERSION, sizeof(struct ip) >> 2,
137 sizeof(struct ip), /* total length */
140 ENCAP_TTL, ENCAP_PROTO,
145 * Bandwidth meter variables and constants
147 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters");
149 * Pending timeouts are stored in a hash table, the key being the
150 * expiration time. Periodically, the entries are analysed and processed.
152 #define BW_METER_BUCKETS 1024
153 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS];
154 static struct callout bw_meter_ch;
155 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */
158 * Pending upcalls are stored in a vector which is flushed when
159 * full, or periodically
161 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX];
162 static u_int bw_upcalls_n; /* # of pending upcalls */
163 static struct callout bw_upcalls_ch;
164 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */
167 static struct pimstat pimstat;
168 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD,
170 "PIM Statistics (struct pimstat, netinet/pim_var.h)");
173 * Note: the PIM Register encapsulation adds the following in front of a
176 * struct pim_encap_hdr {
178 * struct pim_encap_pimhdr pim;
183 struct pim_encap_pimhdr {
188 static struct ip pim_encap_iphdr = {
189 #if BYTE_ORDER == LITTLE_ENDIAN
190 sizeof(struct ip) >> 2,
194 sizeof(struct ip) >> 2,
197 sizeof(struct ip), /* total length */
205 static struct pim_encap_pimhdr pim_encap_pimhdr = {
207 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */
214 static struct ifnet multicast_register_if;
215 static vifi_t reg_vif_num = VIFI_INVALID;
221 static vifi_t numvifs;
222 static int have_encap_tunnel;
225 * one-back cache used by ipip_input to locate a tunnel's vif
226 * given a datagram's src ip address.
228 static u_long last_encap_src;
229 static struct vif *last_encap_vif;
231 static u_long X_ip_mcast_src(int vifi);
232 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp,
233 struct mbuf *m, struct ip_moptions *imo);
234 static int X_ip_mrouter_done(void);
235 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m);
236 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m);
237 static int X_legal_vif_num(int vif);
238 static int X_mrt_ioctl(int cmd, caddr_t data);
240 static int get_sg_cnt(struct sioc_sg_req *);
241 static int get_vif_cnt(struct sioc_vif_req *);
242 static int ip_mrouter_init(struct socket *, int);
243 static int add_vif(struct vifctl *);
244 static int del_vif(vifi_t);
245 static int add_mfc(struct mfcctl2 *);
246 static int del_mfc(struct mfcctl2 *);
247 static int set_api_config(uint32_t *); /* chose API capabilities */
248 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *);
249 static int set_assert(int);
250 static void expire_upcalls(void *);
251 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t);
252 static void phyint_send(struct ip *, struct vif *, struct mbuf *);
253 static void encap_send(struct ip *, struct vif *, struct mbuf *);
254 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long);
255 static void tbf_queue(struct vif *, struct mbuf *);
256 static void tbf_process_q(struct vif *);
257 static void tbf_reprocess_q(void *);
258 static int tbf_dq_sel(struct vif *, struct ip *);
259 static void tbf_send_packet(struct vif *, struct mbuf *);
260 static void tbf_update_tokens(struct vif *);
261 static int priority(struct vif *, struct ip *);
264 * Bandwidth monitoring
266 static void free_bw_list(struct bw_meter *list);
267 static int add_bw_upcall(struct bw_upcall *);
268 static int del_bw_upcall(struct bw_upcall *);
269 static void bw_meter_receive_packet(struct bw_meter *x, int plen,
270 struct timeval *nowp);
271 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp);
272 static void bw_upcalls_send(void);
273 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp);
274 static void unschedule_bw_meter(struct bw_meter *x);
275 static void bw_meter_process(void);
276 static void expire_bw_upcalls_send(void *);
277 static void expire_bw_meter_process(void *);
280 static int pim_register_send(struct ip *, struct vif *,
281 struct mbuf *, struct mfc *);
282 static int pim_register_send_rp(struct ip *, struct vif *,
283 struct mbuf *, struct mfc *);
284 static int pim_register_send_upcall(struct ip *, struct vif *,
285 struct mbuf *, struct mfc *);
286 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *);
290 * whether or not special PIM assert processing is enabled.
292 static int pim_assert;
294 * Rate limit for assert notification messages, in usec
296 #define ASSERT_MSG_TIME 3000000
299 * Kernel multicast routing API capabilities and setup.
300 * If more API capabilities are added to the kernel, they should be
301 * recorded in `mrt_api_support'.
303 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF |
304 MRT_MFC_FLAGS_BORDER_VIF |
307 static uint32_t mrt_api_config = 0;
310 * Hash function for a source, group entry
312 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \
313 ((g) >> 20) ^ ((g) >> 10) ^ (g))
316 * Find a route for a given origin IP address and Multicast group address
317 * Type of service parameter to be added in the future!!!
318 * Statistics are updated by the caller if needed
319 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses)
322 mfc_find(in_addr_t o, in_addr_t g)
326 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next)
327 if ((rt->mfc_origin.s_addr == o) &&
328 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL))
334 * Macros to compute elapsed time efficiently
335 * Borrowed from Van Jacobson's scheduling code
337 #define TV_DELTA(a, b, delta) { \
339 delta = (a).tv_usec - (b).tv_usec; \
340 if ((xxs = (a).tv_sec - (b).tv_sec)) { \
349 delta += (1000000 * xxs); \
354 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \
355 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec)
358 * Handle MRT setsockopt commands to modify the multicast routing tables.
361 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt)
367 struct bw_upcall bw_upcall;
370 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT)
374 switch (sopt->sopt_name) {
376 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
379 error = ip_mrouter_init(so, optval);
383 error = ip_mrouter_done();
387 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc);
390 error = add_vif(&vifc);
394 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
397 error = del_vif(vifi);
403 * select data size depending on API version.
405 if (sopt->sopt_name == MRT_ADD_MFC &&
406 mrt_api_config & MRT_API_FLAGS_ALL) {
407 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2),
408 sizeof(struct mfcctl2));
410 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl),
411 sizeof(struct mfcctl));
412 bzero((caddr_t)&mfc + sizeof(struct mfcctl),
413 sizeof(mfc) - sizeof(struct mfcctl));
417 if (sopt->sopt_name == MRT_ADD_MFC)
418 error = add_mfc(&mfc);
420 error = del_mfc(&mfc);
424 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval);
431 error = sooptcopyin(sopt, &i, sizeof i, sizeof i);
433 error = set_api_config(&i);
435 error = sooptcopyout(sopt, &i, sizeof i);
438 case MRT_ADD_BW_UPCALL:
439 case MRT_DEL_BW_UPCALL:
440 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall,
444 if (sopt->sopt_name == MRT_ADD_BW_UPCALL)
445 error = add_bw_upcall(&bw_upcall);
447 error = del_bw_upcall(&bw_upcall);
458 * Handle MRT getsockopt commands
461 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt)
464 static int version = 0x0305; /* !!! why is this here? XXX */
466 switch (sopt->sopt_name) {
468 error = sooptcopyout(sopt, &version, sizeof version);
472 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert);
475 case MRT_API_SUPPORT:
476 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support);
480 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config);
491 * Handle ioctl commands to obtain information from the cache
494 X_mrt_ioctl(int cmd, caddr_t data)
500 error = get_vif_cnt((struct sioc_vif_req *)data);
504 error = get_sg_cnt((struct sioc_sg_req *)data);
515 * returns the packet, byte, rpf-failure count for the source group provided
518 get_sg_cnt(struct sioc_sg_req *req)
524 rt = mfc_find(req->src.s_addr, req->grp.s_addr);
527 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff;
528 return EADDRNOTAVAIL;
530 req->pktcnt = rt->mfc_pkt_cnt;
531 req->bytecnt = rt->mfc_byte_cnt;
532 req->wrong_if = rt->mfc_wrong_if;
537 * returns the input and output packet and byte counts on the vif provided
540 get_vif_cnt(struct sioc_vif_req *req)
542 vifi_t vifi = req->vifi;
547 req->icount = viftable[vifi].v_pkt_in;
548 req->ocount = viftable[vifi].v_pkt_out;
549 req->ibytes = viftable[vifi].v_bytes_in;
550 req->obytes = viftable[vifi].v_bytes_out;
556 * Enable multicast routing
559 ip_mrouter_init(struct socket *so, int version)
562 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n",
563 so->so_type, so->so_proto->pr_protocol);
565 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP)
571 if (ip_mrouter != NULL)
576 bzero((caddr_t)mfctable, sizeof(mfctable));
577 bzero((caddr_t)nexpire, sizeof(nexpire));
581 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers));
583 callout_init(&expire_upcalls_ch);
584 callout_init(&bw_upcalls_ch);
585 callout_init(&bw_meter_ch);
586 callout_init(&tbf_reprocess_q_ch);
588 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
589 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
590 expire_bw_upcalls_send, NULL);
591 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL);
596 log(LOG_DEBUG, "ip_mrouter_init\n");
602 * Disable multicast routing
605 X_ip_mrouter_done(void)
618 * For each phyint in use, disable promiscuous reception of all IP
621 for (vifi = 0; vifi < numvifs; vifi++) {
622 if (viftable[vifi].v_lcl_addr.s_addr != 0 &&
623 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) {
624 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr);
626 so->sin_len = sizeof(struct sockaddr_in);
627 so->sin_family = AF_INET;
628 so->sin_addr.s_addr = INADDR_ANY;
629 ifp = viftable[vifi].v_ifp;
633 bzero((caddr_t)tbftable, sizeof(tbftable));
634 bzero((caddr_t)viftable, sizeof(viftable));
638 callout_stop(&expire_upcalls_ch);
642 callout_stop(&bw_upcalls_ch);
643 callout_stop(&bw_meter_ch);
644 callout_stop(&tbf_reprocess_q_ch);
647 * Free all multicast forwarding cache entries.
649 for (i = 0; i < MFCTBLSIZ; i++) {
650 for (rt = mfctable[i]; rt != NULL; ) {
651 struct mfc *nr = rt->mfc_next;
653 for (rte = rt->mfc_stall; rte != NULL; ) {
654 struct rtdetq *n = rte->next;
657 free(rte, M_MRTABLE);
660 free_bw_list(rt->mfc_bw_meter);
666 bzero((caddr_t)mfctable, sizeof(mfctable));
668 bzero(bw_meter_timers, sizeof(bw_meter_timers));
671 * Reset de-encapsulation cache
673 last_encap_src = INADDR_ANY;
674 last_encap_vif = NULL;
676 reg_vif_num = VIFI_INVALID;
678 have_encap_tunnel = 0;
685 log(LOG_DEBUG, "ip_mrouter_done\n");
691 * Set PIM assert processing global
696 if ((i != 1) && (i != 0))
705 * Configure API capabilities
708 set_api_config(uint32_t *apival)
713 * We can set the API capabilities only if it is the first operation
714 * after MRT_INIT. I.e.:
715 * - there are no vifs installed
716 * - pim_assert is not enabled
717 * - the MFC table is empty
727 for (i = 0; i < MFCTBLSIZ; i++) {
728 if (mfctable[i] != NULL) {
734 mrt_api_config = *apival & mrt_api_support;
735 *apival = mrt_api_config;
741 * Add a vif to the vif table
744 add_vif(struct vifctl *vifcp)
746 struct vif *vifp = viftable + vifcp->vifc_vifi;
747 struct sockaddr_in sin = {sizeof sin, AF_INET};
751 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi;
753 if (vifcp->vifc_vifi >= MAXVIFS)
755 if (vifp->v_lcl_addr.s_addr != INADDR_ANY)
757 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY)
758 return EADDRNOTAVAIL;
760 /* Find the interface with an address in AF_INET family */
762 if (vifcp->vifc_flags & VIFF_REGISTER) {
764 * XXX: Because VIFF_REGISTER does not really need a valid
765 * local interface (e.g. it could be 127.0.0.2), we don't
772 sin.sin_addr = vifcp->vifc_lcl_addr;
773 ifa = ifa_ifwithaddr((struct sockaddr *)&sin);
775 return EADDRNOTAVAIL;
779 if (vifcp->vifc_flags & VIFF_TUNNEL) {
780 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) {
782 * An encapsulating tunnel is wanted. Tell ipip_input() to
783 * start paying attention to encapsulated packets.
785 if (have_encap_tunnel == 0) {
786 have_encap_tunnel = 1;
787 for (s = 0; s < MAXVIFS; ++s) {
788 if_initname(&multicast_decap_if[s], "mdecap", s);
792 * Set interface to fake encapsulator interface
794 ifp = &multicast_decap_if[vifcp->vifc_vifi];
796 * Prepare cached route entry
798 bzero(&vifp->v_route, sizeof(vifp->v_route));
800 log(LOG_ERR, "source routed tunnels not supported\n");
804 } else if (vifcp->vifc_flags & VIFF_REGISTER) {
805 ifp = &multicast_register_if;
807 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n",
808 (void *)&multicast_register_if);
809 if (reg_vif_num == VIFI_INVALID) {
810 if_initname(&multicast_register_if, "register_vif", 0);
811 multicast_register_if.if_flags = IFF_LOOPBACK;
812 bzero(&vifp->v_route, sizeof(vifp->v_route));
813 reg_vif_num = vifcp->vifc_vifi;
816 } else { /* Make sure the interface supports multicast */
817 if ((ifp->if_flags & IFF_MULTICAST) == 0)
820 /* Enable promiscuous reception of all IP multicasts from the if */
822 error = if_allmulti(ifp, 1);
829 /* define parameters for the tbf structure */
831 GET_TIME(vifp->v_tbf->tbf_last_pkt_t);
832 vifp->v_tbf->tbf_n_tok = 0;
833 vifp->v_tbf->tbf_q_len = 0;
834 vifp->v_tbf->tbf_max_q_len = MAXQSIZE;
835 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL;
837 vifp->v_flags = vifcp->vifc_flags;
838 vifp->v_threshold = vifcp->vifc_threshold;
839 vifp->v_lcl_addr = vifcp->vifc_lcl_addr;
840 vifp->v_rmt_addr = vifcp->vifc_rmt_addr;
842 /* scaling up here allows division by 1024 in critical code */
843 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000;
845 vifp->v_rsvpd = NULL;
846 /* initialize per vif pkt counters */
849 vifp->v_bytes_in = 0;
850 vifp->v_bytes_out = 0;
853 /* Adjust numvifs up if the vifi is higher than numvifs */
854 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1;
857 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n",
859 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr),
860 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask",
861 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr),
862 vifcp->vifc_threshold,
863 vifcp->vifc_rate_limit);
869 * Delete a vif from the vif table
879 vifp = &viftable[vifi];
880 if (vifp->v_lcl_addr.s_addr == INADDR_ANY)
881 return EADDRNOTAVAIL;
885 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER)))
886 if_allmulti(vifp->v_ifp, 0);
888 if (vifp == last_encap_vif) {
889 last_encap_vif = NULL;
890 last_encap_src = INADDR_ANY;
894 * Free packets queued at the interface
896 while (vifp->v_tbf->tbf_q) {
897 struct mbuf *m = vifp->v_tbf->tbf_q;
899 vifp->v_tbf->tbf_q = m->m_nextpkt;
904 if (vifp->v_flags & VIFF_REGISTER)
905 reg_vif_num = VIFI_INVALID;
908 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf)));
909 bzero((caddr_t)vifp, sizeof (*vifp));
912 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs);
914 /* Adjust numvifs down */
915 for (vifi = numvifs; vifi > 0; vifi--)
916 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY)
926 * update an mfc entry without resetting counters and S,G addresses.
929 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
933 rt->mfc_parent = mfccp->mfcc_parent;
934 for (i = 0; i < numvifs; i++) {
935 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i];
936 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config &
939 /* set the RP address */
940 if (mrt_api_config & MRT_MFC_RP)
941 rt->mfc_rp = mfccp->mfcc_rp;
943 rt->mfc_rp.s_addr = INADDR_ANY;
947 * fully initialize an mfc entry from the parameter.
950 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp)
952 rt->mfc_origin = mfccp->mfcc_origin;
953 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp;
955 update_mfc_params(rt, mfccp);
957 /* initialize pkt counters per src-grp */
959 rt->mfc_byte_cnt = 0;
960 rt->mfc_wrong_if = 0;
961 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0;
969 add_mfc(struct mfcctl2 *mfccp)
977 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
979 /* If an entry already exists, just update the fields */
981 if (mrtdebug & DEBUG_MFC)
982 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n",
983 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
984 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
988 update_mfc_params(rt, mfccp);
994 * Find the entry for which the upcall was made and update
997 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr);
998 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) {
1000 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1001 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) &&
1002 (rt->mfc_stall != NULL)) {
1005 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n",
1006 "multiple kernel entries",
1007 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1008 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1009 mfccp->mfcc_parent, (void *)rt->mfc_stall);
1011 if (mrtdebug & DEBUG_MFC)
1012 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n",
1013 (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1014 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1015 mfccp->mfcc_parent, (void *)rt->mfc_stall);
1017 init_mfc_params(rt, mfccp);
1019 rt->mfc_expire = 0; /* Don't clean this guy up */
1022 /* free packets Qed at the end of this entry */
1023 for (rte = rt->mfc_stall; rte != NULL; ) {
1024 struct rtdetq *n = rte->next;
1026 ip_mdq(rte->m, rte->ifp, rt, -1);
1028 free(rte, M_MRTABLE);
1031 rt->mfc_stall = NULL;
1036 * It is possible that an entry is being inserted without an upcall
1039 if (mrtdebug & DEBUG_MFC)
1040 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n",
1041 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr),
1042 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr),
1043 mfccp->mfcc_parent);
1045 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) {
1046 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) &&
1047 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) {
1048 init_mfc_params(rt, mfccp);
1055 if (rt == NULL) { /* no upcall, so make a new entry */
1056 rt = malloc(sizeof(*rt), M_MRTABLE, M_INTWAIT | M_NULLOK);
1062 init_mfc_params(rt, mfccp);
1064 rt->mfc_stall = NULL;
1066 rt->mfc_bw_meter = NULL;
1067 /* insert new entry at head of hash chain */
1068 rt->mfc_next = mfctable[hash];
1069 mfctable[hash] = rt;
1077 * Delete an mfc entry
1080 del_mfc(struct mfcctl2 *mfccp)
1082 struct in_addr origin;
1083 struct in_addr mcastgrp;
1088 struct bw_meter *list;
1090 origin = mfccp->mfcc_origin;
1091 mcastgrp = mfccp->mfcc_mcastgrp;
1093 if (mrtdebug & DEBUG_MFC)
1094 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n",
1095 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr));
1099 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr);
1100 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next)
1101 if (origin.s_addr == rt->mfc_origin.s_addr &&
1102 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr &&
1103 rt->mfc_stall == NULL)
1107 return EADDRNOTAVAIL;
1110 *nptr = rt->mfc_next;
1113 * free the bw_meter entries
1115 list = rt->mfc_bw_meter;
1116 rt->mfc_bw_meter = NULL;
1118 free(rt, M_MRTABLE);
1128 * Send a message to mrouted on the multicast routing socket
1131 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src)
1134 if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) {
1144 * IP multicast forwarding function. This function assumes that the packet
1145 * pointed to by "ip" has arrived on (or is about to be sent to) the interface
1146 * pointed to by "ifp", and the packet is to be relayed to other networks
1147 * that have members of the packet's destination IP multicast group.
1149 * The packet is returned unscathed to the caller, unless it is
1150 * erroneous, in which case a non-zero return value tells the caller to
1154 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */
1157 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m,
1158 struct ip_moptions *imo)
1164 if (mrtdebug & DEBUG_FORWARD)
1165 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n",
1166 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr),
1169 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 ||
1170 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) {
1172 * Packet arrived via a physical interface or
1173 * an encapsulated tunnel or a register_vif.
1177 * Packet arrived through a source-route tunnel.
1178 * Source-route tunnels are no longer supported.
1180 static int last_log;
1181 if (last_log != time_second) {
1182 last_log = time_second;
1184 "ip_mforward: received source-routed packet from %lx\n",
1185 (u_long)ntohl(ip->ip_src.s_addr));
1190 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) {
1191 if (ip->ip_ttl < 255)
1192 ip->ip_ttl++; /* compensate for -1 in *_send routines */
1193 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1194 struct vif *vifp = viftable + vifi;
1196 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n",
1197 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr),
1199 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "",
1200 vifp->v_ifp->if_xname);
1202 return ip_mdq(m, ifp, NULL, vifi);
1204 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) {
1205 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n",
1206 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr));
1208 printf("In fact, no options were specified at all\n");
1212 * Don't forward a packet with time-to-live of zero or one,
1213 * or a packet destined to a local-only group.
1215 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP)
1219 * Determine forwarding vifs from the forwarding cache table
1222 ++mrtstat.mrts_mfc_lookups;
1223 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1225 /* Entry exists, so forward if necessary */
1228 return ip_mdq(m, ifp, rt, -1);
1231 * If we don't have a route for packet's origin,
1232 * Make a copy of the packet & send message to routing daemon
1238 int hlen = ip->ip_hl << 2;
1240 ++mrtstat.mrts_mfc_misses;
1242 mrtstat.mrts_no_route++;
1243 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC))
1244 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n",
1245 (u_long)ntohl(ip->ip_src.s_addr),
1246 (u_long)ntohl(ip->ip_dst.s_addr));
1249 * Allocate mbufs early so that we don't do extra work if we are
1250 * just going to fail anyway. Make sure to pullup the header so
1251 * that other people can't step on it.
1253 rte = malloc((sizeof *rte), M_MRTABLE, M_INTWAIT | M_NULLOK);
1259 mb0 = m_copypacket(m, MB_DONTWAIT);
1260 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen))
1261 mb0 = m_pullup(mb0, hlen);
1263 free(rte, M_MRTABLE);
1268 /* is there an upcall waiting for this flow ? */
1269 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr);
1270 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) {
1271 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) &&
1272 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) &&
1273 (rt->mfc_stall != NULL))
1280 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1284 * Locate the vifi for the incoming interface for this packet.
1285 * If none found, drop packet.
1287 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1289 if (vifi >= numvifs) /* vif not found, drop packet */
1292 /* no upcall, so make a new entry */
1293 rt = malloc(sizeof(*rt), M_MRTABLE, M_INTWAIT | M_NULLOK);
1297 /* Make a copy of the header to send to the user level process */
1298 mm = m_copy(mb0, 0, hlen);
1303 * Send message to routing daemon to install
1304 * a route into the kernel table
1307 im = mtod(mm, struct igmpmsg *);
1308 im->im_msgtype = IGMPMSG_NOCACHE;
1312 mrtstat.mrts_upcalls++;
1314 k_igmpsrc.sin_addr = ip->ip_src;
1315 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1316 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n");
1317 ++mrtstat.mrts_upq_sockfull;
1319 free(rt, M_MRTABLE);
1321 free(rte, M_MRTABLE);
1327 /* insert new entry at head of hash chain */
1328 rt->mfc_origin.s_addr = ip->ip_src.s_addr;
1329 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr;
1330 rt->mfc_expire = UPCALL_EXPIRE;
1332 for (i = 0; i < numvifs; i++) {
1333 rt->mfc_ttls[i] = 0;
1334 rt->mfc_flags[i] = 0;
1336 rt->mfc_parent = -1;
1338 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */
1340 rt->mfc_bw_meter = NULL;
1342 /* link into table */
1343 rt->mfc_next = mfctable[hash];
1344 mfctable[hash] = rt;
1345 rt->mfc_stall = rte;
1348 /* determine if q has overflowed */
1353 * XXX ouch! we need to append to the list, but we
1354 * only have a pointer to the front, so we have to
1355 * scan the entire list every time.
1357 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next)
1360 if (npkts > MAX_UPQ) {
1361 mrtstat.mrts_upq_ovflw++;
1363 free(rte, M_MRTABLE);
1369 /* Add this entry to the end of the queue */
1384 * Clean up the cache entry if upcall is not serviced
1387 expire_upcalls(void *unused)
1390 struct mfc *mfc, **nptr;
1395 for (i = 0; i < MFCTBLSIZ; i++) {
1396 if (nexpire[i] == 0)
1398 nptr = &mfctable[i];
1399 for (mfc = *nptr; mfc != NULL; mfc = *nptr) {
1401 * Skip real cache entries
1402 * Make sure it wasn't marked to not expire (shouldn't happen)
1405 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 &&
1406 --mfc->mfc_expire == 0) {
1407 if (mrtdebug & DEBUG_EXPIRE)
1408 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n",
1409 (u_long)ntohl(mfc->mfc_origin.s_addr),
1410 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr));
1412 * drop all the packets
1413 * free the mbuf with the pkt, if, timing info
1415 for (rte = mfc->mfc_stall; rte; ) {
1416 struct rtdetq *n = rte->next;
1419 free(rte, M_MRTABLE);
1422 ++mrtstat.mrts_cache_cleanups;
1426 * free the bw_meter entries
1428 while (mfc->mfc_bw_meter != NULL) {
1429 struct bw_meter *x = mfc->mfc_bw_meter;
1431 mfc->mfc_bw_meter = x->bm_mfc_next;
1435 *nptr = mfc->mfc_next;
1436 free(mfc, M_MRTABLE);
1438 nptr = &mfc->mfc_next;
1443 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL);
1447 * Packet forwarding routine once entry in the cache is made
1450 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif)
1452 struct ip *ip = mtod(m, struct ip *);
1454 int plen = ip->ip_len;
1457 * Macro to send packet on vif. Since RSVP packets don't get counted on
1458 * input, they shouldn't get counted on output, so statistics keeping is
1461 #define MC_SEND(ip,vifp,m) { \
1462 if ((vifp)->v_flags & VIFF_TUNNEL) \
1463 encap_send((ip), (vifp), (m)); \
1465 phyint_send((ip), (vifp), (m)); \
1469 * If xmt_vif is not -1, send on only the requested vif.
1471 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.)
1473 if (xmt_vif < numvifs) {
1475 if (viftable[xmt_vif].v_flags & VIFF_REGISTER)
1476 pim_register_send(ip, viftable + xmt_vif, m, rt);
1479 MC_SEND(ip, viftable + xmt_vif, m);
1484 * Don't forward if it didn't arrive from the parent vif for its origin.
1486 vifi = rt->mfc_parent;
1487 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) {
1488 /* came in the wrong interface */
1489 if (mrtdebug & DEBUG_FORWARD)
1490 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n",
1491 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp);
1492 ++mrtstat.mrts_wrong_if;
1495 * If we are doing PIM assert processing, send a message
1496 * to the routing daemon.
1498 * XXX: A PIM-SM router needs the WRONGVIF detection so it
1499 * can complete the SPT switch, regardless of the type
1500 * of the iif (broadcast media, GRE tunnel, etc).
1502 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) {
1507 if (ifp == &multicast_register_if)
1508 pimstat.pims_rcv_registers_wrongiif++;
1511 /* Get vifi for the incoming packet */
1512 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++)
1514 if (vifi >= numvifs)
1515 return 0; /* The iif is not found: ignore the packet. */
1517 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF)
1518 return 0; /* WRONGVIF disabled: ignore the packet */
1522 TV_DELTA(rt->mfc_last_assert, now, delta);
1524 if (delta > ASSERT_MSG_TIME) {
1525 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
1527 int hlen = ip->ip_hl << 2;
1528 struct mbuf *mm = m_copy(m, 0, hlen);
1530 if (mm && (M_HASCL(mm) || mm->m_len < hlen))
1531 mm = m_pullup(mm, hlen);
1535 rt->mfc_last_assert = now;
1537 im = mtod(mm, struct igmpmsg *);
1538 im->im_msgtype = IGMPMSG_WRONGVIF;
1542 mrtstat.mrts_upcalls++;
1544 k_igmpsrc.sin_addr = im->im_src;
1545 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) {
1547 "ip_mforward: ip_mrouter socket queue full\n");
1548 ++mrtstat.mrts_upq_sockfull;
1556 /* If I sourced this packet, it counts as output, else it was input. */
1557 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) {
1558 viftable[vifi].v_pkt_out++;
1559 viftable[vifi].v_bytes_out += plen;
1561 viftable[vifi].v_pkt_in++;
1562 viftable[vifi].v_bytes_in += plen;
1565 rt->mfc_byte_cnt += plen;
1568 * For each vif, decide if a copy of the packet should be forwarded.
1570 * - the ttl exceeds the vif's threshold
1571 * - there are group members downstream on interface
1573 for (vifi = 0; vifi < numvifs; vifi++)
1574 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) {
1575 viftable[vifi].v_pkt_out++;
1576 viftable[vifi].v_bytes_out += plen;
1578 if (viftable[vifi].v_flags & VIFF_REGISTER)
1579 pim_register_send(ip, viftable + vifi, m, rt);
1582 MC_SEND(ip, viftable+vifi, m);
1586 * Perform upcall-related bw measuring.
1588 if (rt->mfc_bw_meter != NULL) {
1593 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next)
1594 bw_meter_receive_packet(x, plen, &now);
1601 * check if a vif number is legal/ok. This is used by ip_output.
1604 X_legal_vif_num(int vif)
1606 return (vif >= 0 && vif < numvifs);
1610 * Return the local address used by this vif
1613 X_ip_mcast_src(int vifi)
1615 if (vifi >= 0 && vifi < numvifs)
1616 return viftable[vifi].v_lcl_addr.s_addr;
1622 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1624 struct mbuf *mb_copy;
1625 int hlen = ip->ip_hl << 2;
1628 * Make a new reference to the packet; make sure that
1629 * the IP header is actually copied, not just referenced,
1630 * so that ip_output() only scribbles on the copy.
1632 mb_copy = m_copypacket(m, MB_DONTWAIT);
1633 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen))
1634 mb_copy = m_pullup(mb_copy, hlen);
1635 if (mb_copy == NULL)
1638 if (vifp->v_rate_limit == 0)
1639 tbf_send_packet(vifp, mb_copy);
1641 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len);
1645 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m)
1647 struct mbuf *mb_copy;
1649 int i, len = ip->ip_len;
1651 /* Take care of delayed checksums */
1652 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
1653 in_delayed_cksum(m);
1654 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
1658 * copy the old packet & pullup its IP header into the
1659 * new mbuf so we can modify it. Try to fill the new
1660 * mbuf since if we don't the ethernet driver will.
1662 MGETHDR(mb_copy, MB_DONTWAIT, MT_HEADER);
1663 if (mb_copy == NULL)
1665 mb_copy->m_data += max_linkhdr;
1666 mb_copy->m_len = sizeof(multicast_encap_iphdr);
1668 if ((mb_copy->m_next = m_copypacket(m, MB_DONTWAIT)) == NULL) {
1672 i = MHLEN - M_LEADINGSPACE(mb_copy);
1675 mb_copy = m_pullup(mb_copy, i);
1676 if (mb_copy == NULL)
1678 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr);
1681 * fill in the encapsulating IP header.
1683 ip_copy = mtod(mb_copy, struct ip *);
1684 *ip_copy = multicast_encap_iphdr;
1686 ip_copy->ip_id = ip_randomid();
1688 ip_copy->ip_id = htons(ip_id++);
1690 ip_copy->ip_len += len;
1691 ip_copy->ip_src = vifp->v_lcl_addr;
1692 ip_copy->ip_dst = vifp->v_rmt_addr;
1695 * turn the encapsulated IP header back into a valid one.
1697 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr));
1699 ip->ip_len = htons(ip->ip_len);
1700 ip->ip_off = htons(ip->ip_off);
1702 mb_copy->m_data += sizeof(multicast_encap_iphdr);
1703 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
1704 mb_copy->m_data -= sizeof(multicast_encap_iphdr);
1706 if (vifp->v_rate_limit == 0)
1707 tbf_send_packet(vifp, mb_copy);
1709 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len);
1713 * De-encapsulate a packet and feed it back through ip input (this
1714 * routine is called whenever IP gets a packet with proto type
1715 * ENCAP_PROTO and a local destination address).
1717 * This is similar to mroute_encapcheck() + mroute_encap_input() in -current.
1720 X_ipip_input(struct mbuf *m, int off, int proto)
1722 struct ip *ip = mtod(m, struct ip *);
1723 int hlen = ip->ip_hl << 2;
1725 if (!have_encap_tunnel) {
1726 rip_input(m, off, proto);
1730 * dump the packet if it's not to a multicast destination or if
1731 * we don't have an encapsulating tunnel with the source.
1732 * Note: This code assumes that the remote site IP address
1733 * uniquely identifies the tunnel (i.e., that this site has
1734 * at most one tunnel with the remote site).
1736 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) {
1737 ++mrtstat.mrts_bad_tunnel;
1741 if (ip->ip_src.s_addr != last_encap_src) {
1742 struct vif *vifp = viftable;
1743 struct vif *vife = vifp + numvifs;
1745 last_encap_src = ip->ip_src.s_addr;
1746 last_encap_vif = NULL;
1747 for ( ; vifp < vife; ++vifp)
1748 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) {
1749 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT))
1751 last_encap_vif = vifp;
1755 if (last_encap_vif == NULL) {
1756 last_encap_src = INADDR_ANY;
1757 mrtstat.mrts_cant_tunnel++; /*XXX*/
1760 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n",
1761 (u_long)ntohl(ip->ip_src.s_addr));
1765 if (hlen > sizeof(struct ip))
1767 m->m_data += sizeof(struct ip);
1768 m->m_len -= sizeof(struct ip);
1769 m->m_pkthdr.len -= sizeof(struct ip);
1770 m->m_pkthdr.rcvif = last_encap_vif->v_ifp;
1772 netisr_queue(NETISR_IP, m);
1776 * Token bucket filter module
1780 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len)
1782 struct tbf *t = vifp->v_tbf;
1784 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */
1785 mrtstat.mrts_pkt2large++;
1790 tbf_update_tokens(vifp);
1792 if (t->tbf_q_len == 0) { /* queue empty... */
1793 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */
1794 t->tbf_n_tok -= p_len;
1795 tbf_send_packet(vifp, m);
1796 } else { /* no, queue packet and try later */
1798 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS,
1799 tbf_reprocess_q, vifp);
1801 } else if (t->tbf_q_len < t->tbf_max_q_len) {
1802 /* finite queue length, so queue pkts and process queue */
1804 tbf_process_q(vifp);
1806 /* queue full, try to dq and queue and process */
1807 if (!tbf_dq_sel(vifp, ip)) {
1808 mrtstat.mrts_q_overflow++;
1812 tbf_process_q(vifp);
1818 * adds a packet to the queue at the interface
1821 tbf_queue(struct vif *vifp, struct mbuf *m)
1824 struct tbf *t = vifp->v_tbf;
1826 if (t->tbf_t == NULL) /* Queue was empty */
1828 else /* Insert at tail */
1829 t->tbf_t->m_nextpkt = m;
1831 t->tbf_t = m; /* Set new tail pointer */
1834 /* Make sure we didn't get fed a bogus mbuf */
1836 panic("tbf_queue: m_nextpkt");
1838 m->m_nextpkt = NULL;
1846 * processes the queue at the interface
1849 tbf_process_q(struct vif *vifp)
1852 struct tbf *t = vifp->v_tbf;
1854 /* loop through the queue at the interface and send as many packets
1857 while (t->tbf_q_len > 0) {
1858 struct mbuf *m = t->tbf_q;
1859 int len = mtod(m, struct ip *)->ip_len;
1861 /* determine if the packet can be sent */
1862 if (len > t->tbf_n_tok) /* not enough tokens, we are done */
1864 /* ok, reduce no of tokens, dequeue and send the packet. */
1865 t->tbf_n_tok -= len;
1867 t->tbf_q = m->m_nextpkt;
1868 if (--t->tbf_q_len == 0)
1871 m->m_nextpkt = NULL;
1872 tbf_send_packet(vifp, m);
1878 tbf_reprocess_q(void *xvifp)
1880 struct vif *vifp = xvifp;
1882 if (ip_mrouter == NULL)
1884 tbf_update_tokens(vifp);
1885 tbf_process_q(vifp);
1886 if (vifp->v_tbf->tbf_q_len)
1887 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS,
1888 tbf_reprocess_q, vifp);
1891 /* function that will selectively discard a member of the queue
1892 * based on the precedence value and the priority
1895 tbf_dq_sel(struct vif *vifp, struct ip *ip)
1899 struct mbuf *m, *last;
1901 struct tbf *t = vifp->v_tbf;
1903 p = priority(vifp, ip);
1907 while ((m = *np) != NULL) {
1908 if (p > priority(vifp, mtod(m, struct ip *))) {
1910 /* If we're removing the last packet, fix the tail pointer */
1914 /* It's impossible for the queue to be empty, but check anyways. */
1915 if (--t->tbf_q_len == 0)
1918 mrtstat.mrts_drop_sel++;
1929 tbf_send_packet(struct vif *vifp, struct mbuf *m)
1933 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */
1934 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL);
1936 struct ip_moptions imo;
1938 static struct route ro; /* XXX check this */
1940 imo.imo_multicast_ifp = vifp->v_ifp;
1941 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1;
1942 imo.imo_multicast_loop = 1;
1943 imo.imo_multicast_vif = -1;
1946 * Re-entrancy should not be a problem here, because
1947 * the packets that we send out and are looped back at us
1948 * should get rejected because they appear to come from
1949 * the loopback interface, thus preventing looping.
1951 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL);
1953 if (mrtdebug & DEBUG_XMIT)
1954 log(LOG_DEBUG, "phyint_send on vif %d err %d\n",
1955 (int)(vifp - viftable), error);
1960 /* determine the current time and then
1961 * the elapsed time (between the last time and time now)
1962 * in milliseconds & update the no. of tokens in the bucket
1965 tbf_update_tokens(struct vif *vifp)
1970 struct tbf *t = vifp->v_tbf;
1974 TV_DELTA(tp, t->tbf_last_pkt_t, tm);
1977 * This formula is actually
1978 * "time in seconds" * "bytes/second".
1980 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8)
1982 * The (1000/1024) was introduced in add_vif to optimize
1983 * this divide into a shift.
1985 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8;
1986 t->tbf_last_pkt_t = tp;
1988 if (t->tbf_n_tok > MAX_BKT_SIZE)
1989 t->tbf_n_tok = MAX_BKT_SIZE;
1995 priority(struct vif *vifp, struct ip *ip)
1997 int prio = 50; /* the lowest priority -- default case */
1999 /* temporary hack; may add general packet classifier some day */
2002 * The UDP port space is divided up into four priority ranges:
2003 * [0, 16384) : unclassified - lowest priority
2004 * [16384, 32768) : audio - highest priority
2005 * [32768, 49152) : whiteboard - medium priority
2006 * [49152, 65536) : video - low priority
2008 * Everything else gets lowest priority.
2010 if (ip->ip_p == IPPROTO_UDP) {
2011 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2));
2012 switch (ntohs(udp->uh_dport) & 0xc000) {
2028 * End of token bucket filter modifications
2032 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt)
2036 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2039 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi);
2045 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */
2047 return EADDRNOTAVAIL;
2050 if (sopt->sopt_name == IP_RSVP_VIF_ON) {
2051 /* Check if socket is available. */
2052 if (viftable[vifi].v_rsvpd != NULL) {
2057 viftable[vifi].v_rsvpd = so;
2058 /* This may seem silly, but we need to be sure we don't over-increment
2059 * the RSVP counter, in case something slips up.
2061 if (!viftable[vifi].v_rsvp_on) {
2062 viftable[vifi].v_rsvp_on = 1;
2065 } else { /* must be VIF_OFF */
2067 * XXX as an additional consistency check, one could make sure
2068 * that viftable[vifi].v_rsvpd == so, otherwise passing so as
2069 * first parameter is pretty useless.
2071 viftable[vifi].v_rsvpd = NULL;
2073 * This may seem silly, but we need to be sure we don't over-decrement
2074 * the RSVP counter, in case something slips up.
2076 if (viftable[vifi].v_rsvp_on) {
2077 viftable[vifi].v_rsvp_on = 0;
2086 X_ip_rsvp_force_done(struct socket *so)
2091 /* Don't bother if it is not the right type of socket. */
2092 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP)
2097 /* The socket may be attached to more than one vif...this
2098 * is perfectly legal.
2100 for (vifi = 0; vifi < numvifs; vifi++) {
2101 if (viftable[vifi].v_rsvpd == so) {
2102 viftable[vifi].v_rsvpd = NULL;
2103 /* This may seem silly, but we need to be sure we don't
2104 * over-decrement the RSVP counter, in case something slips up.
2106 if (viftable[vifi].v_rsvp_on) {
2107 viftable[vifi].v_rsvp_on = 0;
2117 X_rsvp_input(struct mbuf *m, ...)
2120 struct ip *ip = mtod(m, struct ip *);
2121 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET };
2126 /* support IP_RECVIF used by rsvpd rel4.2a1 */
2134 off = __va_arg(ap, int);
2135 proto = __va_arg(ap, int);
2139 printf("rsvp_input: rsvp_on %d\n",rsvp_on);
2141 /* Can still get packets with rsvp_on = 0 if there is a local member
2142 * of the group to which the RSVP packet is addressed. But in this
2143 * case we want to throw the packet away.
2153 printf("rsvp_input: check vifs\n");
2156 if (!(m->m_flags & M_PKTHDR))
2157 panic("rsvp_input no hdr");
2160 ifp = m->m_pkthdr.rcvif;
2161 /* Find which vif the packet arrived on. */
2162 for (vifi = 0; vifi < numvifs; vifi++)
2163 if (viftable[vifi].v_ifp == ifp)
2167 if (vifi == numvifs || (so = viftable[vifi].v_rsvpd) == NULL) {
2169 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) {
2172 * If the old-style non-vif-associated socket is set,
2173 * then use it. Otherwise, drop packet since there
2174 * is no specific socket for this vif.
2176 if (ip_rsvpd != NULL) {
2178 printf("rsvp_input: Sending packet up old-style socket\n");
2179 rip_input(m, off, proto); /* xxx */
2181 if (rsvpdebug && vifi == numvifs)
2182 printf("rsvp_input: Can't find vif for packet.\n");
2183 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL)
2184 printf("rsvp_input: No socket defined for vif %d\n",vifi);
2190 rsvp_src.sin_addr = ip->ip_src;
2193 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n",
2194 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv)));
2198 inp = (struct inpcb *)so->so_pcb;
2199 if (inp->inp_flags & INP_CONTROLOPTS ||
2200 inp->inp_socket->so_options & SO_TIMESTAMP)
2201 ip_savecontrol(inp, &opts, ip, m);
2202 if (sbappendaddr(&so->so_rcv,
2203 (struct sockaddr *)&rsvp_src,m, opts) == 0) {
2208 printf("rsvp_input: Failed to append to socket\n");
2213 printf("rsvp_input: send packet up\n");
2216 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) {
2218 printf("rsvp_input: Failed to append to socket\n");
2221 printf("rsvp_input: send packet up\n");
2229 * Code for bandwidth monitors
2233 * Define common interface for timeval-related methods
2235 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp)
2236 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp))
2237 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp))
2240 compute_bw_meter_flags(struct bw_upcall *req)
2244 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS)
2245 flags |= BW_METER_UNIT_PACKETS;
2246 if (req->bu_flags & BW_UPCALL_UNIT_BYTES)
2247 flags |= BW_METER_UNIT_BYTES;
2248 if (req->bu_flags & BW_UPCALL_GEQ)
2249 flags |= BW_METER_GEQ;
2250 if (req->bu_flags & BW_UPCALL_LEQ)
2251 flags |= BW_METER_LEQ;
2257 * Add a bw_meter entry
2260 add_bw_upcall(struct bw_upcall *req)
2263 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC,
2264 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC };
2270 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2273 /* Test if the flags are valid */
2274 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES)))
2276 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)))
2278 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2279 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ))
2282 /* Test if the threshold time interval is valid */
2283 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <))
2286 flags = compute_bw_meter_flags(req);
2289 * Find if we have already same bw_meter entry
2292 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2295 return EADDRNOTAVAIL;
2297 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) {
2298 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2299 &req->bu_threshold.b_time, ==)) &&
2300 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2301 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2302 (x->bm_flags & BW_METER_USER_FLAGS) == flags) {
2304 return 0; /* XXX Already installed */
2309 /* Allocate the new bw_meter entry */
2310 x = malloc(sizeof(*x), M_BWMETER, M_INTWAIT);
2312 /* Set the new bw_meter entry */
2313 x->bm_threshold.b_time = req->bu_threshold.b_time;
2315 x->bm_start_time = now;
2316 x->bm_threshold.b_packets = req->bu_threshold.b_packets;
2317 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes;
2318 x->bm_measured.b_packets = 0;
2319 x->bm_measured.b_bytes = 0;
2320 x->bm_flags = flags;
2321 x->bm_time_next = NULL;
2322 x->bm_time_hash = BW_METER_BUCKETS;
2324 /* Add the new bw_meter entry to the front of entries for this MFC */
2327 x->bm_mfc_next = mfc->mfc_bw_meter;
2328 mfc->mfc_bw_meter = x;
2329 schedule_bw_meter(x, &now);
2336 free_bw_list(struct bw_meter *list)
2338 while (list != NULL) {
2339 struct bw_meter *x = list;
2341 list = list->bm_mfc_next;
2342 unschedule_bw_meter(x);
2348 * Delete one or multiple bw_meter entries
2351 del_bw_upcall(struct bw_upcall *req)
2357 if (!(mrt_api_config & MRT_MFC_BW_UPCALL))
2361 /* Find the corresponding MFC entry */
2362 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr);
2365 return EADDRNOTAVAIL;
2366 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) {
2368 * Delete all bw_meter entries for this mfc
2370 struct bw_meter *list;
2372 list = mfc->mfc_bw_meter;
2373 mfc->mfc_bw_meter = NULL;
2377 } else { /* Delete a single bw_meter entry */
2378 struct bw_meter *prev;
2381 flags = compute_bw_meter_flags(req);
2383 /* Find the bw_meter entry to delete */
2384 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL;
2385 prev = x, x = x->bm_mfc_next) {
2386 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time,
2387 &req->bu_threshold.b_time, ==)) &&
2388 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) &&
2389 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) &&
2390 (x->bm_flags & BW_METER_USER_FLAGS) == flags)
2393 if (x != NULL) { /* Delete entry from the list for this MFC */
2395 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/
2397 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */
2400 unschedule_bw_meter(x);
2401 /* Free the bw_meter entry */
2413 * Perform bandwidth measurement processing that may result in an upcall
2416 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp)
2418 struct timeval delta;
2423 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2425 if (x->bm_flags & BW_METER_GEQ) {
2427 * Processing for ">=" type of bw_meter entry
2429 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2430 /* Reset the bw_meter entry */
2431 x->bm_start_time = *nowp;
2432 x->bm_measured.b_packets = 0;
2433 x->bm_measured.b_bytes = 0;
2434 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2437 /* Record that a packet is received */
2438 x->bm_measured.b_packets++;
2439 x->bm_measured.b_bytes += plen;
2442 * Test if we should deliver an upcall
2444 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) {
2445 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2446 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) ||
2447 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2448 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) {
2449 /* Prepare an upcall for delivery */
2450 bw_meter_prepare_upcall(x, nowp);
2451 x->bm_flags |= BW_METER_UPCALL_DELIVERED;
2454 } else if (x->bm_flags & BW_METER_LEQ) {
2456 * Processing for "<=" type of bw_meter entry
2458 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) {
2460 * We are behind time with the multicast forwarding table
2461 * scanning for "<=" type of bw_meter entries, so test now
2462 * if we should deliver an upcall.
2464 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2465 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2466 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2467 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2468 /* Prepare an upcall for delivery */
2469 bw_meter_prepare_upcall(x, nowp);
2471 /* Reschedule the bw_meter entry */
2472 unschedule_bw_meter(x);
2473 schedule_bw_meter(x, nowp);
2476 /* Record that a packet is received */
2477 x->bm_measured.b_packets++;
2478 x->bm_measured.b_bytes += plen;
2481 * Test if we should restart the measuring interval
2483 if ((x->bm_flags & BW_METER_UNIT_PACKETS &&
2484 x->bm_measured.b_packets <= x->bm_threshold.b_packets) ||
2485 (x->bm_flags & BW_METER_UNIT_BYTES &&
2486 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) {
2487 /* Don't restart the measuring interval */
2489 /* Do restart the measuring interval */
2491 * XXX: note that we don't unschedule and schedule, because this
2492 * might be too much overhead per packet. Instead, when we process
2493 * all entries for a given timer hash bin, we check whether it is
2494 * really a timeout. If not, we reschedule at that time.
2496 x->bm_start_time = *nowp;
2497 x->bm_measured.b_packets = 0;
2498 x->bm_measured.b_bytes = 0;
2499 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2506 * Prepare a bandwidth-related upcall
2509 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp)
2511 struct timeval delta;
2512 struct bw_upcall *u;
2518 * Compute the measured time interval
2521 BW_TIMEVALDECR(&delta, &x->bm_start_time);
2524 * If there are too many pending upcalls, deliver them now
2526 if (bw_upcalls_n >= BW_UPCALLS_MAX)
2530 * Set the bw_upcall entry
2532 u = &bw_upcalls[bw_upcalls_n++];
2533 u->bu_src = x->bm_mfc->mfc_origin;
2534 u->bu_dst = x->bm_mfc->mfc_mcastgrp;
2535 u->bu_threshold.b_time = x->bm_threshold.b_time;
2536 u->bu_threshold.b_packets = x->bm_threshold.b_packets;
2537 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes;
2538 u->bu_measured.b_time = delta;
2539 u->bu_measured.b_packets = x->bm_measured.b_packets;
2540 u->bu_measured.b_bytes = x->bm_measured.b_bytes;
2542 if (x->bm_flags & BW_METER_UNIT_PACKETS)
2543 u->bu_flags |= BW_UPCALL_UNIT_PACKETS;
2544 if (x->bm_flags & BW_METER_UNIT_BYTES)
2545 u->bu_flags |= BW_UPCALL_UNIT_BYTES;
2546 if (x->bm_flags & BW_METER_GEQ)
2547 u->bu_flags |= BW_UPCALL_GEQ;
2548 if (x->bm_flags & BW_METER_LEQ)
2549 u->bu_flags |= BW_UPCALL_LEQ;
2555 * Send the pending bandwidth-related upcalls
2558 bw_upcalls_send(void)
2561 int len = bw_upcalls_n * sizeof(bw_upcalls[0]);
2562 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2563 static struct igmpmsg igmpmsg = { 0, /* unused1 */
2565 IGMPMSG_BW_UPCALL,/* im_msgtype */
2570 { 0 } }; /* im_dst */
2572 if (bw_upcalls_n == 0)
2573 return; /* No pending upcalls */
2578 * Allocate a new mbuf, initialize it with the header and
2579 * the payload for the pending calls.
2581 MGETHDR(m, MB_DONTWAIT, MT_HEADER);
2583 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n");
2587 m->m_len = m->m_pkthdr.len = 0;
2588 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg);
2589 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]);
2593 * XXX do we need to set the address in k_igmpsrc ?
2595 mrtstat.mrts_upcalls++;
2596 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) {
2597 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n");
2598 ++mrtstat.mrts_upq_sockfull;
2603 * Compute the timeout hash value for the bw_meter entries
2605 #define BW_METER_TIMEHASH(bw_meter, hash) \
2607 struct timeval next_timeval = (bw_meter)->bm_start_time; \
2609 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \
2610 (hash) = next_timeval.tv_sec; \
2611 if (next_timeval.tv_usec) \
2612 (hash)++; /* XXX: make sure we don't timeout early */ \
2613 (hash) %= BW_METER_BUCKETS; \
2617 * Schedule a timer to process periodically bw_meter entry of type "<="
2618 * by linking the entry in the proper hash bucket.
2621 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp)
2625 if (!(x->bm_flags & BW_METER_LEQ))
2626 return; /* XXX: we schedule timers only for "<=" entries */
2629 * Reset the bw_meter entry
2632 x->bm_start_time = *nowp;
2633 x->bm_measured.b_packets = 0;
2634 x->bm_measured.b_bytes = 0;
2635 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED;
2639 * Compute the timeout hash value and insert the entry
2641 BW_METER_TIMEHASH(x, time_hash);
2642 x->bm_time_next = bw_meter_timers[time_hash];
2643 bw_meter_timers[time_hash] = x;
2644 x->bm_time_hash = time_hash;
2648 * Unschedule the periodic timer that processes bw_meter entry of type "<="
2649 * by removing the entry from the proper hash bucket.
2652 unschedule_bw_meter(struct bw_meter *x)
2655 struct bw_meter *prev, *tmp;
2657 if (!(x->bm_flags & BW_METER_LEQ))
2658 return; /* XXX: we schedule timers only for "<=" entries */
2661 * Compute the timeout hash value and delete the entry
2663 time_hash = x->bm_time_hash;
2664 if (time_hash >= BW_METER_BUCKETS)
2665 return; /* Entry was not scheduled */
2667 for (prev = NULL, tmp = bw_meter_timers[time_hash];
2668 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next)
2673 panic("unschedule_bw_meter: bw_meter entry not found");
2676 prev->bm_time_next = x->bm_time_next;
2678 bw_meter_timers[time_hash] = x->bm_time_next;
2680 x->bm_time_next = NULL;
2681 x->bm_time_hash = BW_METER_BUCKETS;
2686 * Process all "<=" type of bw_meter that should be processed now,
2687 * and for each entry prepare an upcall if necessary. Each processed
2688 * entry is rescheduled again for the (periodic) processing.
2690 * This is run periodically (once per second normally). On each round,
2691 * all the potentially matching entries are in the hash slot that we are
2697 static uint32_t last_tv_sec; /* last time we processed this */
2701 struct timeval now, process_endtime;
2704 if (last_tv_sec == now.tv_sec)
2705 return; /* nothing to do */
2708 loops = now.tv_sec - last_tv_sec;
2709 last_tv_sec = now.tv_sec;
2710 if (loops > BW_METER_BUCKETS)
2711 loops = BW_METER_BUCKETS;
2714 * Process all bins of bw_meter entries from the one after the last
2715 * processed to the current one. On entry, i points to the last bucket
2716 * visited, so we need to increment i at the beginning of the loop.
2718 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) {
2719 struct bw_meter *x, *tmp_list;
2721 if (++i >= BW_METER_BUCKETS)
2724 /* Disconnect the list of bw_meter entries from the bin */
2725 tmp_list = bw_meter_timers[i];
2726 bw_meter_timers[i] = NULL;
2728 /* Process the list of bw_meter entries */
2729 while (tmp_list != NULL) {
2731 tmp_list = tmp_list->bm_time_next;
2733 /* Test if the time interval is over */
2734 process_endtime = x->bm_start_time;
2735 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time);
2736 if (BW_TIMEVALCMP(&process_endtime, &now, >)) {
2737 /* Not yet: reschedule, but don't reset */
2740 BW_METER_TIMEHASH(x, time_hash);
2741 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) {
2743 * XXX: somehow the bin processing is a bit ahead of time.
2744 * Put the entry in the next bin.
2746 if (++time_hash >= BW_METER_BUCKETS)
2749 x->bm_time_next = bw_meter_timers[time_hash];
2750 bw_meter_timers[time_hash] = x;
2751 x->bm_time_hash = time_hash;
2757 * Test if we should deliver an upcall
2759 if (((x->bm_flags & BW_METER_UNIT_PACKETS) &&
2760 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) ||
2761 ((x->bm_flags & BW_METER_UNIT_BYTES) &&
2762 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) {
2763 /* Prepare an upcall for delivery */
2764 bw_meter_prepare_upcall(x, &now);
2768 * Reschedule for next processing
2770 schedule_bw_meter(x, &now);
2775 /* Send all upcalls that are pending delivery */
2780 * A periodic function for sending all upcalls that are pending delivery
2783 expire_bw_upcalls_send(void *unused)
2787 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD,
2788 expire_bw_upcalls_send, NULL);
2792 * A periodic function for periodic scanning of the multicast forwarding
2793 * table for processing all "<=" bw_meter entries.
2796 expire_bw_meter_process(void *unused)
2798 if (mrt_api_config & MRT_MFC_BW_UPCALL)
2801 callout_reset(&bw_meter_ch, BW_METER_PERIOD,
2802 expire_bw_meter_process, NULL);
2806 * End of bandwidth monitoring code
2811 * Send the packet up to the user daemon, or eventually do kernel encapsulation
2815 pim_register_send(struct ip *ip, struct vif *vifp,
2816 struct mbuf *m, struct mfc *rt)
2818 struct mbuf *mb_copy, *mm;
2820 if (mrtdebug & DEBUG_PIM)
2821 log(LOG_DEBUG, "pim_register_send: ");
2823 mb_copy = pim_register_prepare(ip, m);
2824 if (mb_copy == NULL)
2828 * Send all the fragments. Note that the mbuf for each fragment
2829 * is freed by the sending machinery.
2831 for (mm = mb_copy; mm; mm = mb_copy) {
2832 mb_copy = mm->m_nextpkt;
2834 mm = m_pullup(mm, sizeof(struct ip));
2836 ip = mtod(mm, struct ip *);
2837 if ((mrt_api_config & MRT_MFC_RP) &&
2838 (rt->mfc_rp.s_addr != INADDR_ANY)) {
2839 pim_register_send_rp(ip, vifp, mm, rt);
2841 pim_register_send_upcall(ip, vifp, mm, rt);
2850 * Return a copy of the data packet that is ready for PIM Register
2852 * XXX: Note that in the returned copy the IP header is a valid one.
2854 static struct mbuf *
2855 pim_register_prepare(struct ip *ip, struct mbuf *m)
2857 struct mbuf *mb_copy = NULL;
2860 /* Take care of delayed checksums */
2861 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) {
2862 in_delayed_cksum(m);
2863 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA;
2867 * Copy the old packet & pullup its IP header into the
2868 * new mbuf so we can modify it.
2870 mb_copy = m_copypacket(m, MB_DONTWAIT);
2871 if (mb_copy == NULL)
2873 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2);
2874 if (mb_copy == NULL)
2877 /* take care of the TTL */
2878 ip = mtod(mb_copy, struct ip *);
2881 /* Compute the MTU after the PIM Register encapsulation */
2882 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr);
2884 if (ip->ip_len <= mtu) {
2885 /* Turn the IP header into a valid one */
2886 ip->ip_len = htons(ip->ip_len);
2887 ip->ip_off = htons(ip->ip_off);
2889 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2);
2891 /* Fragment the packet */
2892 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) {
2901 * Send an upcall with the data packet to the user-level process.
2904 pim_register_send_upcall(struct ip *ip, struct vif *vifp,
2905 struct mbuf *mb_copy, struct mfc *rt)
2907 struct mbuf *mb_first;
2908 int len = ntohs(ip->ip_len);
2910 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET };
2913 * Add a new mbuf with an upcall header
2915 MGETHDR(mb_first, MB_DONTWAIT, MT_HEADER);
2916 if (mb_first == NULL) {
2920 mb_first->m_data += max_linkhdr;
2921 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg);
2922 mb_first->m_len = sizeof(struct igmpmsg);
2923 mb_first->m_next = mb_copy;
2925 /* Send message to routing daemon */
2926 im = mtod(mb_first, struct igmpmsg *);
2927 im->im_msgtype = IGMPMSG_WHOLEPKT;
2929 im->im_vif = vifp - viftable;
2930 im->im_src = ip->ip_src;
2931 im->im_dst = ip->ip_dst;
2933 k_igmpsrc.sin_addr = ip->ip_src;
2935 mrtstat.mrts_upcalls++;
2937 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) {
2938 if (mrtdebug & DEBUG_PIM)
2940 "mcast: pim_register_send_upcall: ip_mrouter socket queue full");
2941 ++mrtstat.mrts_upq_sockfull;
2945 /* Keep statistics */
2946 pimstat.pims_snd_registers_msgs++;
2947 pimstat.pims_snd_registers_bytes += len;
2953 * Encapsulate the data packet in PIM Register message and send it to the RP.
2956 pim_register_send_rp(struct ip *ip, struct vif *vifp,
2957 struct mbuf *mb_copy, struct mfc *rt)
2959 struct mbuf *mb_first;
2960 struct ip *ip_outer;
2961 struct pim_encap_pimhdr *pimhdr;
2962 int len = ntohs(ip->ip_len);
2963 vifi_t vifi = rt->mfc_parent;
2965 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) {
2967 return EADDRNOTAVAIL; /* The iif vif is invalid */
2971 * Add a new mbuf with the encapsulating header
2973 MGETHDR(mb_first, MB_DONTWAIT, MT_HEADER);
2974 if (mb_first == NULL) {
2978 mb_first->m_data += max_linkhdr;
2979 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2980 mb_first->m_next = mb_copy;
2982 mb_first->m_pkthdr.len = len + mb_first->m_len;
2985 * Fill in the encapsulating IP and PIM header
2987 ip_outer = mtod(mb_first, struct ip *);
2988 *ip_outer = pim_encap_iphdr;
2990 ip_outer->ip_id = ip_randomid();
2992 ip_outer->ip_id = htons(ip_id++);
2994 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr);
2995 ip_outer->ip_src = viftable[vifi].v_lcl_addr;
2996 ip_outer->ip_dst = rt->mfc_rp;
2998 * Copy the inner header TOS to the outer header, and take care of the
3001 ip_outer->ip_tos = ip->ip_tos;
3002 if (ntohs(ip->ip_off) & IP_DF)
3003 ip_outer->ip_off |= IP_DF;
3004 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer
3005 + sizeof(pim_encap_iphdr));
3006 *pimhdr = pim_encap_pimhdr;
3007 /* If the iif crosses a border, set the Border-bit */
3008 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config)
3009 pimhdr->flags |= htonl(PIM_BORDER_REGISTER);
3011 mb_first->m_data += sizeof(pim_encap_iphdr);
3012 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr));
3013 mb_first->m_data -= sizeof(pim_encap_iphdr);
3015 if (vifp->v_rate_limit == 0)
3016 tbf_send_packet(vifp, mb_first);
3018 tbf_control(vifp, mb_first, ip, ip_outer->ip_len);
3020 /* Keep statistics */
3021 pimstat.pims_snd_registers_msgs++;
3022 pimstat.pims_snd_registers_bytes += len;
3028 * PIM-SMv2 and PIM-DM messages processing.
3029 * Receives and verifies the PIM control messages, and passes them
3030 * up to the listening socket, using rip_input().
3031 * The only message with special processing is the PIM_REGISTER message
3032 * (used by PIM-SM): the PIM header is stripped off, and the inner packet
3033 * is passed to if_simloop().
3036 pim_input(struct mbuf *m, ...)
3039 struct ip *ip = mtod(m, struct ip *);
3042 int datalen = ip->ip_len;
3048 off = __va_arg(ap, int);
3049 proto = __va_arg(ap, int);
3054 /* Keep statistics */
3055 pimstat.pims_rcv_total_msgs++;
3056 pimstat.pims_rcv_total_bytes += datalen;
3061 if (datalen < PIM_MINLEN) {
3062 pimstat.pims_rcv_tooshort++;
3063 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n",
3064 datalen, (u_long)ip->ip_src.s_addr);
3070 * If the packet is at least as big as a REGISTER, go agead
3071 * and grab the PIM REGISTER header size, to avoid another
3072 * possible m_pullup() later.
3074 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8
3075 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28
3077 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN);
3079 * Get the IP and PIM headers in contiguous memory, and
3080 * possibly the PIM REGISTER header.
3082 if ((m->m_flags & M_EXT || m->m_len < minlen) &&
3083 (m = m_pullup(m, minlen)) == 0) {
3084 log(LOG_ERR, "pim_input: m_pullup failure\n");
3087 /* m_pullup() may have given us a new mbuf so reset ip. */
3088 ip = mtod(m, struct ip *);
3089 ip_tos = ip->ip_tos;
3091 /* adjust mbuf to point to the PIM header */
3092 m->m_data += iphlen;
3094 pim = mtod(m, struct pim *);
3097 * Validate checksum. If PIM REGISTER, exclude the data packet.
3099 * XXX: some older PIMv2 implementations don't make this distinction,
3100 * so for compatibility reason perform the checksum over part of the
3101 * message, and if error, then over the whole message.
3103 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) {
3104 /* do nothing, checksum okay */
3105 } else if (in_cksum(m, datalen)) {
3106 pimstat.pims_rcv_badsum++;
3107 if (mrtdebug & DEBUG_PIM)
3108 log(LOG_DEBUG, "pim_input: invalid checksum");
3113 /* PIM version check */
3114 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) {
3115 pimstat.pims_rcv_badversion++;
3116 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n",
3117 PIM_VT_V(pim->pim_vt), PIM_VERSION);
3122 /* restore mbuf back to the outer IP */
3123 m->m_data -= iphlen;
3126 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) {
3128 * Since this is a REGISTER, we'll make a copy of the register
3129 * headers ip + pim + u_int32 + encap_ip, to be passed up to the
3132 struct sockaddr_in dst = { sizeof(dst), AF_INET };
3134 struct ip *encap_ip;
3137 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) {
3138 if (mrtdebug & DEBUG_PIM)
3140 "pim_input: register vif not set: %d\n", reg_vif_num);
3148 if (datalen < PIM_REG_MINLEN) {
3149 pimstat.pims_rcv_tooshort++;
3150 pimstat.pims_rcv_badregisters++;
3152 "pim_input: register packet size too small %d from %lx\n",
3153 datalen, (u_long)ip->ip_src.s_addr);
3158 reghdr = (u_int32_t *)(pim + 1);
3159 encap_ip = (struct ip *)(reghdr + 1);
3161 if (mrtdebug & DEBUG_PIM) {
3163 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n",
3164 (u_long)ntohl(encap_ip->ip_src.s_addr),
3165 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3166 ntohs(encap_ip->ip_len));
3169 /* verify the version number of the inner packet */
3170 if (encap_ip->ip_v != IPVERSION) {
3171 pimstat.pims_rcv_badregisters++;
3172 if (mrtdebug & DEBUG_PIM) {
3173 log(LOG_DEBUG, "pim_input: invalid IP version (%d) "
3174 "of the inner packet\n", encap_ip->ip_v);
3180 /* verify the inner packet is destined to a mcast group */
3181 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) {
3182 pimstat.pims_rcv_badregisters++;
3183 if (mrtdebug & DEBUG_PIM)
3185 "pim_input: inner packet of register is not "
3187 (u_long)ntohl(encap_ip->ip_dst.s_addr));
3192 /* If a NULL_REGISTER, pass it to the daemon */
3193 if ((ntohl(*reghdr) & PIM_NULL_REGISTER))
3194 goto pim_input_to_daemon;
3197 * Copy the TOS from the outer IP header to the inner IP header.
3199 if (encap_ip->ip_tos != ip_tos) {
3200 /* Outer TOS -> inner TOS */
3201 encap_ip->ip_tos = ip_tos;
3202 /* Recompute the inner header checksum. Sigh... */
3204 /* adjust mbuf to point to the inner IP header */
3205 m->m_data += (iphlen + PIM_MINLEN);
3206 m->m_len -= (iphlen + PIM_MINLEN);
3208 encap_ip->ip_sum = 0;
3209 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2);
3211 /* restore mbuf to point back to the outer IP header */
3212 m->m_data -= (iphlen + PIM_MINLEN);
3213 m->m_len += (iphlen + PIM_MINLEN);
3217 * Decapsulate the inner IP packet and loopback to forward it
3218 * as a normal multicast packet. Also, make a copy of the
3219 * outer_iphdr + pimhdr + reghdr + encap_iphdr
3220 * to pass to the daemon later, so it can take the appropriate
3221 * actions (e.g., send back PIM_REGISTER_STOP).
3222 * XXX: here m->m_data points to the outer IP header.
3224 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN);
3227 "pim_input: pim register: could not copy register head\n");
3232 /* Keep statistics */
3233 /* XXX: registers_bytes include only the encap. mcast pkt */
3234 pimstat.pims_rcv_registers_msgs++;
3235 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len);
3238 * forward the inner ip packet; point m_data at the inner ip.
3240 m_adj(m, iphlen + PIM_MINLEN);
3242 if (mrtdebug & DEBUG_PIM) {
3244 "pim_input: forwarding decapsulated register: "
3245 "src %lx, dst %lx, vif %d\n",
3246 (u_long)ntohl(encap_ip->ip_src.s_addr),
3247 (u_long)ntohl(encap_ip->ip_dst.s_addr),
3250 if_simloop(viftable[reg_vif_num].v_ifp, m, dst.sin_family, 0);
3252 /* prepare the register head to send to the mrouting daemon */
3256 pim_input_to_daemon:
3258 * Pass the PIM message up to the daemon; if it is a Register message,
3259 * pass the 'head' only up to the daemon. This includes the
3260 * outer IP header, PIM header, PIM-Register header and the
3262 * XXX: the outer IP header pkt size of a Register is not adjust to
3263 * reflect the fact that the inner multicast data is truncated.
3265 rip_input(m, iphlen, proto);
3272 ip_mroute_modevent(module_t mod, int type, void *unused)
3279 /* XXX Protect against multiple loading */
3280 ip_mcast_src = X_ip_mcast_src;
3281 ip_mforward = X_ip_mforward;
3282 ip_mrouter_done = X_ip_mrouter_done;
3283 ip_mrouter_get = X_ip_mrouter_get;
3284 ip_mrouter_set = X_ip_mrouter_set;
3285 ip_rsvp_force_done = X_ip_rsvp_force_done;
3286 ip_rsvp_vif = X_ip_rsvp_vif;
3287 ipip_input = X_ipip_input;
3288 legal_vif_num = X_legal_vif_num;
3289 mrt_ioctl = X_mrt_ioctl;
3290 rsvp_input_p = X_rsvp_input;
3299 ip_mcast_src = NULL;
3301 ip_mrouter_done = NULL;
3302 ip_mrouter_get = NULL;
3303 ip_mrouter_set = NULL;
3304 ip_rsvp_force_done = NULL;
3307 legal_vif_num = NULL;
3309 rsvp_input_p = NULL;
3316 static moduledata_t ip_mroutemod = {
3321 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY);