ip_mroute.c revision 119401
1/* 2 * IP multicast forwarding procedures 3 * 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 15 * 16 * MROUTING Revision: 3.5 17 * and PIM-SMv2 and PIM-DM support, advanced API support, 18 * bandwidth metering and signaling 19 * 20 * $FreeBSD: head/sys/netinet/ip_mroute.c 119401 2003-08-24 08:27:57Z hsu $ 21 */ 22 23#include "opt_mac.h" 24#include "opt_mrouting.h" 25#include "opt_random_ip_id.h" 26 27#ifdef PIM 28#define _PIM_VT 1 29#endif 30 31#include <sys/param.h> 32#include <sys/kernel.h> 33#include <sys/lock.h> 34#include <sys/mac.h> 35#include <sys/malloc.h> 36#include <sys/mbuf.h> 37#include <sys/protosw.h> 38#include <sys/signalvar.h> 39#include <sys/socket.h> 40#include <sys/socketvar.h> 41#include <sys/sockio.h> 42#include <sys/sx.h> 43#include <sys/sysctl.h> 44#include <sys/syslog.h> 45#include <sys/systm.h> 46#include <sys/time.h> 47#include <net/if.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 <netinet/ip_encap.h> 56#include <netinet/ip_mroute.h> 57#include <netinet/ip_var.h> 58#ifdef PIM 59#include <netinet/pim.h> 60#include <netinet/pim_var.h> 61#endif 62#include <netinet/udp.h> 63#include <machine/in_cksum.h> 64 65/* 66 * Control debugging code for rsvp and multicast routing code. 67 * Can only set them with the debugger. 68 */ 69static u_int rsvpdebug; /* non-zero enables debugging */ 70 71static u_int mrtdebug; /* any set of the flags below */ 72#define DEBUG_MFC 0x02 73#define DEBUG_FORWARD 0x04 74#define DEBUG_EXPIRE 0x08 75#define DEBUG_XMIT 0x10 76#define DEBUG_PIM 0x20 77 78#define VIFI_INVALID ((vifi_t) -1) 79 80#define M_HASCL(m) ((m)->m_flags & M_EXT) 81 82static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 83 84static struct mrtstat mrtstat; 85SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 86 &mrtstat, mrtstat, 87 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 88 89static struct mfc *mfctable[MFCTBLSIZ]; 90SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 91 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 92 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 93 94static struct vif viftable[MAXVIFS]; 95SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 96 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 97 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 98 99static u_char nexpire[MFCTBLSIZ]; 100 101static struct callout_handle expire_upcalls_ch; 102 103#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 104#define UPCALL_EXPIRE 6 /* number of timeouts */ 105 106/* 107 * Define the token bucket filter structures 108 * tbftable -> each vif has one of these for storing info 109 */ 110 111static struct tbf tbftable[MAXVIFS]; 112#define TBF_REPROCESS (hz / 100) /* 100x / second */ 113 114/* 115 * 'Interfaces' associated with decapsulator (so we can tell 116 * packets that went through it from ones that get reflected 117 * by a broken gateway). These interfaces are never linked into 118 * the system ifnet list & no routes point to them. I.e., packets 119 * can't be sent this way. They only exist as a placeholder for 120 * multicast source verification. 121 */ 122static struct ifnet multicast_decap_if[MAXVIFS]; 123 124#define ENCAP_TTL 64 125#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 126 127/* prototype IP hdr for encapsulated packets */ 128static struct ip multicast_encap_iphdr = { 129#if BYTE_ORDER == LITTLE_ENDIAN 130 sizeof(struct ip) >> 2, IPVERSION, 131#else 132 IPVERSION, sizeof(struct ip) >> 2, 133#endif 134 0, /* tos */ 135 sizeof(struct ip), /* total length */ 136 0, /* id */ 137 0, /* frag offset */ 138 ENCAP_TTL, ENCAP_PROTO, 139 0, /* checksum */ 140}; 141 142/* 143 * Bandwidth meter variables and constants 144 */ 145static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 146/* 147 * Pending timeouts are stored in a hash table, the key being the 148 * expiration time. Periodically, the entries are analysed and processed. 149 */ 150#define BW_METER_BUCKETS 1024 151static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 152static struct callout_handle bw_meter_ch; 153#define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 154 155/* 156 * Pending upcalls are stored in a vector which is flushed when 157 * full, or periodically 158 */ 159static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 160static u_int bw_upcalls_n; /* # of pending upcalls */ 161static struct callout_handle bw_upcalls_ch; 162#define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 163 164#ifdef PIM 165static struct pimstat pimstat; 166SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 167 &pimstat, pimstat, 168 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 169 170/* 171 * Note: the PIM Register encapsulation adds the following in front of a 172 * data packet: 173 * 174 * struct pim_encap_hdr { 175 * struct ip ip; 176 * struct pim_encap_pimhdr pim; 177 * } 178 * 179 */ 180 181struct pim_encap_pimhdr { 182 struct pim pim; 183 uint32_t flags; 184}; 185 186static struct ip pim_encap_iphdr = { 187#if BYTE_ORDER == LITTLE_ENDIAN 188 sizeof(struct ip) >> 2, 189 IPVERSION, 190#else 191 IPVERSION, 192 sizeof(struct ip) >> 2, 193#endif 194 0, /* tos */ 195 sizeof(struct ip), /* total length */ 196 0, /* id */ 197 0, /* frag offset */ 198 ENCAP_TTL, 199 IPPROTO_PIM, 200 0, /* checksum */ 201}; 202 203static struct pim_encap_pimhdr pim_encap_pimhdr = { 204 { 205 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 206 0, /* reserved */ 207 0, /* checksum */ 208 }, 209 0 /* flags */ 210}; 211 212static struct ifnet multicast_register_if; 213static vifi_t reg_vif_num = VIFI_INVALID; 214#endif /* PIM */ 215 216/* 217 * Private variables. 218 */ 219static vifi_t numvifs; 220static const struct encaptab *encap_cookie; 221 222/* 223 * one-back cache used by mroute_encapcheck to locate a tunnel's vif 224 * given a datagram's src ip address. 225 */ 226static u_long last_encap_src; 227static struct vif *last_encap_vif; 228 229static u_long X_ip_mcast_src(int vifi); 230static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 231 struct mbuf *m, struct ip_moptions *imo); 232static int X_ip_mrouter_done(void); 233static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 234static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 235static int X_legal_vif_num(int vif); 236static int X_mrt_ioctl(int cmd, caddr_t data); 237 238static int get_sg_cnt(struct sioc_sg_req *); 239static int get_vif_cnt(struct sioc_vif_req *); 240static int ip_mrouter_init(struct socket *, int); 241static int add_vif(struct vifctl *); 242static int del_vif(vifi_t); 243static int add_mfc(struct mfcctl2 *); 244static int del_mfc(struct mfcctl2 *); 245static int set_api_config(uint32_t *); /* chose API capabilities */ 246static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 247static int set_assert(int); 248static void expire_upcalls(void *); 249static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 250static void phyint_send(struct ip *, struct vif *, struct mbuf *); 251static void encap_send(struct ip *, struct vif *, struct mbuf *); 252static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 253static void tbf_queue(struct vif *, struct mbuf *); 254static void tbf_process_q(struct vif *); 255static void tbf_reprocess_q(void *); 256static int tbf_dq_sel(struct vif *, struct ip *); 257static void tbf_send_packet(struct vif *, struct mbuf *); 258static void tbf_update_tokens(struct vif *); 259static int priority(struct vif *, struct ip *); 260 261/* 262 * Bandwidth monitoring 263 */ 264static void free_bw_list(struct bw_meter *list); 265static int add_bw_upcall(struct bw_upcall *); 266static int del_bw_upcall(struct bw_upcall *); 267static void bw_meter_receive_packet(struct bw_meter *x, int plen, 268 struct timeval *nowp); 269static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 270static void bw_upcalls_send(void); 271static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 272static void unschedule_bw_meter(struct bw_meter *x); 273static void bw_meter_process(void); 274static void expire_bw_upcalls_send(void *); 275static void expire_bw_meter_process(void *); 276 277#ifdef PIM 278static int pim_register_send(struct ip *, struct vif *, 279 struct mbuf *, struct mfc *); 280static int pim_register_send_rp(struct ip *, struct vif *, 281 struct mbuf *, struct mfc *); 282static int pim_register_send_upcall(struct ip *, struct vif *, 283 struct mbuf *, struct mfc *); 284static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 285#endif 286 287/* 288 * whether or not special PIM assert processing is enabled. 289 */ 290static int pim_assert; 291/* 292 * Rate limit for assert notification messages, in usec 293 */ 294#define ASSERT_MSG_TIME 3000000 295 296/* 297 * Kernel multicast routing API capabilities and setup. 298 * If more API capabilities are added to the kernel, they should be 299 * recorded in `mrt_api_support'. 300 */ 301static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 302 MRT_MFC_FLAGS_BORDER_VIF | 303 MRT_MFC_RP | 304 MRT_MFC_BW_UPCALL); 305static uint32_t mrt_api_config = 0; 306 307/* 308 * Hash function for a source, group entry 309 */ 310#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 311 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 312 313/* 314 * Find a route for a given origin IP address and Multicast group address 315 * Type of service parameter to be added in the future!!! 316 * Statistics are updated by the caller if needed 317 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 318 */ 319static struct mfc * 320mfc_find(in_addr_t o, in_addr_t g) 321{ 322 struct mfc *rt; 323 324 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 325 if ((rt->mfc_origin.s_addr == o) && 326 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 327 break; 328 return rt; 329} 330 331/* 332 * Macros to compute elapsed time efficiently 333 * Borrowed from Van Jacobson's scheduling code 334 */ 335#define TV_DELTA(a, b, delta) { \ 336 int xxs; \ 337 delta = (a).tv_usec - (b).tv_usec; \ 338 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 339 switch (xxs) { \ 340 case 2: \ 341 delta += 1000000; \ 342 /* FALLTHROUGH */ \ 343 case 1: \ 344 delta += 1000000; \ 345 break; \ 346 default: \ 347 delta += (1000000 * xxs); \ 348 } \ 349 } \ 350} 351 352#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 353 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 354 355/* 356 * Handle MRT setsockopt commands to modify the multicast routing tables. 357 */ 358static int 359X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 360{ 361 int error, optval; 362 vifi_t vifi; 363 struct vifctl vifc; 364 struct mfcctl2 mfc; 365 struct bw_upcall bw_upcall; 366 uint32_t i; 367 368 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 369 return EPERM; 370 371 error = 0; 372 switch (sopt->sopt_name) { 373 case MRT_INIT: 374 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 375 if (error) 376 break; 377 error = ip_mrouter_init(so, optval); 378 break; 379 380 case MRT_DONE: 381 error = ip_mrouter_done(); 382 break; 383 384 case MRT_ADD_VIF: 385 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 386 if (error) 387 break; 388 error = add_vif(&vifc); 389 break; 390 391 case MRT_DEL_VIF: 392 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 393 if (error) 394 break; 395 error = del_vif(vifi); 396 break; 397 398 case MRT_ADD_MFC: 399 case MRT_DEL_MFC: 400 /* 401 * select data size depending on API version. 402 */ 403 if (sopt->sopt_name == MRT_ADD_MFC && 404 mrt_api_config & MRT_API_FLAGS_ALL) { 405 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 406 sizeof(struct mfcctl2)); 407 } else { 408 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 409 sizeof(struct mfcctl)); 410 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 411 sizeof(mfc) - sizeof(struct mfcctl)); 412 } 413 if (error) 414 break; 415 if (sopt->sopt_name == MRT_ADD_MFC) 416 error = add_mfc(&mfc); 417 else 418 error = del_mfc(&mfc); 419 break; 420 421 case MRT_ASSERT: 422 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 423 if (error) 424 break; 425 set_assert(optval); 426 break; 427 428 case MRT_API_CONFIG: 429 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 430 if (!error) 431 error = set_api_config(&i); 432 if (!error) 433 error = sooptcopyout(sopt, &i, sizeof i); 434 break; 435 436 case MRT_ADD_BW_UPCALL: 437 case MRT_DEL_BW_UPCALL: 438 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 439 sizeof bw_upcall); 440 if (error) 441 break; 442 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 443 error = add_bw_upcall(&bw_upcall); 444 else 445 error = del_bw_upcall(&bw_upcall); 446 break; 447 448 default: 449 error = EOPNOTSUPP; 450 break; 451 } 452 return error; 453} 454 455/* 456 * Handle MRT getsockopt commands 457 */ 458static int 459X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 460{ 461 int error; 462 static int version = 0x0305; /* !!! why is this here? XXX */ 463 464 switch (sopt->sopt_name) { 465 case MRT_VERSION: 466 error = sooptcopyout(sopt, &version, sizeof version); 467 break; 468 469 case MRT_ASSERT: 470 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 471 break; 472 473 case MRT_API_SUPPORT: 474 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 475 break; 476 477 case MRT_API_CONFIG: 478 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 479 break; 480 481 default: 482 error = EOPNOTSUPP; 483 break; 484 } 485 return error; 486} 487 488/* 489 * Handle ioctl commands to obtain information from the cache 490 */ 491static int 492X_mrt_ioctl(int cmd, caddr_t data) 493{ 494 int error = 0; 495 496 switch (cmd) { 497 case (SIOCGETVIFCNT): 498 error = get_vif_cnt((struct sioc_vif_req *)data); 499 break; 500 501 case (SIOCGETSGCNT): 502 error = get_sg_cnt((struct sioc_sg_req *)data); 503 break; 504 505 default: 506 error = EINVAL; 507 break; 508 } 509 return error; 510} 511 512/* 513 * returns the packet, byte, rpf-failure count for the source group provided 514 */ 515static int 516get_sg_cnt(struct sioc_sg_req *req) 517{ 518 int s; 519 struct mfc *rt; 520 521 s = splnet(); 522 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 523 splx(s); 524 if (rt == NULL) { 525 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 526 return EADDRNOTAVAIL; 527 } 528 req->pktcnt = rt->mfc_pkt_cnt; 529 req->bytecnt = rt->mfc_byte_cnt; 530 req->wrong_if = rt->mfc_wrong_if; 531 return 0; 532} 533 534/* 535 * returns the input and output packet and byte counts on the vif provided 536 */ 537static int 538get_vif_cnt(struct sioc_vif_req *req) 539{ 540 vifi_t vifi = req->vifi; 541 542 if (vifi >= numvifs) 543 return EINVAL; 544 545 req->icount = viftable[vifi].v_pkt_in; 546 req->ocount = viftable[vifi].v_pkt_out; 547 req->ibytes = viftable[vifi].v_bytes_in; 548 req->obytes = viftable[vifi].v_bytes_out; 549 550 return 0; 551} 552 553/* 554 * Enable multicast routing 555 */ 556static int 557ip_mrouter_init(struct socket *so, int version) 558{ 559 if (mrtdebug) 560 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 561 so->so_type, so->so_proto->pr_protocol); 562 563 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 564 return EOPNOTSUPP; 565 566 if (version != 1) 567 return ENOPROTOOPT; 568 569 if (ip_mrouter != NULL) 570 return EADDRINUSE; 571 572 ip_mrouter = so; 573 574 bzero((caddr_t)mfctable, sizeof(mfctable)); 575 bzero((caddr_t)nexpire, sizeof(nexpire)); 576 577 pim_assert = 0; 578 579 expire_upcalls_ch = timeout(expire_upcalls, NULL, EXPIRE_TIMEOUT); 580 581 bw_upcalls_n = 0; 582 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 583 bw_upcalls_ch = timeout(expire_bw_upcalls_send, NULL, BW_UPCALLS_PERIOD); 584 bw_meter_ch = timeout(expire_bw_meter_process, NULL, BW_METER_PERIOD); 585 586 mrt_api_config = 0; 587 588 if (mrtdebug) 589 log(LOG_DEBUG, "ip_mrouter_init\n"); 590 591 return 0; 592} 593 594/* 595 * Disable multicast routing 596 */ 597static int 598X_ip_mrouter_done(void) 599{ 600 vifi_t vifi; 601 int i; 602 struct ifnet *ifp; 603 struct ifreq ifr; 604 struct mfc *rt; 605 struct rtdetq *rte; 606 int s; 607 608 s = splnet(); 609 610 /* 611 * For each phyint in use, disable promiscuous reception of all IP 612 * multicasts. 613 */ 614 for (vifi = 0; vifi < numvifs; vifi++) { 615 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 616 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 617 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 618 619 so->sin_len = sizeof(struct sockaddr_in); 620 so->sin_family = AF_INET; 621 so->sin_addr.s_addr = INADDR_ANY; 622 ifp = viftable[vifi].v_ifp; 623 if_allmulti(ifp, 0); 624 } 625 } 626 bzero((caddr_t)tbftable, sizeof(tbftable)); 627 bzero((caddr_t)viftable, sizeof(viftable)); 628 numvifs = 0; 629 pim_assert = 0; 630 631 untimeout(expire_upcalls, NULL, expire_upcalls_ch); 632 633 mrt_api_config = 0; 634 bw_upcalls_n = 0; 635 untimeout(expire_bw_upcalls_send, NULL, bw_upcalls_ch); 636 untimeout(expire_bw_meter_process, NULL, bw_meter_ch); 637 638 /* 639 * Free all multicast forwarding cache entries. 640 */ 641 for (i = 0; i < MFCTBLSIZ; i++) { 642 for (rt = mfctable[i]; rt != NULL; ) { 643 struct mfc *nr = rt->mfc_next; 644 645 for (rte = rt->mfc_stall; rte != NULL; ) { 646 struct rtdetq *n = rte->next; 647 648 m_freem(rte->m); 649 free(rte, M_MRTABLE); 650 rte = n; 651 } 652 free_bw_list(rt->mfc_bw_meter); 653 free(rt, M_MRTABLE); 654 rt = nr; 655 } 656 } 657 658 bzero((caddr_t)mfctable, sizeof(mfctable)); 659 660 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 661 662 /* 663 * Reset de-encapsulation cache 664 */ 665 last_encap_src = INADDR_ANY; 666 last_encap_vif = NULL; 667#ifdef PIM 668 reg_vif_num = VIFI_INVALID; 669#endif 670 671 if (encap_cookie) { 672 encap_detach(encap_cookie); 673 encap_cookie = NULL; 674 } 675 676 ip_mrouter = NULL; 677 678 splx(s); 679 680 if (mrtdebug) 681 log(LOG_DEBUG, "ip_mrouter_done\n"); 682 683 return 0; 684} 685 686/* 687 * Set PIM assert processing global 688 */ 689static int 690set_assert(int i) 691{ 692 if ((i != 1) && (i != 0)) 693 return EINVAL; 694 695 pim_assert = i; 696 697 return 0; 698} 699 700/* 701 * Configure API capabilities 702 */ 703int 704set_api_config(uint32_t *apival) 705{ 706 int i; 707 708 /* 709 * We can set the API capabilities only if it is the first operation 710 * after MRT_INIT. I.e.: 711 * - there are no vifs installed 712 * - pim_assert is not enabled 713 * - the MFC table is empty 714 */ 715 if (numvifs > 0) { 716 *apival = 0; 717 return EPERM; 718 } 719 if (pim_assert) { 720 *apival = 0; 721 return EPERM; 722 } 723 for (i = 0; i < MFCTBLSIZ; i++) { 724 if (mfctable[i] != NULL) { 725 *apival = 0; 726 return EPERM; 727 } 728 } 729 730 mrt_api_config = *apival & mrt_api_support; 731 *apival = mrt_api_config; 732 733 return 0; 734} 735 736/* 737 * Decide if a packet is from a tunnelled peer. 738 * Return 0 if not, 64 if so. XXX yuck.. 64 ??? 739 */ 740static int 741mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 742{ 743 struct ip *ip = mtod(m, struct ip *); 744 int hlen = ip->ip_hl << 2; 745 746 /* 747 * don't claim the packet if it's not to a multicast destination or if 748 * we don't have an encapsulating tunnel with the source. 749 * Note: This code assumes that the remote site IP address 750 * uniquely identifies the tunnel (i.e., that this site has 751 * at most one tunnel with the remote site). 752 */ 753 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) 754 return 0; 755 if (ip->ip_src.s_addr != last_encap_src) { 756 struct vif *vifp = viftable; 757 struct vif *vife = vifp + numvifs; 758 759 last_encap_src = ip->ip_src.s_addr; 760 last_encap_vif = NULL; 761 for ( ; vifp < vife; ++vifp) 762 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 763 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL) 764 last_encap_vif = vifp; 765 break; 766 } 767 } 768 if (last_encap_vif == NULL) { 769 last_encap_src = INADDR_ANY; 770 return 0; 771 } 772 return 64; 773} 774 775/* 776 * De-encapsulate a packet and feed it back through ip input (this 777 * routine is called whenever IP gets a packet that mroute_encap_func() 778 * claimed). 779 */ 780static void 781mroute_encap_input(struct mbuf *m, int off) 782{ 783 struct ip *ip = mtod(m, struct ip *); 784 int hlen = ip->ip_hl << 2; 785 786 if (hlen > sizeof(struct ip)) 787 ip_stripoptions(m, (struct mbuf *) 0); 788 m->m_data += sizeof(struct ip); 789 m->m_len -= sizeof(struct ip); 790 m->m_pkthdr.len -= sizeof(struct ip); 791 792 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 793 794 netisr_queue(NETISR_IP, m); 795 /* 796 * normally we would need a "schednetisr(NETISR_IP)" 797 * here but we were called by ip_input and it is going 798 * to loop back & try to dequeue the packet we just 799 * queued as soon as we return so we avoid the 800 * unnecessary software interrrupt. 801 * 802 * XXX 803 * This no longer holds - we may have direct-dispatched the packet, 804 * or there may be a queue processing limit. 805 */ 806} 807 808extern struct domain inetdomain; 809static struct protosw mroute_encap_protosw = 810{ SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 811 mroute_encap_input, 0, 0, rip_ctloutput, 812 0, 813 0, 0, 0, 0, 814 &rip_usrreqs 815}; 816 817/* 818 * Add a vif to the vif table 819 */ 820static int 821add_vif(struct vifctl *vifcp) 822{ 823 struct vif *vifp = viftable + vifcp->vifc_vifi; 824 struct sockaddr_in sin = {sizeof sin, AF_INET}; 825 struct ifaddr *ifa; 826 struct ifnet *ifp; 827 int error, s; 828 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 829 830 if (vifcp->vifc_vifi >= MAXVIFS) 831 return EINVAL; 832 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) 833 return EADDRINUSE; 834 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) 835 return EADDRNOTAVAIL; 836 837 /* Find the interface with an address in AF_INET family */ 838#ifdef PIM 839 if (vifcp->vifc_flags & VIFF_REGISTER) { 840 /* 841 * XXX: Because VIFF_REGISTER does not really need a valid 842 * local interface (e.g. it could be 127.0.0.2), we don't 843 * check its address. 844 */ 845 ifp = NULL; 846 } else 847#endif 848 { 849 sin.sin_addr = vifcp->vifc_lcl_addr; 850 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 851 if (ifa == NULL) 852 return EADDRNOTAVAIL; 853 ifp = ifa->ifa_ifp; 854 } 855 856 if (vifcp->vifc_flags & VIFF_TUNNEL) { 857 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 858 /* 859 * An encapsulating tunnel is wanted. Tell 860 * mroute_encap_input() to start paying attention 861 * to encapsulated packets. 862 */ 863 if (encap_cookie == NULL) { 864 encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 865 mroute_encapcheck, 866 (struct protosw *)&mroute_encap_protosw, NULL); 867 868 if (encap_cookie == NULL) { 869 printf("ip_mroute: unable to attach encap\n"); 870 return EIO; /* XXX */ 871 } 872 for (s = 0; s < MAXVIFS; ++s) { 873 multicast_decap_if[s].if_name = "mdecap"; 874 multicast_decap_if[s].if_unit = s; 875 } 876 } 877 /* 878 * Set interface to fake encapsulator interface 879 */ 880 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 881 /* 882 * Prepare cached route entry 883 */ 884 bzero(&vifp->v_route, sizeof(vifp->v_route)); 885 } else { 886 log(LOG_ERR, "source routed tunnels not supported\n"); 887 return EOPNOTSUPP; 888 } 889#ifdef PIM 890 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 891 ifp = &multicast_register_if; 892 if (mrtdebug) 893 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 894 (void *)&multicast_register_if); 895 if (reg_vif_num == VIFI_INVALID) { 896 multicast_register_if.if_name = "register_vif"; 897 multicast_register_if.if_unit = 0; 898 multicast_register_if.if_flags = IFF_LOOPBACK; 899 bzero(&vifp->v_route, sizeof(vifp->v_route)); 900 reg_vif_num = vifcp->vifc_vifi; 901 } 902#endif 903 } else { /* Make sure the interface supports multicast */ 904 if ((ifp->if_flags & IFF_MULTICAST) == 0) 905 return EOPNOTSUPP; 906 907 /* Enable promiscuous reception of all IP multicasts from the if */ 908 s = splnet(); 909 error = if_allmulti(ifp, 1); 910 splx(s); 911 if (error) 912 return error; 913 } 914 915 s = splnet(); 916 /* define parameters for the tbf structure */ 917 vifp->v_tbf = v_tbf; 918 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 919 vifp->v_tbf->tbf_n_tok = 0; 920 vifp->v_tbf->tbf_q_len = 0; 921 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 922 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 923 924 vifp->v_flags = vifcp->vifc_flags; 925 vifp->v_threshold = vifcp->vifc_threshold; 926 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 927 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 928 vifp->v_ifp = ifp; 929 /* scaling up here allows division by 1024 in critical code */ 930 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 931 vifp->v_rsvp_on = 0; 932 vifp->v_rsvpd = NULL; 933 /* initialize per vif pkt counters */ 934 vifp->v_pkt_in = 0; 935 vifp->v_pkt_out = 0; 936 vifp->v_bytes_in = 0; 937 vifp->v_bytes_out = 0; 938 splx(s); 939 940 /* Adjust numvifs up if the vifi is higher than numvifs */ 941 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 942 943 if (mrtdebug) 944 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 945 vifcp->vifc_vifi, 946 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 947 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 948 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 949 vifcp->vifc_threshold, 950 vifcp->vifc_rate_limit); 951 952 return 0; 953} 954 955/* 956 * Delete a vif from the vif table 957 */ 958static int 959del_vif(vifi_t vifi) 960{ 961 struct vif *vifp; 962 int s; 963 964 if (vifi >= numvifs) 965 return EINVAL; 966 vifp = &viftable[vifi]; 967 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) 968 return EADDRNOTAVAIL; 969 970 s = splnet(); 971 972 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 973 if_allmulti(vifp->v_ifp, 0); 974 975 if (vifp == last_encap_vif) { 976 last_encap_vif = NULL; 977 last_encap_src = INADDR_ANY; 978 } 979 980 /* 981 * Free packets queued at the interface 982 */ 983 while (vifp->v_tbf->tbf_q) { 984 struct mbuf *m = vifp->v_tbf->tbf_q; 985 986 vifp->v_tbf->tbf_q = m->m_act; 987 m_freem(m); 988 } 989 990#ifdef PIM 991 if (vifp->v_flags & VIFF_REGISTER) 992 reg_vif_num = VIFI_INVALID; 993#endif 994 995 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 996 bzero((caddr_t)vifp, sizeof (*vifp)); 997 998 if (mrtdebug) 999 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1000 1001 /* Adjust numvifs down */ 1002 for (vifi = numvifs; vifi > 0; vifi--) 1003 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1004 break; 1005 numvifs = vifi; 1006 1007 splx(s); 1008 1009 return 0; 1010} 1011 1012/* 1013 * update an mfc entry without resetting counters and S,G addresses. 1014 */ 1015static void 1016update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1017{ 1018 int i; 1019 1020 rt->mfc_parent = mfccp->mfcc_parent; 1021 for (i = 0; i < numvifs; i++) { 1022 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1023 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1024 MRT_MFC_FLAGS_ALL; 1025 } 1026 /* set the RP address */ 1027 if (mrt_api_config & MRT_MFC_RP) 1028 rt->mfc_rp = mfccp->mfcc_rp; 1029 else 1030 rt->mfc_rp.s_addr = INADDR_ANY; 1031} 1032 1033/* 1034 * fully initialize an mfc entry from the parameter. 1035 */ 1036static void 1037init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1038{ 1039 rt->mfc_origin = mfccp->mfcc_origin; 1040 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1041 1042 update_mfc_params(rt, mfccp); 1043 1044 /* initialize pkt counters per src-grp */ 1045 rt->mfc_pkt_cnt = 0; 1046 rt->mfc_byte_cnt = 0; 1047 rt->mfc_wrong_if = 0; 1048 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1049} 1050 1051 1052/* 1053 * Add an mfc entry 1054 */ 1055static int 1056add_mfc(struct mfcctl2 *mfccp) 1057{ 1058 struct mfc *rt; 1059 u_long hash; 1060 struct rtdetq *rte; 1061 u_short nstl; 1062 int s; 1063 1064 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1065 1066 /* If an entry already exists, just update the fields */ 1067 if (rt) { 1068 if (mrtdebug & DEBUG_MFC) 1069 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1070 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1071 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1072 mfccp->mfcc_parent); 1073 1074 s = splnet(); 1075 update_mfc_params(rt, mfccp); 1076 splx(s); 1077 return 0; 1078 } 1079 1080 /* 1081 * Find the entry for which the upcall was made and update 1082 */ 1083 s = splnet(); 1084 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1085 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1086 1087 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1088 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1089 (rt->mfc_stall != NULL)) { 1090 1091 if (nstl++) 1092 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1093 "multiple kernel entries", 1094 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1095 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1096 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1097 1098 if (mrtdebug & DEBUG_MFC) 1099 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1100 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1101 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1102 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1103 1104 init_mfc_params(rt, mfccp); 1105 1106 rt->mfc_expire = 0; /* Don't clean this guy up */ 1107 nexpire[hash]--; 1108 1109 /* free packets Qed at the end of this entry */ 1110 for (rte = rt->mfc_stall; rte != NULL; ) { 1111 struct rtdetq *n = rte->next; 1112 1113 ip_mdq(rte->m, rte->ifp, rt, -1); 1114 m_freem(rte->m); 1115 free(rte, M_MRTABLE); 1116 rte = n; 1117 } 1118 rt->mfc_stall = NULL; 1119 } 1120 } 1121 1122 /* 1123 * It is possible that an entry is being inserted without an upcall 1124 */ 1125 if (nstl == 0) { 1126 if (mrtdebug & DEBUG_MFC) 1127 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1128 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1129 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1130 mfccp->mfcc_parent); 1131 1132 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1133 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1134 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1135 init_mfc_params(rt, mfccp); 1136 if (rt->mfc_expire) 1137 nexpire[hash]--; 1138 rt->mfc_expire = 0; 1139 break; /* XXX */ 1140 } 1141 } 1142 if (rt == NULL) { /* no upcall, so make a new entry */ 1143 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1144 if (rt == NULL) { 1145 splx(s); 1146 return ENOBUFS; 1147 } 1148 1149 init_mfc_params(rt, mfccp); 1150 rt->mfc_expire = 0; 1151 rt->mfc_stall = NULL; 1152 1153 rt->mfc_bw_meter = NULL; 1154 /* insert new entry at head of hash chain */ 1155 rt->mfc_next = mfctable[hash]; 1156 mfctable[hash] = rt; 1157 } 1158 } 1159 splx(s); 1160 return 0; 1161} 1162 1163/* 1164 * Delete an mfc entry 1165 */ 1166static int 1167del_mfc(struct mfcctl2 *mfccp) 1168{ 1169 struct in_addr origin; 1170 struct in_addr mcastgrp; 1171 struct mfc *rt; 1172 struct mfc **nptr; 1173 u_long hash; 1174 int s; 1175 struct bw_meter *list; 1176 1177 origin = mfccp->mfcc_origin; 1178 mcastgrp = mfccp->mfcc_mcastgrp; 1179 1180 if (mrtdebug & DEBUG_MFC) 1181 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1182 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1183 1184 s = splnet(); 1185 1186 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1187 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1188 if (origin.s_addr == rt->mfc_origin.s_addr && 1189 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1190 rt->mfc_stall == NULL) 1191 break; 1192 if (rt == NULL) { 1193 splx(s); 1194 return EADDRNOTAVAIL; 1195 } 1196 1197 *nptr = rt->mfc_next; 1198 1199 /* 1200 * free the bw_meter entries 1201 */ 1202 list = rt->mfc_bw_meter; 1203 rt->mfc_bw_meter = NULL; 1204 1205 free(rt, M_MRTABLE); 1206 1207 splx(s); 1208 1209 free_bw_list(list); 1210 1211 return 0; 1212} 1213 1214/* 1215 * Send a message to mrouted on the multicast routing socket 1216 */ 1217static int 1218socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1219{ 1220 if (s) { 1221 if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) { 1222 sorwakeup(s); 1223 return 0; 1224 } 1225 } 1226 m_freem(mm); 1227 return -1; 1228} 1229 1230/* 1231 * IP multicast forwarding function. This function assumes that the packet 1232 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1233 * pointed to by "ifp", and the packet is to be relayed to other networks 1234 * that have members of the packet's destination IP multicast group. 1235 * 1236 * The packet is returned unscathed to the caller, unless it is 1237 * erroneous, in which case a non-zero return value tells the caller to 1238 * discard it. 1239 */ 1240 1241#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1242 1243static int 1244X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1245 struct ip_moptions *imo) 1246{ 1247 struct mfc *rt; 1248 int s; 1249 vifi_t vifi; 1250 1251 if (mrtdebug & DEBUG_FORWARD) 1252 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1253 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1254 (void *)ifp); 1255 1256 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1257 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1258 /* 1259 * Packet arrived via a physical interface or 1260 * an encapsulated tunnel or a register_vif. 1261 */ 1262 } else { 1263 /* 1264 * Packet arrived through a source-route tunnel. 1265 * Source-route tunnels are no longer supported. 1266 */ 1267 static int last_log; 1268 if (last_log != time_second) { 1269 last_log = time_second; 1270 log(LOG_ERR, 1271 "ip_mforward: received source-routed packet from %lx\n", 1272 (u_long)ntohl(ip->ip_src.s_addr)); 1273 } 1274 return 1; 1275 } 1276 1277 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1278 if (ip->ip_ttl < 255) 1279 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1280 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1281 struct vif *vifp = viftable + vifi; 1282 1283 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1284 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1285 vifi, 1286 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1287 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1288 } 1289 return ip_mdq(m, ifp, NULL, vifi); 1290 } 1291 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1292 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1293 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1294 if (!imo) 1295 printf("In fact, no options were specified at all\n"); 1296 } 1297 1298 /* 1299 * Don't forward a packet with time-to-live of zero or one, 1300 * or a packet destined to a local-only group. 1301 */ 1302 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1303 return 0; 1304 1305 /* 1306 * Determine forwarding vifs from the forwarding cache table 1307 */ 1308 s = splnet(); 1309 ++mrtstat.mrts_mfc_lookups; 1310 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1311 1312 /* Entry exists, so forward if necessary */ 1313 if (rt != NULL) { 1314 splx(s); 1315 return ip_mdq(m, ifp, rt, -1); 1316 } else { 1317 /* 1318 * If we don't have a route for packet's origin, 1319 * Make a copy of the packet & send message to routing daemon 1320 */ 1321 1322 struct mbuf *mb0; 1323 struct rtdetq *rte; 1324 u_long hash; 1325 int hlen = ip->ip_hl << 2; 1326 1327 ++mrtstat.mrts_mfc_misses; 1328 1329 mrtstat.mrts_no_route++; 1330 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1331 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1332 (u_long)ntohl(ip->ip_src.s_addr), 1333 (u_long)ntohl(ip->ip_dst.s_addr)); 1334 1335 /* 1336 * Allocate mbufs early so that we don't do extra work if we are 1337 * just going to fail anyway. Make sure to pullup the header so 1338 * that other people can't step on it. 1339 */ 1340 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1341 if (rte == NULL) { 1342 splx(s); 1343 return ENOBUFS; 1344 } 1345 mb0 = m_copypacket(m, M_DONTWAIT); 1346 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1347 mb0 = m_pullup(mb0, hlen); 1348 if (mb0 == NULL) { 1349 free(rte, M_MRTABLE); 1350 splx(s); 1351 return ENOBUFS; 1352 } 1353 1354 /* is there an upcall waiting for this flow ? */ 1355 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1356 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1357 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1358 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1359 (rt->mfc_stall != NULL)) 1360 break; 1361 } 1362 1363 if (rt == NULL) { 1364 int i; 1365 struct igmpmsg *im; 1366 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1367 struct mbuf *mm; 1368 1369 /* 1370 * Locate the vifi for the incoming interface for this packet. 1371 * If none found, drop packet. 1372 */ 1373 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1374 ; 1375 if (vifi >= numvifs) /* vif not found, drop packet */ 1376 goto non_fatal; 1377 1378 /* no upcall, so make a new entry */ 1379 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1380 if (rt == NULL) 1381 goto fail; 1382 /* Make a copy of the header to send to the user level process */ 1383 mm = m_copy(mb0, 0, hlen); 1384 if (mm == NULL) 1385 goto fail1; 1386 1387 /* 1388 * Send message to routing daemon to install 1389 * a route into the kernel table 1390 */ 1391 1392 im = mtod(mm, struct igmpmsg *); 1393 im->im_msgtype = IGMPMSG_NOCACHE; 1394 im->im_mbz = 0; 1395 im->im_vif = vifi; 1396 1397 mrtstat.mrts_upcalls++; 1398 1399 k_igmpsrc.sin_addr = ip->ip_src; 1400 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1401 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1402 ++mrtstat.mrts_upq_sockfull; 1403fail1: 1404 free(rt, M_MRTABLE); 1405fail: 1406 free(rte, M_MRTABLE); 1407 m_freem(mb0); 1408 splx(s); 1409 return ENOBUFS; 1410 } 1411 1412 /* insert new entry at head of hash chain */ 1413 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1414 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1415 rt->mfc_expire = UPCALL_EXPIRE; 1416 nexpire[hash]++; 1417 for (i = 0; i < numvifs; i++) { 1418 rt->mfc_ttls[i] = 0; 1419 rt->mfc_flags[i] = 0; 1420 } 1421 rt->mfc_parent = -1; 1422 1423 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1424 1425 rt->mfc_bw_meter = NULL; 1426 1427 /* link into table */ 1428 rt->mfc_next = mfctable[hash]; 1429 mfctable[hash] = rt; 1430 rt->mfc_stall = rte; 1431 1432 } else { 1433 /* determine if q has overflowed */ 1434 int npkts = 0; 1435 struct rtdetq **p; 1436 1437 /* 1438 * XXX ouch! we need to append to the list, but we 1439 * only have a pointer to the front, so we have to 1440 * scan the entire list every time. 1441 */ 1442 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1443 npkts++; 1444 1445 if (npkts > MAX_UPQ) { 1446 mrtstat.mrts_upq_ovflw++; 1447non_fatal: 1448 free(rte, M_MRTABLE); 1449 m_freem(mb0); 1450 splx(s); 1451 return 0; 1452 } 1453 1454 /* Add this entry to the end of the queue */ 1455 *p = rte; 1456 } 1457 1458 rte->m = mb0; 1459 rte->ifp = ifp; 1460 rte->next = NULL; 1461 1462 splx(s); 1463 1464 return 0; 1465 } 1466} 1467 1468/* 1469 * Clean up the cache entry if upcall is not serviced 1470 */ 1471static void 1472expire_upcalls(void *unused) 1473{ 1474 struct rtdetq *rte; 1475 struct mfc *mfc, **nptr; 1476 int i; 1477 int s; 1478 1479 s = splnet(); 1480 for (i = 0; i < MFCTBLSIZ; i++) { 1481 if (nexpire[i] == 0) 1482 continue; 1483 nptr = &mfctable[i]; 1484 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1485 /* 1486 * Skip real cache entries 1487 * Make sure it wasn't marked to not expire (shouldn't happen) 1488 * If it expires now 1489 */ 1490 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1491 --mfc->mfc_expire == 0) { 1492 if (mrtdebug & DEBUG_EXPIRE) 1493 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1494 (u_long)ntohl(mfc->mfc_origin.s_addr), 1495 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1496 /* 1497 * drop all the packets 1498 * free the mbuf with the pkt, if, timing info 1499 */ 1500 for (rte = mfc->mfc_stall; rte; ) { 1501 struct rtdetq *n = rte->next; 1502 1503 m_freem(rte->m); 1504 free(rte, M_MRTABLE); 1505 rte = n; 1506 } 1507 ++mrtstat.mrts_cache_cleanups; 1508 nexpire[i]--; 1509 1510 /* 1511 * free the bw_meter entries 1512 */ 1513 while (mfc->mfc_bw_meter != NULL) { 1514 struct bw_meter *x = mfc->mfc_bw_meter; 1515 1516 mfc->mfc_bw_meter = x->bm_mfc_next; 1517 free(x, M_BWMETER); 1518 } 1519 1520 *nptr = mfc->mfc_next; 1521 free(mfc, M_MRTABLE); 1522 } else { 1523 nptr = &mfc->mfc_next; 1524 } 1525 } 1526 } 1527 splx(s); 1528 expire_upcalls_ch = timeout(expire_upcalls, NULL, EXPIRE_TIMEOUT); 1529} 1530 1531/* 1532 * Packet forwarding routine once entry in the cache is made 1533 */ 1534static int 1535ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1536{ 1537 struct ip *ip = mtod(m, struct ip *); 1538 vifi_t vifi; 1539 int plen = ip->ip_len; 1540 1541/* 1542 * Macro to send packet on vif. Since RSVP packets don't get counted on 1543 * input, they shouldn't get counted on output, so statistics keeping is 1544 * separate. 1545 */ 1546#define MC_SEND(ip,vifp,m) { \ 1547 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1548 encap_send((ip), (vifp), (m)); \ 1549 else \ 1550 phyint_send((ip), (vifp), (m)); \ 1551} 1552 1553 /* 1554 * If xmt_vif is not -1, send on only the requested vif. 1555 * 1556 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1557 */ 1558 if (xmt_vif < numvifs) { 1559#ifdef PIM 1560 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1561 pim_register_send(ip, viftable + xmt_vif, m, rt); 1562 else 1563#endif 1564 MC_SEND(ip, viftable + xmt_vif, m); 1565 return 1; 1566 } 1567 1568 /* 1569 * Don't forward if it didn't arrive from the parent vif for its origin. 1570 */ 1571 vifi = rt->mfc_parent; 1572 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1573 /* came in the wrong interface */ 1574 if (mrtdebug & DEBUG_FORWARD) 1575 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1576 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1577 ++mrtstat.mrts_wrong_if; 1578 ++rt->mfc_wrong_if; 1579 /* 1580 * If we are doing PIM assert processing, send a message 1581 * to the routing daemon. 1582 * 1583 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1584 * can complete the SPT switch, regardless of the type 1585 * of the iif (broadcast media, GRE tunnel, etc). 1586 */ 1587 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1588 struct timeval now; 1589 u_long delta; 1590 1591#ifdef PIM 1592 if (ifp == &multicast_register_if) 1593 pimstat.pims_rcv_registers_wrongiif++; 1594#endif 1595 1596 /* Get vifi for the incoming packet */ 1597 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1598 ; 1599 if (vifi >= numvifs) 1600 return 0; /* The iif is not found: ignore the packet. */ 1601 1602 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1603 return 0; /* WRONGVIF disabled: ignore the packet */ 1604 1605 GET_TIME(now); 1606 1607 TV_DELTA(rt->mfc_last_assert, now, delta); 1608 1609 if (delta > ASSERT_MSG_TIME) { 1610 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1611 struct igmpmsg *im; 1612 int hlen = ip->ip_hl << 2; 1613 struct mbuf *mm = m_copy(m, 0, hlen); 1614 1615 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1616 mm = m_pullup(mm, hlen); 1617 if (mm == NULL) 1618 return ENOBUFS; 1619 1620 rt->mfc_last_assert = now; 1621 1622 im = mtod(mm, struct igmpmsg *); 1623 im->im_msgtype = IGMPMSG_WRONGVIF; 1624 im->im_mbz = 0; 1625 im->im_vif = vifi; 1626 1627 mrtstat.mrts_upcalls++; 1628 1629 k_igmpsrc.sin_addr = im->im_src; 1630 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1631 log(LOG_WARNING, 1632 "ip_mforward: ip_mrouter socket queue full\n"); 1633 ++mrtstat.mrts_upq_sockfull; 1634 return ENOBUFS; 1635 } 1636 } 1637 } 1638 return 0; 1639 } 1640 1641 /* If I sourced this packet, it counts as output, else it was input. */ 1642 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1643 viftable[vifi].v_pkt_out++; 1644 viftable[vifi].v_bytes_out += plen; 1645 } else { 1646 viftable[vifi].v_pkt_in++; 1647 viftable[vifi].v_bytes_in += plen; 1648 } 1649 rt->mfc_pkt_cnt++; 1650 rt->mfc_byte_cnt += plen; 1651 1652 /* 1653 * For each vif, decide if a copy of the packet should be forwarded. 1654 * Forward if: 1655 * - the ttl exceeds the vif's threshold 1656 * - there are group members downstream on interface 1657 */ 1658 for (vifi = 0; vifi < numvifs; vifi++) 1659 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1660 viftable[vifi].v_pkt_out++; 1661 viftable[vifi].v_bytes_out += plen; 1662#ifdef PIM 1663 if (viftable[vifi].v_flags & VIFF_REGISTER) 1664 pim_register_send(ip, viftable + vifi, m, rt); 1665 else 1666#endif 1667 MC_SEND(ip, viftable+vifi, m); 1668 } 1669 1670 /* 1671 * Perform upcall-related bw measuring. 1672 */ 1673 if (rt->mfc_bw_meter != NULL) { 1674 struct bw_meter *x; 1675 struct timeval now; 1676 1677 GET_TIME(now); 1678 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1679 bw_meter_receive_packet(x, plen, &now); 1680 } 1681 1682 return 0; 1683} 1684 1685/* 1686 * check if a vif number is legal/ok. This is used by ip_output. 1687 */ 1688static int 1689X_legal_vif_num(int vif) 1690{ 1691 return (vif >= 0 && vif < numvifs); 1692} 1693 1694/* 1695 * Return the local address used by this vif 1696 */ 1697static u_long 1698X_ip_mcast_src(int vifi) 1699{ 1700 if (vifi >= 0 && vifi < numvifs) 1701 return viftable[vifi].v_lcl_addr.s_addr; 1702 else 1703 return INADDR_ANY; 1704} 1705 1706static void 1707phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1708{ 1709 struct mbuf *mb_copy; 1710 int hlen = ip->ip_hl << 2; 1711 1712 /* 1713 * Make a new reference to the packet; make sure that 1714 * the IP header is actually copied, not just referenced, 1715 * so that ip_output() only scribbles on the copy. 1716 */ 1717 mb_copy = m_copypacket(m, M_DONTWAIT); 1718 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1719 mb_copy = m_pullup(mb_copy, hlen); 1720 if (mb_copy == NULL) 1721 return; 1722 1723 if (vifp->v_rate_limit == 0) 1724 tbf_send_packet(vifp, mb_copy); 1725 else 1726 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1727} 1728 1729static void 1730encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1731{ 1732 struct mbuf *mb_copy; 1733 struct ip *ip_copy; 1734 int i, len = ip->ip_len; 1735 1736 /* Take care of delayed checksums */ 1737 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1738 in_delayed_cksum(m); 1739 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1740 } 1741 1742 /* 1743 * copy the old packet & pullup its IP header into the 1744 * new mbuf so we can modify it. Try to fill the new 1745 * mbuf since if we don't the ethernet driver will. 1746 */ 1747 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1748 if (mb_copy == NULL) 1749 return; 1750#ifdef MAC 1751 mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy); 1752#endif 1753 mb_copy->m_data += max_linkhdr; 1754 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1755 1756 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) { 1757 m_freem(mb_copy); 1758 return; 1759 } 1760 i = MHLEN - M_LEADINGSPACE(mb_copy); 1761 if (i > len) 1762 i = len; 1763 mb_copy = m_pullup(mb_copy, i); 1764 if (mb_copy == NULL) 1765 return; 1766 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1767 1768 /* 1769 * fill in the encapsulating IP header. 1770 */ 1771 ip_copy = mtod(mb_copy, struct ip *); 1772 *ip_copy = multicast_encap_iphdr; 1773#ifdef RANDOM_IP_ID 1774 ip_copy->ip_id = ip_randomid(); 1775#else 1776 ip_copy->ip_id = htons(ip_id++); 1777#endif 1778 ip_copy->ip_len += len; 1779 ip_copy->ip_src = vifp->v_lcl_addr; 1780 ip_copy->ip_dst = vifp->v_rmt_addr; 1781 1782 /* 1783 * turn the encapsulated IP header back into a valid one. 1784 */ 1785 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1786 --ip->ip_ttl; 1787 ip->ip_len = htons(ip->ip_len); 1788 ip->ip_off = htons(ip->ip_off); 1789 ip->ip_sum = 0; 1790 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1791 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1792 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1793 1794 if (vifp->v_rate_limit == 0) 1795 tbf_send_packet(vifp, mb_copy); 1796 else 1797 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1798} 1799 1800/* 1801 * Token bucket filter module 1802 */ 1803 1804static void 1805tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len) 1806{ 1807 struct tbf *t = vifp->v_tbf; 1808 1809 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */ 1810 mrtstat.mrts_pkt2large++; 1811 m_freem(m); 1812 return; 1813 } 1814 1815 tbf_update_tokens(vifp); 1816 1817 if (t->tbf_q_len == 0) { /* queue empty... */ 1818 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */ 1819 t->tbf_n_tok -= p_len; 1820 tbf_send_packet(vifp, m); 1821 } else { /* no, queue packet and try later */ 1822 tbf_queue(vifp, m); 1823 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1824 } 1825 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1826 /* finite queue length, so queue pkts and process queue */ 1827 tbf_queue(vifp, m); 1828 tbf_process_q(vifp); 1829 } else { 1830 /* queue full, try to dq and queue and process */ 1831 if (!tbf_dq_sel(vifp, ip)) { 1832 mrtstat.mrts_q_overflow++; 1833 m_freem(m); 1834 } else { 1835 tbf_queue(vifp, m); 1836 tbf_process_q(vifp); 1837 } 1838 } 1839} 1840 1841/* 1842 * adds a packet to the queue at the interface 1843 */ 1844static void 1845tbf_queue(struct vif *vifp, struct mbuf *m) 1846{ 1847 int s = splnet(); 1848 struct tbf *t = vifp->v_tbf; 1849 1850 if (t->tbf_t == NULL) /* Queue was empty */ 1851 t->tbf_q = m; 1852 else /* Insert at tail */ 1853 t->tbf_t->m_act = m; 1854 1855 t->tbf_t = m; /* Set new tail pointer */ 1856 1857#ifdef DIAGNOSTIC 1858 /* Make sure we didn't get fed a bogus mbuf */ 1859 if (m->m_act) 1860 panic("tbf_queue: m_act"); 1861#endif 1862 m->m_act = NULL; 1863 1864 t->tbf_q_len++; 1865 1866 splx(s); 1867} 1868 1869/* 1870 * processes the queue at the interface 1871 */ 1872static void 1873tbf_process_q(struct vif *vifp) 1874{ 1875 int s = splnet(); 1876 struct tbf *t = vifp->v_tbf; 1877 1878 /* loop through the queue at the interface and send as many packets 1879 * as possible 1880 */ 1881 while (t->tbf_q_len > 0) { 1882 struct mbuf *m = t->tbf_q; 1883 int len = mtod(m, struct ip *)->ip_len; 1884 1885 /* determine if the packet can be sent */ 1886 if (len > t->tbf_n_tok) /* not enough tokens, we are done */ 1887 break; 1888 /* ok, reduce no of tokens, dequeue and send the packet. */ 1889 t->tbf_n_tok -= len; 1890 1891 t->tbf_q = m->m_act; 1892 if (--t->tbf_q_len == 0) 1893 t->tbf_t = NULL; 1894 1895 m->m_act = NULL; 1896 tbf_send_packet(vifp, m); 1897 } 1898 splx(s); 1899} 1900 1901static void 1902tbf_reprocess_q(void *xvifp) 1903{ 1904 struct vif *vifp = xvifp; 1905 1906 if (ip_mrouter == NULL) 1907 return; 1908 tbf_update_tokens(vifp); 1909 tbf_process_q(vifp); 1910 if (vifp->v_tbf->tbf_q_len) 1911 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1912} 1913 1914/* function that will selectively discard a member of the queue 1915 * based on the precedence value and the priority 1916 */ 1917static int 1918tbf_dq_sel(struct vif *vifp, struct ip *ip) 1919{ 1920 int s = splnet(); 1921 u_int p; 1922 struct mbuf *m, *last; 1923 struct mbuf **np; 1924 struct tbf *t = vifp->v_tbf; 1925 1926 p = priority(vifp, ip); 1927 1928 np = &t->tbf_q; 1929 last = NULL; 1930 while ((m = *np) != NULL) { 1931 if (p > priority(vifp, mtod(m, struct ip *))) { 1932 *np = m->m_act; 1933 /* If we're removing the last packet, fix the tail pointer */ 1934 if (m == t->tbf_t) 1935 t->tbf_t = last; 1936 m_freem(m); 1937 /* It's impossible for the queue to be empty, but check anyways. */ 1938 if (--t->tbf_q_len == 0) 1939 t->tbf_t = NULL; 1940 splx(s); 1941 mrtstat.mrts_drop_sel++; 1942 return 1; 1943 } 1944 np = &m->m_act; 1945 last = m; 1946 } 1947 splx(s); 1948 return 0; 1949} 1950 1951static void 1952tbf_send_packet(struct vif *vifp, struct mbuf *m) 1953{ 1954 int s = splnet(); 1955 1956 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */ 1957 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 1958 else { 1959 struct ip_moptions imo; 1960 int error; 1961 static struct route ro; /* XXX check this */ 1962 1963 imo.imo_multicast_ifp = vifp->v_ifp; 1964 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1965 imo.imo_multicast_loop = 1; 1966 imo.imo_multicast_vif = -1; 1967 1968 /* 1969 * Re-entrancy should not be a problem here, because 1970 * the packets that we send out and are looped back at us 1971 * should get rejected because they appear to come from 1972 * the loopback interface, thus preventing looping. 1973 */ 1974 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL); 1975 1976 if (mrtdebug & DEBUG_XMIT) 1977 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1978 (int)(vifp - viftable), error); 1979 } 1980 splx(s); 1981} 1982 1983/* determine the current time and then 1984 * the elapsed time (between the last time and time now) 1985 * in milliseconds & update the no. of tokens in the bucket 1986 */ 1987static void 1988tbf_update_tokens(struct vif *vifp) 1989{ 1990 struct timeval tp; 1991 u_long tm; 1992 int s = splnet(); 1993 struct tbf *t = vifp->v_tbf; 1994 1995 GET_TIME(tp); 1996 1997 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1998 1999 /* 2000 * This formula is actually 2001 * "time in seconds" * "bytes/second". 2002 * 2003 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2004 * 2005 * The (1000/1024) was introduced in add_vif to optimize 2006 * this divide into a shift. 2007 */ 2008 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 2009 t->tbf_last_pkt_t = tp; 2010 2011 if (t->tbf_n_tok > MAX_BKT_SIZE) 2012 t->tbf_n_tok = MAX_BKT_SIZE; 2013 2014 splx(s); 2015} 2016 2017static int 2018priority(struct vif *vifp, struct ip *ip) 2019{ 2020 int prio = 50; /* the lowest priority -- default case */ 2021 2022 /* temporary hack; may add general packet classifier some day */ 2023 2024 /* 2025 * The UDP port space is divided up into four priority ranges: 2026 * [0, 16384) : unclassified - lowest priority 2027 * [16384, 32768) : audio - highest priority 2028 * [32768, 49152) : whiteboard - medium priority 2029 * [49152, 65536) : video - low priority 2030 * 2031 * Everything else gets lowest priority. 2032 */ 2033 if (ip->ip_p == IPPROTO_UDP) { 2034 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2035 switch (ntohs(udp->uh_dport) & 0xc000) { 2036 case 0x4000: 2037 prio = 70; 2038 break; 2039 case 0x8000: 2040 prio = 60; 2041 break; 2042 case 0xc000: 2043 prio = 55; 2044 break; 2045 } 2046 } 2047 return prio; 2048} 2049 2050/* 2051 * End of token bucket filter modifications 2052 */ 2053 2054static int 2055X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 2056{ 2057 int error, vifi, s; 2058 2059 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2060 return EOPNOTSUPP; 2061 2062 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 2063 if (error) 2064 return error; 2065 2066 s = splnet(); 2067 2068 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 2069 splx(s); 2070 return EADDRNOTAVAIL; 2071 } 2072 2073 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 2074 /* Check if socket is available. */ 2075 if (viftable[vifi].v_rsvpd != NULL) { 2076 splx(s); 2077 return EADDRINUSE; 2078 } 2079 2080 viftable[vifi].v_rsvpd = so; 2081 /* This may seem silly, but we need to be sure we don't over-increment 2082 * the RSVP counter, in case something slips up. 2083 */ 2084 if (!viftable[vifi].v_rsvp_on) { 2085 viftable[vifi].v_rsvp_on = 1; 2086 rsvp_on++; 2087 } 2088 } else { /* must be VIF_OFF */ 2089 /* 2090 * XXX as an additional consistency check, one could make sure 2091 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 2092 * first parameter is pretty useless. 2093 */ 2094 viftable[vifi].v_rsvpd = NULL; 2095 /* 2096 * This may seem silly, but we need to be sure we don't over-decrement 2097 * the RSVP counter, in case something slips up. 2098 */ 2099 if (viftable[vifi].v_rsvp_on) { 2100 viftable[vifi].v_rsvp_on = 0; 2101 rsvp_on--; 2102 } 2103 } 2104 splx(s); 2105 return 0; 2106} 2107 2108static void 2109X_ip_rsvp_force_done(struct socket *so) 2110{ 2111 int vifi; 2112 int s; 2113 2114 /* Don't bother if it is not the right type of socket. */ 2115 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2116 return; 2117 2118 s = splnet(); 2119 2120 /* The socket may be attached to more than one vif...this 2121 * is perfectly legal. 2122 */ 2123 for (vifi = 0; vifi < numvifs; vifi++) { 2124 if (viftable[vifi].v_rsvpd == so) { 2125 viftable[vifi].v_rsvpd = NULL; 2126 /* This may seem silly, but we need to be sure we don't 2127 * over-decrement the RSVP counter, in case something slips up. 2128 */ 2129 if (viftable[vifi].v_rsvp_on) { 2130 viftable[vifi].v_rsvp_on = 0; 2131 rsvp_on--; 2132 } 2133 } 2134 } 2135 2136 splx(s); 2137} 2138 2139static void 2140X_rsvp_input(struct mbuf *m, int off) 2141{ 2142 int vifi; 2143 struct ip *ip = mtod(m, struct ip *); 2144 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2145 int s; 2146 struct ifnet *ifp; 2147 2148 if (rsvpdebug) 2149 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2150 2151 /* Can still get packets with rsvp_on = 0 if there is a local member 2152 * of the group to which the RSVP packet is addressed. But in this 2153 * case we want to throw the packet away. 2154 */ 2155 if (!rsvp_on) { 2156 m_freem(m); 2157 return; 2158 } 2159 2160 s = splnet(); 2161 2162 if (rsvpdebug) 2163 printf("rsvp_input: check vifs\n"); 2164 2165#ifdef DIAGNOSTIC 2166 M_ASSERTPKTHDR(m); 2167#endif 2168 2169 ifp = m->m_pkthdr.rcvif; 2170 /* Find which vif the packet arrived on. */ 2171 for (vifi = 0; vifi < numvifs; vifi++) 2172 if (viftable[vifi].v_ifp == ifp) 2173 break; 2174 2175 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2176 /* 2177 * If the old-style non-vif-associated socket is set, 2178 * then use it. Otherwise, drop packet since there 2179 * is no specific socket for this vif. 2180 */ 2181 if (ip_rsvpd != NULL) { 2182 if (rsvpdebug) 2183 printf("rsvp_input: Sending packet up old-style socket\n"); 2184 rip_input(m, off); /* xxx */ 2185 } else { 2186 if (rsvpdebug && vifi == numvifs) 2187 printf("rsvp_input: Can't find vif for packet.\n"); 2188 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2189 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2190 m_freem(m); 2191 } 2192 splx(s); 2193 return; 2194 } 2195 rsvp_src.sin_addr = ip->ip_src; 2196 2197 if (rsvpdebug && m) 2198 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2199 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2200 2201 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2202 if (rsvpdebug) 2203 printf("rsvp_input: Failed to append to socket\n"); 2204 } else { 2205 if (rsvpdebug) 2206 printf("rsvp_input: send packet up\n"); 2207 } 2208 2209 splx(s); 2210} 2211 2212/* 2213 * Code for bandwidth monitors 2214 */ 2215 2216/* 2217 * Define common interface for timeval-related methods 2218 */ 2219#define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 2220#define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 2221#define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 2222 2223static uint32_t 2224compute_bw_meter_flags(struct bw_upcall *req) 2225{ 2226 uint32_t flags = 0; 2227 2228 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2229 flags |= BW_METER_UNIT_PACKETS; 2230 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2231 flags |= BW_METER_UNIT_BYTES; 2232 if (req->bu_flags & BW_UPCALL_GEQ) 2233 flags |= BW_METER_GEQ; 2234 if (req->bu_flags & BW_UPCALL_LEQ) 2235 flags |= BW_METER_LEQ; 2236 2237 return flags; 2238} 2239 2240/* 2241 * Add a bw_meter entry 2242 */ 2243static int 2244add_bw_upcall(struct bw_upcall *req) 2245{ 2246 struct mfc *mfc; 2247 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2248 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2249 struct timeval now; 2250 struct bw_meter *x; 2251 uint32_t flags; 2252 int s; 2253 2254 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2255 return EOPNOTSUPP; 2256 2257 /* Test if the flags are valid */ 2258 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2259 return EINVAL; 2260 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2261 return EINVAL; 2262 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2263 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2264 return EINVAL; 2265 2266 /* Test if the threshold time interval is valid */ 2267 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2268 return EINVAL; 2269 2270 flags = compute_bw_meter_flags(req); 2271 2272 /* 2273 * Find if we have already same bw_meter entry 2274 */ 2275 s = splnet(); 2276 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2277 if (mfc == NULL) { 2278 splx(s); 2279 return EADDRNOTAVAIL; 2280 } 2281 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2282 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2283 &req->bu_threshold.b_time, ==)) && 2284 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2285 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2286 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2287 splx(s); 2288 return 0; /* XXX Already installed */ 2289 } 2290 } 2291 splx(s); 2292 2293 /* Allocate the new bw_meter entry */ 2294 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2295 if (x == NULL) 2296 return ENOBUFS; 2297 2298 /* Set the new bw_meter entry */ 2299 x->bm_threshold.b_time = req->bu_threshold.b_time; 2300 GET_TIME(now); 2301 x->bm_start_time = now; 2302 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2303 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2304 x->bm_measured.b_packets = 0; 2305 x->bm_measured.b_bytes = 0; 2306 x->bm_flags = flags; 2307 x->bm_time_next = NULL; 2308 x->bm_time_hash = BW_METER_BUCKETS; 2309 2310 /* Add the new bw_meter entry to the front of entries for this MFC */ 2311 s = splnet(); 2312 x->bm_mfc = mfc; 2313 x->bm_mfc_next = mfc->mfc_bw_meter; 2314 mfc->mfc_bw_meter = x; 2315 schedule_bw_meter(x, &now); 2316 splx(s); 2317 2318 return 0; 2319} 2320 2321static void 2322free_bw_list(struct bw_meter *list) 2323{ 2324 while (list != NULL) { 2325 struct bw_meter *x = list; 2326 2327 list = list->bm_mfc_next; 2328 unschedule_bw_meter(x); 2329 free(x, M_BWMETER); 2330 } 2331} 2332 2333/* 2334 * Delete one or multiple bw_meter entries 2335 */ 2336static int 2337del_bw_upcall(struct bw_upcall *req) 2338{ 2339 struct mfc *mfc; 2340 struct bw_meter *x; 2341 int s; 2342 2343 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2344 return EOPNOTSUPP; 2345 2346 s = splnet(); 2347 /* Find the corresponding MFC entry */ 2348 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2349 if (mfc == NULL) { 2350 splx(s); 2351 return EADDRNOTAVAIL; 2352 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2353 /* 2354 * Delete all bw_meter entries for this mfc 2355 */ 2356 struct bw_meter *list; 2357 2358 list = mfc->mfc_bw_meter; 2359 mfc->mfc_bw_meter = NULL; 2360 splx(s); 2361 free_bw_list(list); 2362 return 0; 2363 } else { /* Delete a single bw_meter entry */ 2364 struct bw_meter *prev; 2365 uint32_t flags = 0; 2366 2367 flags = compute_bw_meter_flags(req); 2368 2369 /* Find the bw_meter entry to delete */ 2370 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2371 x = x->bm_mfc_next) { 2372 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2373 &req->bu_threshold.b_time, ==)) && 2374 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2375 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2376 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2377 break; 2378 } 2379 if (x != NULL) { /* Delete entry from the list for this MFC */ 2380 if (prev != NULL) 2381 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2382 else 2383 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2384 splx(s); 2385 2386 unschedule_bw_meter(x); 2387 /* Free the bw_meter entry */ 2388 free(x, M_BWMETER); 2389 return 0; 2390 } else { 2391 splx(s); 2392 return EINVAL; 2393 } 2394 } 2395 /* NOTREACHED */ 2396} 2397 2398/* 2399 * Perform bandwidth measurement processing that may result in an upcall 2400 */ 2401static void 2402bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2403{ 2404 struct timeval delta; 2405 int s; 2406 2407 s = splnet(); 2408 delta = *nowp; 2409 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2410 2411 if (x->bm_flags & BW_METER_GEQ) { 2412 /* 2413 * Processing for ">=" type of bw_meter entry 2414 */ 2415 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2416 /* Reset the bw_meter entry */ 2417 x->bm_start_time = *nowp; 2418 x->bm_measured.b_packets = 0; 2419 x->bm_measured.b_bytes = 0; 2420 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2421 } 2422 2423 /* Record that a packet is received */ 2424 x->bm_measured.b_packets++; 2425 x->bm_measured.b_bytes += plen; 2426 2427 /* 2428 * Test if we should deliver an upcall 2429 */ 2430 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2431 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2432 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2433 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2434 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2435 /* Prepare an upcall for delivery */ 2436 bw_meter_prepare_upcall(x, nowp); 2437 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2438 } 2439 } 2440 } else if (x->bm_flags & BW_METER_LEQ) { 2441 /* 2442 * Processing for "<=" type of bw_meter entry 2443 */ 2444 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2445 /* 2446 * We are behind time with the multicast forwarding table 2447 * scanning for "<=" type of bw_meter entries, so test now 2448 * if we should deliver an upcall. 2449 */ 2450 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2451 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2452 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2453 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2454 /* Prepare an upcall for delivery */ 2455 bw_meter_prepare_upcall(x, nowp); 2456 } 2457 /* Reschedule the bw_meter entry */ 2458 unschedule_bw_meter(x); 2459 schedule_bw_meter(x, nowp); 2460 } 2461 2462 /* Record that a packet is received */ 2463 x->bm_measured.b_packets++; 2464 x->bm_measured.b_bytes += plen; 2465 2466 /* 2467 * Test if we should restart the measuring interval 2468 */ 2469 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2470 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2471 (x->bm_flags & BW_METER_UNIT_BYTES && 2472 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2473 /* Don't restart the measuring interval */ 2474 } else { 2475 /* Do restart the measuring interval */ 2476 /* 2477 * XXX: note that we don't unschedule and schedule, because this 2478 * might be too much overhead per packet. Instead, when we process 2479 * all entries for a given timer hash bin, we check whether it is 2480 * really a timeout. If not, we reschedule at that time. 2481 */ 2482 x->bm_start_time = *nowp; 2483 x->bm_measured.b_packets = 0; 2484 x->bm_measured.b_bytes = 0; 2485 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2486 } 2487 } 2488 splx(s); 2489} 2490 2491/* 2492 * Prepare a bandwidth-related upcall 2493 */ 2494static void 2495bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2496{ 2497 struct timeval delta; 2498 struct bw_upcall *u; 2499 int s; 2500 2501 s = splnet(); 2502 2503 /* 2504 * Compute the measured time interval 2505 */ 2506 delta = *nowp; 2507 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2508 2509 /* 2510 * If there are too many pending upcalls, deliver them now 2511 */ 2512 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2513 bw_upcalls_send(); 2514 2515 /* 2516 * Set the bw_upcall entry 2517 */ 2518 u = &bw_upcalls[bw_upcalls_n++]; 2519 u->bu_src = x->bm_mfc->mfc_origin; 2520 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2521 u->bu_threshold.b_time = x->bm_threshold.b_time; 2522 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2523 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2524 u->bu_measured.b_time = delta; 2525 u->bu_measured.b_packets = x->bm_measured.b_packets; 2526 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2527 u->bu_flags = 0; 2528 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2529 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2530 if (x->bm_flags & BW_METER_UNIT_BYTES) 2531 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2532 if (x->bm_flags & BW_METER_GEQ) 2533 u->bu_flags |= BW_UPCALL_GEQ; 2534 if (x->bm_flags & BW_METER_LEQ) 2535 u->bu_flags |= BW_UPCALL_LEQ; 2536 2537 splx(s); 2538} 2539 2540/* 2541 * Send the pending bandwidth-related upcalls 2542 */ 2543static void 2544bw_upcalls_send(void) 2545{ 2546 struct mbuf *m; 2547 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2548 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2549 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2550 0, /* unused2 */ 2551 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2552 0, /* im_mbz */ 2553 0, /* im_vif */ 2554 0, /* unused3 */ 2555 { 0 }, /* im_src */ 2556 { 0 } }; /* im_dst */ 2557 2558 if (bw_upcalls_n == 0) 2559 return; /* No pending upcalls */ 2560 2561 bw_upcalls_n = 0; 2562 2563 /* 2564 * Allocate a new mbuf, initialize it with the header and 2565 * the payload for the pending calls. 2566 */ 2567 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2568 if (m == NULL) { 2569 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2570 return; 2571 } 2572 2573 m->m_len = m->m_pkthdr.len = 0; 2574 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2575 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2576 2577 /* 2578 * Send the upcalls 2579 * XXX do we need to set the address in k_igmpsrc ? 2580 */ 2581 mrtstat.mrts_upcalls++; 2582 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2583 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2584 ++mrtstat.mrts_upq_sockfull; 2585 } 2586} 2587 2588/* 2589 * Compute the timeout hash value for the bw_meter entries 2590 */ 2591#define BW_METER_TIMEHASH(bw_meter, hash) \ 2592 do { \ 2593 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2594 \ 2595 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2596 (hash) = next_timeval.tv_sec; \ 2597 if (next_timeval.tv_usec) \ 2598 (hash)++; /* XXX: make sure we don't timeout early */ \ 2599 (hash) %= BW_METER_BUCKETS; \ 2600 } while (0) 2601 2602/* 2603 * Schedule a timer to process periodically bw_meter entry of type "<=" 2604 * by linking the entry in the proper hash bucket. 2605 */ 2606static void 2607schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2608{ 2609 int time_hash, s; 2610 2611 if (!(x->bm_flags & BW_METER_LEQ)) 2612 return; /* XXX: we schedule timers only for "<=" entries */ 2613 2614 /* 2615 * Reset the bw_meter entry 2616 */ 2617 s = splnet(); 2618 x->bm_start_time = *nowp; 2619 x->bm_measured.b_packets = 0; 2620 x->bm_measured.b_bytes = 0; 2621 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2622 splx(s); 2623 2624 /* 2625 * Compute the timeout hash value and insert the entry 2626 */ 2627 BW_METER_TIMEHASH(x, time_hash); 2628 x->bm_time_next = bw_meter_timers[time_hash]; 2629 bw_meter_timers[time_hash] = x; 2630 x->bm_time_hash = time_hash; 2631} 2632 2633/* 2634 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2635 * by removing the entry from the proper hash bucket. 2636 */ 2637static void 2638unschedule_bw_meter(struct bw_meter *x) 2639{ 2640 int time_hash; 2641 struct bw_meter *prev, *tmp; 2642 2643 if (!(x->bm_flags & BW_METER_LEQ)) 2644 return; /* XXX: we schedule timers only for "<=" entries */ 2645 2646 /* 2647 * Compute the timeout hash value and delete the entry 2648 */ 2649 time_hash = x->bm_time_hash; 2650 if (time_hash >= BW_METER_BUCKETS) 2651 return; /* Entry was not scheduled */ 2652 2653 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2654 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2655 if (tmp == x) 2656 break; 2657 2658 if (tmp == NULL) 2659 panic("unschedule_bw_meter: bw_meter entry not found"); 2660 2661 if (prev != NULL) 2662 prev->bm_time_next = x->bm_time_next; 2663 else 2664 bw_meter_timers[time_hash] = x->bm_time_next; 2665 2666 x->bm_time_next = NULL; 2667 x->bm_time_hash = BW_METER_BUCKETS; 2668} 2669 2670 2671/* 2672 * Process all "<=" type of bw_meter that should be processed now, 2673 * and for each entry prepare an upcall if necessary. Each processed 2674 * entry is rescheduled again for the (periodic) processing. 2675 * 2676 * This is run periodically (once per second normally). On each round, 2677 * all the potentially matching entries are in the hash slot that we are 2678 * looking at. 2679 */ 2680static void 2681bw_meter_process() 2682{ 2683 static uint32_t last_tv_sec; /* last time we processed this */ 2684 2685 uint32_t loops; 2686 int i, s; 2687 struct timeval now, process_endtime; 2688 2689 GET_TIME(now); 2690 if (last_tv_sec == now.tv_sec) 2691 return; /* nothing to do */ 2692 2693 s = splnet(); 2694 loops = now.tv_sec - last_tv_sec; 2695 last_tv_sec = now.tv_sec; 2696 if (loops > BW_METER_BUCKETS) 2697 loops = BW_METER_BUCKETS; 2698 2699 /* 2700 * Process all bins of bw_meter entries from the one after the last 2701 * processed to the current one. On entry, i points to the last bucket 2702 * visited, so we need to increment i at the beginning of the loop. 2703 */ 2704 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2705 struct bw_meter *x, *tmp_list; 2706 2707 if (++i >= BW_METER_BUCKETS) 2708 i = 0; 2709 2710 /* Disconnect the list of bw_meter entries from the bin */ 2711 tmp_list = bw_meter_timers[i]; 2712 bw_meter_timers[i] = NULL; 2713 2714 /* Process the list of bw_meter entries */ 2715 while (tmp_list != NULL) { 2716 x = tmp_list; 2717 tmp_list = tmp_list->bm_time_next; 2718 2719 /* Test if the time interval is over */ 2720 process_endtime = x->bm_start_time; 2721 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2722 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2723 /* Not yet: reschedule, but don't reset */ 2724 int time_hash; 2725 2726 BW_METER_TIMEHASH(x, time_hash); 2727 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2728 /* 2729 * XXX: somehow the bin processing is a bit ahead of time. 2730 * Put the entry in the next bin. 2731 */ 2732 if (++time_hash >= BW_METER_BUCKETS) 2733 time_hash = 0; 2734 } 2735 x->bm_time_next = bw_meter_timers[time_hash]; 2736 bw_meter_timers[time_hash] = x; 2737 x->bm_time_hash = time_hash; 2738 2739 continue; 2740 } 2741 2742 /* 2743 * Test if we should deliver an upcall 2744 */ 2745 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2746 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2747 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2748 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2749 /* Prepare an upcall for delivery */ 2750 bw_meter_prepare_upcall(x, &now); 2751 } 2752 2753 /* 2754 * Reschedule for next processing 2755 */ 2756 schedule_bw_meter(x, &now); 2757 } 2758 } 2759 splx(s); 2760 2761 /* Send all upcalls that are pending delivery */ 2762 bw_upcalls_send(); 2763} 2764 2765/* 2766 * A periodic function for sending all upcalls that are pending delivery 2767 */ 2768static void 2769expire_bw_upcalls_send(void *unused) 2770{ 2771 bw_upcalls_send(); 2772 2773 bw_upcalls_ch = timeout(expire_bw_upcalls_send, NULL, BW_UPCALLS_PERIOD); 2774} 2775 2776/* 2777 * A periodic function for periodic scanning of the multicast forwarding 2778 * table for processing all "<=" bw_meter entries. 2779 */ 2780static void 2781expire_bw_meter_process(void *unused) 2782{ 2783 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2784 bw_meter_process(); 2785 2786 bw_meter_ch = timeout(expire_bw_meter_process, NULL, BW_METER_PERIOD); 2787} 2788 2789/* 2790 * End of bandwidth monitoring code 2791 */ 2792 2793#ifdef PIM 2794/* 2795 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2796 * 2797 */ 2798static int 2799pim_register_send(struct ip *ip, struct vif *vifp, 2800 struct mbuf *m, struct mfc *rt) 2801{ 2802 struct mbuf *mb_copy, *mm; 2803 2804 if (mrtdebug & DEBUG_PIM) 2805 log(LOG_DEBUG, "pim_register_send: "); 2806 2807 mb_copy = pim_register_prepare(ip, m); 2808 if (mb_copy == NULL) 2809 return ENOBUFS; 2810 2811 /* 2812 * Send all the fragments. Note that the mbuf for each fragment 2813 * is freed by the sending machinery. 2814 */ 2815 for (mm = mb_copy; mm; mm = mb_copy) { 2816 mb_copy = mm->m_nextpkt; 2817 mm->m_nextpkt = 0; 2818 mm = m_pullup(mm, sizeof(struct ip)); 2819 if (mm != NULL) { 2820 ip = mtod(mm, struct ip *); 2821 if ((mrt_api_config & MRT_MFC_RP) && 2822 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2823 pim_register_send_rp(ip, vifp, mm, rt); 2824 } else { 2825 pim_register_send_upcall(ip, vifp, mm, rt); 2826 } 2827 } 2828 } 2829 2830 return 0; 2831} 2832 2833/* 2834 * Return a copy of the data packet that is ready for PIM Register 2835 * encapsulation. 2836 * XXX: Note that in the returned copy the IP header is a valid one. 2837 */ 2838static struct mbuf * 2839pim_register_prepare(struct ip *ip, struct mbuf *m) 2840{ 2841 struct mbuf *mb_copy = NULL; 2842 int mtu; 2843 2844 /* Take care of delayed checksums */ 2845 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2846 in_delayed_cksum(m); 2847 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2848 } 2849 2850 /* 2851 * Copy the old packet & pullup its IP header into the 2852 * new mbuf so we can modify it. 2853 */ 2854 mb_copy = m_copypacket(m, M_DONTWAIT); 2855 if (mb_copy == NULL) 2856 return NULL; 2857 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2858 if (mb_copy == NULL) 2859 return NULL; 2860 2861 /* take care of the TTL */ 2862 ip = mtod(mb_copy, struct ip *); 2863 --ip->ip_ttl; 2864 2865 /* Compute the MTU after the PIM Register encapsulation */ 2866 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2867 2868 if (ip->ip_len <= mtu) { 2869 /* Turn the IP header into a valid one */ 2870 ip->ip_len = htons(ip->ip_len); 2871 ip->ip_off = htons(ip->ip_off); 2872 ip->ip_sum = 0; 2873 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2874 } else { 2875 /* Fragment the packet */ 2876 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2877 m_freem(mb_copy); 2878 return NULL; 2879 } 2880 } 2881 return mb_copy; 2882} 2883 2884/* 2885 * Send an upcall with the data packet to the user-level process. 2886 */ 2887static int 2888pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2889 struct mbuf *mb_copy, struct mfc *rt) 2890{ 2891 struct mbuf *mb_first; 2892 int len = ntohs(ip->ip_len); 2893 struct igmpmsg *im; 2894 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2895 2896 /* 2897 * Add a new mbuf with an upcall header 2898 */ 2899 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 2900 if (mb_first == NULL) { 2901 m_freem(mb_copy); 2902 return ENOBUFS; 2903 } 2904 mb_first->m_data += max_linkhdr; 2905 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2906 mb_first->m_len = sizeof(struct igmpmsg); 2907 mb_first->m_next = mb_copy; 2908 2909 /* Send message to routing daemon */ 2910 im = mtod(mb_first, struct igmpmsg *); 2911 im->im_msgtype = IGMPMSG_WHOLEPKT; 2912 im->im_mbz = 0; 2913 im->im_vif = vifp - viftable; 2914 im->im_src = ip->ip_src; 2915 im->im_dst = ip->ip_dst; 2916 2917 k_igmpsrc.sin_addr = ip->ip_src; 2918 2919 mrtstat.mrts_upcalls++; 2920 2921 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2922 if (mrtdebug & DEBUG_PIM) 2923 log(LOG_WARNING, 2924 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2925 ++mrtstat.mrts_upq_sockfull; 2926 return ENOBUFS; 2927 } 2928 2929 /* Keep statistics */ 2930 pimstat.pims_snd_registers_msgs++; 2931 pimstat.pims_snd_registers_bytes += len; 2932 2933 return 0; 2934} 2935 2936/* 2937 * Encapsulate the data packet in PIM Register message and send it to the RP. 2938 */ 2939static int 2940pim_register_send_rp(struct ip *ip, struct vif *vifp, 2941 struct mbuf *mb_copy, struct mfc *rt) 2942{ 2943 struct mbuf *mb_first; 2944 struct ip *ip_outer; 2945 struct pim_encap_pimhdr *pimhdr; 2946 int len = ntohs(ip->ip_len); 2947 vifi_t vifi = rt->mfc_parent; 2948 2949 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2950 m_freem(mb_copy); 2951 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2952 } 2953 2954 /* 2955 * Add a new mbuf with the encapsulating header 2956 */ 2957 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 2958 if (mb_first == NULL) { 2959 m_freem(mb_copy); 2960 return ENOBUFS; 2961 } 2962 mb_first->m_data += max_linkhdr; 2963 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2964 mb_first->m_next = mb_copy; 2965 2966 mb_first->m_pkthdr.len = len + mb_first->m_len; 2967 2968 /* 2969 * Fill in the encapsulating IP and PIM header 2970 */ 2971 ip_outer = mtod(mb_first, struct ip *); 2972 *ip_outer = pim_encap_iphdr; 2973#ifdef RANDOM_IP_ID 2974 ip_outer->ip_id = ip_randomid(); 2975#else 2976 ip_outer->ip_id = htons(ip_id++); 2977#endif 2978 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2979 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2980 ip_outer->ip_dst = rt->mfc_rp; 2981 /* 2982 * Copy the inner header TOS to the outer header, and take care of the 2983 * IP_DF bit. 2984 */ 2985 ip_outer->ip_tos = ip->ip_tos; 2986 if (ntohs(ip->ip_off) & IP_DF) 2987 ip_outer->ip_off |= IP_DF; 2988 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2989 + sizeof(pim_encap_iphdr)); 2990 *pimhdr = pim_encap_pimhdr; 2991 /* If the iif crosses a border, set the Border-bit */ 2992 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2993 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2994 2995 mb_first->m_data += sizeof(pim_encap_iphdr); 2996 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2997 mb_first->m_data -= sizeof(pim_encap_iphdr); 2998 2999 if (vifp->v_rate_limit == 0) 3000 tbf_send_packet(vifp, mb_first); 3001 else 3002 tbf_control(vifp, mb_first, ip, ip_outer->ip_len); 3003 3004 /* Keep statistics */ 3005 pimstat.pims_snd_registers_msgs++; 3006 pimstat.pims_snd_registers_bytes += len; 3007 3008 return 0; 3009} 3010 3011/* 3012 * PIM-SMv2 and PIM-DM messages processing. 3013 * Receives and verifies the PIM control messages, and passes them 3014 * up to the listening socket, using rip_input(). 3015 * The only message with special processing is the PIM_REGISTER message 3016 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3017 * is passed to if_simloop(). 3018 */ 3019void 3020pim_input(struct mbuf *m, int off) 3021{ 3022 struct ip *ip = mtod(m, struct ip *); 3023 struct pim *pim; 3024 int minlen; 3025 int datalen = ip->ip_len; 3026 int ip_tos; 3027 int iphlen = off; 3028 3029 /* Keep statistics */ 3030 pimstat.pims_rcv_total_msgs++; 3031 pimstat.pims_rcv_total_bytes += datalen; 3032 3033 /* 3034 * Validate lengths 3035 */ 3036 if (datalen < PIM_MINLEN) { 3037 pimstat.pims_rcv_tooshort++; 3038 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3039 datalen, (u_long)ip->ip_src.s_addr); 3040 m_freem(m); 3041 return; 3042 } 3043 3044 /* 3045 * If the packet is at least as big as a REGISTER, go agead 3046 * and grab the PIM REGISTER header size, to avoid another 3047 * possible m_pullup() later. 3048 * 3049 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3050 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3051 */ 3052 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3053 /* 3054 * Get the IP and PIM headers in contiguous memory, and 3055 * possibly the PIM REGISTER header. 3056 */ 3057 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3058 (m = m_pullup(m, minlen)) == 0) { 3059 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3060 return; 3061 } 3062 /* m_pullup() may have given us a new mbuf so reset ip. */ 3063 ip = mtod(m, struct ip *); 3064 ip_tos = ip->ip_tos; 3065 3066 /* adjust mbuf to point to the PIM header */ 3067 m->m_data += iphlen; 3068 m->m_len -= iphlen; 3069 pim = mtod(m, struct pim *); 3070 3071 /* 3072 * Validate checksum. If PIM REGISTER, exclude the data packet. 3073 * 3074 * XXX: some older PIMv2 implementations don't make this distinction, 3075 * so for compatibility reason perform the checksum over part of the 3076 * message, and if error, then over the whole message. 3077 */ 3078 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3079 /* do nothing, checksum okay */ 3080 } else if (in_cksum(m, datalen)) { 3081 pimstat.pims_rcv_badsum++; 3082 if (mrtdebug & DEBUG_PIM) 3083 log(LOG_DEBUG, "pim_input: invalid checksum"); 3084 m_freem(m); 3085 return; 3086 } 3087 3088 /* PIM version check */ 3089 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3090 pimstat.pims_rcv_badversion++; 3091 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3092 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3093 m_freem(m); 3094 return; 3095 } 3096 3097 /* restore mbuf back to the outer IP */ 3098 m->m_data -= iphlen; 3099 m->m_len += iphlen; 3100 3101 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3102 /* 3103 * Since this is a REGISTER, we'll make a copy of the register 3104 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3105 * routing daemon. 3106 */ 3107 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3108 struct mbuf *mcp; 3109 struct ip *encap_ip; 3110 u_int32_t *reghdr; 3111 3112 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3113 if (mrtdebug & DEBUG_PIM) 3114 log(LOG_DEBUG, 3115 "pim_input: register vif not set: %d\n", reg_vif_num); 3116 m_freem(m); 3117 return; 3118 } 3119 3120 /* 3121 * Validate length 3122 */ 3123 if (datalen < PIM_REG_MINLEN) { 3124 pimstat.pims_rcv_tooshort++; 3125 pimstat.pims_rcv_badregisters++; 3126 log(LOG_ERR, 3127 "pim_input: register packet size too small %d from %lx\n", 3128 datalen, (u_long)ip->ip_src.s_addr); 3129 m_freem(m); 3130 return; 3131 } 3132 3133 reghdr = (u_int32_t *)(pim + 1); 3134 encap_ip = (struct ip *)(reghdr + 1); 3135 3136 if (mrtdebug & DEBUG_PIM) { 3137 log(LOG_DEBUG, 3138 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3139 (u_long)ntohl(encap_ip->ip_src.s_addr), 3140 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3141 ntohs(encap_ip->ip_len)); 3142 } 3143 3144 /* verify the version number of the inner packet */ 3145 if (encap_ip->ip_v != IPVERSION) { 3146 pimstat.pims_rcv_badregisters++; 3147 if (mrtdebug & DEBUG_PIM) { 3148 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3149 "of the inner packet\n", encap_ip->ip_v); 3150 } 3151 m_freem(m); 3152 return; 3153 } 3154 3155 /* verify the inner packet is destined to a mcast group */ 3156 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 3157 pimstat.pims_rcv_badregisters++; 3158 if (mrtdebug & DEBUG_PIM) 3159 log(LOG_DEBUG, 3160 "pim_input: inner packet of register is not " 3161 "multicast %lx\n", 3162 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3163 m_freem(m); 3164 return; 3165 } 3166 3167 /* 3168 * Copy the TOS from the outer IP header to the inner IP header. 3169 */ 3170 if (encap_ip->ip_tos != ip_tos) { 3171 /* Outer TOS -> inner TOS */ 3172 encap_ip->ip_tos = ip_tos; 3173 /* Recompute the inner header checksum. Sigh... */ 3174 3175 /* adjust mbuf to point to the inner IP header */ 3176 m->m_data += (iphlen + PIM_MINLEN); 3177 m->m_len -= (iphlen + PIM_MINLEN); 3178 3179 encap_ip->ip_sum = 0; 3180 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3181 3182 /* restore mbuf to point back to the outer IP header */ 3183 m->m_data -= (iphlen + PIM_MINLEN); 3184 m->m_len += (iphlen + PIM_MINLEN); 3185 } 3186 3187 /* If a NULL_REGISTER, pass it to the daemon */ 3188 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3189 goto pim_input_to_daemon; 3190 3191 /* 3192 * Decapsulate the inner IP packet and loopback to forward it 3193 * as a normal multicast packet. Also, make a copy of the 3194 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3195 * to pass to the daemon later, so it can take the appropriate 3196 * actions (e.g., send back PIM_REGISTER_STOP). 3197 * XXX: here m->m_data points to the outer IP header. 3198 */ 3199 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3200 if (mcp == NULL) { 3201 log(LOG_ERR, 3202 "pim_input: pim register: could not copy register head\n"); 3203 m_freem(m); 3204 return; 3205 } 3206 3207 /* Keep statistics */ 3208 /* XXX: registers_bytes include only the encap. mcast pkt */ 3209 pimstat.pims_rcv_registers_msgs++; 3210 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3211 3212 /* 3213 * forward the inner ip packet; point m_data at the inner ip. 3214 */ 3215 m_adj(m, iphlen + PIM_MINLEN); 3216 3217 if (mrtdebug & DEBUG_PIM) { 3218 log(LOG_DEBUG, 3219 "pim_input: forwarding decapsulated register: " 3220 "src %lx, dst %lx, vif %d\n", 3221 (u_long)ntohl(encap_ip->ip_src.s_addr), 3222 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3223 reg_vif_num); 3224 } 3225 if_simloop(viftable[reg_vif_num].v_ifp, m, dst.sin_family, 0); 3226 3227 /* prepare the register head to send to the mrouting daemon */ 3228 m = mcp; 3229 } 3230 3231pim_input_to_daemon: 3232 /* 3233 * Pass the PIM message up to the daemon; if it is a Register message, 3234 * pass the 'head' only up to the daemon. This includes the 3235 * outer IP header, PIM header, PIM-Register header and the 3236 * inner IP header. 3237 * XXX: the outer IP header pkt size of a Register is not adjust to 3238 * reflect the fact that the inner multicast data is truncated. 3239 */ 3240 rip_input(m, iphlen); 3241 3242 return; 3243} 3244#endif /* PIM */ 3245 3246static int 3247ip_mroute_modevent(module_t mod, int type, void *unused) 3248{ 3249 int s; 3250 3251 switch (type) { 3252 case MOD_LOAD: 3253 s = splnet(); 3254 /* XXX Protect against multiple loading */ 3255 ip_mcast_src = X_ip_mcast_src; 3256 ip_mforward = X_ip_mforward; 3257 ip_mrouter_done = X_ip_mrouter_done; 3258 ip_mrouter_get = X_ip_mrouter_get; 3259 ip_mrouter_set = X_ip_mrouter_set; 3260 ip_rsvp_force_done = X_ip_rsvp_force_done; 3261 ip_rsvp_vif = X_ip_rsvp_vif; 3262 legal_vif_num = X_legal_vif_num; 3263 mrt_ioctl = X_mrt_ioctl; 3264 rsvp_input_p = X_rsvp_input; 3265 splx(s); 3266 break; 3267 3268 case MOD_UNLOAD: 3269 if (ip_mrouter) 3270 return EINVAL; 3271 3272 s = splnet(); 3273 ip_mcast_src = NULL; 3274 ip_mforward = NULL; 3275 ip_mrouter_done = NULL; 3276 ip_mrouter_get = NULL; 3277 ip_mrouter_set = NULL; 3278 ip_rsvp_force_done = NULL; 3279 ip_rsvp_vif = NULL; 3280 legal_vif_num = NULL; 3281 mrt_ioctl = NULL; 3282 rsvp_input_p = NULL; 3283 splx(s); 3284 break; 3285 } 3286 return 0; 3287} 3288 3289static moduledata_t ip_mroutemod = { 3290 "ip_mroute", 3291 ip_mroute_modevent, 3292 0 3293}; 3294DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3295