ip_mroute.c revision 121446
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 121446 2003-10-24 00:09:18Z sam $ 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 84/* 85 * Locking. We use two locks: one for the virtual interface table and 86 * one for the forwarding table. These locks may be nested in which case 87 * the VIF lock must always be taken first. Note that each lock is used 88 * to cover not only the specific data structure but also related data 89 * structures. It may be better to add more fine-grained locking later; 90 * it's not clear how performance-critical this code is. 91 */ 92 93static struct mrtstat mrtstat; 94SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 95 &mrtstat, mrtstat, 96 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 97 98static struct mfc *mfctable[MFCTBLSIZ]; 99SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 100 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 101 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 102 103static struct mtx mfc_mtx; 104#define MFC_LOCK() mtx_lock(&mfc_mtx) 105#define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 106#define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED) 107#define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF) 108#define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 109 110static struct vif viftable[MAXVIFS]; 111SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 112 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 113 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 114 115static struct mtx vif_mtx; 116#define VIF_LOCK() mtx_lock(&vif_mtx) 117#define VIF_UNLOCK() mtx_unlock(&vif_mtx) 118#define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 119#define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF) 120#define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 121 122static u_char nexpire[MFCTBLSIZ]; 123 124static struct callout expire_upcalls_ch; 125 126#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 127#define UPCALL_EXPIRE 6 /* number of timeouts */ 128 129/* 130 * Define the token bucket filter structures 131 * tbftable -> each vif has one of these for storing info 132 */ 133 134static struct tbf tbftable[MAXVIFS]; 135#define TBF_REPROCESS (hz / 100) /* 100x / second */ 136 137/* 138 * 'Interfaces' associated with decapsulator (so we can tell 139 * packets that went through it from ones that get reflected 140 * by a broken gateway). These interfaces are never linked into 141 * the system ifnet list & no routes point to them. I.e., packets 142 * can't be sent this way. They only exist as a placeholder for 143 * multicast source verification. 144 */ 145static struct ifnet multicast_decap_if[MAXVIFS]; 146 147#define ENCAP_TTL 64 148#define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 149 150/* prototype IP hdr for encapsulated packets */ 151static struct ip multicast_encap_iphdr = { 152#if BYTE_ORDER == LITTLE_ENDIAN 153 sizeof(struct ip) >> 2, IPVERSION, 154#else 155 IPVERSION, sizeof(struct ip) >> 2, 156#endif 157 0, /* tos */ 158 sizeof(struct ip), /* total length */ 159 0, /* id */ 160 0, /* frag offset */ 161 ENCAP_TTL, ENCAP_PROTO, 162 0, /* checksum */ 163}; 164 165/* 166 * Bandwidth meter variables and constants 167 */ 168static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 169/* 170 * Pending timeouts are stored in a hash table, the key being the 171 * expiration time. Periodically, the entries are analysed and processed. 172 */ 173#define BW_METER_BUCKETS 1024 174static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 175static struct callout bw_meter_ch; 176#define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 177 178/* 179 * Pending upcalls are stored in a vector which is flushed when 180 * full, or periodically 181 */ 182static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 183static u_int bw_upcalls_n; /* # of pending upcalls */ 184static struct callout bw_upcalls_ch; 185#define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 186 187#ifdef PIM 188static struct pimstat pimstat; 189SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 190 &pimstat, pimstat, 191 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 192 193/* 194 * Note: the PIM Register encapsulation adds the following in front of a 195 * data packet: 196 * 197 * struct pim_encap_hdr { 198 * struct ip ip; 199 * struct pim_encap_pimhdr pim; 200 * } 201 * 202 */ 203 204struct pim_encap_pimhdr { 205 struct pim pim; 206 uint32_t flags; 207}; 208 209static struct ip pim_encap_iphdr = { 210#if BYTE_ORDER == LITTLE_ENDIAN 211 sizeof(struct ip) >> 2, 212 IPVERSION, 213#else 214 IPVERSION, 215 sizeof(struct ip) >> 2, 216#endif 217 0, /* tos */ 218 sizeof(struct ip), /* total length */ 219 0, /* id */ 220 0, /* frag offset */ 221 ENCAP_TTL, 222 IPPROTO_PIM, 223 0, /* checksum */ 224}; 225 226static struct pim_encap_pimhdr pim_encap_pimhdr = { 227 { 228 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 229 0, /* reserved */ 230 0, /* checksum */ 231 }, 232 0 /* flags */ 233}; 234 235static struct ifnet multicast_register_if; 236static vifi_t reg_vif_num = VIFI_INVALID; 237#endif /* PIM */ 238 239/* 240 * Private variables. 241 */ 242static vifi_t numvifs; 243static const struct encaptab *encap_cookie; 244 245/* 246 * one-back cache used by mroute_encapcheck to locate a tunnel's vif 247 * given a datagram's src ip address. 248 */ 249static u_long last_encap_src; 250static struct vif *last_encap_vif; 251 252/* 253 * Callout for queue processing. 254 */ 255static struct callout tbf_reprocess_ch; 256 257static u_long X_ip_mcast_src(int vifi); 258static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 259 struct mbuf *m, struct ip_moptions *imo); 260static int X_ip_mrouter_done(void); 261static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 262static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 263static int X_legal_vif_num(int vif); 264static int X_mrt_ioctl(int cmd, caddr_t data); 265 266static int get_sg_cnt(struct sioc_sg_req *); 267static int get_vif_cnt(struct sioc_vif_req *); 268static int ip_mrouter_init(struct socket *, int); 269static int add_vif(struct vifctl *); 270static int del_vif(vifi_t); 271static int add_mfc(struct mfcctl2 *); 272static int del_mfc(struct mfcctl2 *); 273static int set_api_config(uint32_t *); /* chose API capabilities */ 274static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 275static int set_assert(int); 276static void expire_upcalls(void *); 277static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 278static void phyint_send(struct ip *, struct vif *, struct mbuf *); 279static void encap_send(struct ip *, struct vif *, struct mbuf *); 280static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 281static void tbf_queue(struct vif *, struct mbuf *); 282static void tbf_process_q(struct vif *); 283static void tbf_reprocess_q(void *); 284static int tbf_dq_sel(struct vif *, struct ip *); 285static void tbf_send_packet(struct vif *, struct mbuf *); 286static void tbf_update_tokens(struct vif *); 287static int priority(struct vif *, struct ip *); 288 289/* 290 * Bandwidth monitoring 291 */ 292static void free_bw_list(struct bw_meter *list); 293static int add_bw_upcall(struct bw_upcall *); 294static int del_bw_upcall(struct bw_upcall *); 295static void bw_meter_receive_packet(struct bw_meter *x, int plen, 296 struct timeval *nowp); 297static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 298static void bw_upcalls_send(void); 299static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 300static void unschedule_bw_meter(struct bw_meter *x); 301static void bw_meter_process(void); 302static void expire_bw_upcalls_send(void *); 303static void expire_bw_meter_process(void *); 304 305#ifdef PIM 306static int pim_register_send(struct ip *, struct vif *, 307 struct mbuf *, struct mfc *); 308static int pim_register_send_rp(struct ip *, struct vif *, 309 struct mbuf *, struct mfc *); 310static int pim_register_send_upcall(struct ip *, struct vif *, 311 struct mbuf *, struct mfc *); 312static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 313#endif 314 315/* 316 * whether or not special PIM assert processing is enabled. 317 */ 318static int pim_assert; 319/* 320 * Rate limit for assert notification messages, in usec 321 */ 322#define ASSERT_MSG_TIME 3000000 323 324/* 325 * Kernel multicast routing API capabilities and setup. 326 * If more API capabilities are added to the kernel, they should be 327 * recorded in `mrt_api_support'. 328 */ 329static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 330 MRT_MFC_FLAGS_BORDER_VIF | 331 MRT_MFC_RP | 332 MRT_MFC_BW_UPCALL); 333static uint32_t mrt_api_config = 0; 334 335/* 336 * Hash function for a source, group entry 337 */ 338#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 339 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 340 341/* 342 * Find a route for a given origin IP address and Multicast group address 343 * Type of service parameter to be added in the future!!! 344 * Statistics are updated by the caller if needed 345 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 346 */ 347static struct mfc * 348mfc_find(in_addr_t o, in_addr_t g) 349{ 350 struct mfc *rt; 351 352 MFC_LOCK_ASSERT(); 353 354 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 355 if ((rt->mfc_origin.s_addr == o) && 356 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 357 break; 358 return rt; 359} 360 361/* 362 * Macros to compute elapsed time efficiently 363 * Borrowed from Van Jacobson's scheduling code 364 */ 365#define TV_DELTA(a, b, delta) { \ 366 int xxs; \ 367 delta = (a).tv_usec - (b).tv_usec; \ 368 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 369 switch (xxs) { \ 370 case 2: \ 371 delta += 1000000; \ 372 /* FALLTHROUGH */ \ 373 case 1: \ 374 delta += 1000000; \ 375 break; \ 376 default: \ 377 delta += (1000000 * xxs); \ 378 } \ 379 } \ 380} 381 382#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 383 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 384 385/* 386 * Handle MRT setsockopt commands to modify the multicast routing tables. 387 */ 388static int 389X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 390{ 391 int error, optval; 392 vifi_t vifi; 393 struct vifctl vifc; 394 struct mfcctl2 mfc; 395 struct bw_upcall bw_upcall; 396 uint32_t i; 397 398 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 399 return EPERM; 400 401 error = 0; 402 switch (sopt->sopt_name) { 403 case MRT_INIT: 404 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 405 if (error) 406 break; 407 error = ip_mrouter_init(so, optval); 408 break; 409 410 case MRT_DONE: 411 error = ip_mrouter_done(); 412 break; 413 414 case MRT_ADD_VIF: 415 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 416 if (error) 417 break; 418 error = add_vif(&vifc); 419 break; 420 421 case MRT_DEL_VIF: 422 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 423 if (error) 424 break; 425 error = del_vif(vifi); 426 break; 427 428 case MRT_ADD_MFC: 429 case MRT_DEL_MFC: 430 /* 431 * select data size depending on API version. 432 */ 433 if (sopt->sopt_name == MRT_ADD_MFC && 434 mrt_api_config & MRT_API_FLAGS_ALL) { 435 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 436 sizeof(struct mfcctl2)); 437 } else { 438 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 439 sizeof(struct mfcctl)); 440 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 441 sizeof(mfc) - sizeof(struct mfcctl)); 442 } 443 if (error) 444 break; 445 if (sopt->sopt_name == MRT_ADD_MFC) 446 error = add_mfc(&mfc); 447 else 448 error = del_mfc(&mfc); 449 break; 450 451 case MRT_ASSERT: 452 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 453 if (error) 454 break; 455 set_assert(optval); 456 break; 457 458 case MRT_API_CONFIG: 459 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 460 if (!error) 461 error = set_api_config(&i); 462 if (!error) 463 error = sooptcopyout(sopt, &i, sizeof i); 464 break; 465 466 case MRT_ADD_BW_UPCALL: 467 case MRT_DEL_BW_UPCALL: 468 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 469 sizeof bw_upcall); 470 if (error) 471 break; 472 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 473 error = add_bw_upcall(&bw_upcall); 474 else 475 error = del_bw_upcall(&bw_upcall); 476 break; 477 478 default: 479 error = EOPNOTSUPP; 480 break; 481 } 482 return error; 483} 484 485/* 486 * Handle MRT getsockopt commands 487 */ 488static int 489X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 490{ 491 int error; 492 static int version = 0x0305; /* !!! why is this here? XXX */ 493 494 switch (sopt->sopt_name) { 495 case MRT_VERSION: 496 error = sooptcopyout(sopt, &version, sizeof version); 497 break; 498 499 case MRT_ASSERT: 500 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 501 break; 502 503 case MRT_API_SUPPORT: 504 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 505 break; 506 507 case MRT_API_CONFIG: 508 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 509 break; 510 511 default: 512 error = EOPNOTSUPP; 513 break; 514 } 515 return error; 516} 517 518/* 519 * Handle ioctl commands to obtain information from the cache 520 */ 521static int 522X_mrt_ioctl(int cmd, caddr_t data) 523{ 524 int error = 0; 525 526 switch (cmd) { 527 case (SIOCGETVIFCNT): 528 error = get_vif_cnt((struct sioc_vif_req *)data); 529 break; 530 531 case (SIOCGETSGCNT): 532 error = get_sg_cnt((struct sioc_sg_req *)data); 533 break; 534 535 default: 536 error = EINVAL; 537 break; 538 } 539 return error; 540} 541 542/* 543 * returns the packet, byte, rpf-failure count for the source group provided 544 */ 545static int 546get_sg_cnt(struct sioc_sg_req *req) 547{ 548 struct mfc *rt; 549 550 MFC_LOCK(); 551 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 552 if (rt == NULL) { 553 MFC_UNLOCK(); 554 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 555 return EADDRNOTAVAIL; 556 } 557 req->pktcnt = rt->mfc_pkt_cnt; 558 req->bytecnt = rt->mfc_byte_cnt; 559 req->wrong_if = rt->mfc_wrong_if; 560 MFC_UNLOCK(); 561 return 0; 562} 563 564/* 565 * returns the input and output packet and byte counts on the vif provided 566 */ 567static int 568get_vif_cnt(struct sioc_vif_req *req) 569{ 570 vifi_t vifi = req->vifi; 571 572 VIF_LOCK(); 573 if (vifi >= numvifs) { 574 VIF_UNLOCK(); 575 return EINVAL; 576 } 577 578 req->icount = viftable[vifi].v_pkt_in; 579 req->ocount = viftable[vifi].v_pkt_out; 580 req->ibytes = viftable[vifi].v_bytes_in; 581 req->obytes = viftable[vifi].v_bytes_out; 582 VIF_UNLOCK(); 583 584 return 0; 585} 586 587static void 588ip_mrouter_reset(void) 589{ 590 bzero((caddr_t)mfctable, sizeof(mfctable)); 591 MFC_LOCK_INIT(); 592 VIF_LOCK_INIT(); 593 bzero((caddr_t)nexpire, sizeof(nexpire)); 594 595 pim_assert = 0; 596 mrt_api_config = 0; 597 598 callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE); 599 600 bw_upcalls_n = 0; 601 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 602 callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE); 603 callout_init(&bw_meter_ch, CALLOUT_MPSAFE); 604 605 callout_init(&tbf_reprocess_ch, CALLOUT_MPSAFE); 606} 607 608/* 609 * Enable multicast routing 610 */ 611static int 612ip_mrouter_init(struct socket *so, int version) 613{ 614 if (mrtdebug) 615 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 616 so->so_type, so->so_proto->pr_protocol); 617 618 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 619 return EOPNOTSUPP; 620 621 if (version != 1) 622 return ENOPROTOOPT; 623 624 if (ip_mrouter != NULL) 625 return EADDRINUSE; 626 627 ip_mrouter_reset(); 628 629 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 630 631 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 632 expire_bw_upcalls_send, NULL); 633 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 634 635 ip_mrouter = so; 636 637 if (mrtdebug) 638 log(LOG_DEBUG, "ip_mrouter_init\n"); 639 640 return 0; 641} 642 643/* 644 * Disable multicast routing 645 */ 646static int 647X_ip_mrouter_done(void) 648{ 649 vifi_t vifi; 650 int i; 651 struct ifnet *ifp; 652 struct ifreq ifr; 653 struct mfc *rt; 654 struct rtdetq *rte; 655 656 /* 657 * Detach/disable hooks to the reset of the system. 658 */ 659 ip_mrouter = NULL; 660 mrt_api_config = 0; 661 662 VIF_LOCK(); 663 if (encap_cookie) { 664 encap_detach(encap_cookie); 665 encap_cookie = NULL; 666 } 667 callout_stop(&tbf_reprocess_ch); 668 669 /* 670 * For each phyint in use, disable promiscuous reception of all IP 671 * multicasts. 672 */ 673 for (vifi = 0; vifi < numvifs; vifi++) { 674 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 675 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 676 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 677 678 so->sin_len = sizeof(struct sockaddr_in); 679 so->sin_family = AF_INET; 680 so->sin_addr.s_addr = INADDR_ANY; 681 ifp = viftable[vifi].v_ifp; 682 if_allmulti(ifp, 0); 683 } 684 } 685 bzero((caddr_t)tbftable, sizeof(tbftable)); 686 bzero((caddr_t)viftable, sizeof(viftable)); 687 numvifs = 0; 688 pim_assert = 0; 689 VIF_LOCK_DESTROY(); 690 691 /* 692 * Free all multicast forwarding cache entries. 693 */ 694 MFC_LOCK(); 695 callout_stop(&expire_upcalls_ch); 696 callout_stop(&bw_upcalls_ch); 697 callout_stop(&bw_meter_ch); 698 699 for (i = 0; i < MFCTBLSIZ; i++) { 700 for (rt = mfctable[i]; rt != NULL; ) { 701 struct mfc *nr = rt->mfc_next; 702 703 for (rte = rt->mfc_stall; rte != NULL; ) { 704 struct rtdetq *n = rte->next; 705 706 m_freem(rte->m); 707 free(rte, M_MRTABLE); 708 rte = n; 709 } 710 free_bw_list(rt->mfc_bw_meter); 711 free(rt, M_MRTABLE); 712 rt = nr; 713 } 714 } 715 bzero((caddr_t)mfctable, sizeof(mfctable)); 716 bw_upcalls_n = 0; 717 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 718 MFC_LOCK_DESTROY(); 719 720 /* 721 * Reset de-encapsulation cache 722 */ 723 last_encap_src = INADDR_ANY; 724 last_encap_vif = NULL; 725#ifdef PIM 726 reg_vif_num = VIFI_INVALID; 727#endif 728 729 if (mrtdebug) 730 log(LOG_DEBUG, "ip_mrouter_done\n"); 731 732 return 0; 733} 734 735/* 736 * Set PIM assert processing global 737 */ 738static int 739set_assert(int i) 740{ 741 if ((i != 1) && (i != 0)) 742 return EINVAL; 743 744 pim_assert = i; 745 746 return 0; 747} 748 749/* 750 * Configure API capabilities 751 */ 752int 753set_api_config(uint32_t *apival) 754{ 755 int i; 756 757 /* 758 * We can set the API capabilities only if it is the first operation 759 * after MRT_INIT. I.e.: 760 * - there are no vifs installed 761 * - pim_assert is not enabled 762 * - the MFC table is empty 763 */ 764 if (numvifs > 0) { 765 *apival = 0; 766 return EPERM; 767 } 768 if (pim_assert) { 769 *apival = 0; 770 return EPERM; 771 } 772 for (i = 0; i < MFCTBLSIZ; i++) { 773 if (mfctable[i] != NULL) { 774 *apival = 0; 775 return EPERM; 776 } 777 } 778 779 mrt_api_config = *apival & mrt_api_support; 780 *apival = mrt_api_config; 781 782 return 0; 783} 784 785/* 786 * Decide if a packet is from a tunnelled peer. 787 * Return 0 if not, 64 if so. XXX yuck.. 64 ??? 788 */ 789static int 790mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 791{ 792 struct ip *ip = mtod(m, struct ip *); 793 int hlen = ip->ip_hl << 2; 794 795 /* 796 * don't claim the packet if it's not to a multicast destination or if 797 * we don't have an encapsulating tunnel with the source. 798 * Note: This code assumes that the remote site IP address 799 * uniquely identifies the tunnel (i.e., that this site has 800 * at most one tunnel with the remote site). 801 */ 802 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) 803 return 0; 804 if (ip->ip_src.s_addr != last_encap_src) { 805 struct vif *vifp = viftable; 806 struct vif *vife = vifp + numvifs; 807 808 last_encap_src = ip->ip_src.s_addr; 809 last_encap_vif = NULL; 810 for ( ; vifp < vife; ++vifp) 811 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 812 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL) 813 last_encap_vif = vifp; 814 break; 815 } 816 } 817 if (last_encap_vif == NULL) { 818 last_encap_src = INADDR_ANY; 819 return 0; 820 } 821 return 64; 822} 823 824/* 825 * De-encapsulate a packet and feed it back through ip input (this 826 * routine is called whenever IP gets a packet that mroute_encap_func() 827 * claimed). 828 */ 829static void 830mroute_encap_input(struct mbuf *m, int off) 831{ 832 struct ip *ip = mtod(m, struct ip *); 833 int hlen = ip->ip_hl << 2; 834 835 if (hlen > sizeof(struct ip)) 836 ip_stripoptions(m, (struct mbuf *) 0); 837 m->m_data += sizeof(struct ip); 838 m->m_len -= sizeof(struct ip); 839 m->m_pkthdr.len -= sizeof(struct ip); 840 841 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 842 843 netisr_queue(NETISR_IP, m); 844 /* 845 * normally we would need a "schednetisr(NETISR_IP)" 846 * here but we were called by ip_input and it is going 847 * to loop back & try to dequeue the packet we just 848 * queued as soon as we return so we avoid the 849 * unnecessary software interrrupt. 850 * 851 * XXX 852 * This no longer holds - we may have direct-dispatched the packet, 853 * or there may be a queue processing limit. 854 */ 855} 856 857extern struct domain inetdomain; 858static struct protosw mroute_encap_protosw = 859{ SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 860 mroute_encap_input, 0, 0, rip_ctloutput, 861 0, 862 0, 0, 0, 0, 863 &rip_usrreqs 864}; 865 866/* 867 * Add a vif to the vif table 868 */ 869static int 870add_vif(struct vifctl *vifcp) 871{ 872 struct vif *vifp = viftable + vifcp->vifc_vifi; 873 struct sockaddr_in sin = {sizeof sin, AF_INET}; 874 struct ifaddr *ifa; 875 struct ifnet *ifp; 876 int error; 877 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 878 879 VIF_LOCK(); 880 if (vifcp->vifc_vifi >= MAXVIFS) { 881 VIF_UNLOCK(); 882 return EINVAL; 883 } 884 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) { 885 VIF_UNLOCK(); 886 return EADDRINUSE; 887 } 888 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) { 889 VIF_UNLOCK(); 890 return EADDRNOTAVAIL; 891 } 892 893 /* Find the interface with an address in AF_INET family */ 894#ifdef PIM 895 if (vifcp->vifc_flags & VIFF_REGISTER) { 896 /* 897 * XXX: Because VIFF_REGISTER does not really need a valid 898 * local interface (e.g. it could be 127.0.0.2), we don't 899 * check its address. 900 */ 901 ifp = NULL; 902 } else 903#endif 904 { 905 sin.sin_addr = vifcp->vifc_lcl_addr; 906 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 907 if (ifa == NULL) { 908 VIF_UNLOCK(); 909 return EADDRNOTAVAIL; 910 } 911 ifp = ifa->ifa_ifp; 912 } 913 914 if (vifcp->vifc_flags & VIFF_TUNNEL) { 915 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 916 /* 917 * An encapsulating tunnel is wanted. Tell 918 * mroute_encap_input() to start paying attention 919 * to encapsulated packets. 920 */ 921 if (encap_cookie == NULL) { 922 int i; 923 924 encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 925 mroute_encapcheck, 926 (struct protosw *)&mroute_encap_protosw, NULL); 927 928 if (encap_cookie == NULL) { 929 printf("ip_mroute: unable to attach encap\n"); 930 VIF_UNLOCK(); 931 return EIO; /* XXX */ 932 } 933 for (i = 0; i < MAXVIFS; ++i) { 934 multicast_decap_if[i].if_name = "mdecap"; 935 multicast_decap_if[i].if_unit = i; 936 } 937 } 938 /* 939 * Set interface to fake encapsulator interface 940 */ 941 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 942 /* 943 * Prepare cached route entry 944 */ 945 bzero(&vifp->v_route, sizeof(vifp->v_route)); 946 } else { 947 log(LOG_ERR, "source routed tunnels not supported\n"); 948 VIF_UNLOCK(); 949 return EOPNOTSUPP; 950 } 951#ifdef PIM 952 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 953 ifp = &multicast_register_if; 954 if (mrtdebug) 955 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 956 (void *)&multicast_register_if); 957 if (reg_vif_num == VIFI_INVALID) { 958 multicast_register_if.if_name = "register_vif"; 959 multicast_register_if.if_unit = 0; 960 multicast_register_if.if_flags = IFF_LOOPBACK; 961 bzero(&vifp->v_route, sizeof(vifp->v_route)); 962 reg_vif_num = vifcp->vifc_vifi; 963 } 964#endif 965 } else { /* Make sure the interface supports multicast */ 966 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 967 VIF_UNLOCK(); 968 return EOPNOTSUPP; 969 } 970 971 /* Enable promiscuous reception of all IP multicasts from the if */ 972 error = if_allmulti(ifp, 1); 973 if (error) { 974 VIF_UNLOCK(); 975 return error; 976 } 977 } 978 979 /* define parameters for the tbf structure */ 980 vifp->v_tbf = v_tbf; 981 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 982 vifp->v_tbf->tbf_n_tok = 0; 983 vifp->v_tbf->tbf_q_len = 0; 984 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 985 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 986 987 vifp->v_flags = vifcp->vifc_flags; 988 vifp->v_threshold = vifcp->vifc_threshold; 989 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 990 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 991 vifp->v_ifp = ifp; 992 /* scaling up here allows division by 1024 in critical code */ 993 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 994 vifp->v_rsvp_on = 0; 995 vifp->v_rsvpd = NULL; 996 /* initialize per vif pkt counters */ 997 vifp->v_pkt_in = 0; 998 vifp->v_pkt_out = 0; 999 vifp->v_bytes_in = 0; 1000 vifp->v_bytes_out = 0; 1001 1002 /* Adjust numvifs up if the vifi is higher than numvifs */ 1003 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 1004 1005 VIF_UNLOCK(); 1006 1007 if (mrtdebug) 1008 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 1009 vifcp->vifc_vifi, 1010 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 1011 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 1012 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 1013 vifcp->vifc_threshold, 1014 vifcp->vifc_rate_limit); 1015 1016 return 0; 1017} 1018 1019/* 1020 * Delete a vif from the vif table 1021 */ 1022static int 1023del_vif(vifi_t vifi) 1024{ 1025 struct vif *vifp; 1026 1027 VIF_LOCK(); 1028 1029 if (vifi >= numvifs) { 1030 VIF_UNLOCK(); 1031 return EINVAL; 1032 } 1033 vifp = &viftable[vifi]; 1034 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) { 1035 VIF_UNLOCK(); 1036 return EADDRNOTAVAIL; 1037 } 1038 1039 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 1040 if_allmulti(vifp->v_ifp, 0); 1041 1042 if (vifp == last_encap_vif) { 1043 last_encap_vif = NULL; 1044 last_encap_src = INADDR_ANY; 1045 } 1046 1047 /* 1048 * Free packets queued at the interface 1049 */ 1050 while (vifp->v_tbf->tbf_q) { 1051 struct mbuf *m = vifp->v_tbf->tbf_q; 1052 1053 vifp->v_tbf->tbf_q = m->m_act; 1054 m_freem(m); 1055 } 1056 1057#ifdef PIM 1058 if (vifp->v_flags & VIFF_REGISTER) 1059 reg_vif_num = VIFI_INVALID; 1060#endif 1061 1062 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 1063 bzero((caddr_t)vifp, sizeof (*vifp)); 1064 1065 if (mrtdebug) 1066 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1067 1068 /* Adjust numvifs down */ 1069 for (vifi = numvifs; vifi > 0; vifi--) 1070 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1071 break; 1072 numvifs = vifi; 1073 1074 VIF_UNLOCK(); 1075 1076 return 0; 1077} 1078 1079/* 1080 * update an mfc entry without resetting counters and S,G addresses. 1081 */ 1082static void 1083update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1084{ 1085 int i; 1086 1087 rt->mfc_parent = mfccp->mfcc_parent; 1088 for (i = 0; i < numvifs; i++) { 1089 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1090 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1091 MRT_MFC_FLAGS_ALL; 1092 } 1093 /* set the RP address */ 1094 if (mrt_api_config & MRT_MFC_RP) 1095 rt->mfc_rp = mfccp->mfcc_rp; 1096 else 1097 rt->mfc_rp.s_addr = INADDR_ANY; 1098} 1099 1100/* 1101 * fully initialize an mfc entry from the parameter. 1102 */ 1103static void 1104init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1105{ 1106 rt->mfc_origin = mfccp->mfcc_origin; 1107 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1108 1109 update_mfc_params(rt, mfccp); 1110 1111 /* initialize pkt counters per src-grp */ 1112 rt->mfc_pkt_cnt = 0; 1113 rt->mfc_byte_cnt = 0; 1114 rt->mfc_wrong_if = 0; 1115 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1116} 1117 1118 1119/* 1120 * Add an mfc entry 1121 */ 1122static int 1123add_mfc(struct mfcctl2 *mfccp) 1124{ 1125 struct mfc *rt; 1126 u_long hash; 1127 struct rtdetq *rte; 1128 u_short nstl; 1129 1130 VIF_LOCK(); 1131 MFC_LOCK(); 1132 1133 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1134 1135 /* If an entry already exists, just update the fields */ 1136 if (rt) { 1137 if (mrtdebug & DEBUG_MFC) 1138 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1139 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1140 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1141 mfccp->mfcc_parent); 1142 1143 update_mfc_params(rt, mfccp); 1144 MFC_UNLOCK(); 1145 VIF_UNLOCK(); 1146 return 0; 1147 } 1148 1149 /* 1150 * Find the entry for which the upcall was made and update 1151 */ 1152 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1153 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1154 1155 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1156 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1157 (rt->mfc_stall != NULL)) { 1158 1159 if (nstl++) 1160 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1161 "multiple kernel entries", 1162 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1163 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1164 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1165 1166 if (mrtdebug & DEBUG_MFC) 1167 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1168 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1169 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1170 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1171 1172 init_mfc_params(rt, mfccp); 1173 1174 rt->mfc_expire = 0; /* Don't clean this guy up */ 1175 nexpire[hash]--; 1176 1177 /* free packets Qed at the end of this entry */ 1178 for (rte = rt->mfc_stall; rte != NULL; ) { 1179 struct rtdetq *n = rte->next; 1180 1181 ip_mdq(rte->m, rte->ifp, rt, -1); 1182 m_freem(rte->m); 1183 free(rte, M_MRTABLE); 1184 rte = n; 1185 } 1186 rt->mfc_stall = NULL; 1187 } 1188 } 1189 1190 /* 1191 * It is possible that an entry is being inserted without an upcall 1192 */ 1193 if (nstl == 0) { 1194 if (mrtdebug & DEBUG_MFC) 1195 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1196 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1197 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1198 mfccp->mfcc_parent); 1199 1200 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1201 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1202 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1203 init_mfc_params(rt, mfccp); 1204 if (rt->mfc_expire) 1205 nexpire[hash]--; 1206 rt->mfc_expire = 0; 1207 break; /* XXX */ 1208 } 1209 } 1210 if (rt == NULL) { /* no upcall, so make a new entry */ 1211 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1212 if (rt == NULL) { 1213 MFC_UNLOCK(); 1214 VIF_UNLOCK(); 1215 return ENOBUFS; 1216 } 1217 1218 init_mfc_params(rt, mfccp); 1219 rt->mfc_expire = 0; 1220 rt->mfc_stall = NULL; 1221 1222 rt->mfc_bw_meter = NULL; 1223 /* insert new entry at head of hash chain */ 1224 rt->mfc_next = mfctable[hash]; 1225 mfctable[hash] = rt; 1226 } 1227 } 1228 MFC_UNLOCK(); 1229 VIF_UNLOCK(); 1230 return 0; 1231} 1232 1233/* 1234 * Delete an mfc entry 1235 */ 1236static int 1237del_mfc(struct mfcctl2 *mfccp) 1238{ 1239 struct in_addr origin; 1240 struct in_addr mcastgrp; 1241 struct mfc *rt; 1242 struct mfc **nptr; 1243 u_long hash; 1244 struct bw_meter *list; 1245 1246 origin = mfccp->mfcc_origin; 1247 mcastgrp = mfccp->mfcc_mcastgrp; 1248 1249 if (mrtdebug & DEBUG_MFC) 1250 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1251 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1252 1253 MFC_LOCK(); 1254 1255 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1256 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1257 if (origin.s_addr == rt->mfc_origin.s_addr && 1258 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1259 rt->mfc_stall == NULL) 1260 break; 1261 if (rt == NULL) { 1262 MFC_UNLOCK(); 1263 return EADDRNOTAVAIL; 1264 } 1265 1266 *nptr = rt->mfc_next; 1267 1268 /* 1269 * free the bw_meter entries 1270 */ 1271 list = rt->mfc_bw_meter; 1272 rt->mfc_bw_meter = NULL; 1273 1274 free(rt, M_MRTABLE); 1275 1276 free_bw_list(list); 1277 1278 MFC_UNLOCK(); 1279 1280 return 0; 1281} 1282 1283/* 1284 * Send a message to mrouted on the multicast routing socket 1285 */ 1286static int 1287socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1288{ 1289 if (s) { 1290 if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) { 1291 sorwakeup(s); 1292 return 0; 1293 } 1294 } 1295 m_freem(mm); 1296 return -1; 1297} 1298 1299/* 1300 * IP multicast forwarding function. This function assumes that the packet 1301 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1302 * pointed to by "ifp", and the packet is to be relayed to other networks 1303 * that have members of the packet's destination IP multicast group. 1304 * 1305 * The packet is returned unscathed to the caller, unless it is 1306 * erroneous, in which case a non-zero return value tells the caller to 1307 * discard it. 1308 */ 1309 1310#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1311 1312static int 1313X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1314 struct ip_moptions *imo) 1315{ 1316 struct mfc *rt; 1317 int error; 1318 vifi_t vifi; 1319 1320 if (mrtdebug & DEBUG_FORWARD) 1321 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1322 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1323 (void *)ifp); 1324 1325 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1326 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1327 /* 1328 * Packet arrived via a physical interface or 1329 * an encapsulated tunnel or a register_vif. 1330 */ 1331 } else { 1332 /* 1333 * Packet arrived through a source-route tunnel. 1334 * Source-route tunnels are no longer supported. 1335 */ 1336 static int last_log; 1337 if (last_log != time_second) { 1338 last_log = time_second; 1339 log(LOG_ERR, 1340 "ip_mforward: received source-routed packet from %lx\n", 1341 (u_long)ntohl(ip->ip_src.s_addr)); 1342 } 1343 return 1; 1344 } 1345 1346 VIF_LOCK(); 1347 MFC_LOCK(); 1348 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1349 if (ip->ip_ttl < 255) 1350 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1351 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1352 struct vif *vifp = viftable + vifi; 1353 1354 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1355 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1356 vifi, 1357 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1358 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1359 } 1360 error = ip_mdq(m, ifp, NULL, vifi); 1361 MFC_UNLOCK(); 1362 VIF_UNLOCK(); 1363 return error; 1364 } 1365 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1366 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1367 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1368 if (!imo) 1369 printf("In fact, no options were specified at all\n"); 1370 } 1371 1372 /* 1373 * Don't forward a packet with time-to-live of zero or one, 1374 * or a packet destined to a local-only group. 1375 */ 1376 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) { 1377 MFC_UNLOCK(); 1378 VIF_UNLOCK(); 1379 return 0; 1380 } 1381 1382 /* 1383 * Determine forwarding vifs from the forwarding cache table 1384 */ 1385 ++mrtstat.mrts_mfc_lookups; 1386 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1387 1388 /* Entry exists, so forward if necessary */ 1389 if (rt != NULL) { 1390 error = ip_mdq(m, ifp, rt, -1); 1391 MFC_UNLOCK(); 1392 VIF_UNLOCK(); 1393 return error; 1394 } else { 1395 /* 1396 * If we don't have a route for packet's origin, 1397 * Make a copy of the packet & send message to routing daemon 1398 */ 1399 1400 struct mbuf *mb0; 1401 struct rtdetq *rte; 1402 u_long hash; 1403 int hlen = ip->ip_hl << 2; 1404 1405 ++mrtstat.mrts_mfc_misses; 1406 1407 mrtstat.mrts_no_route++; 1408 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1409 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1410 (u_long)ntohl(ip->ip_src.s_addr), 1411 (u_long)ntohl(ip->ip_dst.s_addr)); 1412 1413 /* 1414 * Allocate mbufs early so that we don't do extra work if we are 1415 * just going to fail anyway. Make sure to pullup the header so 1416 * that other people can't step on it. 1417 */ 1418 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1419 if (rte == NULL) { 1420 MFC_UNLOCK(); 1421 VIF_UNLOCK(); 1422 return ENOBUFS; 1423 } 1424 mb0 = m_copypacket(m, M_DONTWAIT); 1425 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1426 mb0 = m_pullup(mb0, hlen); 1427 if (mb0 == NULL) { 1428 free(rte, M_MRTABLE); 1429 MFC_UNLOCK(); 1430 VIF_UNLOCK(); 1431 return ENOBUFS; 1432 } 1433 1434 /* is there an upcall waiting for this flow ? */ 1435 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1436 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1437 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1438 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1439 (rt->mfc_stall != NULL)) 1440 break; 1441 } 1442 1443 if (rt == NULL) { 1444 int i; 1445 struct igmpmsg *im; 1446 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1447 struct mbuf *mm; 1448 1449 /* 1450 * Locate the vifi for the incoming interface for this packet. 1451 * If none found, drop packet. 1452 */ 1453 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1454 ; 1455 if (vifi >= numvifs) /* vif not found, drop packet */ 1456 goto non_fatal; 1457 1458 /* no upcall, so make a new entry */ 1459 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1460 if (rt == NULL) 1461 goto fail; 1462 /* Make a copy of the header to send to the user level process */ 1463 mm = m_copy(mb0, 0, hlen); 1464 if (mm == NULL) 1465 goto fail1; 1466 1467 /* 1468 * Send message to routing daemon to install 1469 * a route into the kernel table 1470 */ 1471 1472 im = mtod(mm, struct igmpmsg *); 1473 im->im_msgtype = IGMPMSG_NOCACHE; 1474 im->im_mbz = 0; 1475 im->im_vif = vifi; 1476 1477 mrtstat.mrts_upcalls++; 1478 1479 k_igmpsrc.sin_addr = ip->ip_src; 1480 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1481 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1482 ++mrtstat.mrts_upq_sockfull; 1483fail1: 1484 free(rt, M_MRTABLE); 1485fail: 1486 free(rte, M_MRTABLE); 1487 m_freem(mb0); 1488 MFC_UNLOCK(); 1489 VIF_UNLOCK(); 1490 return ENOBUFS; 1491 } 1492 1493 /* insert new entry at head of hash chain */ 1494 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1495 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1496 rt->mfc_expire = UPCALL_EXPIRE; 1497 nexpire[hash]++; 1498 for (i = 0; i < numvifs; i++) { 1499 rt->mfc_ttls[i] = 0; 1500 rt->mfc_flags[i] = 0; 1501 } 1502 rt->mfc_parent = -1; 1503 1504 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1505 1506 rt->mfc_bw_meter = NULL; 1507 1508 /* link into table */ 1509 rt->mfc_next = mfctable[hash]; 1510 mfctable[hash] = rt; 1511 rt->mfc_stall = rte; 1512 1513 } else { 1514 /* determine if q has overflowed */ 1515 int npkts = 0; 1516 struct rtdetq **p; 1517 1518 /* 1519 * XXX ouch! we need to append to the list, but we 1520 * only have a pointer to the front, so we have to 1521 * scan the entire list every time. 1522 */ 1523 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1524 npkts++; 1525 1526 if (npkts > MAX_UPQ) { 1527 mrtstat.mrts_upq_ovflw++; 1528non_fatal: 1529 free(rte, M_MRTABLE); 1530 m_freem(mb0); 1531 MFC_UNLOCK(); 1532 VIF_UNLOCK(); 1533 return 0; 1534 } 1535 1536 /* Add this entry to the end of the queue */ 1537 *p = rte; 1538 } 1539 1540 rte->m = mb0; 1541 rte->ifp = ifp; 1542 rte->next = NULL; 1543 1544 MFC_UNLOCK(); 1545 VIF_UNLOCK(); 1546 1547 return 0; 1548 } 1549} 1550 1551/* 1552 * Clean up the cache entry if upcall is not serviced 1553 */ 1554static void 1555expire_upcalls(void *unused) 1556{ 1557 struct rtdetq *rte; 1558 struct mfc *mfc, **nptr; 1559 int i; 1560 1561 MFC_LOCK(); 1562 for (i = 0; i < MFCTBLSIZ; i++) { 1563 if (nexpire[i] == 0) 1564 continue; 1565 nptr = &mfctable[i]; 1566 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1567 /* 1568 * Skip real cache entries 1569 * Make sure it wasn't marked to not expire (shouldn't happen) 1570 * If it expires now 1571 */ 1572 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1573 --mfc->mfc_expire == 0) { 1574 if (mrtdebug & DEBUG_EXPIRE) 1575 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1576 (u_long)ntohl(mfc->mfc_origin.s_addr), 1577 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1578 /* 1579 * drop all the packets 1580 * free the mbuf with the pkt, if, timing info 1581 */ 1582 for (rte = mfc->mfc_stall; rte; ) { 1583 struct rtdetq *n = rte->next; 1584 1585 m_freem(rte->m); 1586 free(rte, M_MRTABLE); 1587 rte = n; 1588 } 1589 ++mrtstat.mrts_cache_cleanups; 1590 nexpire[i]--; 1591 1592 /* 1593 * free the bw_meter entries 1594 */ 1595 while (mfc->mfc_bw_meter != NULL) { 1596 struct bw_meter *x = mfc->mfc_bw_meter; 1597 1598 mfc->mfc_bw_meter = x->bm_mfc_next; 1599 free(x, M_BWMETER); 1600 } 1601 1602 *nptr = mfc->mfc_next; 1603 free(mfc, M_MRTABLE); 1604 } else { 1605 nptr = &mfc->mfc_next; 1606 } 1607 } 1608 } 1609 MFC_UNLOCK(); 1610 1611 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1612} 1613 1614/* 1615 * Packet forwarding routine once entry in the cache is made 1616 */ 1617static int 1618ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1619{ 1620 struct ip *ip = mtod(m, struct ip *); 1621 vifi_t vifi; 1622 int plen = ip->ip_len; 1623 1624 VIF_LOCK_ASSERT(); 1625/* 1626 * Macro to send packet on vif. Since RSVP packets don't get counted on 1627 * input, they shouldn't get counted on output, so statistics keeping is 1628 * separate. 1629 */ 1630#define MC_SEND(ip,vifp,m) { \ 1631 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1632 encap_send((ip), (vifp), (m)); \ 1633 else \ 1634 phyint_send((ip), (vifp), (m)); \ 1635} 1636 1637 /* 1638 * If xmt_vif is not -1, send on only the requested vif. 1639 * 1640 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1641 */ 1642 if (xmt_vif < numvifs) { 1643#ifdef PIM 1644 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1645 pim_register_send(ip, viftable + xmt_vif, m, rt); 1646 else 1647#endif 1648 MC_SEND(ip, viftable + xmt_vif, m); 1649 return 1; 1650 } 1651 1652 /* 1653 * Don't forward if it didn't arrive from the parent vif for its origin. 1654 */ 1655 vifi = rt->mfc_parent; 1656 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1657 /* came in the wrong interface */ 1658 if (mrtdebug & DEBUG_FORWARD) 1659 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1660 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1661 ++mrtstat.mrts_wrong_if; 1662 ++rt->mfc_wrong_if; 1663 /* 1664 * If we are doing PIM assert processing, send a message 1665 * to the routing daemon. 1666 * 1667 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1668 * can complete the SPT switch, regardless of the type 1669 * of the iif (broadcast media, GRE tunnel, etc). 1670 */ 1671 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1672 struct timeval now; 1673 u_long delta; 1674 1675#ifdef PIM 1676 if (ifp == &multicast_register_if) 1677 pimstat.pims_rcv_registers_wrongiif++; 1678#endif 1679 1680 /* Get vifi for the incoming packet */ 1681 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1682 ; 1683 if (vifi >= numvifs) 1684 return 0; /* The iif is not found: ignore the packet. */ 1685 1686 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1687 return 0; /* WRONGVIF disabled: ignore the packet */ 1688 1689 GET_TIME(now); 1690 1691 TV_DELTA(rt->mfc_last_assert, now, delta); 1692 1693 if (delta > ASSERT_MSG_TIME) { 1694 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1695 struct igmpmsg *im; 1696 int hlen = ip->ip_hl << 2; 1697 struct mbuf *mm = m_copy(m, 0, hlen); 1698 1699 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1700 mm = m_pullup(mm, hlen); 1701 if (mm == NULL) 1702 return ENOBUFS; 1703 1704 rt->mfc_last_assert = now; 1705 1706 im = mtod(mm, struct igmpmsg *); 1707 im->im_msgtype = IGMPMSG_WRONGVIF; 1708 im->im_mbz = 0; 1709 im->im_vif = vifi; 1710 1711 mrtstat.mrts_upcalls++; 1712 1713 k_igmpsrc.sin_addr = im->im_src; 1714 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1715 log(LOG_WARNING, 1716 "ip_mforward: ip_mrouter socket queue full\n"); 1717 ++mrtstat.mrts_upq_sockfull; 1718 return ENOBUFS; 1719 } 1720 } 1721 } 1722 return 0; 1723 } 1724 1725 /* If I sourced this packet, it counts as output, else it was input. */ 1726 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1727 viftable[vifi].v_pkt_out++; 1728 viftable[vifi].v_bytes_out += plen; 1729 } else { 1730 viftable[vifi].v_pkt_in++; 1731 viftable[vifi].v_bytes_in += plen; 1732 } 1733 rt->mfc_pkt_cnt++; 1734 rt->mfc_byte_cnt += plen; 1735 1736 /* 1737 * For each vif, decide if a copy of the packet should be forwarded. 1738 * Forward if: 1739 * - the ttl exceeds the vif's threshold 1740 * - there are group members downstream on interface 1741 */ 1742 for (vifi = 0; vifi < numvifs; vifi++) 1743 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1744 viftable[vifi].v_pkt_out++; 1745 viftable[vifi].v_bytes_out += plen; 1746#ifdef PIM 1747 if (viftable[vifi].v_flags & VIFF_REGISTER) 1748 pim_register_send(ip, viftable + vifi, m, rt); 1749 else 1750#endif 1751 MC_SEND(ip, viftable+vifi, m); 1752 } 1753 1754 /* 1755 * Perform upcall-related bw measuring. 1756 */ 1757 if (rt->mfc_bw_meter != NULL) { 1758 struct bw_meter *x; 1759 struct timeval now; 1760 1761 GET_TIME(now); 1762 MFC_LOCK_ASSERT(); 1763 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1764 bw_meter_receive_packet(x, plen, &now); 1765 } 1766 1767 return 0; 1768} 1769 1770/* 1771 * check if a vif number is legal/ok. This is used by ip_output. 1772 */ 1773static int 1774X_legal_vif_num(int vif) 1775{ 1776 /* XXX unlocked, matter? */ 1777 return (vif >= 0 && vif < numvifs); 1778} 1779 1780/* 1781 * Return the local address used by this vif 1782 */ 1783static u_long 1784X_ip_mcast_src(int vifi) 1785{ 1786 /* XXX unlocked, matter? */ 1787 if (vifi >= 0 && vifi < numvifs) 1788 return viftable[vifi].v_lcl_addr.s_addr; 1789 else 1790 return INADDR_ANY; 1791} 1792 1793static void 1794phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1795{ 1796 struct mbuf *mb_copy; 1797 int hlen = ip->ip_hl << 2; 1798 1799 VIF_LOCK_ASSERT(); 1800 1801 /* 1802 * Make a new reference to the packet; make sure that 1803 * the IP header is actually copied, not just referenced, 1804 * so that ip_output() only scribbles on the copy. 1805 */ 1806 mb_copy = m_copypacket(m, M_DONTWAIT); 1807 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1808 mb_copy = m_pullup(mb_copy, hlen); 1809 if (mb_copy == NULL) 1810 return; 1811 1812 if (vifp->v_rate_limit == 0) 1813 tbf_send_packet(vifp, mb_copy); 1814 else 1815 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1816} 1817 1818static void 1819encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1820{ 1821 struct mbuf *mb_copy; 1822 struct ip *ip_copy; 1823 int i, len = ip->ip_len; 1824 1825 VIF_LOCK_ASSERT(); 1826 1827 /* Take care of delayed checksums */ 1828 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1829 in_delayed_cksum(m); 1830 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1831 } 1832 1833 /* 1834 * copy the old packet & pullup its IP header into the 1835 * new mbuf so we can modify it. Try to fill the new 1836 * mbuf since if we don't the ethernet driver will. 1837 */ 1838 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1839 if (mb_copy == NULL) 1840 return; 1841#ifdef MAC 1842 mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy); 1843#endif 1844 mb_copy->m_data += max_linkhdr; 1845 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1846 1847 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) { 1848 m_freem(mb_copy); 1849 return; 1850 } 1851 i = MHLEN - M_LEADINGSPACE(mb_copy); 1852 if (i > len) 1853 i = len; 1854 mb_copy = m_pullup(mb_copy, i); 1855 if (mb_copy == NULL) 1856 return; 1857 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1858 1859 /* 1860 * fill in the encapsulating IP header. 1861 */ 1862 ip_copy = mtod(mb_copy, struct ip *); 1863 *ip_copy = multicast_encap_iphdr; 1864#ifdef RANDOM_IP_ID 1865 ip_copy->ip_id = ip_randomid(); 1866#else 1867 ip_copy->ip_id = htons(ip_id++); 1868#endif 1869 ip_copy->ip_len += len; 1870 ip_copy->ip_src = vifp->v_lcl_addr; 1871 ip_copy->ip_dst = vifp->v_rmt_addr; 1872 1873 /* 1874 * turn the encapsulated IP header back into a valid one. 1875 */ 1876 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1877 --ip->ip_ttl; 1878 ip->ip_len = htons(ip->ip_len); 1879 ip->ip_off = htons(ip->ip_off); 1880 ip->ip_sum = 0; 1881 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1882 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1883 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1884 1885 if (vifp->v_rate_limit == 0) 1886 tbf_send_packet(vifp, mb_copy); 1887 else 1888 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1889} 1890 1891/* 1892 * Token bucket filter module 1893 */ 1894 1895static void 1896tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len) 1897{ 1898 struct tbf *t = vifp->v_tbf; 1899 1900 VIF_LOCK_ASSERT(); 1901 1902 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */ 1903 mrtstat.mrts_pkt2large++; 1904 m_freem(m); 1905 return; 1906 } 1907 1908 tbf_update_tokens(vifp); 1909 1910 if (t->tbf_q_len == 0) { /* queue empty... */ 1911 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */ 1912 t->tbf_n_tok -= p_len; 1913 tbf_send_packet(vifp, m); 1914 } else { /* no, queue packet and try later */ 1915 tbf_queue(vifp, m); 1916 callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, 1917 tbf_reprocess_q, vifp); 1918 } 1919 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1920 /* finite queue length, so queue pkts and process queue */ 1921 tbf_queue(vifp, m); 1922 tbf_process_q(vifp); 1923 } else { 1924 /* queue full, try to dq and queue and process */ 1925 if (!tbf_dq_sel(vifp, ip)) { 1926 mrtstat.mrts_q_overflow++; 1927 m_freem(m); 1928 } else { 1929 tbf_queue(vifp, m); 1930 tbf_process_q(vifp); 1931 } 1932 } 1933} 1934 1935/* 1936 * adds a packet to the queue at the interface 1937 */ 1938static void 1939tbf_queue(struct vif *vifp, struct mbuf *m) 1940{ 1941 struct tbf *t = vifp->v_tbf; 1942 1943 VIF_LOCK_ASSERT(); 1944 1945 if (t->tbf_t == NULL) /* Queue was empty */ 1946 t->tbf_q = m; 1947 else /* Insert at tail */ 1948 t->tbf_t->m_act = m; 1949 1950 t->tbf_t = m; /* Set new tail pointer */ 1951 1952#ifdef DIAGNOSTIC 1953 /* Make sure we didn't get fed a bogus mbuf */ 1954 if (m->m_act) 1955 panic("tbf_queue: m_act"); 1956#endif 1957 m->m_act = NULL; 1958 1959 t->tbf_q_len++; 1960} 1961 1962/* 1963 * processes the queue at the interface 1964 */ 1965static void 1966tbf_process_q(struct vif *vifp) 1967{ 1968 struct tbf *t = vifp->v_tbf; 1969 1970 VIF_LOCK_ASSERT(); 1971 1972 /* loop through the queue at the interface and send as many packets 1973 * as possible 1974 */ 1975 while (t->tbf_q_len > 0) { 1976 struct mbuf *m = t->tbf_q; 1977 int len = mtod(m, struct ip *)->ip_len; 1978 1979 /* determine if the packet can be sent */ 1980 if (len > t->tbf_n_tok) /* not enough tokens, we are done */ 1981 break; 1982 /* ok, reduce no of tokens, dequeue and send the packet. */ 1983 t->tbf_n_tok -= len; 1984 1985 t->tbf_q = m->m_act; 1986 if (--t->tbf_q_len == 0) 1987 t->tbf_t = NULL; 1988 1989 m->m_act = NULL; 1990 tbf_send_packet(vifp, m); 1991 } 1992} 1993 1994static void 1995tbf_reprocess_q(void *xvifp) 1996{ 1997 struct vif *vifp = xvifp; 1998 1999 if (ip_mrouter == NULL) 2000 return; 2001 VIF_LOCK(); 2002 tbf_update_tokens(vifp); 2003 tbf_process_q(vifp); 2004 if (vifp->v_tbf->tbf_q_len) 2005 callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, tbf_reprocess_q, vifp); 2006 VIF_UNLOCK(); 2007} 2008 2009/* function that will selectively discard a member of the queue 2010 * based on the precedence value and the priority 2011 */ 2012static int 2013tbf_dq_sel(struct vif *vifp, struct ip *ip) 2014{ 2015 u_int p; 2016 struct mbuf *m, *last; 2017 struct mbuf **np; 2018 struct tbf *t = vifp->v_tbf; 2019 2020 VIF_LOCK_ASSERT(); 2021 2022 p = priority(vifp, ip); 2023 2024 np = &t->tbf_q; 2025 last = NULL; 2026 while ((m = *np) != NULL) { 2027 if (p > priority(vifp, mtod(m, struct ip *))) { 2028 *np = m->m_act; 2029 /* If we're removing the last packet, fix the tail pointer */ 2030 if (m == t->tbf_t) 2031 t->tbf_t = last; 2032 m_freem(m); 2033 /* It's impossible for the queue to be empty, but check anyways. */ 2034 if (--t->tbf_q_len == 0) 2035 t->tbf_t = NULL; 2036 mrtstat.mrts_drop_sel++; 2037 return 1; 2038 } 2039 np = &m->m_act; 2040 last = m; 2041 } 2042 return 0; 2043} 2044 2045static void 2046tbf_send_packet(struct vif *vifp, struct mbuf *m) 2047{ 2048 VIF_LOCK_ASSERT(); 2049 2050 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */ 2051 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 2052 else { 2053 struct ip_moptions imo; 2054 int error; 2055 static struct route ro; /* XXX check this */ 2056 2057 imo.imo_multicast_ifp = vifp->v_ifp; 2058 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 2059 imo.imo_multicast_loop = 1; 2060 imo.imo_multicast_vif = -1; 2061 2062 /* 2063 * Re-entrancy should not be a problem here, because 2064 * the packets that we send out and are looped back at us 2065 * should get rejected because they appear to come from 2066 * the loopback interface, thus preventing looping. 2067 */ 2068 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL); 2069 2070 if (mrtdebug & DEBUG_XMIT) 2071 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 2072 (int)(vifp - viftable), error); 2073 } 2074} 2075 2076/* determine the current time and then 2077 * the elapsed time (between the last time and time now) 2078 * in milliseconds & update the no. of tokens in the bucket 2079 */ 2080static void 2081tbf_update_tokens(struct vif *vifp) 2082{ 2083 struct timeval tp; 2084 u_long tm; 2085 struct tbf *t = vifp->v_tbf; 2086 2087 VIF_LOCK_ASSERT(); 2088 2089 GET_TIME(tp); 2090 2091 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 2092 2093 /* 2094 * This formula is actually 2095 * "time in seconds" * "bytes/second". 2096 * 2097 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2098 * 2099 * The (1000/1024) was introduced in add_vif to optimize 2100 * this divide into a shift. 2101 */ 2102 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 2103 t->tbf_last_pkt_t = tp; 2104 2105 if (t->tbf_n_tok > MAX_BKT_SIZE) 2106 t->tbf_n_tok = MAX_BKT_SIZE; 2107} 2108 2109static int 2110priority(struct vif *vifp, struct ip *ip) 2111{ 2112 int prio = 50; /* the lowest priority -- default case */ 2113 2114 /* temporary hack; may add general packet classifier some day */ 2115 2116 /* 2117 * The UDP port space is divided up into four priority ranges: 2118 * [0, 16384) : unclassified - lowest priority 2119 * [16384, 32768) : audio - highest priority 2120 * [32768, 49152) : whiteboard - medium priority 2121 * [49152, 65536) : video - low priority 2122 * 2123 * Everything else gets lowest priority. 2124 */ 2125 if (ip->ip_p == IPPROTO_UDP) { 2126 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2127 switch (ntohs(udp->uh_dport) & 0xc000) { 2128 case 0x4000: 2129 prio = 70; 2130 break; 2131 case 0x8000: 2132 prio = 60; 2133 break; 2134 case 0xc000: 2135 prio = 55; 2136 break; 2137 } 2138 } 2139 return prio; 2140} 2141 2142/* 2143 * End of token bucket filter modifications 2144 */ 2145 2146static int 2147X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 2148{ 2149 int error, vifi; 2150 2151 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2152 return EOPNOTSUPP; 2153 2154 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 2155 if (error) 2156 return error; 2157 2158 VIF_LOCK(); 2159 2160 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 2161 VIF_UNLOCK(); 2162 return EADDRNOTAVAIL; 2163 } 2164 2165 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 2166 /* Check if socket is available. */ 2167 if (viftable[vifi].v_rsvpd != NULL) { 2168 VIF_UNLOCK(); 2169 return EADDRINUSE; 2170 } 2171 2172 viftable[vifi].v_rsvpd = so; 2173 /* This may seem silly, but we need to be sure we don't over-increment 2174 * the RSVP counter, in case something slips up. 2175 */ 2176 if (!viftable[vifi].v_rsvp_on) { 2177 viftable[vifi].v_rsvp_on = 1; 2178 rsvp_on++; 2179 } 2180 } else { /* must be VIF_OFF */ 2181 /* 2182 * XXX as an additional consistency check, one could make sure 2183 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 2184 * first parameter is pretty useless. 2185 */ 2186 viftable[vifi].v_rsvpd = NULL; 2187 /* 2188 * This may seem silly, but we need to be sure we don't over-decrement 2189 * the RSVP counter, in case something slips up. 2190 */ 2191 if (viftable[vifi].v_rsvp_on) { 2192 viftable[vifi].v_rsvp_on = 0; 2193 rsvp_on--; 2194 } 2195 } 2196 VIF_UNLOCK(); 2197 return 0; 2198} 2199 2200static void 2201X_ip_rsvp_force_done(struct socket *so) 2202{ 2203 int vifi; 2204 2205 /* Don't bother if it is not the right type of socket. */ 2206 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2207 return; 2208 2209 VIF_LOCK(); 2210 2211 /* The socket may be attached to more than one vif...this 2212 * is perfectly legal. 2213 */ 2214 for (vifi = 0; vifi < numvifs; vifi++) { 2215 if (viftable[vifi].v_rsvpd == so) { 2216 viftable[vifi].v_rsvpd = NULL; 2217 /* This may seem silly, but we need to be sure we don't 2218 * over-decrement the RSVP counter, in case something slips up. 2219 */ 2220 if (viftable[vifi].v_rsvp_on) { 2221 viftable[vifi].v_rsvp_on = 0; 2222 rsvp_on--; 2223 } 2224 } 2225 } 2226 2227 VIF_UNLOCK(); 2228} 2229 2230static void 2231X_rsvp_input(struct mbuf *m, int off) 2232{ 2233 int vifi; 2234 struct ip *ip = mtod(m, struct ip *); 2235 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2236 struct ifnet *ifp; 2237 2238 if (rsvpdebug) 2239 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2240 2241 /* Can still get packets with rsvp_on = 0 if there is a local member 2242 * of the group to which the RSVP packet is addressed. But in this 2243 * case we want to throw the packet away. 2244 */ 2245 if (!rsvp_on) { 2246 m_freem(m); 2247 return; 2248 } 2249 2250 if (rsvpdebug) 2251 printf("rsvp_input: check vifs\n"); 2252 2253#ifdef DIAGNOSTIC 2254 M_ASSERTPKTHDR(m); 2255#endif 2256 2257 ifp = m->m_pkthdr.rcvif; 2258 2259 VIF_LOCK(); 2260 /* Find which vif the packet arrived on. */ 2261 for (vifi = 0; vifi < numvifs; vifi++) 2262 if (viftable[vifi].v_ifp == ifp) 2263 break; 2264 2265 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2266 /* 2267 * Drop the lock here to avoid holding it across rip_input. 2268 * This could make rsvpdebug printfs wrong. If you care, 2269 * record the state of stuff before dropping the lock. 2270 */ 2271 VIF_UNLOCK(); 2272 /* 2273 * If the old-style non-vif-associated socket is set, 2274 * then use it. Otherwise, drop packet since there 2275 * is no specific socket for this vif. 2276 */ 2277 if (ip_rsvpd != NULL) { 2278 if (rsvpdebug) 2279 printf("rsvp_input: Sending packet up old-style socket\n"); 2280 rip_input(m, off); /* xxx */ 2281 } else { 2282 if (rsvpdebug && vifi == numvifs) 2283 printf("rsvp_input: Can't find vif for packet.\n"); 2284 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2285 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2286 m_freem(m); 2287 } 2288 return; 2289 } 2290 rsvp_src.sin_addr = ip->ip_src; 2291 2292 if (rsvpdebug && m) 2293 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2294 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2295 2296 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2297 if (rsvpdebug) 2298 printf("rsvp_input: Failed to append to socket\n"); 2299 } else { 2300 if (rsvpdebug) 2301 printf("rsvp_input: send packet up\n"); 2302 } 2303 VIF_UNLOCK(); 2304} 2305 2306/* 2307 * Code for bandwidth monitors 2308 */ 2309 2310/* 2311 * Define common interface for timeval-related methods 2312 */ 2313#define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 2314#define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 2315#define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 2316 2317static uint32_t 2318compute_bw_meter_flags(struct bw_upcall *req) 2319{ 2320 uint32_t flags = 0; 2321 2322 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2323 flags |= BW_METER_UNIT_PACKETS; 2324 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2325 flags |= BW_METER_UNIT_BYTES; 2326 if (req->bu_flags & BW_UPCALL_GEQ) 2327 flags |= BW_METER_GEQ; 2328 if (req->bu_flags & BW_UPCALL_LEQ) 2329 flags |= BW_METER_LEQ; 2330 2331 return flags; 2332} 2333 2334/* 2335 * Add a bw_meter entry 2336 */ 2337static int 2338add_bw_upcall(struct bw_upcall *req) 2339{ 2340 struct mfc *mfc; 2341 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2342 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2343 struct timeval now; 2344 struct bw_meter *x; 2345 uint32_t flags; 2346 2347 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2348 return EOPNOTSUPP; 2349 2350 /* Test if the flags are valid */ 2351 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2352 return EINVAL; 2353 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2354 return EINVAL; 2355 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2356 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2357 return EINVAL; 2358 2359 /* Test if the threshold time interval is valid */ 2360 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2361 return EINVAL; 2362 2363 flags = compute_bw_meter_flags(req); 2364 2365 /* 2366 * Find if we have already same bw_meter entry 2367 */ 2368 MFC_LOCK(); 2369 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2370 if (mfc == NULL) { 2371 MFC_UNLOCK(); 2372 return EADDRNOTAVAIL; 2373 } 2374 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2375 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2376 &req->bu_threshold.b_time, ==)) && 2377 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2378 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2379 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2380 MFC_UNLOCK(); 2381 return 0; /* XXX Already installed */ 2382 } 2383 } 2384 2385 /* Allocate the new bw_meter entry */ 2386 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2387 if (x == NULL) { 2388 MFC_UNLOCK(); 2389 return ENOBUFS; 2390 } 2391 2392 /* Set the new bw_meter entry */ 2393 x->bm_threshold.b_time = req->bu_threshold.b_time; 2394 GET_TIME(now); 2395 x->bm_start_time = now; 2396 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2397 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2398 x->bm_measured.b_packets = 0; 2399 x->bm_measured.b_bytes = 0; 2400 x->bm_flags = flags; 2401 x->bm_time_next = NULL; 2402 x->bm_time_hash = BW_METER_BUCKETS; 2403 2404 /* Add the new bw_meter entry to the front of entries for this MFC */ 2405 x->bm_mfc = mfc; 2406 x->bm_mfc_next = mfc->mfc_bw_meter; 2407 mfc->mfc_bw_meter = x; 2408 schedule_bw_meter(x, &now); 2409 MFC_UNLOCK(); 2410 2411 return 0; 2412} 2413 2414static void 2415free_bw_list(struct bw_meter *list) 2416{ 2417 while (list != NULL) { 2418 struct bw_meter *x = list; 2419 2420 list = list->bm_mfc_next; 2421 unschedule_bw_meter(x); 2422 free(x, M_BWMETER); 2423 } 2424} 2425 2426/* 2427 * Delete one or multiple bw_meter entries 2428 */ 2429static int 2430del_bw_upcall(struct bw_upcall *req) 2431{ 2432 struct mfc *mfc; 2433 struct bw_meter *x; 2434 2435 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2436 return EOPNOTSUPP; 2437 2438 MFC_LOCK(); 2439 /* Find the corresponding MFC entry */ 2440 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2441 if (mfc == NULL) { 2442 MFC_UNLOCK(); 2443 return EADDRNOTAVAIL; 2444 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2445 /* 2446 * Delete all bw_meter entries for this mfc 2447 */ 2448 struct bw_meter *list; 2449 2450 list = mfc->mfc_bw_meter; 2451 mfc->mfc_bw_meter = NULL; 2452 free_bw_list(list); 2453 MFC_UNLOCK(); 2454 return 0; 2455 } else { /* Delete a single bw_meter entry */ 2456 struct bw_meter *prev; 2457 uint32_t flags = 0; 2458 2459 flags = compute_bw_meter_flags(req); 2460 2461 /* Find the bw_meter entry to delete */ 2462 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2463 x = x->bm_mfc_next) { 2464 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2465 &req->bu_threshold.b_time, ==)) && 2466 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2467 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2468 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2469 break; 2470 } 2471 if (x != NULL) { /* Delete entry from the list for this MFC */ 2472 if (prev != NULL) 2473 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2474 else 2475 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2476 2477 unschedule_bw_meter(x); 2478 MFC_UNLOCK(); 2479 /* Free the bw_meter entry */ 2480 free(x, M_BWMETER); 2481 return 0; 2482 } else { 2483 MFC_UNLOCK(); 2484 return EINVAL; 2485 } 2486 } 2487 /* NOTREACHED */ 2488} 2489 2490/* 2491 * Perform bandwidth measurement processing that may result in an upcall 2492 */ 2493static void 2494bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2495{ 2496 struct timeval delta; 2497 2498 MFC_LOCK_ASSERT(); 2499 2500 delta = *nowp; 2501 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2502 2503 if (x->bm_flags & BW_METER_GEQ) { 2504 /* 2505 * Processing for ">=" type of bw_meter entry 2506 */ 2507 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2508 /* Reset the bw_meter entry */ 2509 x->bm_start_time = *nowp; 2510 x->bm_measured.b_packets = 0; 2511 x->bm_measured.b_bytes = 0; 2512 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2513 } 2514 2515 /* Record that a packet is received */ 2516 x->bm_measured.b_packets++; 2517 x->bm_measured.b_bytes += plen; 2518 2519 /* 2520 * Test if we should deliver an upcall 2521 */ 2522 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2523 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2524 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2525 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2526 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2527 /* Prepare an upcall for delivery */ 2528 bw_meter_prepare_upcall(x, nowp); 2529 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2530 } 2531 } 2532 } else if (x->bm_flags & BW_METER_LEQ) { 2533 /* 2534 * Processing for "<=" type of bw_meter entry 2535 */ 2536 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2537 /* 2538 * We are behind time with the multicast forwarding table 2539 * scanning for "<=" type of bw_meter entries, so test now 2540 * if we should deliver an upcall. 2541 */ 2542 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2543 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2544 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2545 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2546 /* Prepare an upcall for delivery */ 2547 bw_meter_prepare_upcall(x, nowp); 2548 } 2549 /* Reschedule the bw_meter entry */ 2550 unschedule_bw_meter(x); 2551 schedule_bw_meter(x, nowp); 2552 } 2553 2554 /* Record that a packet is received */ 2555 x->bm_measured.b_packets++; 2556 x->bm_measured.b_bytes += plen; 2557 2558 /* 2559 * Test if we should restart the measuring interval 2560 */ 2561 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2562 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2563 (x->bm_flags & BW_METER_UNIT_BYTES && 2564 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2565 /* Don't restart the measuring interval */ 2566 } else { 2567 /* Do restart the measuring interval */ 2568 /* 2569 * XXX: note that we don't unschedule and schedule, because this 2570 * might be too much overhead per packet. Instead, when we process 2571 * all entries for a given timer hash bin, we check whether it is 2572 * really a timeout. If not, we reschedule at that time. 2573 */ 2574 x->bm_start_time = *nowp; 2575 x->bm_measured.b_packets = 0; 2576 x->bm_measured.b_bytes = 0; 2577 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2578 } 2579 } 2580} 2581 2582/* 2583 * Prepare a bandwidth-related upcall 2584 */ 2585static void 2586bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2587{ 2588 struct timeval delta; 2589 struct bw_upcall *u; 2590 2591 MFC_LOCK_ASSERT(); 2592 2593 /* 2594 * Compute the measured time interval 2595 */ 2596 delta = *nowp; 2597 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2598 2599 /* 2600 * If there are too many pending upcalls, deliver them now 2601 */ 2602 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2603 bw_upcalls_send(); 2604 2605 /* 2606 * Set the bw_upcall entry 2607 */ 2608 u = &bw_upcalls[bw_upcalls_n++]; 2609 u->bu_src = x->bm_mfc->mfc_origin; 2610 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2611 u->bu_threshold.b_time = x->bm_threshold.b_time; 2612 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2613 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2614 u->bu_measured.b_time = delta; 2615 u->bu_measured.b_packets = x->bm_measured.b_packets; 2616 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2617 u->bu_flags = 0; 2618 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2619 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2620 if (x->bm_flags & BW_METER_UNIT_BYTES) 2621 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2622 if (x->bm_flags & BW_METER_GEQ) 2623 u->bu_flags |= BW_UPCALL_GEQ; 2624 if (x->bm_flags & BW_METER_LEQ) 2625 u->bu_flags |= BW_UPCALL_LEQ; 2626} 2627 2628/* 2629 * Send the pending bandwidth-related upcalls 2630 */ 2631static void 2632bw_upcalls_send(void) 2633{ 2634 struct mbuf *m; 2635 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2636 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2637 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2638 0, /* unused2 */ 2639 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2640 0, /* im_mbz */ 2641 0, /* im_vif */ 2642 0, /* unused3 */ 2643 { 0 }, /* im_src */ 2644 { 0 } }; /* im_dst */ 2645 2646 MFC_LOCK_ASSERT(); 2647 2648 if (bw_upcalls_n == 0) 2649 return; /* No pending upcalls */ 2650 2651 bw_upcalls_n = 0; 2652 2653 /* 2654 * Allocate a new mbuf, initialize it with the header and 2655 * the payload for the pending calls. 2656 */ 2657 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2658 if (m == NULL) { 2659 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2660 return; 2661 } 2662 2663 m->m_len = m->m_pkthdr.len = 0; 2664 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2665 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2666 2667 /* 2668 * Send the upcalls 2669 * XXX do we need to set the address in k_igmpsrc ? 2670 */ 2671 mrtstat.mrts_upcalls++; 2672 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2673 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2674 ++mrtstat.mrts_upq_sockfull; 2675 } 2676} 2677 2678/* 2679 * Compute the timeout hash value for the bw_meter entries 2680 */ 2681#define BW_METER_TIMEHASH(bw_meter, hash) \ 2682 do { \ 2683 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2684 \ 2685 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2686 (hash) = next_timeval.tv_sec; \ 2687 if (next_timeval.tv_usec) \ 2688 (hash)++; /* XXX: make sure we don't timeout early */ \ 2689 (hash) %= BW_METER_BUCKETS; \ 2690 } while (0) 2691 2692/* 2693 * Schedule a timer to process periodically bw_meter entry of type "<=" 2694 * by linking the entry in the proper hash bucket. 2695 */ 2696static void 2697schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2698{ 2699 int time_hash; 2700 2701 MFC_LOCK_ASSERT(); 2702 2703 if (!(x->bm_flags & BW_METER_LEQ)) 2704 return; /* XXX: we schedule timers only for "<=" entries */ 2705 2706 /* 2707 * Reset the bw_meter entry 2708 */ 2709 x->bm_start_time = *nowp; 2710 x->bm_measured.b_packets = 0; 2711 x->bm_measured.b_bytes = 0; 2712 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2713 2714 /* 2715 * Compute the timeout hash value and insert the entry 2716 */ 2717 BW_METER_TIMEHASH(x, time_hash); 2718 x->bm_time_next = bw_meter_timers[time_hash]; 2719 bw_meter_timers[time_hash] = x; 2720 x->bm_time_hash = time_hash; 2721} 2722 2723/* 2724 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2725 * by removing the entry from the proper hash bucket. 2726 */ 2727static void 2728unschedule_bw_meter(struct bw_meter *x) 2729{ 2730 int time_hash; 2731 struct bw_meter *prev, *tmp; 2732 2733 MFC_LOCK_ASSERT(); 2734 2735 if (!(x->bm_flags & BW_METER_LEQ)) 2736 return; /* XXX: we schedule timers only for "<=" entries */ 2737 2738 /* 2739 * Compute the timeout hash value and delete the entry 2740 */ 2741 time_hash = x->bm_time_hash; 2742 if (time_hash >= BW_METER_BUCKETS) 2743 return; /* Entry was not scheduled */ 2744 2745 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2746 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2747 if (tmp == x) 2748 break; 2749 2750 if (tmp == NULL) 2751 panic("unschedule_bw_meter: bw_meter entry not found"); 2752 2753 if (prev != NULL) 2754 prev->bm_time_next = x->bm_time_next; 2755 else 2756 bw_meter_timers[time_hash] = x->bm_time_next; 2757 2758 x->bm_time_next = NULL; 2759 x->bm_time_hash = BW_METER_BUCKETS; 2760} 2761 2762 2763/* 2764 * Process all "<=" type of bw_meter that should be processed now, 2765 * and for each entry prepare an upcall if necessary. Each processed 2766 * entry is rescheduled again for the (periodic) processing. 2767 * 2768 * This is run periodically (once per second normally). On each round, 2769 * all the potentially matching entries are in the hash slot that we are 2770 * looking at. 2771 */ 2772static void 2773bw_meter_process() 2774{ 2775 static uint32_t last_tv_sec; /* last time we processed this */ 2776 2777 uint32_t loops; 2778 int i; 2779 struct timeval now, process_endtime; 2780 2781 GET_TIME(now); 2782 if (last_tv_sec == now.tv_sec) 2783 return; /* nothing to do */ 2784 2785 loops = now.tv_sec - last_tv_sec; 2786 last_tv_sec = now.tv_sec; 2787 if (loops > BW_METER_BUCKETS) 2788 loops = BW_METER_BUCKETS; 2789 2790 MFC_LOCK(); 2791 /* 2792 * Process all bins of bw_meter entries from the one after the last 2793 * processed to the current one. On entry, i points to the last bucket 2794 * visited, so we need to increment i at the beginning of the loop. 2795 */ 2796 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2797 struct bw_meter *x, *tmp_list; 2798 2799 if (++i >= BW_METER_BUCKETS) 2800 i = 0; 2801 2802 /* Disconnect the list of bw_meter entries from the bin */ 2803 tmp_list = bw_meter_timers[i]; 2804 bw_meter_timers[i] = NULL; 2805 2806 /* Process the list of bw_meter entries */ 2807 while (tmp_list != NULL) { 2808 x = tmp_list; 2809 tmp_list = tmp_list->bm_time_next; 2810 2811 /* Test if the time interval is over */ 2812 process_endtime = x->bm_start_time; 2813 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2814 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2815 /* Not yet: reschedule, but don't reset */ 2816 int time_hash; 2817 2818 BW_METER_TIMEHASH(x, time_hash); 2819 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2820 /* 2821 * XXX: somehow the bin processing is a bit ahead of time. 2822 * Put the entry in the next bin. 2823 */ 2824 if (++time_hash >= BW_METER_BUCKETS) 2825 time_hash = 0; 2826 } 2827 x->bm_time_next = bw_meter_timers[time_hash]; 2828 bw_meter_timers[time_hash] = x; 2829 x->bm_time_hash = time_hash; 2830 2831 continue; 2832 } 2833 2834 /* 2835 * Test if we should deliver an upcall 2836 */ 2837 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2838 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2839 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2840 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2841 /* Prepare an upcall for delivery */ 2842 bw_meter_prepare_upcall(x, &now); 2843 } 2844 2845 /* 2846 * Reschedule for next processing 2847 */ 2848 schedule_bw_meter(x, &now); 2849 } 2850 } 2851 2852 /* Send all upcalls that are pending delivery */ 2853 bw_upcalls_send(); 2854 2855 MFC_UNLOCK(); 2856} 2857 2858/* 2859 * A periodic function for sending all upcalls that are pending delivery 2860 */ 2861static void 2862expire_bw_upcalls_send(void *unused) 2863{ 2864 MFC_LOCK(); 2865 bw_upcalls_send(); 2866 MFC_UNLOCK(); 2867 2868 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2869 expire_bw_upcalls_send, NULL); 2870} 2871 2872/* 2873 * A periodic function for periodic scanning of the multicast forwarding 2874 * table for processing all "<=" bw_meter entries. 2875 */ 2876static void 2877expire_bw_meter_process(void *unused) 2878{ 2879 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2880 bw_meter_process(); 2881 2882 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2883} 2884 2885/* 2886 * End of bandwidth monitoring code 2887 */ 2888 2889#ifdef PIM 2890/* 2891 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2892 * 2893 */ 2894static int 2895pim_register_send(struct ip *ip, struct vif *vifp, 2896 struct mbuf *m, struct mfc *rt) 2897{ 2898 struct mbuf *mb_copy, *mm; 2899 2900 if (mrtdebug & DEBUG_PIM) 2901 log(LOG_DEBUG, "pim_register_send: "); 2902 2903 mb_copy = pim_register_prepare(ip, m); 2904 if (mb_copy == NULL) 2905 return ENOBUFS; 2906 2907 /* 2908 * Send all the fragments. Note that the mbuf for each fragment 2909 * is freed by the sending machinery. 2910 */ 2911 for (mm = mb_copy; mm; mm = mb_copy) { 2912 mb_copy = mm->m_nextpkt; 2913 mm->m_nextpkt = 0; 2914 mm = m_pullup(mm, sizeof(struct ip)); 2915 if (mm != NULL) { 2916 ip = mtod(mm, struct ip *); 2917 if ((mrt_api_config & MRT_MFC_RP) && 2918 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2919 pim_register_send_rp(ip, vifp, mm, rt); 2920 } else { 2921 pim_register_send_upcall(ip, vifp, mm, rt); 2922 } 2923 } 2924 } 2925 2926 return 0; 2927} 2928 2929/* 2930 * Return a copy of the data packet that is ready for PIM Register 2931 * encapsulation. 2932 * XXX: Note that in the returned copy the IP header is a valid one. 2933 */ 2934static struct mbuf * 2935pim_register_prepare(struct ip *ip, struct mbuf *m) 2936{ 2937 struct mbuf *mb_copy = NULL; 2938 int mtu; 2939 2940 /* Take care of delayed checksums */ 2941 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2942 in_delayed_cksum(m); 2943 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2944 } 2945 2946 /* 2947 * Copy the old packet & pullup its IP header into the 2948 * new mbuf so we can modify it. 2949 */ 2950 mb_copy = m_copypacket(m, M_DONTWAIT); 2951 if (mb_copy == NULL) 2952 return NULL; 2953 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2954 if (mb_copy == NULL) 2955 return NULL; 2956 2957 /* take care of the TTL */ 2958 ip = mtod(mb_copy, struct ip *); 2959 --ip->ip_ttl; 2960 2961 /* Compute the MTU after the PIM Register encapsulation */ 2962 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2963 2964 if (ip->ip_len <= mtu) { 2965 /* Turn the IP header into a valid one */ 2966 ip->ip_len = htons(ip->ip_len); 2967 ip->ip_off = htons(ip->ip_off); 2968 ip->ip_sum = 0; 2969 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2970 } else { 2971 /* Fragment the packet */ 2972 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2973 m_freem(mb_copy); 2974 return NULL; 2975 } 2976 } 2977 return mb_copy; 2978} 2979 2980/* 2981 * Send an upcall with the data packet to the user-level process. 2982 */ 2983static int 2984pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2985 struct mbuf *mb_copy, struct mfc *rt) 2986{ 2987 struct mbuf *mb_first; 2988 int len = ntohs(ip->ip_len); 2989 struct igmpmsg *im; 2990 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2991 2992 VIF_LOCK_ASSERT(); 2993 2994 /* 2995 * Add a new mbuf with an upcall header 2996 */ 2997 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 2998 if (mb_first == NULL) { 2999 m_freem(mb_copy); 3000 return ENOBUFS; 3001 } 3002 mb_first->m_data += max_linkhdr; 3003 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3004 mb_first->m_len = sizeof(struct igmpmsg); 3005 mb_first->m_next = mb_copy; 3006 3007 /* Send message to routing daemon */ 3008 im = mtod(mb_first, struct igmpmsg *); 3009 im->im_msgtype = IGMPMSG_WHOLEPKT; 3010 im->im_mbz = 0; 3011 im->im_vif = vifp - viftable; 3012 im->im_src = ip->ip_src; 3013 im->im_dst = ip->ip_dst; 3014 3015 k_igmpsrc.sin_addr = ip->ip_src; 3016 3017 mrtstat.mrts_upcalls++; 3018 3019 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3020 if (mrtdebug & DEBUG_PIM) 3021 log(LOG_WARNING, 3022 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3023 ++mrtstat.mrts_upq_sockfull; 3024 return ENOBUFS; 3025 } 3026 3027 /* Keep statistics */ 3028 pimstat.pims_snd_registers_msgs++; 3029 pimstat.pims_snd_registers_bytes += len; 3030 3031 return 0; 3032} 3033 3034/* 3035 * Encapsulate the data packet in PIM Register message and send it to the RP. 3036 */ 3037static int 3038pim_register_send_rp(struct ip *ip, struct vif *vifp, 3039 struct mbuf *mb_copy, struct mfc *rt) 3040{ 3041 struct mbuf *mb_first; 3042 struct ip *ip_outer; 3043 struct pim_encap_pimhdr *pimhdr; 3044 int len = ntohs(ip->ip_len); 3045 vifi_t vifi = rt->mfc_parent; 3046 3047 VIF_LOCK_ASSERT(); 3048 3049 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 3050 m_freem(mb_copy); 3051 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3052 } 3053 3054 /* 3055 * Add a new mbuf with the encapsulating header 3056 */ 3057 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3058 if (mb_first == NULL) { 3059 m_freem(mb_copy); 3060 return ENOBUFS; 3061 } 3062 mb_first->m_data += max_linkhdr; 3063 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3064 mb_first->m_next = mb_copy; 3065 3066 mb_first->m_pkthdr.len = len + mb_first->m_len; 3067 3068 /* 3069 * Fill in the encapsulating IP and PIM header 3070 */ 3071 ip_outer = mtod(mb_first, struct ip *); 3072 *ip_outer = pim_encap_iphdr; 3073#ifdef RANDOM_IP_ID 3074 ip_outer->ip_id = ip_randomid(); 3075#else 3076 ip_outer->ip_id = htons(ip_id++); 3077#endif 3078 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3079 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3080 ip_outer->ip_dst = rt->mfc_rp; 3081 /* 3082 * Copy the inner header TOS to the outer header, and take care of the 3083 * IP_DF bit. 3084 */ 3085 ip_outer->ip_tos = ip->ip_tos; 3086 if (ntohs(ip->ip_off) & IP_DF) 3087 ip_outer->ip_off |= IP_DF; 3088 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 3089 + sizeof(pim_encap_iphdr)); 3090 *pimhdr = pim_encap_pimhdr; 3091 /* If the iif crosses a border, set the Border-bit */ 3092 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3093 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3094 3095 mb_first->m_data += sizeof(pim_encap_iphdr); 3096 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3097 mb_first->m_data -= sizeof(pim_encap_iphdr); 3098 3099 if (vifp->v_rate_limit == 0) 3100 tbf_send_packet(vifp, mb_first); 3101 else 3102 tbf_control(vifp, mb_first, ip, ip_outer->ip_len); 3103 3104 /* Keep statistics */ 3105 pimstat.pims_snd_registers_msgs++; 3106 pimstat.pims_snd_registers_bytes += len; 3107 3108 return 0; 3109} 3110 3111/* 3112 * PIM-SMv2 and PIM-DM messages processing. 3113 * Receives and verifies the PIM control messages, and passes them 3114 * up to the listening socket, using rip_input(). 3115 * The only message with special processing is the PIM_REGISTER message 3116 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3117 * is passed to if_simloop(). 3118 */ 3119void 3120pim_input(struct mbuf *m, int off) 3121{ 3122 struct ip *ip = mtod(m, struct ip *); 3123 struct pim *pim; 3124 int minlen; 3125 int datalen = ip->ip_len; 3126 int ip_tos; 3127 int iphlen = off; 3128 3129 /* Keep statistics */ 3130 pimstat.pims_rcv_total_msgs++; 3131 pimstat.pims_rcv_total_bytes += datalen; 3132 3133 /* 3134 * Validate lengths 3135 */ 3136 if (datalen < PIM_MINLEN) { 3137 pimstat.pims_rcv_tooshort++; 3138 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3139 datalen, (u_long)ip->ip_src.s_addr); 3140 m_freem(m); 3141 return; 3142 } 3143 3144 /* 3145 * If the packet is at least as big as a REGISTER, go agead 3146 * and grab the PIM REGISTER header size, to avoid another 3147 * possible m_pullup() later. 3148 * 3149 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3150 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3151 */ 3152 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3153 /* 3154 * Get the IP and PIM headers in contiguous memory, and 3155 * possibly the PIM REGISTER header. 3156 */ 3157 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3158 (m = m_pullup(m, minlen)) == 0) { 3159 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3160 return; 3161 } 3162 /* m_pullup() may have given us a new mbuf so reset ip. */ 3163 ip = mtod(m, struct ip *); 3164 ip_tos = ip->ip_tos; 3165 3166 /* adjust mbuf to point to the PIM header */ 3167 m->m_data += iphlen; 3168 m->m_len -= iphlen; 3169 pim = mtod(m, struct pim *); 3170 3171 /* 3172 * Validate checksum. If PIM REGISTER, exclude the data packet. 3173 * 3174 * XXX: some older PIMv2 implementations don't make this distinction, 3175 * so for compatibility reason perform the checksum over part of the 3176 * message, and if error, then over the whole message. 3177 */ 3178 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3179 /* do nothing, checksum okay */ 3180 } else if (in_cksum(m, datalen)) { 3181 pimstat.pims_rcv_badsum++; 3182 if (mrtdebug & DEBUG_PIM) 3183 log(LOG_DEBUG, "pim_input: invalid checksum"); 3184 m_freem(m); 3185 return; 3186 } 3187 3188 /* PIM version check */ 3189 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3190 pimstat.pims_rcv_badversion++; 3191 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3192 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3193 m_freem(m); 3194 return; 3195 } 3196 3197 /* restore mbuf back to the outer IP */ 3198 m->m_data -= iphlen; 3199 m->m_len += iphlen; 3200 3201 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3202 /* 3203 * Since this is a REGISTER, we'll make a copy of the register 3204 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3205 * routing daemon. 3206 */ 3207 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3208 struct mbuf *mcp; 3209 struct ip *encap_ip; 3210 u_int32_t *reghdr; 3211 struct ifnet *vifp; 3212 3213 VIF_LOCK(); 3214 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3215 VIF_UNLOCK(); 3216 if (mrtdebug & DEBUG_PIM) 3217 log(LOG_DEBUG, 3218 "pim_input: register vif not set: %d\n", reg_vif_num); 3219 m_freem(m); 3220 return; 3221 } 3222 /* XXX need refcnt? */ 3223 vifp = viftable[reg_vif_num].v_ifp; 3224 VIF_UNLOCK(); 3225 3226 /* 3227 * Validate length 3228 */ 3229 if (datalen < PIM_REG_MINLEN) { 3230 pimstat.pims_rcv_tooshort++; 3231 pimstat.pims_rcv_badregisters++; 3232 log(LOG_ERR, 3233 "pim_input: register packet size too small %d from %lx\n", 3234 datalen, (u_long)ip->ip_src.s_addr); 3235 m_freem(m); 3236 return; 3237 } 3238 3239 reghdr = (u_int32_t *)(pim + 1); 3240 encap_ip = (struct ip *)(reghdr + 1); 3241 3242 if (mrtdebug & DEBUG_PIM) { 3243 log(LOG_DEBUG, 3244 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3245 (u_long)ntohl(encap_ip->ip_src.s_addr), 3246 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3247 ntohs(encap_ip->ip_len)); 3248 } 3249 3250 /* verify the version number of the inner packet */ 3251 if (encap_ip->ip_v != IPVERSION) { 3252 pimstat.pims_rcv_badregisters++; 3253 if (mrtdebug & DEBUG_PIM) { 3254 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3255 "of the inner packet\n", encap_ip->ip_v); 3256 } 3257 m_freem(m); 3258 return; 3259 } 3260 3261 /* verify the inner packet is destined to a mcast group */ 3262 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 3263 pimstat.pims_rcv_badregisters++; 3264 if (mrtdebug & DEBUG_PIM) 3265 log(LOG_DEBUG, 3266 "pim_input: inner packet of register is not " 3267 "multicast %lx\n", 3268 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3269 m_freem(m); 3270 return; 3271 } 3272 3273 /* 3274 * Copy the TOS from the outer IP header to the inner IP header. 3275 */ 3276 if (encap_ip->ip_tos != ip_tos) { 3277 /* Outer TOS -> inner TOS */ 3278 encap_ip->ip_tos = ip_tos; 3279 /* Recompute the inner header checksum. Sigh... */ 3280 3281 /* adjust mbuf to point to the inner IP header */ 3282 m->m_data += (iphlen + PIM_MINLEN); 3283 m->m_len -= (iphlen + PIM_MINLEN); 3284 3285 encap_ip->ip_sum = 0; 3286 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3287 3288 /* restore mbuf to point back to the outer IP header */ 3289 m->m_data -= (iphlen + PIM_MINLEN); 3290 m->m_len += (iphlen + PIM_MINLEN); 3291 } 3292 3293 /* If a NULL_REGISTER, pass it to the daemon */ 3294 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3295 goto pim_input_to_daemon; 3296 3297 /* 3298 * Decapsulate the inner IP packet and loopback to forward it 3299 * as a normal multicast packet. Also, make a copy of the 3300 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3301 * to pass to the daemon later, so it can take the appropriate 3302 * actions (e.g., send back PIM_REGISTER_STOP). 3303 * XXX: here m->m_data points to the outer IP header. 3304 */ 3305 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3306 if (mcp == NULL) { 3307 log(LOG_ERR, 3308 "pim_input: pim register: could not copy register head\n"); 3309 m_freem(m); 3310 return; 3311 } 3312 3313 /* Keep statistics */ 3314 /* XXX: registers_bytes include only the encap. mcast pkt */ 3315 pimstat.pims_rcv_registers_msgs++; 3316 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3317 3318 /* 3319 * forward the inner ip packet; point m_data at the inner ip. 3320 */ 3321 m_adj(m, iphlen + PIM_MINLEN); 3322 3323 if (mrtdebug & DEBUG_PIM) { 3324 log(LOG_DEBUG, 3325 "pim_input: forwarding decapsulated register: " 3326 "src %lx, dst %lx, vif %d\n", 3327 (u_long)ntohl(encap_ip->ip_src.s_addr), 3328 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3329 reg_vif_num); 3330 } 3331 /* NB: vifp was collected above; can it change on us? */ 3332 if_simloop(vifp, m, dst.sin_family, 0); 3333 3334 /* prepare the register head to send to the mrouting daemon */ 3335 m = mcp; 3336 } 3337 3338pim_input_to_daemon: 3339 /* 3340 * Pass the PIM message up to the daemon; if it is a Register message, 3341 * pass the 'head' only up to the daemon. This includes the 3342 * outer IP header, PIM header, PIM-Register header and the 3343 * inner IP header. 3344 * XXX: the outer IP header pkt size of a Register is not adjust to 3345 * reflect the fact that the inner multicast data is truncated. 3346 */ 3347 rip_input(m, iphlen); 3348 3349 return; 3350} 3351#endif /* PIM */ 3352 3353static int 3354ip_mroute_modevent(module_t mod, int type, void *unused) 3355{ 3356 int s; 3357 3358 switch (type) { 3359 case MOD_LOAD: 3360 s = splnet(); 3361 ip_mrouter_reset(); 3362 /* XXX synchronize setup */ 3363 ip_mcast_src = X_ip_mcast_src; 3364 ip_mforward = X_ip_mforward; 3365 ip_mrouter_done = X_ip_mrouter_done; 3366 ip_mrouter_get = X_ip_mrouter_get; 3367 ip_mrouter_set = X_ip_mrouter_set; 3368 ip_rsvp_force_done = X_ip_rsvp_force_done; 3369 ip_rsvp_vif = X_ip_rsvp_vif; 3370 legal_vif_num = X_legal_vif_num; 3371 mrt_ioctl = X_mrt_ioctl; 3372 rsvp_input_p = X_rsvp_input; 3373 break; 3374 3375 case MOD_UNLOAD: 3376 /* 3377 * Typically module unload happens after the user-level 3378 * process has shutdown the kernel services (the check 3379 * below insures someone can't just yank the module out 3380 * from under a running process). But if the module is 3381 * just loaded and then unloaded w/o starting up a user 3382 * process we still need to cleanup. 3383 */ 3384 if (ip_mrouter) 3385 return EINVAL; 3386 3387 X_ip_mrouter_done(); 3388 ip_mcast_src = NULL; 3389 ip_mforward = NULL; 3390 ip_mrouter_done = NULL; 3391 ip_mrouter_get = NULL; 3392 ip_mrouter_set = NULL; 3393 ip_rsvp_force_done = NULL; 3394 ip_rsvp_vif = NULL; 3395 legal_vif_num = NULL; 3396 mrt_ioctl = NULL; 3397 rsvp_input_p = NULL; 3398 break; 3399 } 3400 return 0; 3401} 3402 3403static moduledata_t ip_mroutemod = { 3404 "ip_mroute", 3405 ip_mroute_modevent, 3406 0 3407}; 3408DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3409