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: src/sys/netinet/ip_mroute.c,v 1.56.2.10 2003/08/24 21:37:34 hsu Exp $ 21 * $DragonFly: src/sys/net/ip_mroute/ip_mroute.c,v 1.17 2005/06/15 18:29:30 joerg Exp $ 22 */ 23 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/malloc.h> 34 #include <sys/mbuf.h> 35 #include <sys/protosw.h> 36 #include <sys/socket.h> 37 #include <sys/socketvar.h> 38 #include <sys/sockio.h> 39 #include <sys/sysctl.h> 40 #include <sys/syslog.h> 41 #include <sys/systm.h> 42 #include <sys/thread2.h> 43 #include <sys/time.h> 44 #include <sys/in_cksum.h> 45 46 #include <machine/stdarg.h> 47 48 #include <net/if.h> 49 #include <net/netisr.h> 50 #include <net/route.h> 51 #include <netinet/in.h> 52 #include <netinet/igmp.h> 53 #include <netinet/in_systm.h> 54 #include <netinet/in_var.h> 55 #include <netinet/ip.h> 56 #include "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 #ifdef ALTQ 63 #include <netinet/in_pcb.h> 64 #endif 65 #include <netinet/udp.h> 66 67 /* 68 * Control debugging code for rsvp and multicast routing code. 69 * Can only set them with the debugger. 70 */ 71 static u_int rsvpdebug; /* non-zero enables debugging */ 72 73 static u_int mrtdebug; /* any set of the flags below */ 74 75 #define DEBUG_MFC 0x02 76 #define DEBUG_FORWARD 0x04 77 #define DEBUG_EXPIRE 0x08 78 #define DEBUG_XMIT 0x10 79 #define DEBUG_PIM 0x20 80 81 #define VIFI_INVALID ((vifi_t) -1) 82 83 #define M_HASCL(m) ((m)->m_flags & M_EXT) 84 85 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 86 87 static struct mrtstat mrtstat; 88 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 89 &mrtstat, mrtstat, 90 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 91 92 static struct mfc *mfctable[MFCTBLSIZ]; 93 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 94 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 95 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 96 97 static struct vif viftable[MAXVIFS]; 98 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 99 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 100 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 101 102 static u_char nexpire[MFCTBLSIZ]; 103 104 static struct callout expire_upcalls_ch; 105 static struct callout tbf_reprocess_q_ch; 106 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 107 #define UPCALL_EXPIRE 6 /* number of timeouts */ 108 109 /* 110 * Define the token bucket filter structures 111 * tbftable -> each vif has one of these for storing info 112 */ 113 114 static struct tbf tbftable[MAXVIFS]; 115 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 116 117 /* 118 * 'Interfaces' associated with decapsulator (so we can tell 119 * packets that went through it from ones that get reflected 120 * by a broken gateway). These interfaces are never linked into 121 * the system ifnet list & no routes point to them. I.e., packets 122 * can't be sent this way. They only exist as a placeholder for 123 * multicast source verification. 124 */ 125 static struct ifnet multicast_decap_if[MAXVIFS]; 126 127 #define ENCAP_TTL 64 128 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 129 130 /* prototype IP hdr for encapsulated packets */ 131 static struct ip multicast_encap_iphdr = { 132 #if BYTE_ORDER == LITTLE_ENDIAN 133 sizeof(struct ip) >> 2, IPVERSION, 134 #else 135 IPVERSION, sizeof(struct ip) >> 2, 136 #endif 137 0, /* tos */ 138 sizeof(struct ip), /* total length */ 139 0, /* id */ 140 0, /* frag offset */ 141 ENCAP_TTL, ENCAP_PROTO, 142 0, /* checksum */ 143 }; 144 145 /* 146 * Bandwidth meter variables and constants 147 */ 148 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 149 /* 150 * Pending timeouts are stored in a hash table, the key being the 151 * expiration time. Periodically, the entries are analysed and processed. 152 */ 153 #define BW_METER_BUCKETS 1024 154 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 155 static struct callout bw_meter_ch; 156 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 157 158 /* 159 * Pending upcalls are stored in a vector which is flushed when 160 * full, or periodically 161 */ 162 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 163 static u_int bw_upcalls_n; /* # of pending upcalls */ 164 static struct callout bw_upcalls_ch; 165 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 166 167 #ifdef PIM 168 static struct pimstat pimstat; 169 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 170 &pimstat, pimstat, 171 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 172 173 /* 174 * Note: the PIM Register encapsulation adds the following in front of a 175 * data packet: 176 * 177 * struct pim_encap_hdr { 178 * struct ip ip; 179 * struct pim_encap_pimhdr pim; 180 * } 181 * 182 */ 183 184 struct pim_encap_pimhdr { 185 struct pim pim; 186 uint32_t flags; 187 }; 188 189 static struct ip pim_encap_iphdr = { 190 #if BYTE_ORDER == LITTLE_ENDIAN 191 sizeof(struct ip) >> 2, 192 IPVERSION, 193 #else 194 IPVERSION, 195 sizeof(struct ip) >> 2, 196 #endif 197 0, /* tos */ 198 sizeof(struct ip), /* total length */ 199 0, /* id */ 200 0, /* frag offset */ 201 ENCAP_TTL, 202 IPPROTO_PIM, 203 0, /* checksum */ 204 }; 205 206 static struct pim_encap_pimhdr pim_encap_pimhdr = { 207 { 208 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 209 0, /* reserved */ 210 0, /* checksum */ 211 }, 212 0 /* flags */ 213 }; 214 215 static struct ifnet multicast_register_if; 216 static vifi_t reg_vif_num = VIFI_INVALID; 217 #endif /* PIM */ 218 219 /* 220 * Private variables. 221 */ 222 static vifi_t numvifs; 223 static int have_encap_tunnel; 224 225 /* 226 * one-back cache used by ipip_input to locate a tunnel's vif 227 * given a datagram's src ip address. 228 */ 229 static u_long last_encap_src; 230 static struct vif *last_encap_vif; 231 232 static u_long X_ip_mcast_src(int vifi); 233 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 234 struct mbuf *m, struct ip_moptions *imo); 235 static int X_ip_mrouter_done(void); 236 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 237 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 238 static int X_legal_vif_num(int vif); 239 static int X_mrt_ioctl(int cmd, caddr_t data); 240 241 static int get_sg_cnt(struct sioc_sg_req *); 242 static int get_vif_cnt(struct sioc_vif_req *); 243 static int ip_mrouter_init(struct socket *, int); 244 static int add_vif(struct vifctl *); 245 static int del_vif(vifi_t); 246 static int add_mfc(struct mfcctl2 *); 247 static int del_mfc(struct mfcctl2 *); 248 static int set_api_config(uint32_t *); /* chose API capabilities */ 249 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 250 static int set_assert(int); 251 static void expire_upcalls(void *); 252 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 253 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 254 static void encap_send(struct ip *, struct vif *, struct mbuf *); 255 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 256 static void tbf_queue(struct vif *, struct mbuf *); 257 static void tbf_process_q(struct vif *); 258 static void tbf_reprocess_q(void *); 259 static int tbf_dq_sel(struct vif *, struct ip *); 260 static void tbf_send_packet(struct vif *, struct mbuf *); 261 static void tbf_update_tokens(struct vif *); 262 static int priority(struct vif *, struct ip *); 263 264 /* 265 * Bandwidth monitoring 266 */ 267 static void free_bw_list(struct bw_meter *list); 268 static int add_bw_upcall(struct bw_upcall *); 269 static int del_bw_upcall(struct bw_upcall *); 270 static void bw_meter_receive_packet(struct bw_meter *x, int plen, 271 struct timeval *nowp); 272 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 273 static void bw_upcalls_send(void); 274 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 275 static void unschedule_bw_meter(struct bw_meter *x); 276 static void bw_meter_process(void); 277 static void expire_bw_upcalls_send(void *); 278 static void expire_bw_meter_process(void *); 279 280 #ifdef PIM 281 static int pim_register_send(struct ip *, struct vif *, 282 struct mbuf *, struct mfc *); 283 static int pim_register_send_rp(struct ip *, struct vif *, 284 struct mbuf *, struct mfc *); 285 static int pim_register_send_upcall(struct ip *, struct vif *, 286 struct mbuf *, struct mfc *); 287 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 288 #endif 289 290 /* 291 * whether or not special PIM assert processing is enabled. 292 */ 293 static int pim_assert; 294 /* 295 * Rate limit for assert notification messages, in usec 296 */ 297 #define ASSERT_MSG_TIME 3000000 298 299 /* 300 * Kernel multicast routing API capabilities and setup. 301 * If more API capabilities are added to the kernel, they should be 302 * recorded in `mrt_api_support'. 303 */ 304 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 305 MRT_MFC_FLAGS_BORDER_VIF | 306 MRT_MFC_RP | 307 MRT_MFC_BW_UPCALL); 308 static uint32_t mrt_api_config = 0; 309 310 /* 311 * Hash function for a source, group entry 312 */ 313 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 314 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 315 316 /* 317 * Find a route for a given origin IP address and Multicast group address 318 * Type of service parameter to be added in the future!!! 319 * Statistics are updated by the caller if needed 320 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 321 */ 322 static struct mfc * 323 mfc_find(in_addr_t o, in_addr_t g) 324 { 325 struct mfc *rt; 326 327 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 328 if ((rt->mfc_origin.s_addr == o) && 329 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 330 break; 331 return rt; 332 } 333 334 /* 335 * Macros to compute elapsed time efficiently 336 * Borrowed from Van Jacobson's scheduling code 337 */ 338 #define TV_DELTA(a, b, delta) { \ 339 int xxs; \ 340 delta = (a).tv_usec - (b).tv_usec; \ 341 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 342 switch (xxs) { \ 343 case 2: \ 344 delta += 1000000; \ 345 /* FALLTHROUGH */ \ 346 case 1: \ 347 delta += 1000000; \ 348 break; \ 349 default: \ 350 delta += (1000000 * xxs); \ 351 } \ 352 } \ 353 } 354 355 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 356 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 357 358 /* 359 * Handle MRT setsockopt commands to modify the multicast routing tables. 360 */ 361 static int 362 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 363 { 364 int error, optval; 365 vifi_t vifi; 366 struct vifctl vifc; 367 struct mfcctl2 mfc; 368 struct bw_upcall bw_upcall; 369 uint32_t i; 370 371 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 372 return EPERM; 373 374 error = 0; 375 switch (sopt->sopt_name) { 376 case MRT_INIT: 377 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 378 if (error) 379 break; 380 error = ip_mrouter_init(so, optval); 381 break; 382 383 case MRT_DONE: 384 error = ip_mrouter_done(); 385 break; 386 387 case MRT_ADD_VIF: 388 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 389 if (error) 390 break; 391 error = add_vif(&vifc); 392 break; 393 394 case MRT_DEL_VIF: 395 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 396 if (error) 397 break; 398 error = del_vif(vifi); 399 break; 400 401 case MRT_ADD_MFC: 402 case MRT_DEL_MFC: 403 /* 404 * select data size depending on API version. 405 */ 406 if (sopt->sopt_name == MRT_ADD_MFC && 407 mrt_api_config & MRT_API_FLAGS_ALL) { 408 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 409 sizeof(struct mfcctl2)); 410 } else { 411 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 412 sizeof(struct mfcctl)); 413 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 414 sizeof(mfc) - sizeof(struct mfcctl)); 415 } 416 if (error) 417 break; 418 if (sopt->sopt_name == MRT_ADD_MFC) 419 error = add_mfc(&mfc); 420 else 421 error = del_mfc(&mfc); 422 break; 423 424 case MRT_ASSERT: 425 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 426 if (error) 427 break; 428 set_assert(optval); 429 break; 430 431 case MRT_API_CONFIG: 432 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 433 if (!error) 434 error = set_api_config(&i); 435 if (!error) 436 error = sooptcopyout(sopt, &i, sizeof i); 437 break; 438 439 case MRT_ADD_BW_UPCALL: 440 case MRT_DEL_BW_UPCALL: 441 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 442 sizeof bw_upcall); 443 if (error) 444 break; 445 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 446 error = add_bw_upcall(&bw_upcall); 447 else 448 error = del_bw_upcall(&bw_upcall); 449 break; 450 451 default: 452 error = EOPNOTSUPP; 453 break; 454 } 455 return error; 456 } 457 458 /* 459 * Handle MRT getsockopt commands 460 */ 461 static int 462 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 463 { 464 int error; 465 static int version = 0x0305; /* !!! why is this here? XXX */ 466 467 switch (sopt->sopt_name) { 468 case MRT_VERSION: 469 error = sooptcopyout(sopt, &version, sizeof version); 470 break; 471 472 case MRT_ASSERT: 473 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 474 break; 475 476 case MRT_API_SUPPORT: 477 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 478 break; 479 480 case MRT_API_CONFIG: 481 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 482 break; 483 484 default: 485 error = EOPNOTSUPP; 486 break; 487 } 488 return error; 489 } 490 491 /* 492 * Handle ioctl commands to obtain information from the cache 493 */ 494 static int 495 X_mrt_ioctl(int cmd, caddr_t data) 496 { 497 int error = 0; 498 499 switch (cmd) { 500 case SIOCGETVIFCNT: 501 error = get_vif_cnt((struct sioc_vif_req *)data); 502 break; 503 504 case SIOCGETSGCNT: 505 error = get_sg_cnt((struct sioc_sg_req *)data); 506 break; 507 508 default: 509 error = EINVAL; 510 break; 511 } 512 return error; 513 } 514 515 /* 516 * returns the packet, byte, rpf-failure count for the source group provided 517 */ 518 static int 519 get_sg_cnt(struct sioc_sg_req *req) 520 { 521 struct mfc *rt; 522 523 crit_enter(); 524 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 525 crit_exit(); 526 if (rt == NULL) { 527 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 528 return EADDRNOTAVAIL; 529 } 530 req->pktcnt = rt->mfc_pkt_cnt; 531 req->bytecnt = rt->mfc_byte_cnt; 532 req->wrong_if = rt->mfc_wrong_if; 533 return 0; 534 } 535 536 /* 537 * returns the input and output packet and byte counts on the vif provided 538 */ 539 static int 540 get_vif_cnt(struct sioc_vif_req *req) 541 { 542 vifi_t vifi = req->vifi; 543 544 if (vifi >= numvifs) 545 return EINVAL; 546 547 req->icount = viftable[vifi].v_pkt_in; 548 req->ocount = viftable[vifi].v_pkt_out; 549 req->ibytes = viftable[vifi].v_bytes_in; 550 req->obytes = viftable[vifi].v_bytes_out; 551 552 return 0; 553 } 554 555 /* 556 * Enable multicast routing 557 */ 558 static int 559 ip_mrouter_init(struct socket *so, int version) 560 { 561 if (mrtdebug) 562 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 563 so->so_type, so->so_proto->pr_protocol); 564 565 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 566 return EOPNOTSUPP; 567 568 if (version != 1) 569 return ENOPROTOOPT; 570 571 if (ip_mrouter != NULL) 572 return EADDRINUSE; 573 574 ip_mrouter = so; 575 576 bzero((caddr_t)mfctable, sizeof(mfctable)); 577 bzero((caddr_t)nexpire, sizeof(nexpire)); 578 579 pim_assert = 0; 580 bw_upcalls_n = 0; 581 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 582 583 callout_init(&expire_upcalls_ch); 584 callout_init(&bw_upcalls_ch); 585 callout_init(&bw_meter_ch); 586 callout_init(&tbf_reprocess_q_ch); 587 588 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 589 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 590 expire_bw_upcalls_send, NULL); 591 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 592 593 mrt_api_config = 0; 594 595 if (mrtdebug) 596 log(LOG_DEBUG, "ip_mrouter_init\n"); 597 598 return 0; 599 } 600 601 /* 602 * Disable multicast routing 603 */ 604 static int 605 X_ip_mrouter_done(void) 606 { 607 vifi_t vifi; 608 int i; 609 struct ifnet *ifp; 610 struct ifreq ifr; 611 struct mfc *rt; 612 struct rtdetq *rte; 613 614 crit_enter(); 615 616 /* 617 * For each phyint in use, disable promiscuous reception of all IP 618 * multicasts. 619 */ 620 for (vifi = 0; vifi < numvifs; vifi++) { 621 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 622 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 623 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 624 625 so->sin_len = sizeof(struct sockaddr_in); 626 so->sin_family = AF_INET; 627 so->sin_addr.s_addr = INADDR_ANY; 628 ifp = viftable[vifi].v_ifp; 629 if_allmulti(ifp, 0); 630 } 631 } 632 bzero((caddr_t)tbftable, sizeof(tbftable)); 633 bzero((caddr_t)viftable, sizeof(viftable)); 634 numvifs = 0; 635 pim_assert = 0; 636 637 callout_stop(&expire_upcalls_ch); 638 639 mrt_api_config = 0; 640 bw_upcalls_n = 0; 641 callout_stop(&bw_upcalls_ch); 642 callout_stop(&bw_meter_ch); 643 callout_stop(&tbf_reprocess_q_ch); 644 645 /* 646 * Free all multicast forwarding cache entries. 647 */ 648 for (i = 0; i < MFCTBLSIZ; i++) { 649 for (rt = mfctable[i]; rt != NULL; ) { 650 struct mfc *nr = rt->mfc_next; 651 652 for (rte = rt->mfc_stall; rte != NULL; ) { 653 struct rtdetq *n = rte->next; 654 655 m_freem(rte->m); 656 free(rte, M_MRTABLE); 657 rte = n; 658 } 659 free_bw_list(rt->mfc_bw_meter); 660 free(rt, M_MRTABLE); 661 rt = nr; 662 } 663 } 664 665 bzero((caddr_t)mfctable, sizeof(mfctable)); 666 667 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 668 669 /* 670 * Reset de-encapsulation cache 671 */ 672 last_encap_src = INADDR_ANY; 673 last_encap_vif = NULL; 674 #ifdef PIM 675 reg_vif_num = VIFI_INVALID; 676 #endif 677 have_encap_tunnel = 0; 678 679 ip_mrouter = NULL; 680 681 crit_exit(); 682 683 if (mrtdebug) 684 log(LOG_DEBUG, "ip_mrouter_done\n"); 685 686 return 0; 687 } 688 689 /* 690 * Set PIM assert processing global 691 */ 692 static int 693 set_assert(int i) 694 { 695 if ((i != 1) && (i != 0)) 696 return EINVAL; 697 698 pim_assert = i; 699 700 return 0; 701 } 702 703 /* 704 * Configure API capabilities 705 */ 706 int 707 set_api_config(uint32_t *apival) 708 { 709 int i; 710 711 /* 712 * We can set the API capabilities only if it is the first operation 713 * after MRT_INIT. I.e.: 714 * - there are no vifs installed 715 * - pim_assert is not enabled 716 * - the MFC table is empty 717 */ 718 if (numvifs > 0) { 719 *apival = 0; 720 return EPERM; 721 } 722 if (pim_assert) { 723 *apival = 0; 724 return EPERM; 725 } 726 for (i = 0; i < MFCTBLSIZ; i++) { 727 if (mfctable[i] != NULL) { 728 *apival = 0; 729 return EPERM; 730 } 731 } 732 733 mrt_api_config = *apival & mrt_api_support; 734 *apival = mrt_api_config; 735 736 return 0; 737 } 738 739 /* 740 * Add a vif to the vif table 741 */ 742 static int 743 add_vif(struct vifctl *vifcp) 744 { 745 struct vif *vifp = viftable + vifcp->vifc_vifi; 746 struct sockaddr_in sin = {sizeof sin, AF_INET}; 747 struct ifaddr *ifa; 748 struct ifnet *ifp; 749 int error, i; 750 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 751 752 if (vifcp->vifc_vifi >= MAXVIFS) 753 return EINVAL; 754 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) 755 return EADDRINUSE; 756 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) 757 return EADDRNOTAVAIL; 758 759 /* Find the interface with an address in AF_INET family */ 760 #ifdef PIM 761 if (vifcp->vifc_flags & VIFF_REGISTER) { 762 /* 763 * XXX: Because VIFF_REGISTER does not really need a valid 764 * local interface (e.g. it could be 127.0.0.2), we don't 765 * check its address. 766 */ 767 ifp = NULL; 768 } else 769 #endif 770 { 771 sin.sin_addr = vifcp->vifc_lcl_addr; 772 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 773 if (ifa == NULL) 774 return EADDRNOTAVAIL; 775 ifp = ifa->ifa_ifp; 776 } 777 778 if (vifcp->vifc_flags & VIFF_TUNNEL) { 779 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 780 /* 781 * An encapsulating tunnel is wanted. Tell ipip_input() to 782 * start paying attention to encapsulated packets. 783 */ 784 if (have_encap_tunnel == 0) { 785 have_encap_tunnel = 1; 786 for (i = 0; i < MAXVIFS; i++) { 787 if_initname(&multicast_decap_if[i], "mdecap", i); 788 } 789 } 790 /* 791 * Set interface to fake encapsulator interface 792 */ 793 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 794 /* 795 * Prepare cached route entry 796 */ 797 bzero(&vifp->v_route, sizeof(vifp->v_route)); 798 } else { 799 log(LOG_ERR, "source routed tunnels not supported\n"); 800 return EOPNOTSUPP; 801 } 802 #ifdef PIM 803 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 804 ifp = &multicast_register_if; 805 if (mrtdebug) 806 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 807 (void *)&multicast_register_if); 808 if (reg_vif_num == VIFI_INVALID) { 809 if_initname(&multicast_register_if, "register_vif", 0); 810 multicast_register_if.if_flags = IFF_LOOPBACK; 811 bzero(&vifp->v_route, sizeof(vifp->v_route)); 812 reg_vif_num = vifcp->vifc_vifi; 813 } 814 #endif 815 } else { /* Make sure the interface supports multicast */ 816 if ((ifp->if_flags & IFF_MULTICAST) == 0) 817 return EOPNOTSUPP; 818 819 /* Enable promiscuous reception of all IP multicasts from the if */ 820 crit_enter(); 821 error = if_allmulti(ifp, 1); 822 crit_exit(); 823 if (error) 824 return error; 825 } 826 827 crit_enter(); 828 /* define parameters for the tbf structure */ 829 vifp->v_tbf = v_tbf; 830 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 831 vifp->v_tbf->tbf_n_tok = 0; 832 vifp->v_tbf->tbf_q_len = 0; 833 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 834 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 835 836 vifp->v_flags = vifcp->vifc_flags; 837 vifp->v_threshold = vifcp->vifc_threshold; 838 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 839 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 840 vifp->v_ifp = ifp; 841 /* scaling up here allows division by 1024 in critical code */ 842 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 843 vifp->v_rsvp_on = 0; 844 vifp->v_rsvpd = NULL; 845 /* initialize per vif pkt counters */ 846 vifp->v_pkt_in = 0; 847 vifp->v_pkt_out = 0; 848 vifp->v_bytes_in = 0; 849 vifp->v_bytes_out = 0; 850 crit_exit(); 851 852 /* Adjust numvifs up if the vifi is higher than numvifs */ 853 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 854 855 if (mrtdebug) 856 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 857 vifcp->vifc_vifi, 858 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 859 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 860 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 861 vifcp->vifc_threshold, 862 vifcp->vifc_rate_limit); 863 864 return 0; 865 } 866 867 /* 868 * Delete a vif from the vif table 869 */ 870 static int 871 del_vif(vifi_t vifi) 872 { 873 struct vif *vifp; 874 875 if (vifi >= numvifs) 876 return EINVAL; 877 vifp = &viftable[vifi]; 878 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) 879 return EADDRNOTAVAIL; 880 881 crit_enter(); 882 883 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 884 if_allmulti(vifp->v_ifp, 0); 885 886 if (vifp == last_encap_vif) { 887 last_encap_vif = NULL; 888 last_encap_src = INADDR_ANY; 889 } 890 891 /* 892 * Free packets queued at the interface 893 */ 894 while (vifp->v_tbf->tbf_q) { 895 struct mbuf *m = vifp->v_tbf->tbf_q; 896 897 vifp->v_tbf->tbf_q = m->m_nextpkt; 898 m_freem(m); 899 } 900 901 #ifdef PIM 902 if (vifp->v_flags & VIFF_REGISTER) 903 reg_vif_num = VIFI_INVALID; 904 #endif 905 906 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 907 bzero((caddr_t)vifp, sizeof (*vifp)); 908 909 if (mrtdebug) 910 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 911 912 /* Adjust numvifs down */ 913 for (vifi = numvifs; vifi > 0; vifi--) 914 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 915 break; 916 numvifs = vifi; 917 918 crit_exit(); 919 920 return 0; 921 } 922 923 /* 924 * update an mfc entry without resetting counters and S,G addresses. 925 */ 926 static void 927 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 928 { 929 int i; 930 931 rt->mfc_parent = mfccp->mfcc_parent; 932 for (i = 0; i < numvifs; i++) { 933 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 934 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 935 MRT_MFC_FLAGS_ALL; 936 } 937 /* set the RP address */ 938 if (mrt_api_config & MRT_MFC_RP) 939 rt->mfc_rp = mfccp->mfcc_rp; 940 else 941 rt->mfc_rp.s_addr = INADDR_ANY; 942 } 943 944 /* 945 * fully initialize an mfc entry from the parameter. 946 */ 947 static void 948 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 949 { 950 rt->mfc_origin = mfccp->mfcc_origin; 951 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 952 953 update_mfc_params(rt, mfccp); 954 955 /* initialize pkt counters per src-grp */ 956 rt->mfc_pkt_cnt = 0; 957 rt->mfc_byte_cnt = 0; 958 rt->mfc_wrong_if = 0; 959 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 960 } 961 962 963 /* 964 * Add an mfc entry 965 */ 966 static int 967 add_mfc(struct mfcctl2 *mfccp) 968 { 969 struct mfc *rt; 970 u_long hash; 971 struct rtdetq *rte; 972 u_short nstl; 973 974 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 975 976 /* If an entry already exists, just update the fields */ 977 if (rt) { 978 if (mrtdebug & DEBUG_MFC) 979 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 980 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 981 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 982 mfccp->mfcc_parent); 983 984 crit_enter(); 985 update_mfc_params(rt, mfccp); 986 crit_exit(); 987 return 0; 988 } 989 990 /* 991 * Find the entry for which the upcall was made and update 992 */ 993 crit_enter(); 994 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 995 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 996 997 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 998 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 999 (rt->mfc_stall != NULL)) { 1000 1001 if (nstl++) 1002 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1003 "multiple kernel entries", 1004 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1005 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1006 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1007 1008 if (mrtdebug & DEBUG_MFC) 1009 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1010 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1011 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1012 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1013 1014 init_mfc_params(rt, mfccp); 1015 1016 rt->mfc_expire = 0; /* Don't clean this guy up */ 1017 nexpire[hash]--; 1018 1019 /* free packets Qed at the end of this entry */ 1020 for (rte = rt->mfc_stall; rte != NULL; ) { 1021 struct rtdetq *n = rte->next; 1022 1023 ip_mdq(rte->m, rte->ifp, rt, -1); 1024 m_freem(rte->m); 1025 free(rte, M_MRTABLE); 1026 rte = n; 1027 } 1028 rt->mfc_stall = NULL; 1029 } 1030 } 1031 1032 /* 1033 * It is possible that an entry is being inserted without an upcall 1034 */ 1035 if (nstl == 0) { 1036 if (mrtdebug & DEBUG_MFC) 1037 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1038 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1039 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1040 mfccp->mfcc_parent); 1041 1042 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1043 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1044 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1045 init_mfc_params(rt, mfccp); 1046 if (rt->mfc_expire) 1047 nexpire[hash]--; 1048 rt->mfc_expire = 0; 1049 break; /* XXX */ 1050 } 1051 } 1052 if (rt == NULL) { /* no upcall, so make a new entry */ 1053 rt = malloc(sizeof(*rt), M_MRTABLE, M_INTWAIT | M_NULLOK); 1054 if (rt == NULL) { 1055 crit_exit(); 1056 return ENOBUFS; 1057 } 1058 1059 init_mfc_params(rt, mfccp); 1060 rt->mfc_expire = 0; 1061 rt->mfc_stall = NULL; 1062 1063 rt->mfc_bw_meter = NULL; 1064 /* insert new entry at head of hash chain */ 1065 rt->mfc_next = mfctable[hash]; 1066 mfctable[hash] = rt; 1067 } 1068 } 1069 crit_exit(); 1070 return 0; 1071 } 1072 1073 /* 1074 * Delete an mfc entry 1075 */ 1076 static int 1077 del_mfc(struct mfcctl2 *mfccp) 1078 { 1079 struct in_addr origin; 1080 struct in_addr mcastgrp; 1081 struct mfc *rt; 1082 struct mfc **nptr; 1083 u_long hash; 1084 struct bw_meter *list; 1085 1086 origin = mfccp->mfcc_origin; 1087 mcastgrp = mfccp->mfcc_mcastgrp; 1088 1089 if (mrtdebug & DEBUG_MFC) 1090 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1091 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1092 1093 crit_enter(); 1094 1095 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1096 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1097 if (origin.s_addr == rt->mfc_origin.s_addr && 1098 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1099 rt->mfc_stall == NULL) 1100 break; 1101 if (rt == NULL) { 1102 crit_exit(); 1103 return EADDRNOTAVAIL; 1104 } 1105 1106 *nptr = rt->mfc_next; 1107 1108 /* 1109 * free the bw_meter entries 1110 */ 1111 list = rt->mfc_bw_meter; 1112 rt->mfc_bw_meter = NULL; 1113 1114 free(rt, M_MRTABLE); 1115 1116 crit_exit(); 1117 1118 free_bw_list(list); 1119 1120 return 0; 1121 } 1122 1123 /* 1124 * Send a message to mrouted on the multicast routing socket 1125 */ 1126 static int 1127 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1128 { 1129 if (s) { 1130 if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) { 1131 sorwakeup(s); 1132 return 0; 1133 } 1134 } 1135 m_freem(mm); 1136 return -1; 1137 } 1138 1139 /* 1140 * IP multicast forwarding function. This function assumes that the packet 1141 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1142 * pointed to by "ifp", and the packet is to be relayed to other networks 1143 * that have members of the packet's destination IP multicast group. 1144 * 1145 * The packet is returned unscathed to the caller, unless it is 1146 * erroneous, in which case a non-zero return value tells the caller to 1147 * discard it. 1148 */ 1149 1150 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1151 1152 static int 1153 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1154 struct ip_moptions *imo) 1155 { 1156 struct mfc *rt; 1157 vifi_t vifi; 1158 1159 if (mrtdebug & DEBUG_FORWARD) 1160 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1161 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1162 (void *)ifp); 1163 1164 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1165 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1166 /* 1167 * Packet arrived via a physical interface or 1168 * an encapsulated tunnel or a register_vif. 1169 */ 1170 } else { 1171 /* 1172 * Packet arrived through a source-route tunnel. 1173 * Source-route tunnels are no longer supported. 1174 */ 1175 static int last_log; 1176 if (last_log != time_second) { 1177 last_log = time_second; 1178 log(LOG_ERR, 1179 "ip_mforward: received source-routed packet from %lx\n", 1180 (u_long)ntohl(ip->ip_src.s_addr)); 1181 } 1182 return 1; 1183 } 1184 1185 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1186 if (ip->ip_ttl < 255) 1187 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1188 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1189 struct vif *vifp = viftable + vifi; 1190 1191 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1192 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1193 vifi, 1194 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1195 vifp->v_ifp->if_xname); 1196 } 1197 return ip_mdq(m, ifp, NULL, vifi); 1198 } 1199 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1200 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1201 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1202 if (!imo) 1203 printf("In fact, no options were specified at all\n"); 1204 } 1205 1206 /* 1207 * Don't forward a packet with time-to-live of zero or one, 1208 * or a packet destined to a local-only group. 1209 */ 1210 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1211 return 0; 1212 1213 /* 1214 * Determine forwarding vifs from the forwarding cache table 1215 */ 1216 crit_enter(); 1217 ++mrtstat.mrts_mfc_lookups; 1218 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1219 1220 /* Entry exists, so forward if necessary */ 1221 if (rt != NULL) { 1222 crit_exit(); 1223 return ip_mdq(m, ifp, rt, -1); 1224 } else { 1225 /* 1226 * If we don't have a route for packet's origin, 1227 * Make a copy of the packet & send message to routing daemon 1228 */ 1229 1230 struct mbuf *mb0; 1231 struct rtdetq *rte; 1232 u_long hash; 1233 int hlen = ip->ip_hl << 2; 1234 1235 ++mrtstat.mrts_mfc_misses; 1236 1237 mrtstat.mrts_no_route++; 1238 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1239 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1240 (u_long)ntohl(ip->ip_src.s_addr), 1241 (u_long)ntohl(ip->ip_dst.s_addr)); 1242 1243 /* 1244 * Allocate mbufs early so that we don't do extra work if we are 1245 * just going to fail anyway. Make sure to pullup the header so 1246 * that other people can't step on it. 1247 */ 1248 rte = malloc((sizeof *rte), M_MRTABLE, M_INTWAIT | M_NULLOK); 1249 if (rte == NULL) { 1250 crit_exit(); 1251 return ENOBUFS; 1252 } 1253 1254 mb0 = m_copypacket(m, MB_DONTWAIT); 1255 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1256 mb0 = m_pullup(mb0, hlen); 1257 if (mb0 == NULL) { 1258 free(rte, M_MRTABLE); 1259 crit_exit(); 1260 return ENOBUFS; 1261 } 1262 1263 /* is there an upcall waiting for this flow ? */ 1264 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1265 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1266 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1267 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1268 (rt->mfc_stall != NULL)) 1269 break; 1270 } 1271 1272 if (rt == NULL) { 1273 int i; 1274 struct igmpmsg *im; 1275 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1276 struct mbuf *mm; 1277 1278 /* 1279 * Locate the vifi for the incoming interface for this packet. 1280 * If none found, drop packet. 1281 */ 1282 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1283 ; 1284 if (vifi >= numvifs) /* vif not found, drop packet */ 1285 goto non_fatal; 1286 1287 /* no upcall, so make a new entry */ 1288 rt = malloc(sizeof(*rt), M_MRTABLE, M_INTWAIT | M_NULLOK); 1289 if (rt == NULL) 1290 goto fail; 1291 1292 /* Make a copy of the header to send to the user level process */ 1293 mm = m_copy(mb0, 0, hlen); 1294 if (mm == NULL) 1295 goto fail1; 1296 1297 /* 1298 * Send message to routing daemon to install 1299 * a route into the kernel table 1300 */ 1301 1302 im = mtod(mm, struct igmpmsg *); 1303 im->im_msgtype = IGMPMSG_NOCACHE; 1304 im->im_mbz = 0; 1305 im->im_vif = vifi; 1306 1307 mrtstat.mrts_upcalls++; 1308 1309 k_igmpsrc.sin_addr = ip->ip_src; 1310 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1311 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1312 ++mrtstat.mrts_upq_sockfull; 1313 fail1: 1314 free(rt, M_MRTABLE); 1315 fail: 1316 free(rte, M_MRTABLE); 1317 m_freem(mb0); 1318 crit_exit(); 1319 return ENOBUFS; 1320 } 1321 1322 /* insert new entry at head of hash chain */ 1323 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1324 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1325 rt->mfc_expire = UPCALL_EXPIRE; 1326 nexpire[hash]++; 1327 for (i = 0; i < numvifs; i++) { 1328 rt->mfc_ttls[i] = 0; 1329 rt->mfc_flags[i] = 0; 1330 } 1331 rt->mfc_parent = -1; 1332 1333 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1334 1335 rt->mfc_bw_meter = NULL; 1336 1337 /* link into table */ 1338 rt->mfc_next = mfctable[hash]; 1339 mfctable[hash] = rt; 1340 rt->mfc_stall = rte; 1341 1342 } else { 1343 /* determine if q has overflowed */ 1344 int npkts = 0; 1345 struct rtdetq **p; 1346 1347 /* 1348 * XXX ouch! we need to append to the list, but we 1349 * only have a pointer to the front, so we have to 1350 * scan the entire list every time. 1351 */ 1352 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1353 npkts++; 1354 1355 if (npkts > MAX_UPQ) { 1356 mrtstat.mrts_upq_ovflw++; 1357 non_fatal: 1358 free(rte, M_MRTABLE); 1359 m_freem(mb0); 1360 crit_exit(); 1361 return 0; 1362 } 1363 1364 /* Add this entry to the end of the queue */ 1365 *p = rte; 1366 } 1367 1368 rte->m = mb0; 1369 rte->ifp = ifp; 1370 rte->next = NULL; 1371 1372 crit_exit(); 1373 return 0; 1374 } 1375 } 1376 1377 /* 1378 * Clean up the cache entry if upcall is not serviced 1379 */ 1380 static void 1381 expire_upcalls(void *unused) 1382 { 1383 struct rtdetq *rte; 1384 struct mfc *mfc, **nptr; 1385 int i; 1386 1387 crit_enter(); 1388 for (i = 0; i < MFCTBLSIZ; i++) { 1389 if (nexpire[i] == 0) 1390 continue; 1391 nptr = &mfctable[i]; 1392 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1393 /* 1394 * Skip real cache entries 1395 * Make sure it wasn't marked to not expire (shouldn't happen) 1396 * If it expires now 1397 */ 1398 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1399 --mfc->mfc_expire == 0) { 1400 if (mrtdebug & DEBUG_EXPIRE) 1401 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1402 (u_long)ntohl(mfc->mfc_origin.s_addr), 1403 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1404 /* 1405 * drop all the packets 1406 * free the mbuf with the pkt, if, timing info 1407 */ 1408 for (rte = mfc->mfc_stall; rte; ) { 1409 struct rtdetq *n = rte->next; 1410 1411 m_freem(rte->m); 1412 free(rte, M_MRTABLE); 1413 rte = n; 1414 } 1415 ++mrtstat.mrts_cache_cleanups; 1416 nexpire[i]--; 1417 1418 /* 1419 * free the bw_meter entries 1420 */ 1421 while (mfc->mfc_bw_meter != NULL) { 1422 struct bw_meter *x = mfc->mfc_bw_meter; 1423 1424 mfc->mfc_bw_meter = x->bm_mfc_next; 1425 free(x, M_BWMETER); 1426 } 1427 1428 *nptr = mfc->mfc_next; 1429 free(mfc, M_MRTABLE); 1430 } else { 1431 nptr = &mfc->mfc_next; 1432 } 1433 } 1434 } 1435 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1436 crit_exit(); 1437 } 1438 1439 /* 1440 * Packet forwarding routine once entry in the cache is made 1441 */ 1442 static int 1443 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1444 { 1445 struct ip *ip = mtod(m, struct ip *); 1446 vifi_t vifi; 1447 int plen = ip->ip_len; 1448 1449 /* 1450 * Macro to send packet on vif. Since RSVP packets don't get counted on 1451 * input, they shouldn't get counted on output, so statistics keeping is 1452 * separate. 1453 */ 1454 #define MC_SEND(ip,vifp,m) { \ 1455 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1456 encap_send((ip), (vifp), (m)); \ 1457 else \ 1458 phyint_send((ip), (vifp), (m)); \ 1459 } 1460 1461 /* 1462 * If xmt_vif is not -1, send on only the requested vif. 1463 * 1464 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1465 */ 1466 if (xmt_vif < numvifs) { 1467 #ifdef PIM 1468 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1469 pim_register_send(ip, viftable + xmt_vif, m, rt); 1470 else 1471 #endif 1472 MC_SEND(ip, viftable + xmt_vif, m); 1473 return 1; 1474 } 1475 1476 /* 1477 * Don't forward if it didn't arrive from the parent vif for its origin. 1478 */ 1479 vifi = rt->mfc_parent; 1480 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1481 /* came in the wrong interface */ 1482 if (mrtdebug & DEBUG_FORWARD) 1483 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1484 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1485 ++mrtstat.mrts_wrong_if; 1486 ++rt->mfc_wrong_if; 1487 /* 1488 * If we are doing PIM assert processing, send a message 1489 * to the routing daemon. 1490 * 1491 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1492 * can complete the SPT switch, regardless of the type 1493 * of the iif (broadcast media, GRE tunnel, etc). 1494 */ 1495 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1496 struct timeval now; 1497 u_long delta; 1498 1499 #ifdef PIM 1500 if (ifp == &multicast_register_if) 1501 pimstat.pims_rcv_registers_wrongiif++; 1502 #endif 1503 1504 /* Get vifi for the incoming packet */ 1505 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1506 ; 1507 if (vifi >= numvifs) 1508 return 0; /* The iif is not found: ignore the packet. */ 1509 1510 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1511 return 0; /* WRONGVIF disabled: ignore the packet */ 1512 1513 GET_TIME(now); 1514 1515 TV_DELTA(rt->mfc_last_assert, now, delta); 1516 1517 if (delta > ASSERT_MSG_TIME) { 1518 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1519 struct igmpmsg *im; 1520 int hlen = ip->ip_hl << 2; 1521 struct mbuf *mm = m_copy(m, 0, hlen); 1522 1523 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1524 mm = m_pullup(mm, hlen); 1525 if (mm == NULL) 1526 return ENOBUFS; 1527 1528 rt->mfc_last_assert = now; 1529 1530 im = mtod(mm, struct igmpmsg *); 1531 im->im_msgtype = IGMPMSG_WRONGVIF; 1532 im->im_mbz = 0; 1533 im->im_vif = vifi; 1534 1535 mrtstat.mrts_upcalls++; 1536 1537 k_igmpsrc.sin_addr = im->im_src; 1538 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1539 log(LOG_WARNING, 1540 "ip_mforward: ip_mrouter socket queue full\n"); 1541 ++mrtstat.mrts_upq_sockfull; 1542 return ENOBUFS; 1543 } 1544 } 1545 } 1546 return 0; 1547 } 1548 1549 /* If I sourced this packet, it counts as output, else it was input. */ 1550 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1551 viftable[vifi].v_pkt_out++; 1552 viftable[vifi].v_bytes_out += plen; 1553 } else { 1554 viftable[vifi].v_pkt_in++; 1555 viftable[vifi].v_bytes_in += plen; 1556 } 1557 rt->mfc_pkt_cnt++; 1558 rt->mfc_byte_cnt += plen; 1559 1560 /* 1561 * For each vif, decide if a copy of the packet should be forwarded. 1562 * Forward if: 1563 * - the ttl exceeds the vif's threshold 1564 * - there are group members downstream on interface 1565 */ 1566 for (vifi = 0; vifi < numvifs; vifi++) 1567 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1568 viftable[vifi].v_pkt_out++; 1569 viftable[vifi].v_bytes_out += plen; 1570 #ifdef PIM 1571 if (viftable[vifi].v_flags & VIFF_REGISTER) 1572 pim_register_send(ip, viftable + vifi, m, rt); 1573 else 1574 #endif 1575 MC_SEND(ip, viftable+vifi, m); 1576 } 1577 1578 /* 1579 * Perform upcall-related bw measuring. 1580 */ 1581 if (rt->mfc_bw_meter != NULL) { 1582 struct bw_meter *x; 1583 struct timeval now; 1584 1585 GET_TIME(now); 1586 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1587 bw_meter_receive_packet(x, plen, &now); 1588 } 1589 1590 return 0; 1591 } 1592 1593 /* 1594 * check if a vif number is legal/ok. This is used by ip_output. 1595 */ 1596 static int 1597 X_legal_vif_num(int vif) 1598 { 1599 return (vif >= 0 && vif < numvifs); 1600 } 1601 1602 /* 1603 * Return the local address used by this vif 1604 */ 1605 static u_long 1606 X_ip_mcast_src(int vifi) 1607 { 1608 if (vifi >= 0 && vifi < numvifs) 1609 return viftable[vifi].v_lcl_addr.s_addr; 1610 else 1611 return INADDR_ANY; 1612 } 1613 1614 static void 1615 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1616 { 1617 struct mbuf *mb_copy; 1618 int hlen = ip->ip_hl << 2; 1619 1620 /* 1621 * Make a new reference to the packet; make sure that 1622 * the IP header is actually copied, not just referenced, 1623 * so that ip_output() only scribbles on the copy. 1624 */ 1625 mb_copy = m_copypacket(m, MB_DONTWAIT); 1626 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1627 mb_copy = m_pullup(mb_copy, hlen); 1628 if (mb_copy == NULL) 1629 return; 1630 1631 if (vifp->v_rate_limit == 0) 1632 tbf_send_packet(vifp, mb_copy); 1633 else 1634 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1635 } 1636 1637 static void 1638 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1639 { 1640 struct mbuf *mb_copy; 1641 struct ip *ip_copy; 1642 int i, len = ip->ip_len; 1643 1644 /* Take care of delayed checksums */ 1645 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1646 in_delayed_cksum(m); 1647 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1648 } 1649 1650 /* 1651 * copy the old packet & pullup its IP header into the 1652 * new mbuf so we can modify it. Try to fill the new 1653 * mbuf since if we don't the ethernet driver will. 1654 */ 1655 MGETHDR(mb_copy, MB_DONTWAIT, MT_HEADER); 1656 if (mb_copy == NULL) 1657 return; 1658 mb_copy->m_data += max_linkhdr; 1659 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1660 1661 if ((mb_copy->m_next = m_copypacket(m, MB_DONTWAIT)) == NULL) { 1662 m_freem(mb_copy); 1663 return; 1664 } 1665 i = MHLEN - M_LEADINGSPACE(mb_copy); 1666 if (i > len) 1667 i = len; 1668 mb_copy = m_pullup(mb_copy, i); 1669 if (mb_copy == NULL) 1670 return; 1671 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1672 1673 /* 1674 * fill in the encapsulating IP header. 1675 */ 1676 ip_copy = mtod(mb_copy, struct ip *); 1677 *ip_copy = multicast_encap_iphdr; 1678 #ifdef RANDOM_IP_ID 1679 ip_copy->ip_id = ip_randomid(); 1680 #else 1681 ip_copy->ip_id = htons(ip_id++); 1682 #endif 1683 ip_copy->ip_len += len; 1684 ip_copy->ip_src = vifp->v_lcl_addr; 1685 ip_copy->ip_dst = vifp->v_rmt_addr; 1686 1687 /* 1688 * turn the encapsulated IP header back into a valid one. 1689 */ 1690 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1691 --ip->ip_ttl; 1692 ip->ip_len = htons(ip->ip_len); 1693 ip->ip_off = htons(ip->ip_off); 1694 ip->ip_sum = 0; 1695 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1696 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1697 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1698 1699 if (vifp->v_rate_limit == 0) 1700 tbf_send_packet(vifp, mb_copy); 1701 else 1702 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1703 } 1704 1705 /* 1706 * De-encapsulate a packet and feed it back through ip input (this 1707 * routine is called whenever IP gets a packet with proto type 1708 * ENCAP_PROTO and a local destination address). 1709 * 1710 * This is similar to mroute_encapcheck() + mroute_encap_input() in -current. 1711 */ 1712 static void 1713 X_ipip_input(struct mbuf *m, int off, int proto) 1714 { 1715 struct ip *ip = mtod(m, struct ip *); 1716 int hlen = ip->ip_hl << 2; 1717 1718 if (!have_encap_tunnel) { 1719 rip_input(m, off, proto); 1720 return; 1721 } 1722 /* 1723 * dump the packet if it's not to a multicast destination or if 1724 * we don't have an encapsulating tunnel with the source. 1725 * Note: This code assumes that the remote site IP address 1726 * uniquely identifies the tunnel (i.e., that this site has 1727 * at most one tunnel with the remote site). 1728 */ 1729 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) { 1730 ++mrtstat.mrts_bad_tunnel; 1731 m_freem(m); 1732 return; 1733 } 1734 if (ip->ip_src.s_addr != last_encap_src) { 1735 struct vif *vifp = viftable; 1736 struct vif *vife = vifp + numvifs; 1737 1738 last_encap_src = ip->ip_src.s_addr; 1739 last_encap_vif = NULL; 1740 for ( ; vifp < vife; ++vifp) 1741 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1742 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1743 == VIFF_TUNNEL) 1744 last_encap_vif = vifp; 1745 break; 1746 } 1747 } 1748 if (last_encap_vif == NULL) { 1749 last_encap_src = INADDR_ANY; 1750 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1751 m_freem(m); 1752 if (mrtdebug) 1753 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1754 (u_long)ntohl(ip->ip_src.s_addr)); 1755 return; 1756 } 1757 1758 if (hlen > sizeof(struct ip)) 1759 ip_stripoptions(m); 1760 m->m_data += sizeof(struct ip); 1761 m->m_len -= sizeof(struct ip); 1762 m->m_pkthdr.len -= sizeof(struct ip); 1763 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 1764 1765 netisr_queue(NETISR_IP, m); 1766 } 1767 1768 /* 1769 * Token bucket filter module 1770 */ 1771 1772 static void 1773 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len) 1774 { 1775 struct tbf *t = vifp->v_tbf; 1776 1777 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */ 1778 mrtstat.mrts_pkt2large++; 1779 m_freem(m); 1780 return; 1781 } 1782 1783 tbf_update_tokens(vifp); 1784 1785 if (t->tbf_q_len == 0) { /* queue empty... */ 1786 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */ 1787 t->tbf_n_tok -= p_len; 1788 tbf_send_packet(vifp, m); 1789 } else { /* no, queue packet and try later */ 1790 tbf_queue(vifp, m); 1791 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS, 1792 tbf_reprocess_q, vifp); 1793 } 1794 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1795 /* finite queue length, so queue pkts and process queue */ 1796 tbf_queue(vifp, m); 1797 tbf_process_q(vifp); 1798 } else { 1799 /* queue full, try to dq and queue and process */ 1800 if (!tbf_dq_sel(vifp, ip)) { 1801 mrtstat.mrts_q_overflow++; 1802 m_freem(m); 1803 } else { 1804 tbf_queue(vifp, m); 1805 tbf_process_q(vifp); 1806 } 1807 } 1808 } 1809 1810 /* 1811 * adds a packet to the queue at the interface 1812 */ 1813 static void 1814 tbf_queue(struct vif *vifp, struct mbuf *m) 1815 { 1816 struct tbf *t = vifp->v_tbf; 1817 1818 crit_enter(); 1819 1820 if (t->tbf_t == NULL) /* Queue was empty */ 1821 t->tbf_q = m; 1822 else /* Insert at tail */ 1823 t->tbf_t->m_nextpkt = m; 1824 1825 t->tbf_t = m; /* Set new tail pointer */ 1826 1827 #ifdef DIAGNOSTIC 1828 /* Make sure we didn't get fed a bogus mbuf */ 1829 if (m->m_nextpkt) 1830 panic("tbf_queue: m_nextpkt"); 1831 #endif 1832 m->m_nextpkt = NULL; 1833 1834 t->tbf_q_len++; 1835 1836 crit_exit(); 1837 } 1838 1839 /* 1840 * processes the queue at the interface 1841 */ 1842 static void 1843 tbf_process_q(struct vif *vifp) 1844 { 1845 struct tbf *t = vifp->v_tbf; 1846 1847 crit_enter(); 1848 1849 /* loop through the queue at the interface and send as many packets 1850 * as possible 1851 */ 1852 while (t->tbf_q_len > 0) { 1853 struct mbuf *m = t->tbf_q; 1854 int len = mtod(m, struct ip *)->ip_len; 1855 1856 /* determine if the packet can be sent */ 1857 if (len > t->tbf_n_tok) /* not enough tokens, we are done */ 1858 break; 1859 /* ok, reduce no of tokens, dequeue and send the packet. */ 1860 t->tbf_n_tok -= len; 1861 1862 t->tbf_q = m->m_nextpkt; 1863 if (--t->tbf_q_len == 0) 1864 t->tbf_t = NULL; 1865 1866 m->m_nextpkt = NULL; 1867 tbf_send_packet(vifp, m); 1868 } 1869 crit_exit(); 1870 } 1871 1872 static void 1873 tbf_reprocess_q(void *xvifp) 1874 { 1875 struct vif *vifp = xvifp; 1876 1877 if (ip_mrouter == NULL) 1878 return; 1879 tbf_update_tokens(vifp); 1880 tbf_process_q(vifp); 1881 if (vifp->v_tbf->tbf_q_len) 1882 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS, 1883 tbf_reprocess_q, vifp); 1884 } 1885 1886 /* function that will selectively discard a member of the queue 1887 * based on the precedence value and the priority 1888 */ 1889 static int 1890 tbf_dq_sel(struct vif *vifp, struct ip *ip) 1891 { 1892 u_int p; 1893 struct mbuf *m, *last; 1894 struct mbuf **np; 1895 struct tbf *t = vifp->v_tbf; 1896 1897 crit_enter(); 1898 1899 p = priority(vifp, ip); 1900 1901 np = &t->tbf_q; 1902 last = NULL; 1903 while ((m = *np) != NULL) { 1904 if (p > priority(vifp, mtod(m, struct ip *))) { 1905 *np = m->m_nextpkt; 1906 /* If we're removing the last packet, fix the tail pointer */ 1907 if (m == t->tbf_t) 1908 t->tbf_t = last; 1909 m_freem(m); 1910 /* It's impossible for the queue to be empty, but check anyways. */ 1911 if (--t->tbf_q_len == 0) 1912 t->tbf_t = NULL; 1913 crit_exit(); 1914 mrtstat.mrts_drop_sel++; 1915 return 1; 1916 } 1917 np = &m->m_nextpkt; 1918 last = m; 1919 } 1920 crit_exit(); 1921 return 0; 1922 } 1923 1924 static void 1925 tbf_send_packet(struct vif *vifp, struct mbuf *m) 1926 { 1927 crit_enter(); 1928 1929 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */ 1930 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 1931 else { 1932 struct ip_moptions imo; 1933 int error; 1934 static struct route ro; /* XXX check this */ 1935 1936 imo.imo_multicast_ifp = vifp->v_ifp; 1937 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1938 imo.imo_multicast_loop = 1; 1939 imo.imo_multicast_vif = -1; 1940 1941 /* 1942 * Re-entrancy should not be a problem here, because 1943 * the packets that we send out and are looped back at us 1944 * should get rejected because they appear to come from 1945 * the loopback interface, thus preventing looping. 1946 */ 1947 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL); 1948 1949 if (mrtdebug & DEBUG_XMIT) 1950 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1951 (int)(vifp - viftable), error); 1952 } 1953 crit_exit(); 1954 } 1955 1956 /* determine the current time and then 1957 * the elapsed time (between the last time and time now) 1958 * in milliseconds & update the no. of tokens in the bucket 1959 */ 1960 static void 1961 tbf_update_tokens(struct vif *vifp) 1962 { 1963 struct timeval tp; 1964 u_long tm; 1965 struct tbf *t = vifp->v_tbf; 1966 1967 crit_enter(); 1968 1969 GET_TIME(tp); 1970 1971 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1972 1973 /* 1974 * This formula is actually 1975 * "time in seconds" * "bytes/second". 1976 * 1977 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1978 * 1979 * The (1000/1024) was introduced in add_vif to optimize 1980 * this divide into a shift. 1981 */ 1982 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1983 t->tbf_last_pkt_t = tp; 1984 1985 if (t->tbf_n_tok > MAX_BKT_SIZE) 1986 t->tbf_n_tok = MAX_BKT_SIZE; 1987 1988 crit_exit(); 1989 } 1990 1991 static int 1992 priority(struct vif *vifp, struct ip *ip) 1993 { 1994 int prio = 50; /* the lowest priority -- default case */ 1995 1996 /* temporary hack; may add general packet classifier some day */ 1997 1998 /* 1999 * The UDP port space is divided up into four priority ranges: 2000 * [0, 16384) : unclassified - lowest priority 2001 * [16384, 32768) : audio - highest priority 2002 * [32768, 49152) : whiteboard - medium priority 2003 * [49152, 65536) : video - low priority 2004 * 2005 * Everything else gets lowest priority. 2006 */ 2007 if (ip->ip_p == IPPROTO_UDP) { 2008 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2009 switch (ntohs(udp->uh_dport) & 0xc000) { 2010 case 0x4000: 2011 prio = 70; 2012 break; 2013 case 0x8000: 2014 prio = 60; 2015 break; 2016 case 0xc000: 2017 prio = 55; 2018 break; 2019 } 2020 } 2021 return prio; 2022 } 2023 2024 /* 2025 * End of token bucket filter modifications 2026 */ 2027 2028 static int 2029 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 2030 { 2031 int error, vifi; 2032 2033 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2034 return EOPNOTSUPP; 2035 2036 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 2037 if (error) 2038 return error; 2039 2040 crit_enter(); 2041 2042 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 2043 crit_exit(); 2044 return EADDRNOTAVAIL; 2045 } 2046 2047 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 2048 /* Check if socket is available. */ 2049 if (viftable[vifi].v_rsvpd != NULL) { 2050 crit_exit(); 2051 return EADDRINUSE; 2052 } 2053 2054 viftable[vifi].v_rsvpd = so; 2055 /* This may seem silly, but we need to be sure we don't over-increment 2056 * the RSVP counter, in case something slips up. 2057 */ 2058 if (!viftable[vifi].v_rsvp_on) { 2059 viftable[vifi].v_rsvp_on = 1; 2060 rsvp_on++; 2061 } 2062 } else { /* must be VIF_OFF */ 2063 /* 2064 * XXX as an additional consistency check, one could make sure 2065 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 2066 * first parameter is pretty useless. 2067 */ 2068 viftable[vifi].v_rsvpd = NULL; 2069 /* 2070 * This may seem silly, but we need to be sure we don't over-decrement 2071 * the RSVP counter, in case something slips up. 2072 */ 2073 if (viftable[vifi].v_rsvp_on) { 2074 viftable[vifi].v_rsvp_on = 0; 2075 rsvp_on--; 2076 } 2077 } 2078 crit_exit(); 2079 return 0; 2080 } 2081 2082 static void 2083 X_ip_rsvp_force_done(struct socket *so) 2084 { 2085 int vifi; 2086 2087 /* Don't bother if it is not the right type of socket. */ 2088 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2089 return; 2090 2091 crit_enter(); 2092 2093 /* The socket may be attached to more than one vif...this 2094 * is perfectly legal. 2095 */ 2096 for (vifi = 0; vifi < numvifs; vifi++) { 2097 if (viftable[vifi].v_rsvpd == so) { 2098 viftable[vifi].v_rsvpd = NULL; 2099 /* This may seem silly, but we need to be sure we don't 2100 * over-decrement the RSVP counter, in case something slips up. 2101 */ 2102 if (viftable[vifi].v_rsvp_on) { 2103 viftable[vifi].v_rsvp_on = 0; 2104 rsvp_on--; 2105 } 2106 } 2107 } 2108 2109 crit_exit(); 2110 } 2111 2112 static void 2113 X_rsvp_input(struct mbuf *m, ...) 2114 { 2115 int vifi; 2116 struct ip *ip = mtod(m, struct ip *); 2117 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2118 struct ifnet *ifp; 2119 int off, proto; 2120 #ifdef ALTQ 2121 /* support IP_RECVIF used by rsvpd rel4.2a1 */ 2122 struct inpcb *inp; 2123 struct socket *so; 2124 struct mbuf *opts; 2125 #endif 2126 __va_list ap; 2127 2128 __va_start(ap, m); 2129 off = __va_arg(ap, int); 2130 proto = __va_arg(ap, int); 2131 __va_end(ap); 2132 2133 if (rsvpdebug) 2134 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2135 2136 /* Can still get packets with rsvp_on = 0 if there is a local member 2137 * of the group to which the RSVP packet is addressed. But in this 2138 * case we want to throw the packet away. 2139 */ 2140 if (!rsvp_on) { 2141 m_freem(m); 2142 return; 2143 } 2144 2145 crit_enter(); 2146 2147 if (rsvpdebug) 2148 printf("rsvp_input: check vifs\n"); 2149 2150 #ifdef DIAGNOSTIC 2151 if (!(m->m_flags & M_PKTHDR)) 2152 panic("rsvp_input no hdr"); 2153 #endif 2154 2155 ifp = m->m_pkthdr.rcvif; 2156 /* Find which vif the packet arrived on. */ 2157 for (vifi = 0; vifi < numvifs; vifi++) 2158 if (viftable[vifi].v_ifp == ifp) 2159 break; 2160 2161 #ifdef ALTQ 2162 if (vifi == numvifs || (so = viftable[vifi].v_rsvpd) == NULL) { 2163 #else 2164 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2165 #endif 2166 /* 2167 * If the old-style non-vif-associated socket is set, 2168 * then use it. Otherwise, drop packet since there 2169 * is no specific socket for this vif. 2170 */ 2171 if (ip_rsvpd != NULL) { 2172 if (rsvpdebug) 2173 printf("rsvp_input: Sending packet up old-style socket\n"); 2174 rip_input(m, off, proto); /* xxx */ 2175 } else { 2176 if (rsvpdebug && vifi == numvifs) 2177 printf("rsvp_input: Can't find vif for packet.\n"); 2178 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2179 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2180 m_freem(m); 2181 } 2182 crit_exit(); 2183 return; 2184 } 2185 rsvp_src.sin_addr = ip->ip_src; 2186 2187 if (rsvpdebug && m) 2188 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2189 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2190 2191 #ifdef ALTQ 2192 opts = NULL; 2193 inp = (struct inpcb *)so->so_pcb; 2194 if (inp->inp_flags & INP_CONTROLOPTS || 2195 inp->inp_socket->so_options & SO_TIMESTAMP) 2196 ip_savecontrol(inp, &opts, ip, m); 2197 if (sbappendaddr(&so->so_rcv, 2198 (struct sockaddr *)&rsvp_src,m, opts) == 0) { 2199 m_freem(m); 2200 if (opts) 2201 m_freem(opts); 2202 if (rsvpdebug) 2203 printf("rsvp_input: Failed to append to socket\n"); 2204 } 2205 else { 2206 sorwakeup(so); 2207 if (rsvpdebug) 2208 printf("rsvp_input: send packet up\n"); 2209 } 2210 #else /* !ALTQ */ 2211 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2212 if (rsvpdebug) 2213 printf("rsvp_input: Failed to append to socket\n"); 2214 } else { 2215 if (rsvpdebug) 2216 printf("rsvp_input: send packet up\n"); 2217 } 2218 #endif /* !ALTQ */ 2219 2220 crit_exit(); 2221 } 2222 2223 /* 2224 * Code for bandwidth monitors 2225 */ 2226 2227 /* 2228 * Define common interface for timeval-related methods 2229 */ 2230 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 2231 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 2232 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 2233 2234 static uint32_t 2235 compute_bw_meter_flags(struct bw_upcall *req) 2236 { 2237 uint32_t flags = 0; 2238 2239 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2240 flags |= BW_METER_UNIT_PACKETS; 2241 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2242 flags |= BW_METER_UNIT_BYTES; 2243 if (req->bu_flags & BW_UPCALL_GEQ) 2244 flags |= BW_METER_GEQ; 2245 if (req->bu_flags & BW_UPCALL_LEQ) 2246 flags |= BW_METER_LEQ; 2247 2248 return flags; 2249 } 2250 2251 /* 2252 * Add a bw_meter entry 2253 */ 2254 static int 2255 add_bw_upcall(struct bw_upcall *req) 2256 { 2257 struct mfc *mfc; 2258 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2259 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2260 struct timeval now; 2261 struct bw_meter *x; 2262 uint32_t flags; 2263 2264 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2265 return EOPNOTSUPP; 2266 2267 /* Test if the flags are valid */ 2268 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2269 return EINVAL; 2270 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2271 return EINVAL; 2272 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2273 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2274 return EINVAL; 2275 2276 /* Test if the threshold time interval is valid */ 2277 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2278 return EINVAL; 2279 2280 flags = compute_bw_meter_flags(req); 2281 2282 /* 2283 * Find if we have already same bw_meter entry 2284 */ 2285 crit_enter(); 2286 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2287 if (mfc == NULL) { 2288 crit_exit(); 2289 return EADDRNOTAVAIL; 2290 } 2291 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2292 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2293 &req->bu_threshold.b_time, ==)) && 2294 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2295 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2296 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2297 crit_exit(); 2298 return 0; /* XXX Already installed */ 2299 } 2300 } 2301 crit_exit(); 2302 2303 /* Allocate the new bw_meter entry */ 2304 x = malloc(sizeof(*x), M_BWMETER, M_INTWAIT); 2305 2306 /* Set the new bw_meter entry */ 2307 x->bm_threshold.b_time = req->bu_threshold.b_time; 2308 GET_TIME(now); 2309 x->bm_start_time = now; 2310 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2311 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2312 x->bm_measured.b_packets = 0; 2313 x->bm_measured.b_bytes = 0; 2314 x->bm_flags = flags; 2315 x->bm_time_next = NULL; 2316 x->bm_time_hash = BW_METER_BUCKETS; 2317 2318 /* Add the new bw_meter entry to the front of entries for this MFC */ 2319 crit_enter(); 2320 x->bm_mfc = mfc; 2321 x->bm_mfc_next = mfc->mfc_bw_meter; 2322 mfc->mfc_bw_meter = x; 2323 schedule_bw_meter(x, &now); 2324 crit_exit(); 2325 2326 return 0; 2327 } 2328 2329 static void 2330 free_bw_list(struct bw_meter *list) 2331 { 2332 while (list != NULL) { 2333 struct bw_meter *x = list; 2334 2335 list = list->bm_mfc_next; 2336 unschedule_bw_meter(x); 2337 free(x, M_BWMETER); 2338 } 2339 } 2340 2341 /* 2342 * Delete one or multiple bw_meter entries 2343 */ 2344 static int 2345 del_bw_upcall(struct bw_upcall *req) 2346 { 2347 struct mfc *mfc; 2348 struct bw_meter *x; 2349 2350 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2351 return EOPNOTSUPP; 2352 2353 crit_enter(); 2354 /* Find the corresponding MFC entry */ 2355 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2356 if (mfc == NULL) { 2357 crit_exit(); 2358 return EADDRNOTAVAIL; 2359 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2360 /* 2361 * Delete all bw_meter entries for this mfc 2362 */ 2363 struct bw_meter *list; 2364 2365 list = mfc->mfc_bw_meter; 2366 mfc->mfc_bw_meter = NULL; 2367 crit_exit(); 2368 free_bw_list(list); 2369 return 0; 2370 } else { /* Delete a single bw_meter entry */ 2371 struct bw_meter *prev; 2372 uint32_t flags = 0; 2373 2374 flags = compute_bw_meter_flags(req); 2375 2376 /* Find the bw_meter entry to delete */ 2377 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2378 prev = x, x = x->bm_mfc_next) { 2379 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2380 &req->bu_threshold.b_time, ==)) && 2381 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2382 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2383 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2384 break; 2385 } 2386 if (x != NULL) { /* Delete entry from the list for this MFC */ 2387 if (prev != NULL) 2388 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2389 else 2390 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2391 crit_exit(); 2392 2393 unschedule_bw_meter(x); 2394 /* Free the bw_meter entry */ 2395 free(x, M_BWMETER); 2396 return 0; 2397 } else { 2398 crit_exit(); 2399 return EINVAL; 2400 } 2401 } 2402 /* NOTREACHED */ 2403 } 2404 2405 /* 2406 * Perform bandwidth measurement processing that may result in an upcall 2407 */ 2408 static void 2409 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2410 { 2411 struct timeval delta; 2412 2413 crit_enter(); 2414 delta = *nowp; 2415 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2416 2417 if (x->bm_flags & BW_METER_GEQ) { 2418 /* 2419 * Processing for ">=" type of bw_meter entry 2420 */ 2421 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2422 /* Reset the bw_meter entry */ 2423 x->bm_start_time = *nowp; 2424 x->bm_measured.b_packets = 0; 2425 x->bm_measured.b_bytes = 0; 2426 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2427 } 2428 2429 /* Record that a packet is received */ 2430 x->bm_measured.b_packets++; 2431 x->bm_measured.b_bytes += plen; 2432 2433 /* 2434 * Test if we should deliver an upcall 2435 */ 2436 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2437 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2438 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2439 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2440 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2441 /* Prepare an upcall for delivery */ 2442 bw_meter_prepare_upcall(x, nowp); 2443 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2444 } 2445 } 2446 } else if (x->bm_flags & BW_METER_LEQ) { 2447 /* 2448 * Processing for "<=" type of bw_meter entry 2449 */ 2450 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2451 /* 2452 * We are behind time with the multicast forwarding table 2453 * scanning for "<=" type of bw_meter entries, so test now 2454 * if we should deliver an upcall. 2455 */ 2456 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2457 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2458 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2459 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2460 /* Prepare an upcall for delivery */ 2461 bw_meter_prepare_upcall(x, nowp); 2462 } 2463 /* Reschedule the bw_meter entry */ 2464 unschedule_bw_meter(x); 2465 schedule_bw_meter(x, nowp); 2466 } 2467 2468 /* Record that a packet is received */ 2469 x->bm_measured.b_packets++; 2470 x->bm_measured.b_bytes += plen; 2471 2472 /* 2473 * Test if we should restart the measuring interval 2474 */ 2475 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2476 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2477 (x->bm_flags & BW_METER_UNIT_BYTES && 2478 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2479 /* Don't restart the measuring interval */ 2480 } else { 2481 /* Do restart the measuring interval */ 2482 /* 2483 * XXX: note that we don't unschedule and schedule, because this 2484 * might be too much overhead per packet. Instead, when we process 2485 * all entries for a given timer hash bin, we check whether it is 2486 * really a timeout. If not, we reschedule at that time. 2487 */ 2488 x->bm_start_time = *nowp; 2489 x->bm_measured.b_packets = 0; 2490 x->bm_measured.b_bytes = 0; 2491 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2492 } 2493 } 2494 crit_exit(); 2495 } 2496 2497 /* 2498 * Prepare a bandwidth-related upcall 2499 */ 2500 static void 2501 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2502 { 2503 struct timeval delta; 2504 struct bw_upcall *u; 2505 2506 crit_enter(); 2507 2508 /* 2509 * Compute the measured time interval 2510 */ 2511 delta = *nowp; 2512 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2513 2514 /* 2515 * If there are too many pending upcalls, deliver them now 2516 */ 2517 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2518 bw_upcalls_send(); 2519 2520 /* 2521 * Set the bw_upcall entry 2522 */ 2523 u = &bw_upcalls[bw_upcalls_n++]; 2524 u->bu_src = x->bm_mfc->mfc_origin; 2525 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2526 u->bu_threshold.b_time = x->bm_threshold.b_time; 2527 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2528 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2529 u->bu_measured.b_time = delta; 2530 u->bu_measured.b_packets = x->bm_measured.b_packets; 2531 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2532 u->bu_flags = 0; 2533 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2534 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2535 if (x->bm_flags & BW_METER_UNIT_BYTES) 2536 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2537 if (x->bm_flags & BW_METER_GEQ) 2538 u->bu_flags |= BW_UPCALL_GEQ; 2539 if (x->bm_flags & BW_METER_LEQ) 2540 u->bu_flags |= BW_UPCALL_LEQ; 2541 2542 crit_exit(); 2543 } 2544 2545 /* 2546 * Send the pending bandwidth-related upcalls 2547 */ 2548 static void 2549 bw_upcalls_send(void) 2550 { 2551 struct mbuf *m; 2552 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2553 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2554 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2555 0, /* unused2 */ 2556 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2557 0, /* im_mbz */ 2558 0, /* im_vif */ 2559 0, /* unused3 */ 2560 { 0 }, /* im_src */ 2561 { 0 } }; /* im_dst */ 2562 2563 if (bw_upcalls_n == 0) 2564 return; /* No pending upcalls */ 2565 2566 bw_upcalls_n = 0; 2567 2568 /* 2569 * Allocate a new mbuf, initialize it with the header and 2570 * the payload for the pending calls. 2571 */ 2572 MGETHDR(m, MB_DONTWAIT, MT_HEADER); 2573 if (m == NULL) { 2574 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2575 return; 2576 } 2577 2578 m->m_len = m->m_pkthdr.len = 0; 2579 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2580 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2581 2582 /* 2583 * Send the upcalls 2584 * XXX do we need to set the address in k_igmpsrc ? 2585 */ 2586 mrtstat.mrts_upcalls++; 2587 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2588 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2589 ++mrtstat.mrts_upq_sockfull; 2590 } 2591 } 2592 2593 /* 2594 * Compute the timeout hash value for the bw_meter entries 2595 */ 2596 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2597 do { \ 2598 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2599 \ 2600 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2601 (hash) = next_timeval.tv_sec; \ 2602 if (next_timeval.tv_usec) \ 2603 (hash)++; /* XXX: make sure we don't timeout early */ \ 2604 (hash) %= BW_METER_BUCKETS; \ 2605 } while (0) 2606 2607 /* 2608 * Schedule a timer to process periodically bw_meter entry of type "<=" 2609 * by linking the entry in the proper hash bucket. 2610 */ 2611 static void 2612 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2613 { 2614 int time_hash; 2615 2616 if (!(x->bm_flags & BW_METER_LEQ)) 2617 return; /* XXX: we schedule timers only for "<=" entries */ 2618 2619 /* 2620 * Reset the bw_meter entry 2621 */ 2622 crit_enter(); 2623 x->bm_start_time = *nowp; 2624 x->bm_measured.b_packets = 0; 2625 x->bm_measured.b_bytes = 0; 2626 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2627 crit_exit(); 2628 2629 /* 2630 * Compute the timeout hash value and insert the entry 2631 */ 2632 BW_METER_TIMEHASH(x, time_hash); 2633 x->bm_time_next = bw_meter_timers[time_hash]; 2634 bw_meter_timers[time_hash] = x; 2635 x->bm_time_hash = time_hash; 2636 } 2637 2638 /* 2639 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2640 * by removing the entry from the proper hash bucket. 2641 */ 2642 static void 2643 unschedule_bw_meter(struct bw_meter *x) 2644 { 2645 int time_hash; 2646 struct bw_meter *prev, *tmp; 2647 2648 if (!(x->bm_flags & BW_METER_LEQ)) 2649 return; /* XXX: we schedule timers only for "<=" entries */ 2650 2651 /* 2652 * Compute the timeout hash value and delete the entry 2653 */ 2654 time_hash = x->bm_time_hash; 2655 if (time_hash >= BW_METER_BUCKETS) 2656 return; /* Entry was not scheduled */ 2657 2658 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2659 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2660 if (tmp == x) 2661 break; 2662 2663 if (tmp == NULL) 2664 panic("unschedule_bw_meter: bw_meter entry not found"); 2665 2666 if (prev != NULL) 2667 prev->bm_time_next = x->bm_time_next; 2668 else 2669 bw_meter_timers[time_hash] = x->bm_time_next; 2670 2671 x->bm_time_next = NULL; 2672 x->bm_time_hash = BW_METER_BUCKETS; 2673 } 2674 2675 2676 /* 2677 * Process all "<=" type of bw_meter that should be processed now, 2678 * and for each entry prepare an upcall if necessary. Each processed 2679 * entry is rescheduled again for the (periodic) processing. 2680 * 2681 * This is run periodically (once per second normally). On each round, 2682 * all the potentially matching entries are in the hash slot that we are 2683 * looking at. 2684 */ 2685 static void 2686 bw_meter_process() 2687 { 2688 static uint32_t last_tv_sec; /* last time we processed this */ 2689 2690 uint32_t loops; 2691 int i; 2692 struct timeval now, process_endtime; 2693 2694 GET_TIME(now); 2695 if (last_tv_sec == now.tv_sec) 2696 return; /* nothing to do */ 2697 2698 crit_enter(); 2699 loops = now.tv_sec - last_tv_sec; 2700 last_tv_sec = now.tv_sec; 2701 if (loops > BW_METER_BUCKETS) 2702 loops = BW_METER_BUCKETS; 2703 2704 /* 2705 * Process all bins of bw_meter entries from the one after the last 2706 * processed to the current one. On entry, i points to the last bucket 2707 * visited, so we need to increment i at the beginning of the loop. 2708 */ 2709 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2710 struct bw_meter *x, *tmp_list; 2711 2712 if (++i >= BW_METER_BUCKETS) 2713 i = 0; 2714 2715 /* Disconnect the list of bw_meter entries from the bin */ 2716 tmp_list = bw_meter_timers[i]; 2717 bw_meter_timers[i] = NULL; 2718 2719 /* Process the list of bw_meter entries */ 2720 while (tmp_list != NULL) { 2721 x = tmp_list; 2722 tmp_list = tmp_list->bm_time_next; 2723 2724 /* Test if the time interval is over */ 2725 process_endtime = x->bm_start_time; 2726 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2727 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2728 /* Not yet: reschedule, but don't reset */ 2729 int time_hash; 2730 2731 BW_METER_TIMEHASH(x, time_hash); 2732 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2733 /* 2734 * XXX: somehow the bin processing is a bit ahead of time. 2735 * Put the entry in the next bin. 2736 */ 2737 if (++time_hash >= BW_METER_BUCKETS) 2738 time_hash = 0; 2739 } 2740 x->bm_time_next = bw_meter_timers[time_hash]; 2741 bw_meter_timers[time_hash] = x; 2742 x->bm_time_hash = time_hash; 2743 2744 continue; 2745 } 2746 2747 /* 2748 * Test if we should deliver an upcall 2749 */ 2750 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2751 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2752 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2753 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2754 /* Prepare an upcall for delivery */ 2755 bw_meter_prepare_upcall(x, &now); 2756 } 2757 2758 /* 2759 * Reschedule for next processing 2760 */ 2761 schedule_bw_meter(x, &now); 2762 } 2763 } 2764 crit_exit(); 2765 2766 /* Send all upcalls that are pending delivery */ 2767 bw_upcalls_send(); 2768 } 2769 2770 /* 2771 * A periodic function for sending all upcalls that are pending delivery 2772 */ 2773 static void 2774 expire_bw_upcalls_send(void *unused) 2775 { 2776 bw_upcalls_send(); 2777 2778 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2779 expire_bw_upcalls_send, NULL); 2780 } 2781 2782 /* 2783 * A periodic function for periodic scanning of the multicast forwarding 2784 * table for processing all "<=" bw_meter entries. 2785 */ 2786 static void 2787 expire_bw_meter_process(void *unused) 2788 { 2789 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2790 bw_meter_process(); 2791 2792 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 2793 expire_bw_meter_process, NULL); 2794 } 2795 2796 /* 2797 * End of bandwidth monitoring code 2798 */ 2799 2800 #ifdef PIM 2801 /* 2802 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2803 * 2804 */ 2805 static int 2806 pim_register_send(struct ip *ip, struct vif *vifp, 2807 struct mbuf *m, struct mfc *rt) 2808 { 2809 struct mbuf *mb_copy, *mm; 2810 2811 if (mrtdebug & DEBUG_PIM) 2812 log(LOG_DEBUG, "pim_register_send: "); 2813 2814 mb_copy = pim_register_prepare(ip, m); 2815 if (mb_copy == NULL) 2816 return ENOBUFS; 2817 2818 /* 2819 * Send all the fragments. Note that the mbuf for each fragment 2820 * is freed by the sending machinery. 2821 */ 2822 for (mm = mb_copy; mm; mm = mb_copy) { 2823 mb_copy = mm->m_nextpkt; 2824 mm->m_nextpkt = 0; 2825 mm = m_pullup(mm, sizeof(struct ip)); 2826 if (mm != NULL) { 2827 ip = mtod(mm, struct ip *); 2828 if ((mrt_api_config & MRT_MFC_RP) && 2829 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2830 pim_register_send_rp(ip, vifp, mm, rt); 2831 } else { 2832 pim_register_send_upcall(ip, vifp, mm, rt); 2833 } 2834 } 2835 } 2836 2837 return 0; 2838 } 2839 2840 /* 2841 * Return a copy of the data packet that is ready for PIM Register 2842 * encapsulation. 2843 * XXX: Note that in the returned copy the IP header is a valid one. 2844 */ 2845 static struct mbuf * 2846 pim_register_prepare(struct ip *ip, struct mbuf *m) 2847 { 2848 struct mbuf *mb_copy = NULL; 2849 int mtu; 2850 2851 /* Take care of delayed checksums */ 2852 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2853 in_delayed_cksum(m); 2854 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2855 } 2856 2857 /* 2858 * Copy the old packet & pullup its IP header into the 2859 * new mbuf so we can modify it. 2860 */ 2861 mb_copy = m_copypacket(m, MB_DONTWAIT); 2862 if (mb_copy == NULL) 2863 return NULL; 2864 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2865 if (mb_copy == NULL) 2866 return NULL; 2867 2868 /* take care of the TTL */ 2869 ip = mtod(mb_copy, struct ip *); 2870 --ip->ip_ttl; 2871 2872 /* Compute the MTU after the PIM Register encapsulation */ 2873 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2874 2875 if (ip->ip_len <= mtu) { 2876 /* Turn the IP header into a valid one */ 2877 ip->ip_len = htons(ip->ip_len); 2878 ip->ip_off = htons(ip->ip_off); 2879 ip->ip_sum = 0; 2880 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2881 } else { 2882 /* Fragment the packet */ 2883 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2884 m_freem(mb_copy); 2885 return NULL; 2886 } 2887 } 2888 return mb_copy; 2889 } 2890 2891 /* 2892 * Send an upcall with the data packet to the user-level process. 2893 */ 2894 static int 2895 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2896 struct mbuf *mb_copy, struct mfc *rt) 2897 { 2898 struct mbuf *mb_first; 2899 int len = ntohs(ip->ip_len); 2900 struct igmpmsg *im; 2901 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2902 2903 /* 2904 * Add a new mbuf with an upcall header 2905 */ 2906 MGETHDR(mb_first, MB_DONTWAIT, MT_HEADER); 2907 if (mb_first == NULL) { 2908 m_freem(mb_copy); 2909 return ENOBUFS; 2910 } 2911 mb_first->m_data += max_linkhdr; 2912 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2913 mb_first->m_len = sizeof(struct igmpmsg); 2914 mb_first->m_next = mb_copy; 2915 2916 /* Send message to routing daemon */ 2917 im = mtod(mb_first, struct igmpmsg *); 2918 im->im_msgtype = IGMPMSG_WHOLEPKT; 2919 im->im_mbz = 0; 2920 im->im_vif = vifp - viftable; 2921 im->im_src = ip->ip_src; 2922 im->im_dst = ip->ip_dst; 2923 2924 k_igmpsrc.sin_addr = ip->ip_src; 2925 2926 mrtstat.mrts_upcalls++; 2927 2928 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2929 if (mrtdebug & DEBUG_PIM) 2930 log(LOG_WARNING, 2931 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2932 ++mrtstat.mrts_upq_sockfull; 2933 return ENOBUFS; 2934 } 2935 2936 /* Keep statistics */ 2937 pimstat.pims_snd_registers_msgs++; 2938 pimstat.pims_snd_registers_bytes += len; 2939 2940 return 0; 2941 } 2942 2943 /* 2944 * Encapsulate the data packet in PIM Register message and send it to the RP. 2945 */ 2946 static int 2947 pim_register_send_rp(struct ip *ip, struct vif *vifp, 2948 struct mbuf *mb_copy, struct mfc *rt) 2949 { 2950 struct mbuf *mb_first; 2951 struct ip *ip_outer; 2952 struct pim_encap_pimhdr *pimhdr; 2953 int len = ntohs(ip->ip_len); 2954 vifi_t vifi = rt->mfc_parent; 2955 2956 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2957 m_freem(mb_copy); 2958 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2959 } 2960 2961 /* 2962 * Add a new mbuf with the encapsulating header 2963 */ 2964 MGETHDR(mb_first, MB_DONTWAIT, MT_HEADER); 2965 if (mb_first == NULL) { 2966 m_freem(mb_copy); 2967 return ENOBUFS; 2968 } 2969 mb_first->m_data += max_linkhdr; 2970 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2971 mb_first->m_next = mb_copy; 2972 2973 mb_first->m_pkthdr.len = len + mb_first->m_len; 2974 2975 /* 2976 * Fill in the encapsulating IP and PIM header 2977 */ 2978 ip_outer = mtod(mb_first, struct ip *); 2979 *ip_outer = pim_encap_iphdr; 2980 #ifdef RANDOM_IP_ID 2981 ip_outer->ip_id = ip_randomid(); 2982 #else 2983 ip_outer->ip_id = htons(ip_id++); 2984 #endif 2985 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2986 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2987 ip_outer->ip_dst = rt->mfc_rp; 2988 /* 2989 * Copy the inner header TOS to the outer header, and take care of the 2990 * IP_DF bit. 2991 */ 2992 ip_outer->ip_tos = ip->ip_tos; 2993 if (ntohs(ip->ip_off) & IP_DF) 2994 ip_outer->ip_off |= IP_DF; 2995 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2996 + sizeof(pim_encap_iphdr)); 2997 *pimhdr = pim_encap_pimhdr; 2998 /* If the iif crosses a border, set the Border-bit */ 2999 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3000 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3001 3002 mb_first->m_data += sizeof(pim_encap_iphdr); 3003 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3004 mb_first->m_data -= sizeof(pim_encap_iphdr); 3005 3006 if (vifp->v_rate_limit == 0) 3007 tbf_send_packet(vifp, mb_first); 3008 else 3009 tbf_control(vifp, mb_first, ip, ip_outer->ip_len); 3010 3011 /* Keep statistics */ 3012 pimstat.pims_snd_registers_msgs++; 3013 pimstat.pims_snd_registers_bytes += len; 3014 3015 return 0; 3016 } 3017 3018 /* 3019 * PIM-SMv2 and PIM-DM messages processing. 3020 * Receives and verifies the PIM control messages, and passes them 3021 * up to the listening socket, using rip_input(). 3022 * The only message with special processing is the PIM_REGISTER message 3023 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3024 * is passed to if_simloop(). 3025 */ 3026 void 3027 pim_input(struct mbuf *m, ...) 3028 { 3029 int off, proto; 3030 struct ip *ip = mtod(m, struct ip *); 3031 struct pim *pim; 3032 int minlen; 3033 int datalen = ip->ip_len; 3034 int ip_tos; 3035 int iphlen; 3036 __va_list ap; 3037 3038 __va_start(ap, m); 3039 off = __va_arg(ap, int); 3040 proto = __va_arg(ap, int); 3041 __va_end(ap); 3042 3043 iphlen = off; 3044 3045 /* Keep statistics */ 3046 pimstat.pims_rcv_total_msgs++; 3047 pimstat.pims_rcv_total_bytes += datalen; 3048 3049 /* 3050 * Validate lengths 3051 */ 3052 if (datalen < PIM_MINLEN) { 3053 pimstat.pims_rcv_tooshort++; 3054 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3055 datalen, (u_long)ip->ip_src.s_addr); 3056 m_freem(m); 3057 return; 3058 } 3059 3060 /* 3061 * If the packet is at least as big as a REGISTER, go agead 3062 * and grab the PIM REGISTER header size, to avoid another 3063 * possible m_pullup() later. 3064 * 3065 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3066 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3067 */ 3068 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3069 /* 3070 * Get the IP and PIM headers in contiguous memory, and 3071 * possibly the PIM REGISTER header. 3072 */ 3073 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3074 (m = m_pullup(m, minlen)) == 0) { 3075 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3076 return; 3077 } 3078 /* m_pullup() may have given us a new mbuf so reset ip. */ 3079 ip = mtod(m, struct ip *); 3080 ip_tos = ip->ip_tos; 3081 3082 /* adjust mbuf to point to the PIM header */ 3083 m->m_data += iphlen; 3084 m->m_len -= iphlen; 3085 pim = mtod(m, struct pim *); 3086 3087 /* 3088 * Validate checksum. If PIM REGISTER, exclude the data packet. 3089 * 3090 * XXX: some older PIMv2 implementations don't make this distinction, 3091 * so for compatibility reason perform the checksum over part of the 3092 * message, and if error, then over the whole message. 3093 */ 3094 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3095 /* do nothing, checksum okay */ 3096 } else if (in_cksum(m, datalen)) { 3097 pimstat.pims_rcv_badsum++; 3098 if (mrtdebug & DEBUG_PIM) 3099 log(LOG_DEBUG, "pim_input: invalid checksum"); 3100 m_freem(m); 3101 return; 3102 } 3103 3104 /* PIM version check */ 3105 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3106 pimstat.pims_rcv_badversion++; 3107 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3108 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3109 m_freem(m); 3110 return; 3111 } 3112 3113 /* restore mbuf back to the outer IP */ 3114 m->m_data -= iphlen; 3115 m->m_len += iphlen; 3116 3117 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3118 /* 3119 * Since this is a REGISTER, we'll make a copy of the register 3120 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3121 * routing daemon. 3122 */ 3123 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3124 struct mbuf *mcp; 3125 struct ip *encap_ip; 3126 u_int32_t *reghdr; 3127 3128 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3129 if (mrtdebug & DEBUG_PIM) 3130 log(LOG_DEBUG, 3131 "pim_input: register vif not set: %d\n", reg_vif_num); 3132 m_freem(m); 3133 return; 3134 } 3135 3136 /* 3137 * Validate length 3138 */ 3139 if (datalen < PIM_REG_MINLEN) { 3140 pimstat.pims_rcv_tooshort++; 3141 pimstat.pims_rcv_badregisters++; 3142 log(LOG_ERR, 3143 "pim_input: register packet size too small %d from %lx\n", 3144 datalen, (u_long)ip->ip_src.s_addr); 3145 m_freem(m); 3146 return; 3147 } 3148 3149 reghdr = (u_int32_t *)(pim + 1); 3150 encap_ip = (struct ip *)(reghdr + 1); 3151 3152 if (mrtdebug & DEBUG_PIM) { 3153 log(LOG_DEBUG, 3154 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3155 (u_long)ntohl(encap_ip->ip_src.s_addr), 3156 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3157 ntohs(encap_ip->ip_len)); 3158 } 3159 3160 /* verify the version number of the inner packet */ 3161 if (encap_ip->ip_v != IPVERSION) { 3162 pimstat.pims_rcv_badregisters++; 3163 if (mrtdebug & DEBUG_PIM) { 3164 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3165 "of the inner packet\n", encap_ip->ip_v); 3166 } 3167 m_freem(m); 3168 return; 3169 } 3170 3171 /* verify the inner packet is destined to a mcast group */ 3172 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 3173 pimstat.pims_rcv_badregisters++; 3174 if (mrtdebug & DEBUG_PIM) 3175 log(LOG_DEBUG, 3176 "pim_input: inner packet of register is not " 3177 "multicast %lx\n", 3178 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3179 m_freem(m); 3180 return; 3181 } 3182 3183 /* If a NULL_REGISTER, pass it to the daemon */ 3184 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3185 goto pim_input_to_daemon; 3186 3187 /* 3188 * Copy the TOS from the outer IP header to the inner IP header. 3189 */ 3190 if (encap_ip->ip_tos != ip_tos) { 3191 /* Outer TOS -> inner TOS */ 3192 encap_ip->ip_tos = ip_tos; 3193 /* Recompute the inner header checksum. Sigh... */ 3194 3195 /* adjust mbuf to point to the inner IP header */ 3196 m->m_data += (iphlen + PIM_MINLEN); 3197 m->m_len -= (iphlen + PIM_MINLEN); 3198 3199 encap_ip->ip_sum = 0; 3200 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3201 3202 /* restore mbuf to point back to the outer IP header */ 3203 m->m_data -= (iphlen + PIM_MINLEN); 3204 m->m_len += (iphlen + PIM_MINLEN); 3205 } 3206 3207 /* 3208 * Decapsulate the inner IP packet and loopback to forward it 3209 * as a normal multicast packet. Also, make a copy of the 3210 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3211 * to pass to the daemon later, so it can take the appropriate 3212 * actions (e.g., send back PIM_REGISTER_STOP). 3213 * XXX: here m->m_data points to the outer IP header. 3214 */ 3215 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3216 if (mcp == NULL) { 3217 log(LOG_ERR, 3218 "pim_input: pim register: could not copy register head\n"); 3219 m_freem(m); 3220 return; 3221 } 3222 3223 /* Keep statistics */ 3224 /* XXX: registers_bytes include only the encap. mcast pkt */ 3225 pimstat.pims_rcv_registers_msgs++; 3226 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3227 3228 /* 3229 * forward the inner ip packet; point m_data at the inner ip. 3230 */ 3231 m_adj(m, iphlen + PIM_MINLEN); 3232 3233 if (mrtdebug & DEBUG_PIM) { 3234 log(LOG_DEBUG, 3235 "pim_input: forwarding decapsulated register: " 3236 "src %lx, dst %lx, vif %d\n", 3237 (u_long)ntohl(encap_ip->ip_src.s_addr), 3238 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3239 reg_vif_num); 3240 } 3241 if_simloop(viftable[reg_vif_num].v_ifp, m, dst.sin_family, 0); 3242 3243 /* prepare the register head to send to the mrouting daemon */ 3244 m = mcp; 3245 } 3246 3247 pim_input_to_daemon: 3248 /* 3249 * Pass the PIM message up to the daemon; if it is a Register message, 3250 * pass the 'head' only up to the daemon. This includes the 3251 * outer IP header, PIM header, PIM-Register header and the 3252 * inner IP header. 3253 * XXX: the outer IP header pkt size of a Register is not adjust to 3254 * reflect the fact that the inner multicast data is truncated. 3255 */ 3256 rip_input(m, iphlen, proto); 3257 3258 return; 3259 } 3260 #endif /* PIM */ 3261 3262 static int 3263 ip_mroute_modevent(module_t mod, int type, void *unused) 3264 { 3265 switch (type) { 3266 case MOD_LOAD: 3267 crit_enter(); 3268 /* XXX Protect against multiple loading */ 3269 ip_mcast_src = X_ip_mcast_src; 3270 ip_mforward = X_ip_mforward; 3271 ip_mrouter_done = X_ip_mrouter_done; 3272 ip_mrouter_get = X_ip_mrouter_get; 3273 ip_mrouter_set = X_ip_mrouter_set; 3274 ip_rsvp_force_done = X_ip_rsvp_force_done; 3275 ip_rsvp_vif = X_ip_rsvp_vif; 3276 ipip_input = X_ipip_input; 3277 legal_vif_num = X_legal_vif_num; 3278 mrt_ioctl = X_mrt_ioctl; 3279 rsvp_input_p = X_rsvp_input; 3280 crit_exit(); 3281 break; 3282 3283 case MOD_UNLOAD: 3284 if (ip_mrouter) 3285 return EINVAL; 3286 3287 crit_enter(); 3288 ip_mcast_src = NULL; 3289 ip_mforward = NULL; 3290 ip_mrouter_done = NULL; 3291 ip_mrouter_get = NULL; 3292 ip_mrouter_set = NULL; 3293 ip_rsvp_force_done = NULL; 3294 ip_rsvp_vif = NULL; 3295 ipip_input = NULL; 3296 legal_vif_num = NULL; 3297 mrt_ioctl = NULL; 3298 rsvp_input_p = NULL; 3299 crit_exit(); 3300 break; 3301 } 3302 return 0; 3303 } 3304 3305 static moduledata_t ip_mroutemod = { 3306 "ip_mroute", 3307 ip_mroute_modevent, 3308 0 3309 }; 3310 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3311