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 1.15 2004/09/16 23:30:10 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/time.h> 43 #include <sys/in_cksum.h> 44 45 #include <machine/stdarg.h> 46 47 #include <net/if.h> 48 #include <net/netisr.h> 49 #include <net/route.h> 50 #include <netinet/in.h> 51 #include <netinet/igmp.h> 52 #include <netinet/in_systm.h> 53 #include <netinet/in_var.h> 54 #include <netinet/ip.h> 55 #include "ip_mroute.h" 56 #include <netinet/ip_var.h> 57 #ifdef PIM 58 #include <netinet/pim.h> 59 #include <netinet/pim_var.h> 60 #endif 61 #include <netinet/udp.h> 62 63 /* 64 * Control debugging code for rsvp and multicast routing code. 65 * Can only set them with the debugger. 66 */ 67 static u_int rsvpdebug; /* non-zero enables debugging */ 68 69 static u_int mrtdebug; /* any set of the flags below */ 70 71 #define DEBUG_MFC 0x02 72 #define DEBUG_FORWARD 0x04 73 #define DEBUG_EXPIRE 0x08 74 #define DEBUG_XMIT 0x10 75 #define DEBUG_PIM 0x20 76 77 #define VIFI_INVALID ((vifi_t) -1) 78 79 #define M_HASCL(m) ((m)->m_flags & M_EXT) 80 81 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 82 83 static struct mrtstat mrtstat; 84 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 85 &mrtstat, mrtstat, 86 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 87 88 static struct mfc *mfctable[MFCTBLSIZ]; 89 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 90 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 91 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 92 93 static struct vif viftable[MAXVIFS]; 94 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 95 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 96 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 97 98 static u_char nexpire[MFCTBLSIZ]; 99 100 static struct callout expire_upcalls_ch; 101 static struct callout tbf_reprocess_q_ch; 102 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 103 #define UPCALL_EXPIRE 6 /* number of timeouts */ 104 105 /* 106 * Define the token bucket filter structures 107 * tbftable -> each vif has one of these for storing info 108 */ 109 110 static struct tbf tbftable[MAXVIFS]; 111 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 112 113 /* 114 * 'Interfaces' associated with decapsulator (so we can tell 115 * packets that went through it from ones that get reflected 116 * by a broken gateway). These interfaces are never linked into 117 * the system ifnet list & no routes point to them. I.e., packets 118 * can't be sent this way. They only exist as a placeholder for 119 * multicast source verification. 120 */ 121 static struct ifnet multicast_decap_if[MAXVIFS]; 122 123 #define ENCAP_TTL 64 124 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 125 126 /* prototype IP hdr for encapsulated packets */ 127 static struct ip multicast_encap_iphdr = { 128 #if BYTE_ORDER == LITTLE_ENDIAN 129 sizeof(struct ip) >> 2, IPVERSION, 130 #else 131 IPVERSION, sizeof(struct ip) >> 2, 132 #endif 133 0, /* tos */ 134 sizeof(struct ip), /* total length */ 135 0, /* id */ 136 0, /* frag offset */ 137 ENCAP_TTL, ENCAP_PROTO, 138 0, /* checksum */ 139 }; 140 141 /* 142 * Bandwidth meter variables and constants 143 */ 144 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 145 /* 146 * Pending timeouts are stored in a hash table, the key being the 147 * expiration time. Periodically, the entries are analysed and processed. 148 */ 149 #define BW_METER_BUCKETS 1024 150 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 151 static struct callout bw_meter_ch; 152 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 153 154 /* 155 * Pending upcalls are stored in a vector which is flushed when 156 * full, or periodically 157 */ 158 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 159 static u_int bw_upcalls_n; /* # of pending upcalls */ 160 static struct callout bw_upcalls_ch; 161 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 162 163 #ifdef PIM 164 static struct pimstat pimstat; 165 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 166 &pimstat, pimstat, 167 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 168 169 /* 170 * Note: the PIM Register encapsulation adds the following in front of a 171 * data packet: 172 * 173 * struct pim_encap_hdr { 174 * struct ip ip; 175 * struct pim_encap_pimhdr pim; 176 * } 177 * 178 */ 179 180 struct pim_encap_pimhdr { 181 struct pim pim; 182 uint32_t flags; 183 }; 184 185 static struct ip pim_encap_iphdr = { 186 #if BYTE_ORDER == LITTLE_ENDIAN 187 sizeof(struct ip) >> 2, 188 IPVERSION, 189 #else 190 IPVERSION, 191 sizeof(struct ip) >> 2, 192 #endif 193 0, /* tos */ 194 sizeof(struct ip), /* total length */ 195 0, /* id */ 196 0, /* frag offset */ 197 ENCAP_TTL, 198 IPPROTO_PIM, 199 0, /* checksum */ 200 }; 201 202 static struct pim_encap_pimhdr pim_encap_pimhdr = { 203 { 204 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 205 0, /* reserved */ 206 0, /* checksum */ 207 }, 208 0 /* flags */ 209 }; 210 211 static struct ifnet multicast_register_if; 212 static vifi_t reg_vif_num = VIFI_INVALID; 213 #endif /* PIM */ 214 215 /* 216 * Private variables. 217 */ 218 static vifi_t numvifs; 219 static int have_encap_tunnel; 220 221 /* 222 * one-back cache used by ipip_input to locate a tunnel's vif 223 * given a datagram's src ip address. 224 */ 225 static u_long last_encap_src; 226 static struct vif *last_encap_vif; 227 228 static u_long X_ip_mcast_src(int vifi); 229 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 230 struct mbuf *m, struct ip_moptions *imo); 231 static int X_ip_mrouter_done(void); 232 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 233 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 234 static int X_legal_vif_num(int vif); 235 static int X_mrt_ioctl(int cmd, caddr_t data); 236 237 static int get_sg_cnt(struct sioc_sg_req *); 238 static int get_vif_cnt(struct sioc_vif_req *); 239 static int ip_mrouter_init(struct socket *, int); 240 static int add_vif(struct vifctl *); 241 static int del_vif(vifi_t); 242 static int add_mfc(struct mfcctl2 *); 243 static int del_mfc(struct mfcctl2 *); 244 static int set_api_config(uint32_t *); /* chose API capabilities */ 245 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 246 static int set_assert(int); 247 static void expire_upcalls(void *); 248 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 249 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 250 static void encap_send(struct ip *, struct vif *, struct mbuf *); 251 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 252 static void tbf_queue(struct vif *, struct mbuf *); 253 static void tbf_process_q(struct vif *); 254 static void tbf_reprocess_q(void *); 255 static int tbf_dq_sel(struct vif *, struct ip *); 256 static void tbf_send_packet(struct vif *, struct mbuf *); 257 static void tbf_update_tokens(struct vif *); 258 static int priority(struct vif *, struct ip *); 259 260 /* 261 * Bandwidth monitoring 262 */ 263 static void free_bw_list(struct bw_meter *list); 264 static int add_bw_upcall(struct bw_upcall *); 265 static int del_bw_upcall(struct bw_upcall *); 266 static void bw_meter_receive_packet(struct bw_meter *x, int plen, 267 struct timeval *nowp); 268 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 269 static void bw_upcalls_send(void); 270 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 271 static void unschedule_bw_meter(struct bw_meter *x); 272 static void bw_meter_process(void); 273 static void expire_bw_upcalls_send(void *); 274 static void expire_bw_meter_process(void *); 275 276 #ifdef PIM 277 static int pim_register_send(struct ip *, struct vif *, 278 struct mbuf *, struct mfc *); 279 static int pim_register_send_rp(struct ip *, struct vif *, 280 struct mbuf *, struct mfc *); 281 static int pim_register_send_upcall(struct ip *, struct vif *, 282 struct mbuf *, struct mfc *); 283 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 284 #endif 285 286 /* 287 * whether or not special PIM assert processing is enabled. 288 */ 289 static int pim_assert; 290 /* 291 * Rate limit for assert notification messages, in usec 292 */ 293 #define ASSERT_MSG_TIME 3000000 294 295 /* 296 * Kernel multicast routing API capabilities and setup. 297 * If more API capabilities are added to the kernel, they should be 298 * recorded in `mrt_api_support'. 299 */ 300 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 301 MRT_MFC_FLAGS_BORDER_VIF | 302 MRT_MFC_RP | 303 MRT_MFC_BW_UPCALL); 304 static uint32_t mrt_api_config = 0; 305 306 /* 307 * Hash function for a source, group entry 308 */ 309 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 310 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 311 312 /* 313 * Find a route for a given origin IP address and Multicast group address 314 * Type of service parameter to be added in the future!!! 315 * Statistics are updated by the caller if needed 316 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 317 */ 318 static struct mfc * 319 mfc_find(in_addr_t o, in_addr_t g) 320 { 321 struct mfc *rt; 322 323 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 324 if ((rt->mfc_origin.s_addr == o) && 325 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 326 break; 327 return rt; 328 } 329 330 /* 331 * Macros to compute elapsed time efficiently 332 * Borrowed from Van Jacobson's scheduling code 333 */ 334 #define TV_DELTA(a, b, delta) { \ 335 int xxs; \ 336 delta = (a).tv_usec - (b).tv_usec; \ 337 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 338 switch (xxs) { \ 339 case 2: \ 340 delta += 1000000; \ 341 /* FALLTHROUGH */ \ 342 case 1: \ 343 delta += 1000000; \ 344 break; \ 345 default: \ 346 delta += (1000000 * xxs); \ 347 } \ 348 } \ 349 } 350 351 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 352 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 353 354 /* 355 * Handle MRT setsockopt commands to modify the multicast routing tables. 356 */ 357 static int 358 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 359 { 360 int error, optval; 361 vifi_t vifi; 362 struct vifctl vifc; 363 struct mfcctl2 mfc; 364 struct bw_upcall bw_upcall; 365 uint32_t i; 366 367 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 368 return EPERM; 369 370 error = 0; 371 switch (sopt->sopt_name) { 372 case MRT_INIT: 373 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 374 if (error) 375 break; 376 error = ip_mrouter_init(so, optval); 377 break; 378 379 case MRT_DONE: 380 error = ip_mrouter_done(); 381 break; 382 383 case MRT_ADD_VIF: 384 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 385 if (error) 386 break; 387 error = add_vif(&vifc); 388 break; 389 390 case MRT_DEL_VIF: 391 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 392 if (error) 393 break; 394 error = del_vif(vifi); 395 break; 396 397 case MRT_ADD_MFC: 398 case MRT_DEL_MFC: 399 /* 400 * select data size depending on API version. 401 */ 402 if (sopt->sopt_name == MRT_ADD_MFC && 403 mrt_api_config & MRT_API_FLAGS_ALL) { 404 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 405 sizeof(struct mfcctl2)); 406 } else { 407 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 408 sizeof(struct mfcctl)); 409 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 410 sizeof(mfc) - sizeof(struct mfcctl)); 411 } 412 if (error) 413 break; 414 if (sopt->sopt_name == MRT_ADD_MFC) 415 error = add_mfc(&mfc); 416 else 417 error = del_mfc(&mfc); 418 break; 419 420 case MRT_ASSERT: 421 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 422 if (error) 423 break; 424 set_assert(optval); 425 break; 426 427 case MRT_API_CONFIG: 428 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 429 if (!error) 430 error = set_api_config(&i); 431 if (!error) 432 error = sooptcopyout(sopt, &i, sizeof i); 433 break; 434 435 case MRT_ADD_BW_UPCALL: 436 case MRT_DEL_BW_UPCALL: 437 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 438 sizeof bw_upcall); 439 if (error) 440 break; 441 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 442 error = add_bw_upcall(&bw_upcall); 443 else 444 error = del_bw_upcall(&bw_upcall); 445 break; 446 447 default: 448 error = EOPNOTSUPP; 449 break; 450 } 451 return error; 452 } 453 454 /* 455 * Handle MRT getsockopt commands 456 */ 457 static int 458 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 459 { 460 int error; 461 static int version = 0x0305; /* !!! why is this here? XXX */ 462 463 switch (sopt->sopt_name) { 464 case MRT_VERSION: 465 error = sooptcopyout(sopt, &version, sizeof version); 466 break; 467 468 case MRT_ASSERT: 469 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 470 break; 471 472 case MRT_API_SUPPORT: 473 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 474 break; 475 476 case MRT_API_CONFIG: 477 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 478 break; 479 480 default: 481 error = EOPNOTSUPP; 482 break; 483 } 484 return error; 485 } 486 487 /* 488 * Handle ioctl commands to obtain information from the cache 489 */ 490 static int 491 X_mrt_ioctl(int cmd, caddr_t data) 492 { 493 int error = 0; 494 495 switch (cmd) { 496 case SIOCGETVIFCNT: 497 error = get_vif_cnt((struct sioc_vif_req *)data); 498 break; 499 500 case SIOCGETSGCNT: 501 error = get_sg_cnt((struct sioc_sg_req *)data); 502 break; 503 504 default: 505 error = EINVAL; 506 break; 507 } 508 return error; 509 } 510 511 /* 512 * returns the packet, byte, rpf-failure count for the source group provided 513 */ 514 static int 515 get_sg_cnt(struct sioc_sg_req *req) 516 { 517 int s; 518 struct mfc *rt; 519 520 s = splnet(); 521 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 522 splx(s); 523 if (rt == NULL) { 524 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 525 return EADDRNOTAVAIL; 526 } 527 req->pktcnt = rt->mfc_pkt_cnt; 528 req->bytecnt = rt->mfc_byte_cnt; 529 req->wrong_if = rt->mfc_wrong_if; 530 return 0; 531 } 532 533 /* 534 * returns the input and output packet and byte counts on the vif provided 535 */ 536 static int 537 get_vif_cnt(struct sioc_vif_req *req) 538 { 539 vifi_t vifi = req->vifi; 540 541 if (vifi >= numvifs) 542 return EINVAL; 543 544 req->icount = viftable[vifi].v_pkt_in; 545 req->ocount = viftable[vifi].v_pkt_out; 546 req->ibytes = viftable[vifi].v_bytes_in; 547 req->obytes = viftable[vifi].v_bytes_out; 548 549 return 0; 550 } 551 552 /* 553 * Enable multicast routing 554 */ 555 static int 556 ip_mrouter_init(struct socket *so, int version) 557 { 558 if (mrtdebug) 559 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 560 so->so_type, so->so_proto->pr_protocol); 561 562 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 563 return EOPNOTSUPP; 564 565 if (version != 1) 566 return ENOPROTOOPT; 567 568 if (ip_mrouter != NULL) 569 return EADDRINUSE; 570 571 ip_mrouter = so; 572 573 bzero((caddr_t)mfctable, sizeof(mfctable)); 574 bzero((caddr_t)nexpire, sizeof(nexpire)); 575 576 pim_assert = 0; 577 bw_upcalls_n = 0; 578 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 579 580 callout_init(&expire_upcalls_ch); 581 callout_init(&bw_upcalls_ch); 582 callout_init(&bw_meter_ch); 583 callout_init(&tbf_reprocess_q_ch); 584 585 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 586 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 587 expire_bw_upcalls_send, NULL); 588 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 589 590 mrt_api_config = 0; 591 592 if (mrtdebug) 593 log(LOG_DEBUG, "ip_mrouter_init\n"); 594 595 return 0; 596 } 597 598 /* 599 * Disable multicast routing 600 */ 601 static int 602 X_ip_mrouter_done(void) 603 { 604 vifi_t vifi; 605 int i; 606 struct ifnet *ifp; 607 struct ifreq ifr; 608 struct mfc *rt; 609 struct rtdetq *rte; 610 int s; 611 612 s = splnet(); 613 614 /* 615 * For each phyint in use, disable promiscuous reception of all IP 616 * multicasts. 617 */ 618 for (vifi = 0; vifi < numvifs; vifi++) { 619 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 620 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 621 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 622 623 so->sin_len = sizeof(struct sockaddr_in); 624 so->sin_family = AF_INET; 625 so->sin_addr.s_addr = INADDR_ANY; 626 ifp = viftable[vifi].v_ifp; 627 if_allmulti(ifp, 0); 628 } 629 } 630 bzero((caddr_t)tbftable, sizeof(tbftable)); 631 bzero((caddr_t)viftable, sizeof(viftable)); 632 numvifs = 0; 633 pim_assert = 0; 634 635 callout_stop(&expire_upcalls_ch); 636 637 mrt_api_config = 0; 638 bw_upcalls_n = 0; 639 callout_stop(&bw_upcalls_ch); 640 callout_stop(&bw_meter_ch); 641 callout_stop(&tbf_reprocess_q_ch); 642 643 /* 644 * Free all multicast forwarding cache entries. 645 */ 646 for (i = 0; i < MFCTBLSIZ; i++) { 647 for (rt = mfctable[i]; rt != NULL; ) { 648 struct mfc *nr = rt->mfc_next; 649 650 for (rte = rt->mfc_stall; rte != NULL; ) { 651 struct rtdetq *n = rte->next; 652 653 m_freem(rte->m); 654 free(rte, M_MRTABLE); 655 rte = n; 656 } 657 free_bw_list(rt->mfc_bw_meter); 658 free(rt, M_MRTABLE); 659 rt = nr; 660 } 661 } 662 663 bzero((caddr_t)mfctable, sizeof(mfctable)); 664 665 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 666 667 /* 668 * Reset de-encapsulation cache 669 */ 670 last_encap_src = INADDR_ANY; 671 last_encap_vif = NULL; 672 #ifdef PIM 673 reg_vif_num = VIFI_INVALID; 674 #endif 675 have_encap_tunnel = 0; 676 677 ip_mrouter = NULL; 678 679 splx(s); 680 681 if (mrtdebug) 682 log(LOG_DEBUG, "ip_mrouter_done\n"); 683 684 return 0; 685 } 686 687 /* 688 * Set PIM assert processing global 689 */ 690 static int 691 set_assert(int i) 692 { 693 if ((i != 1) && (i != 0)) 694 return EINVAL; 695 696 pim_assert = i; 697 698 return 0; 699 } 700 701 /* 702 * Configure API capabilities 703 */ 704 int 705 set_api_config(uint32_t *apival) 706 { 707 int i; 708 709 /* 710 * We can set the API capabilities only if it is the first operation 711 * after MRT_INIT. I.e.: 712 * - there are no vifs installed 713 * - pim_assert is not enabled 714 * - the MFC table is empty 715 */ 716 if (numvifs > 0) { 717 *apival = 0; 718 return EPERM; 719 } 720 if (pim_assert) { 721 *apival = 0; 722 return EPERM; 723 } 724 for (i = 0; i < MFCTBLSIZ; i++) { 725 if (mfctable[i] != NULL) { 726 *apival = 0; 727 return EPERM; 728 } 729 } 730 731 mrt_api_config = *apival & mrt_api_support; 732 *apival = mrt_api_config; 733 734 return 0; 735 } 736 737 /* 738 * Add a vif to the vif table 739 */ 740 static int 741 add_vif(struct vifctl *vifcp) 742 { 743 struct vif *vifp = viftable + vifcp->vifc_vifi; 744 struct sockaddr_in sin = {sizeof sin, AF_INET}; 745 struct ifaddr *ifa; 746 struct ifnet *ifp; 747 int error, s; 748 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 749 750 if (vifcp->vifc_vifi >= MAXVIFS) 751 return EINVAL; 752 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) 753 return EADDRINUSE; 754 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) 755 return EADDRNOTAVAIL; 756 757 /* Find the interface with an address in AF_INET family */ 758 #ifdef PIM 759 if (vifcp->vifc_flags & VIFF_REGISTER) { 760 /* 761 * XXX: Because VIFF_REGISTER does not really need a valid 762 * local interface (e.g. it could be 127.0.0.2), we don't 763 * check its address. 764 */ 765 ifp = NULL; 766 } else 767 #endif 768 { 769 sin.sin_addr = vifcp->vifc_lcl_addr; 770 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 771 if (ifa == NULL) 772 return EADDRNOTAVAIL; 773 ifp = ifa->ifa_ifp; 774 } 775 776 if (vifcp->vifc_flags & VIFF_TUNNEL) { 777 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 778 /* 779 * An encapsulating tunnel is wanted. Tell ipip_input() to 780 * start paying attention to encapsulated packets. 781 */ 782 if (have_encap_tunnel == 0) { 783 have_encap_tunnel = 1; 784 for (s = 0; s < MAXVIFS; ++s) { 785 if_initname(&multicast_decap_if[s], "mdecap", s); 786 } 787 } 788 /* 789 * Set interface to fake encapsulator interface 790 */ 791 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 792 /* 793 * Prepare cached route entry 794 */ 795 bzero(&vifp->v_route, sizeof(vifp->v_route)); 796 } else { 797 log(LOG_ERR, "source routed tunnels not supported\n"); 798 return EOPNOTSUPP; 799 } 800 #ifdef PIM 801 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 802 ifp = &multicast_register_if; 803 if (mrtdebug) 804 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 805 (void *)&multicast_register_if); 806 if (reg_vif_num == VIFI_INVALID) { 807 if_initname(&multicast_register_if, "register_vif", 0); 808 multicast_register_if.if_flags = IFF_LOOPBACK; 809 bzero(&vifp->v_route, sizeof(vifp->v_route)); 810 reg_vif_num = vifcp->vifc_vifi; 811 } 812 #endif 813 } else { /* Make sure the interface supports multicast */ 814 if ((ifp->if_flags & IFF_MULTICAST) == 0) 815 return EOPNOTSUPP; 816 817 /* Enable promiscuous reception of all IP multicasts from the if */ 818 s = splnet(); 819 error = if_allmulti(ifp, 1); 820 splx(s); 821 if (error) 822 return error; 823 } 824 825 s = splnet(); 826 /* define parameters for the tbf structure */ 827 vifp->v_tbf = v_tbf; 828 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 829 vifp->v_tbf->tbf_n_tok = 0; 830 vifp->v_tbf->tbf_q_len = 0; 831 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 832 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 833 834 vifp->v_flags = vifcp->vifc_flags; 835 vifp->v_threshold = vifcp->vifc_threshold; 836 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 837 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 838 vifp->v_ifp = ifp; 839 /* scaling up here allows division by 1024 in critical code */ 840 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 841 vifp->v_rsvp_on = 0; 842 vifp->v_rsvpd = NULL; 843 /* initialize per vif pkt counters */ 844 vifp->v_pkt_in = 0; 845 vifp->v_pkt_out = 0; 846 vifp->v_bytes_in = 0; 847 vifp->v_bytes_out = 0; 848 splx(s); 849 850 /* Adjust numvifs up if the vifi is higher than numvifs */ 851 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 852 853 if (mrtdebug) 854 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 855 vifcp->vifc_vifi, 856 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 857 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 858 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 859 vifcp->vifc_threshold, 860 vifcp->vifc_rate_limit); 861 862 return 0; 863 } 864 865 /* 866 * Delete a vif from the vif table 867 */ 868 static int 869 del_vif(vifi_t vifi) 870 { 871 struct vif *vifp; 872 int s; 873 874 if (vifi >= numvifs) 875 return EINVAL; 876 vifp = &viftable[vifi]; 877 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) 878 return EADDRNOTAVAIL; 879 880 s = splnet(); 881 882 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 883 if_allmulti(vifp->v_ifp, 0); 884 885 if (vifp == last_encap_vif) { 886 last_encap_vif = NULL; 887 last_encap_src = INADDR_ANY; 888 } 889 890 /* 891 * Free packets queued at the interface 892 */ 893 while (vifp->v_tbf->tbf_q) { 894 struct mbuf *m = vifp->v_tbf->tbf_q; 895 896 vifp->v_tbf->tbf_q = m->m_nextpkt; 897 m_freem(m); 898 } 899 900 #ifdef PIM 901 if (vifp->v_flags & VIFF_REGISTER) 902 reg_vif_num = VIFI_INVALID; 903 #endif 904 905 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 906 bzero((caddr_t)vifp, sizeof (*vifp)); 907 908 if (mrtdebug) 909 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 910 911 /* Adjust numvifs down */ 912 for (vifi = numvifs; vifi > 0; vifi--) 913 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 914 break; 915 numvifs = vifi; 916 917 splx(s); 918 919 return 0; 920 } 921 922 /* 923 * update an mfc entry without resetting counters and S,G addresses. 924 */ 925 static void 926 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 927 { 928 int i; 929 930 rt->mfc_parent = mfccp->mfcc_parent; 931 for (i = 0; i < numvifs; i++) { 932 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 933 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 934 MRT_MFC_FLAGS_ALL; 935 } 936 /* set the RP address */ 937 if (mrt_api_config & MRT_MFC_RP) 938 rt->mfc_rp = mfccp->mfcc_rp; 939 else 940 rt->mfc_rp.s_addr = INADDR_ANY; 941 } 942 943 /* 944 * fully initialize an mfc entry from the parameter. 945 */ 946 static void 947 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 948 { 949 rt->mfc_origin = mfccp->mfcc_origin; 950 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 951 952 update_mfc_params(rt, mfccp); 953 954 /* initialize pkt counters per src-grp */ 955 rt->mfc_pkt_cnt = 0; 956 rt->mfc_byte_cnt = 0; 957 rt->mfc_wrong_if = 0; 958 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 959 } 960 961 962 /* 963 * Add an mfc entry 964 */ 965 static int 966 add_mfc(struct mfcctl2 *mfccp) 967 { 968 struct mfc *rt; 969 u_long hash; 970 struct rtdetq *rte; 971 u_short nstl; 972 int s; 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 s = splnet(); 985 update_mfc_params(rt, mfccp); 986 splx(s); 987 return 0; 988 } 989 990 /* 991 * Find the entry for which the upcall was made and update 992 */ 993 s = splnet(); 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 splx(s); 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 splx(s); 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 int s; 1085 struct bw_meter *list; 1086 1087 origin = mfccp->mfcc_origin; 1088 mcastgrp = mfccp->mfcc_mcastgrp; 1089 1090 if (mrtdebug & DEBUG_MFC) 1091 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1092 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1093 1094 s = splnet(); 1095 1096 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1097 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1098 if (origin.s_addr == rt->mfc_origin.s_addr && 1099 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1100 rt->mfc_stall == NULL) 1101 break; 1102 if (rt == NULL) { 1103 splx(s); 1104 return EADDRNOTAVAIL; 1105 } 1106 1107 *nptr = rt->mfc_next; 1108 1109 /* 1110 * free the bw_meter entries 1111 */ 1112 list = rt->mfc_bw_meter; 1113 rt->mfc_bw_meter = NULL; 1114 1115 free(rt, M_MRTABLE); 1116 1117 splx(s); 1118 1119 free_bw_list(list); 1120 1121 return 0; 1122 } 1123 1124 /* 1125 * Send a message to mrouted on the multicast routing socket 1126 */ 1127 static int 1128 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1129 { 1130 if (s) { 1131 if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) { 1132 sorwakeup(s); 1133 return 0; 1134 } 1135 } 1136 m_freem(mm); 1137 return -1; 1138 } 1139 1140 /* 1141 * IP multicast forwarding function. This function assumes that the packet 1142 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1143 * pointed to by "ifp", and the packet is to be relayed to other networks 1144 * that have members of the packet's destination IP multicast group. 1145 * 1146 * The packet is returned unscathed to the caller, unless it is 1147 * erroneous, in which case a non-zero return value tells the caller to 1148 * discard it. 1149 */ 1150 1151 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1152 1153 static int 1154 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1155 struct ip_moptions *imo) 1156 { 1157 struct mfc *rt; 1158 int s; 1159 vifi_t vifi; 1160 1161 if (mrtdebug & DEBUG_FORWARD) 1162 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1163 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1164 (void *)ifp); 1165 1166 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1167 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1168 /* 1169 * Packet arrived via a physical interface or 1170 * an encapsulated tunnel or a register_vif. 1171 */ 1172 } else { 1173 /* 1174 * Packet arrived through a source-route tunnel. 1175 * Source-route tunnels are no longer supported. 1176 */ 1177 static int last_log; 1178 if (last_log != time_second) { 1179 last_log = time_second; 1180 log(LOG_ERR, 1181 "ip_mforward: received source-routed packet from %lx\n", 1182 (u_long)ntohl(ip->ip_src.s_addr)); 1183 } 1184 return 1; 1185 } 1186 1187 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1188 if (ip->ip_ttl < 255) 1189 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1190 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1191 struct vif *vifp = viftable + vifi; 1192 1193 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1194 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1195 vifi, 1196 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1197 vifp->v_ifp->if_xname); 1198 } 1199 return ip_mdq(m, ifp, NULL, vifi); 1200 } 1201 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1202 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1203 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1204 if (!imo) 1205 printf("In fact, no options were specified at all\n"); 1206 } 1207 1208 /* 1209 * Don't forward a packet with time-to-live of zero or one, 1210 * or a packet destined to a local-only group. 1211 */ 1212 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1213 return 0; 1214 1215 /* 1216 * Determine forwarding vifs from the forwarding cache table 1217 */ 1218 s = splnet(); 1219 ++mrtstat.mrts_mfc_lookups; 1220 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1221 1222 /* Entry exists, so forward if necessary */ 1223 if (rt != NULL) { 1224 splx(s); 1225 return ip_mdq(m, ifp, rt, -1); 1226 } else { 1227 /* 1228 * If we don't have a route for packet's origin, 1229 * Make a copy of the packet & send message to routing daemon 1230 */ 1231 1232 struct mbuf *mb0; 1233 struct rtdetq *rte; 1234 u_long hash; 1235 int hlen = ip->ip_hl << 2; 1236 1237 ++mrtstat.mrts_mfc_misses; 1238 1239 mrtstat.mrts_no_route++; 1240 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1241 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1242 (u_long)ntohl(ip->ip_src.s_addr), 1243 (u_long)ntohl(ip->ip_dst.s_addr)); 1244 1245 /* 1246 * Allocate mbufs early so that we don't do extra work if we are 1247 * just going to fail anyway. Make sure to pullup the header so 1248 * that other people can't step on it. 1249 */ 1250 rte = malloc((sizeof *rte), M_MRTABLE, M_INTWAIT | M_NULLOK); 1251 if (rte == NULL) { 1252 splx(s); 1253 return ENOBUFS; 1254 } 1255 1256 mb0 = m_copypacket(m, MB_DONTWAIT); 1257 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1258 mb0 = m_pullup(mb0, hlen); 1259 if (mb0 == NULL) { 1260 free(rte, M_MRTABLE); 1261 splx(s); 1262 return ENOBUFS; 1263 } 1264 1265 /* is there an upcall waiting for this flow ? */ 1266 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1267 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1268 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1269 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1270 (rt->mfc_stall != NULL)) 1271 break; 1272 } 1273 1274 if (rt == NULL) { 1275 int i; 1276 struct igmpmsg *im; 1277 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1278 struct mbuf *mm; 1279 1280 /* 1281 * Locate the vifi for the incoming interface for this packet. 1282 * If none found, drop packet. 1283 */ 1284 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1285 ; 1286 if (vifi >= numvifs) /* vif not found, drop packet */ 1287 goto non_fatal; 1288 1289 /* no upcall, so make a new entry */ 1290 rt = malloc(sizeof(*rt), M_MRTABLE, M_INTWAIT | M_NULLOK); 1291 if (rt == NULL) 1292 goto fail; 1293 1294 /* Make a copy of the header to send to the user level process */ 1295 mm = m_copy(mb0, 0, hlen); 1296 if (mm == NULL) 1297 goto fail1; 1298 1299 /* 1300 * Send message to routing daemon to install 1301 * a route into the kernel table 1302 */ 1303 1304 im = mtod(mm, struct igmpmsg *); 1305 im->im_msgtype = IGMPMSG_NOCACHE; 1306 im->im_mbz = 0; 1307 im->im_vif = vifi; 1308 1309 mrtstat.mrts_upcalls++; 1310 1311 k_igmpsrc.sin_addr = ip->ip_src; 1312 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1313 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1314 ++mrtstat.mrts_upq_sockfull; 1315 fail1: 1316 free(rt, M_MRTABLE); 1317 fail: 1318 free(rte, M_MRTABLE); 1319 m_freem(mb0); 1320 splx(s); 1321 return ENOBUFS; 1322 } 1323 1324 /* insert new entry at head of hash chain */ 1325 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1326 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1327 rt->mfc_expire = UPCALL_EXPIRE; 1328 nexpire[hash]++; 1329 for (i = 0; i < numvifs; i++) { 1330 rt->mfc_ttls[i] = 0; 1331 rt->mfc_flags[i] = 0; 1332 } 1333 rt->mfc_parent = -1; 1334 1335 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1336 1337 rt->mfc_bw_meter = NULL; 1338 1339 /* link into table */ 1340 rt->mfc_next = mfctable[hash]; 1341 mfctable[hash] = rt; 1342 rt->mfc_stall = rte; 1343 1344 } else { 1345 /* determine if q has overflowed */ 1346 int npkts = 0; 1347 struct rtdetq **p; 1348 1349 /* 1350 * XXX ouch! we need to append to the list, but we 1351 * only have a pointer to the front, so we have to 1352 * scan the entire list every time. 1353 */ 1354 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1355 npkts++; 1356 1357 if (npkts > MAX_UPQ) { 1358 mrtstat.mrts_upq_ovflw++; 1359 non_fatal: 1360 free(rte, M_MRTABLE); 1361 m_freem(mb0); 1362 splx(s); 1363 return 0; 1364 } 1365 1366 /* Add this entry to the end of the queue */ 1367 *p = rte; 1368 } 1369 1370 rte->m = mb0; 1371 rte->ifp = ifp; 1372 rte->next = NULL; 1373 1374 splx(s); 1375 1376 return 0; 1377 } 1378 } 1379 1380 /* 1381 * Clean up the cache entry if upcall is not serviced 1382 */ 1383 static void 1384 expire_upcalls(void *unused) 1385 { 1386 struct rtdetq *rte; 1387 struct mfc *mfc, **nptr; 1388 int i; 1389 int s; 1390 1391 s = splnet(); 1392 for (i = 0; i < MFCTBLSIZ; i++) { 1393 if (nexpire[i] == 0) 1394 continue; 1395 nptr = &mfctable[i]; 1396 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1397 /* 1398 * Skip real cache entries 1399 * Make sure it wasn't marked to not expire (shouldn't happen) 1400 * If it expires now 1401 */ 1402 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1403 --mfc->mfc_expire == 0) { 1404 if (mrtdebug & DEBUG_EXPIRE) 1405 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1406 (u_long)ntohl(mfc->mfc_origin.s_addr), 1407 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1408 /* 1409 * drop all the packets 1410 * free the mbuf with the pkt, if, timing info 1411 */ 1412 for (rte = mfc->mfc_stall; rte; ) { 1413 struct rtdetq *n = rte->next; 1414 1415 m_freem(rte->m); 1416 free(rte, M_MRTABLE); 1417 rte = n; 1418 } 1419 ++mrtstat.mrts_cache_cleanups; 1420 nexpire[i]--; 1421 1422 /* 1423 * free the bw_meter entries 1424 */ 1425 while (mfc->mfc_bw_meter != NULL) { 1426 struct bw_meter *x = mfc->mfc_bw_meter; 1427 1428 mfc->mfc_bw_meter = x->bm_mfc_next; 1429 free(x, M_BWMETER); 1430 } 1431 1432 *nptr = mfc->mfc_next; 1433 free(mfc, M_MRTABLE); 1434 } else { 1435 nptr = &mfc->mfc_next; 1436 } 1437 } 1438 } 1439 splx(s); 1440 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1441 } 1442 1443 /* 1444 * Packet forwarding routine once entry in the cache is made 1445 */ 1446 static int 1447 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1448 { 1449 struct ip *ip = mtod(m, struct ip *); 1450 vifi_t vifi; 1451 int plen = ip->ip_len; 1452 1453 /* 1454 * Macro to send packet on vif. Since RSVP packets don't get counted on 1455 * input, they shouldn't get counted on output, so statistics keeping is 1456 * separate. 1457 */ 1458 #define MC_SEND(ip,vifp,m) { \ 1459 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1460 encap_send((ip), (vifp), (m)); \ 1461 else \ 1462 phyint_send((ip), (vifp), (m)); \ 1463 } 1464 1465 /* 1466 * If xmt_vif is not -1, send on only the requested vif. 1467 * 1468 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1469 */ 1470 if (xmt_vif < numvifs) { 1471 #ifdef PIM 1472 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1473 pim_register_send(ip, viftable + xmt_vif, m, rt); 1474 else 1475 #endif 1476 MC_SEND(ip, viftable + xmt_vif, m); 1477 return 1; 1478 } 1479 1480 /* 1481 * Don't forward if it didn't arrive from the parent vif for its origin. 1482 */ 1483 vifi = rt->mfc_parent; 1484 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1485 /* came in the wrong interface */ 1486 if (mrtdebug & DEBUG_FORWARD) 1487 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1488 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1489 ++mrtstat.mrts_wrong_if; 1490 ++rt->mfc_wrong_if; 1491 /* 1492 * If we are doing PIM assert processing, send a message 1493 * to the routing daemon. 1494 * 1495 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1496 * can complete the SPT switch, regardless of the type 1497 * of the iif (broadcast media, GRE tunnel, etc). 1498 */ 1499 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1500 struct timeval now; 1501 u_long delta; 1502 1503 #ifdef PIM 1504 if (ifp == &multicast_register_if) 1505 pimstat.pims_rcv_registers_wrongiif++; 1506 #endif 1507 1508 /* Get vifi for the incoming packet */ 1509 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1510 ; 1511 if (vifi >= numvifs) 1512 return 0; /* The iif is not found: ignore the packet. */ 1513 1514 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1515 return 0; /* WRONGVIF disabled: ignore the packet */ 1516 1517 GET_TIME(now); 1518 1519 TV_DELTA(rt->mfc_last_assert, now, delta); 1520 1521 if (delta > ASSERT_MSG_TIME) { 1522 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1523 struct igmpmsg *im; 1524 int hlen = ip->ip_hl << 2; 1525 struct mbuf *mm = m_copy(m, 0, hlen); 1526 1527 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1528 mm = m_pullup(mm, hlen); 1529 if (mm == NULL) 1530 return ENOBUFS; 1531 1532 rt->mfc_last_assert = now; 1533 1534 im = mtod(mm, struct igmpmsg *); 1535 im->im_msgtype = IGMPMSG_WRONGVIF; 1536 im->im_mbz = 0; 1537 im->im_vif = vifi; 1538 1539 mrtstat.mrts_upcalls++; 1540 1541 k_igmpsrc.sin_addr = im->im_src; 1542 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1543 log(LOG_WARNING, 1544 "ip_mforward: ip_mrouter socket queue full\n"); 1545 ++mrtstat.mrts_upq_sockfull; 1546 return ENOBUFS; 1547 } 1548 } 1549 } 1550 return 0; 1551 } 1552 1553 /* If I sourced this packet, it counts as output, else it was input. */ 1554 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1555 viftable[vifi].v_pkt_out++; 1556 viftable[vifi].v_bytes_out += plen; 1557 } else { 1558 viftable[vifi].v_pkt_in++; 1559 viftable[vifi].v_bytes_in += plen; 1560 } 1561 rt->mfc_pkt_cnt++; 1562 rt->mfc_byte_cnt += plen; 1563 1564 /* 1565 * For each vif, decide if a copy of the packet should be forwarded. 1566 * Forward if: 1567 * - the ttl exceeds the vif's threshold 1568 * - there are group members downstream on interface 1569 */ 1570 for (vifi = 0; vifi < numvifs; vifi++) 1571 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1572 viftable[vifi].v_pkt_out++; 1573 viftable[vifi].v_bytes_out += plen; 1574 #ifdef PIM 1575 if (viftable[vifi].v_flags & VIFF_REGISTER) 1576 pim_register_send(ip, viftable + vifi, m, rt); 1577 else 1578 #endif 1579 MC_SEND(ip, viftable+vifi, m); 1580 } 1581 1582 /* 1583 * Perform upcall-related bw measuring. 1584 */ 1585 if (rt->mfc_bw_meter != NULL) { 1586 struct bw_meter *x; 1587 struct timeval now; 1588 1589 GET_TIME(now); 1590 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1591 bw_meter_receive_packet(x, plen, &now); 1592 } 1593 1594 return 0; 1595 } 1596 1597 /* 1598 * check if a vif number is legal/ok. This is used by ip_output. 1599 */ 1600 static int 1601 X_legal_vif_num(int vif) 1602 { 1603 return (vif >= 0 && vif < numvifs); 1604 } 1605 1606 /* 1607 * Return the local address used by this vif 1608 */ 1609 static u_long 1610 X_ip_mcast_src(int vifi) 1611 { 1612 if (vifi >= 0 && vifi < numvifs) 1613 return viftable[vifi].v_lcl_addr.s_addr; 1614 else 1615 return INADDR_ANY; 1616 } 1617 1618 static void 1619 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1620 { 1621 struct mbuf *mb_copy; 1622 int hlen = ip->ip_hl << 2; 1623 1624 /* 1625 * Make a new reference to the packet; make sure that 1626 * the IP header is actually copied, not just referenced, 1627 * so that ip_output() only scribbles on the copy. 1628 */ 1629 mb_copy = m_copypacket(m, MB_DONTWAIT); 1630 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1631 mb_copy = m_pullup(mb_copy, hlen); 1632 if (mb_copy == NULL) 1633 return; 1634 1635 if (vifp->v_rate_limit == 0) 1636 tbf_send_packet(vifp, mb_copy); 1637 else 1638 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1639 } 1640 1641 static void 1642 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1643 { 1644 struct mbuf *mb_copy; 1645 struct ip *ip_copy; 1646 int i, len = ip->ip_len; 1647 1648 /* Take care of delayed checksums */ 1649 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1650 in_delayed_cksum(m); 1651 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1652 } 1653 1654 /* 1655 * copy the old packet & pullup its IP header into the 1656 * new mbuf so we can modify it. Try to fill the new 1657 * mbuf since if we don't the ethernet driver will. 1658 */ 1659 MGETHDR(mb_copy, MB_DONTWAIT, MT_HEADER); 1660 if (mb_copy == NULL) 1661 return; 1662 mb_copy->m_data += max_linkhdr; 1663 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1664 1665 if ((mb_copy->m_next = m_copypacket(m, MB_DONTWAIT)) == NULL) { 1666 m_freem(mb_copy); 1667 return; 1668 } 1669 i = MHLEN - M_LEADINGSPACE(mb_copy); 1670 if (i > len) 1671 i = len; 1672 mb_copy = m_pullup(mb_copy, i); 1673 if (mb_copy == NULL) 1674 return; 1675 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1676 1677 /* 1678 * fill in the encapsulating IP header. 1679 */ 1680 ip_copy = mtod(mb_copy, struct ip *); 1681 *ip_copy = multicast_encap_iphdr; 1682 #ifdef RANDOM_IP_ID 1683 ip_copy->ip_id = ip_randomid(); 1684 #else 1685 ip_copy->ip_id = htons(ip_id++); 1686 #endif 1687 ip_copy->ip_len += len; 1688 ip_copy->ip_src = vifp->v_lcl_addr; 1689 ip_copy->ip_dst = vifp->v_rmt_addr; 1690 1691 /* 1692 * turn the encapsulated IP header back into a valid one. 1693 */ 1694 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1695 --ip->ip_ttl; 1696 ip->ip_len = htons(ip->ip_len); 1697 ip->ip_off = htons(ip->ip_off); 1698 ip->ip_sum = 0; 1699 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1700 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1701 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1702 1703 if (vifp->v_rate_limit == 0) 1704 tbf_send_packet(vifp, mb_copy); 1705 else 1706 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1707 } 1708 1709 /* 1710 * De-encapsulate a packet and feed it back through ip input (this 1711 * routine is called whenever IP gets a packet with proto type 1712 * ENCAP_PROTO and a local destination address). 1713 * 1714 * This is similar to mroute_encapcheck() + mroute_encap_input() in -current. 1715 */ 1716 static void 1717 X_ipip_input(struct mbuf *m, int off, int proto) 1718 { 1719 struct ip *ip = mtod(m, struct ip *); 1720 int hlen = ip->ip_hl << 2; 1721 1722 if (!have_encap_tunnel) { 1723 rip_input(m, off, proto); 1724 return; 1725 } 1726 /* 1727 * dump the packet if it's not to a multicast destination or if 1728 * we don't have an encapsulating tunnel with the source. 1729 * Note: This code assumes that the remote site IP address 1730 * uniquely identifies the tunnel (i.e., that this site has 1731 * at most one tunnel with the remote site). 1732 */ 1733 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) { 1734 ++mrtstat.mrts_bad_tunnel; 1735 m_freem(m); 1736 return; 1737 } 1738 if (ip->ip_src.s_addr != last_encap_src) { 1739 struct vif *vifp = viftable; 1740 struct vif *vife = vifp + numvifs; 1741 1742 last_encap_src = ip->ip_src.s_addr; 1743 last_encap_vif = NULL; 1744 for ( ; vifp < vife; ++vifp) 1745 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1746 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1747 == VIFF_TUNNEL) 1748 last_encap_vif = vifp; 1749 break; 1750 } 1751 } 1752 if (last_encap_vif == NULL) { 1753 last_encap_src = INADDR_ANY; 1754 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1755 m_freem(m); 1756 if (mrtdebug) 1757 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1758 (u_long)ntohl(ip->ip_src.s_addr)); 1759 return; 1760 } 1761 1762 if (hlen > sizeof(struct ip)) 1763 ip_stripoptions(m); 1764 m->m_data += sizeof(struct ip); 1765 m->m_len -= sizeof(struct ip); 1766 m->m_pkthdr.len -= sizeof(struct ip); 1767 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 1768 1769 netisr_queue(NETISR_IP, m); 1770 } 1771 1772 /* 1773 * Token bucket filter module 1774 */ 1775 1776 static void 1777 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len) 1778 { 1779 struct tbf *t = vifp->v_tbf; 1780 1781 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */ 1782 mrtstat.mrts_pkt2large++; 1783 m_freem(m); 1784 return; 1785 } 1786 1787 tbf_update_tokens(vifp); 1788 1789 if (t->tbf_q_len == 0) { /* queue empty... */ 1790 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */ 1791 t->tbf_n_tok -= p_len; 1792 tbf_send_packet(vifp, m); 1793 } else { /* no, queue packet and try later */ 1794 tbf_queue(vifp, m); 1795 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS, 1796 tbf_reprocess_q, vifp); 1797 } 1798 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1799 /* finite queue length, so queue pkts and process queue */ 1800 tbf_queue(vifp, m); 1801 tbf_process_q(vifp); 1802 } else { 1803 /* queue full, try to dq and queue and process */ 1804 if (!tbf_dq_sel(vifp, ip)) { 1805 mrtstat.mrts_q_overflow++; 1806 m_freem(m); 1807 } else { 1808 tbf_queue(vifp, m); 1809 tbf_process_q(vifp); 1810 } 1811 } 1812 } 1813 1814 /* 1815 * adds a packet to the queue at the interface 1816 */ 1817 static void 1818 tbf_queue(struct vif *vifp, struct mbuf *m) 1819 { 1820 int s = splnet(); 1821 struct tbf *t = vifp->v_tbf; 1822 1823 if (t->tbf_t == NULL) /* Queue was empty */ 1824 t->tbf_q = m; 1825 else /* Insert at tail */ 1826 t->tbf_t->m_nextpkt = m; 1827 1828 t->tbf_t = m; /* Set new tail pointer */ 1829 1830 #ifdef DIAGNOSTIC 1831 /* Make sure we didn't get fed a bogus mbuf */ 1832 if (m->m_nextpkt) 1833 panic("tbf_queue: m_nextpkt"); 1834 #endif 1835 m->m_nextpkt = NULL; 1836 1837 t->tbf_q_len++; 1838 1839 splx(s); 1840 } 1841 1842 /* 1843 * processes the queue at the interface 1844 */ 1845 static void 1846 tbf_process_q(struct vif *vifp) 1847 { 1848 int s = splnet(); 1849 struct tbf *t = vifp->v_tbf; 1850 1851 /* loop through the queue at the interface and send as many packets 1852 * as possible 1853 */ 1854 while (t->tbf_q_len > 0) { 1855 struct mbuf *m = t->tbf_q; 1856 int len = mtod(m, struct ip *)->ip_len; 1857 1858 /* determine if the packet can be sent */ 1859 if (len > t->tbf_n_tok) /* not enough tokens, we are done */ 1860 break; 1861 /* ok, reduce no of tokens, dequeue and send the packet. */ 1862 t->tbf_n_tok -= len; 1863 1864 t->tbf_q = m->m_nextpkt; 1865 if (--t->tbf_q_len == 0) 1866 t->tbf_t = NULL; 1867 1868 m->m_nextpkt = NULL; 1869 tbf_send_packet(vifp, m); 1870 } 1871 splx(s); 1872 } 1873 1874 static void 1875 tbf_reprocess_q(void *xvifp) 1876 { 1877 struct vif *vifp = xvifp; 1878 1879 if (ip_mrouter == NULL) 1880 return; 1881 tbf_update_tokens(vifp); 1882 tbf_process_q(vifp); 1883 if (vifp->v_tbf->tbf_q_len) 1884 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS, 1885 tbf_reprocess_q, vifp); 1886 } 1887 1888 /* function that will selectively discard a member of the queue 1889 * based on the precedence value and the priority 1890 */ 1891 static int 1892 tbf_dq_sel(struct vif *vifp, struct ip *ip) 1893 { 1894 int s = splnet(); 1895 u_int p; 1896 struct mbuf *m, *last; 1897 struct mbuf **np; 1898 struct tbf *t = vifp->v_tbf; 1899 1900 p = priority(vifp, ip); 1901 1902 np = &t->tbf_q; 1903 last = NULL; 1904 while ((m = *np) != NULL) { 1905 if (p > priority(vifp, mtod(m, struct ip *))) { 1906 *np = m->m_nextpkt; 1907 /* If we're removing the last packet, fix the tail pointer */ 1908 if (m == t->tbf_t) 1909 t->tbf_t = last; 1910 m_freem(m); 1911 /* It's impossible for the queue to be empty, but check anyways. */ 1912 if (--t->tbf_q_len == 0) 1913 t->tbf_t = NULL; 1914 splx(s); 1915 mrtstat.mrts_drop_sel++; 1916 return 1; 1917 } 1918 np = &m->m_nextpkt; 1919 last = m; 1920 } 1921 splx(s); 1922 return 0; 1923 } 1924 1925 static void 1926 tbf_send_packet(struct vif *vifp, struct mbuf *m) 1927 { 1928 int s = splnet(); 1929 1930 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */ 1931 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 1932 else { 1933 struct ip_moptions imo; 1934 int error; 1935 static struct route ro; /* XXX check this */ 1936 1937 imo.imo_multicast_ifp = vifp->v_ifp; 1938 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1939 imo.imo_multicast_loop = 1; 1940 imo.imo_multicast_vif = -1; 1941 1942 /* 1943 * Re-entrancy should not be a problem here, because 1944 * the packets that we send out and are looped back at us 1945 * should get rejected because they appear to come from 1946 * the loopback interface, thus preventing looping. 1947 */ 1948 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL); 1949 1950 if (mrtdebug & DEBUG_XMIT) 1951 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1952 (int)(vifp - viftable), error); 1953 } 1954 splx(s); 1955 } 1956 1957 /* determine the current time and then 1958 * the elapsed time (between the last time and time now) 1959 * in milliseconds & update the no. of tokens in the bucket 1960 */ 1961 static void 1962 tbf_update_tokens(struct vif *vifp) 1963 { 1964 struct timeval tp; 1965 u_long tm; 1966 int s = splnet(); 1967 struct tbf *t = vifp->v_tbf; 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 splx(s); 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, s; 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 s = splnet(); 2041 2042 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 2043 splx(s); 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 splx(s); 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 splx(s); 2079 return 0; 2080 } 2081 2082 static void 2083 X_ip_rsvp_force_done(struct socket *so) 2084 { 2085 int vifi; 2086 int s; 2087 2088 /* Don't bother if it is not the right type of socket. */ 2089 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2090 return; 2091 2092 s = splnet(); 2093 2094 /* The socket may be attached to more than one vif...this 2095 * is perfectly legal. 2096 */ 2097 for (vifi = 0; vifi < numvifs; vifi++) { 2098 if (viftable[vifi].v_rsvpd == so) { 2099 viftable[vifi].v_rsvpd = NULL; 2100 /* This may seem silly, but we need to be sure we don't 2101 * over-decrement the RSVP counter, in case something slips up. 2102 */ 2103 if (viftable[vifi].v_rsvp_on) { 2104 viftable[vifi].v_rsvp_on = 0; 2105 rsvp_on--; 2106 } 2107 } 2108 } 2109 2110 splx(s); 2111 } 2112 2113 static void 2114 X_rsvp_input(struct mbuf *m, ...) 2115 { 2116 int vifi; 2117 struct ip *ip = mtod(m, struct ip *); 2118 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2119 int s; 2120 struct ifnet *ifp; 2121 int off, proto; 2122 __va_list ap; 2123 2124 __va_start(ap, m); 2125 off = __va_arg(ap, int); 2126 proto = __va_arg(ap, int); 2127 __va_end(ap); 2128 2129 if (rsvpdebug) 2130 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2131 2132 /* Can still get packets with rsvp_on = 0 if there is a local member 2133 * of the group to which the RSVP packet is addressed. But in this 2134 * case we want to throw the packet away. 2135 */ 2136 if (!rsvp_on) { 2137 m_freem(m); 2138 return; 2139 } 2140 2141 s = splnet(); 2142 2143 if (rsvpdebug) 2144 printf("rsvp_input: check vifs\n"); 2145 2146 #ifdef DIAGNOSTIC 2147 if (!(m->m_flags & M_PKTHDR)) 2148 panic("rsvp_input no hdr"); 2149 #endif 2150 2151 ifp = m->m_pkthdr.rcvif; 2152 /* Find which vif the packet arrived on. */ 2153 for (vifi = 0; vifi < numvifs; vifi++) 2154 if (viftable[vifi].v_ifp == ifp) 2155 break; 2156 2157 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2158 /* 2159 * If the old-style non-vif-associated socket is set, 2160 * then use it. Otherwise, drop packet since there 2161 * is no specific socket for this vif. 2162 */ 2163 if (ip_rsvpd != NULL) { 2164 if (rsvpdebug) 2165 printf("rsvp_input: Sending packet up old-style socket\n"); 2166 rip_input(m, off, proto); /* xxx */ 2167 } else { 2168 if (rsvpdebug && vifi == numvifs) 2169 printf("rsvp_input: Can't find vif for packet.\n"); 2170 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2171 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2172 m_freem(m); 2173 } 2174 splx(s); 2175 return; 2176 } 2177 rsvp_src.sin_addr = ip->ip_src; 2178 2179 if (rsvpdebug && m) 2180 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2181 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2182 2183 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2184 if (rsvpdebug) 2185 printf("rsvp_input: Failed to append to socket\n"); 2186 } else { 2187 if (rsvpdebug) 2188 printf("rsvp_input: send packet up\n"); 2189 } 2190 2191 splx(s); 2192 } 2193 2194 /* 2195 * Code for bandwidth monitors 2196 */ 2197 2198 /* 2199 * Define common interface for timeval-related methods 2200 */ 2201 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 2202 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 2203 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 2204 2205 static uint32_t 2206 compute_bw_meter_flags(struct bw_upcall *req) 2207 { 2208 uint32_t flags = 0; 2209 2210 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2211 flags |= BW_METER_UNIT_PACKETS; 2212 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2213 flags |= BW_METER_UNIT_BYTES; 2214 if (req->bu_flags & BW_UPCALL_GEQ) 2215 flags |= BW_METER_GEQ; 2216 if (req->bu_flags & BW_UPCALL_LEQ) 2217 flags |= BW_METER_LEQ; 2218 2219 return flags; 2220 } 2221 2222 /* 2223 * Add a bw_meter entry 2224 */ 2225 static int 2226 add_bw_upcall(struct bw_upcall *req) 2227 { 2228 struct mfc *mfc; 2229 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2230 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2231 struct timeval now; 2232 struct bw_meter *x; 2233 uint32_t flags; 2234 int s; 2235 2236 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2237 return EOPNOTSUPP; 2238 2239 /* Test if the flags are valid */ 2240 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2241 return EINVAL; 2242 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2243 return EINVAL; 2244 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2245 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2246 return EINVAL; 2247 2248 /* Test if the threshold time interval is valid */ 2249 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2250 return EINVAL; 2251 2252 flags = compute_bw_meter_flags(req); 2253 2254 /* 2255 * Find if we have already same bw_meter entry 2256 */ 2257 s = splnet(); 2258 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2259 if (mfc == NULL) { 2260 splx(s); 2261 return EADDRNOTAVAIL; 2262 } 2263 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2264 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2265 &req->bu_threshold.b_time, ==)) && 2266 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2267 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2268 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2269 splx(s); 2270 return 0; /* XXX Already installed */ 2271 } 2272 } 2273 splx(s); 2274 2275 /* Allocate the new bw_meter entry */ 2276 x = malloc(sizeof(*x), M_BWMETER, M_INTWAIT); 2277 2278 /* Set the new bw_meter entry */ 2279 x->bm_threshold.b_time = req->bu_threshold.b_time; 2280 GET_TIME(now); 2281 x->bm_start_time = now; 2282 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2283 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2284 x->bm_measured.b_packets = 0; 2285 x->bm_measured.b_bytes = 0; 2286 x->bm_flags = flags; 2287 x->bm_time_next = NULL; 2288 x->bm_time_hash = BW_METER_BUCKETS; 2289 2290 /* Add the new bw_meter entry to the front of entries for this MFC */ 2291 s = splnet(); 2292 x->bm_mfc = mfc; 2293 x->bm_mfc_next = mfc->mfc_bw_meter; 2294 mfc->mfc_bw_meter = x; 2295 schedule_bw_meter(x, &now); 2296 splx(s); 2297 2298 return 0; 2299 } 2300 2301 static void 2302 free_bw_list(struct bw_meter *list) 2303 { 2304 while (list != NULL) { 2305 struct bw_meter *x = list; 2306 2307 list = list->bm_mfc_next; 2308 unschedule_bw_meter(x); 2309 free(x, M_BWMETER); 2310 } 2311 } 2312 2313 /* 2314 * Delete one or multiple bw_meter entries 2315 */ 2316 static int 2317 del_bw_upcall(struct bw_upcall *req) 2318 { 2319 struct mfc *mfc; 2320 struct bw_meter *x; 2321 int s; 2322 2323 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2324 return EOPNOTSUPP; 2325 2326 s = splnet(); 2327 /* Find the corresponding MFC entry */ 2328 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2329 if (mfc == NULL) { 2330 splx(s); 2331 return EADDRNOTAVAIL; 2332 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2333 /* 2334 * Delete all bw_meter entries for this mfc 2335 */ 2336 struct bw_meter *list; 2337 2338 list = mfc->mfc_bw_meter; 2339 mfc->mfc_bw_meter = NULL; 2340 splx(s); 2341 free_bw_list(list); 2342 return 0; 2343 } else { /* Delete a single bw_meter entry */ 2344 struct bw_meter *prev; 2345 uint32_t flags = 0; 2346 2347 flags = compute_bw_meter_flags(req); 2348 2349 /* Find the bw_meter entry to delete */ 2350 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2351 prev = x, x = x->bm_mfc_next) { 2352 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2353 &req->bu_threshold.b_time, ==)) && 2354 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2355 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2356 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2357 break; 2358 } 2359 if (x != NULL) { /* Delete entry from the list for this MFC */ 2360 if (prev != NULL) 2361 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2362 else 2363 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2364 splx(s); 2365 2366 unschedule_bw_meter(x); 2367 /* Free the bw_meter entry */ 2368 free(x, M_BWMETER); 2369 return 0; 2370 } else { 2371 splx(s); 2372 return EINVAL; 2373 } 2374 } 2375 /* NOTREACHED */ 2376 } 2377 2378 /* 2379 * Perform bandwidth measurement processing that may result in an upcall 2380 */ 2381 static void 2382 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2383 { 2384 struct timeval delta; 2385 int s; 2386 2387 s = splnet(); 2388 delta = *nowp; 2389 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2390 2391 if (x->bm_flags & BW_METER_GEQ) { 2392 /* 2393 * Processing for ">=" type of bw_meter entry 2394 */ 2395 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2396 /* Reset the bw_meter entry */ 2397 x->bm_start_time = *nowp; 2398 x->bm_measured.b_packets = 0; 2399 x->bm_measured.b_bytes = 0; 2400 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2401 } 2402 2403 /* Record that a packet is received */ 2404 x->bm_measured.b_packets++; 2405 x->bm_measured.b_bytes += plen; 2406 2407 /* 2408 * Test if we should deliver an upcall 2409 */ 2410 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2411 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2412 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2413 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2414 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2415 /* Prepare an upcall for delivery */ 2416 bw_meter_prepare_upcall(x, nowp); 2417 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2418 } 2419 } 2420 } else if (x->bm_flags & BW_METER_LEQ) { 2421 /* 2422 * Processing for "<=" type of bw_meter entry 2423 */ 2424 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2425 /* 2426 * We are behind time with the multicast forwarding table 2427 * scanning for "<=" type of bw_meter entries, so test now 2428 * if we should deliver an upcall. 2429 */ 2430 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2431 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2432 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2433 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2434 /* Prepare an upcall for delivery */ 2435 bw_meter_prepare_upcall(x, nowp); 2436 } 2437 /* Reschedule the bw_meter entry */ 2438 unschedule_bw_meter(x); 2439 schedule_bw_meter(x, nowp); 2440 } 2441 2442 /* Record that a packet is received */ 2443 x->bm_measured.b_packets++; 2444 x->bm_measured.b_bytes += plen; 2445 2446 /* 2447 * Test if we should restart the measuring interval 2448 */ 2449 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2450 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2451 (x->bm_flags & BW_METER_UNIT_BYTES && 2452 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2453 /* Don't restart the measuring interval */ 2454 } else { 2455 /* Do restart the measuring interval */ 2456 /* 2457 * XXX: note that we don't unschedule and schedule, because this 2458 * might be too much overhead per packet. Instead, when we process 2459 * all entries for a given timer hash bin, we check whether it is 2460 * really a timeout. If not, we reschedule at that time. 2461 */ 2462 x->bm_start_time = *nowp; 2463 x->bm_measured.b_packets = 0; 2464 x->bm_measured.b_bytes = 0; 2465 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2466 } 2467 } 2468 splx(s); 2469 } 2470 2471 /* 2472 * Prepare a bandwidth-related upcall 2473 */ 2474 static void 2475 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2476 { 2477 struct timeval delta; 2478 struct bw_upcall *u; 2479 int s; 2480 2481 s = splnet(); 2482 2483 /* 2484 * Compute the measured time interval 2485 */ 2486 delta = *nowp; 2487 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2488 2489 /* 2490 * If there are too many pending upcalls, deliver them now 2491 */ 2492 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2493 bw_upcalls_send(); 2494 2495 /* 2496 * Set the bw_upcall entry 2497 */ 2498 u = &bw_upcalls[bw_upcalls_n++]; 2499 u->bu_src = x->bm_mfc->mfc_origin; 2500 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2501 u->bu_threshold.b_time = x->bm_threshold.b_time; 2502 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2503 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2504 u->bu_measured.b_time = delta; 2505 u->bu_measured.b_packets = x->bm_measured.b_packets; 2506 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2507 u->bu_flags = 0; 2508 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2509 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2510 if (x->bm_flags & BW_METER_UNIT_BYTES) 2511 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2512 if (x->bm_flags & BW_METER_GEQ) 2513 u->bu_flags |= BW_UPCALL_GEQ; 2514 if (x->bm_flags & BW_METER_LEQ) 2515 u->bu_flags |= BW_UPCALL_LEQ; 2516 2517 splx(s); 2518 } 2519 2520 /* 2521 * Send the pending bandwidth-related upcalls 2522 */ 2523 static void 2524 bw_upcalls_send(void) 2525 { 2526 struct mbuf *m; 2527 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2528 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2529 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2530 0, /* unused2 */ 2531 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2532 0, /* im_mbz */ 2533 0, /* im_vif */ 2534 0, /* unused3 */ 2535 { 0 }, /* im_src */ 2536 { 0 } }; /* im_dst */ 2537 2538 if (bw_upcalls_n == 0) 2539 return; /* No pending upcalls */ 2540 2541 bw_upcalls_n = 0; 2542 2543 /* 2544 * Allocate a new mbuf, initialize it with the header and 2545 * the payload for the pending calls. 2546 */ 2547 MGETHDR(m, MB_DONTWAIT, MT_HEADER); 2548 if (m == NULL) { 2549 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2550 return; 2551 } 2552 2553 m->m_len = m->m_pkthdr.len = 0; 2554 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2555 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2556 2557 /* 2558 * Send the upcalls 2559 * XXX do we need to set the address in k_igmpsrc ? 2560 */ 2561 mrtstat.mrts_upcalls++; 2562 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2563 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2564 ++mrtstat.mrts_upq_sockfull; 2565 } 2566 } 2567 2568 /* 2569 * Compute the timeout hash value for the bw_meter entries 2570 */ 2571 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2572 do { \ 2573 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2574 \ 2575 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2576 (hash) = next_timeval.tv_sec; \ 2577 if (next_timeval.tv_usec) \ 2578 (hash)++; /* XXX: make sure we don't timeout early */ \ 2579 (hash) %= BW_METER_BUCKETS; \ 2580 } while (0) 2581 2582 /* 2583 * Schedule a timer to process periodically bw_meter entry of type "<=" 2584 * by linking the entry in the proper hash bucket. 2585 */ 2586 static void 2587 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2588 { 2589 int time_hash, s; 2590 2591 if (!(x->bm_flags & BW_METER_LEQ)) 2592 return; /* XXX: we schedule timers only for "<=" entries */ 2593 2594 /* 2595 * Reset the bw_meter entry 2596 */ 2597 s = splnet(); 2598 x->bm_start_time = *nowp; 2599 x->bm_measured.b_packets = 0; 2600 x->bm_measured.b_bytes = 0; 2601 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2602 splx(s); 2603 2604 /* 2605 * Compute the timeout hash value and insert the entry 2606 */ 2607 BW_METER_TIMEHASH(x, time_hash); 2608 x->bm_time_next = bw_meter_timers[time_hash]; 2609 bw_meter_timers[time_hash] = x; 2610 x->bm_time_hash = time_hash; 2611 } 2612 2613 /* 2614 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2615 * by removing the entry from the proper hash bucket. 2616 */ 2617 static void 2618 unschedule_bw_meter(struct bw_meter *x) 2619 { 2620 int time_hash; 2621 struct bw_meter *prev, *tmp; 2622 2623 if (!(x->bm_flags & BW_METER_LEQ)) 2624 return; /* XXX: we schedule timers only for "<=" entries */ 2625 2626 /* 2627 * Compute the timeout hash value and delete the entry 2628 */ 2629 time_hash = x->bm_time_hash; 2630 if (time_hash >= BW_METER_BUCKETS) 2631 return; /* Entry was not scheduled */ 2632 2633 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2634 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2635 if (tmp == x) 2636 break; 2637 2638 if (tmp == NULL) 2639 panic("unschedule_bw_meter: bw_meter entry not found"); 2640 2641 if (prev != NULL) 2642 prev->bm_time_next = x->bm_time_next; 2643 else 2644 bw_meter_timers[time_hash] = x->bm_time_next; 2645 2646 x->bm_time_next = NULL; 2647 x->bm_time_hash = BW_METER_BUCKETS; 2648 } 2649 2650 2651 /* 2652 * Process all "<=" type of bw_meter that should be processed now, 2653 * and for each entry prepare an upcall if necessary. Each processed 2654 * entry is rescheduled again for the (periodic) processing. 2655 * 2656 * This is run periodically (once per second normally). On each round, 2657 * all the potentially matching entries are in the hash slot that we are 2658 * looking at. 2659 */ 2660 static void 2661 bw_meter_process() 2662 { 2663 static uint32_t last_tv_sec; /* last time we processed this */ 2664 2665 uint32_t loops; 2666 int i, s; 2667 struct timeval now, process_endtime; 2668 2669 GET_TIME(now); 2670 if (last_tv_sec == now.tv_sec) 2671 return; /* nothing to do */ 2672 2673 s = splnet(); 2674 loops = now.tv_sec - last_tv_sec; 2675 last_tv_sec = now.tv_sec; 2676 if (loops > BW_METER_BUCKETS) 2677 loops = BW_METER_BUCKETS; 2678 2679 /* 2680 * Process all bins of bw_meter entries from the one after the last 2681 * processed to the current one. On entry, i points to the last bucket 2682 * visited, so we need to increment i at the beginning of the loop. 2683 */ 2684 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2685 struct bw_meter *x, *tmp_list; 2686 2687 if (++i >= BW_METER_BUCKETS) 2688 i = 0; 2689 2690 /* Disconnect the list of bw_meter entries from the bin */ 2691 tmp_list = bw_meter_timers[i]; 2692 bw_meter_timers[i] = NULL; 2693 2694 /* Process the list of bw_meter entries */ 2695 while (tmp_list != NULL) { 2696 x = tmp_list; 2697 tmp_list = tmp_list->bm_time_next; 2698 2699 /* Test if the time interval is over */ 2700 process_endtime = x->bm_start_time; 2701 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2702 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2703 /* Not yet: reschedule, but don't reset */ 2704 int time_hash; 2705 2706 BW_METER_TIMEHASH(x, time_hash); 2707 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2708 /* 2709 * XXX: somehow the bin processing is a bit ahead of time. 2710 * Put the entry in the next bin. 2711 */ 2712 if (++time_hash >= BW_METER_BUCKETS) 2713 time_hash = 0; 2714 } 2715 x->bm_time_next = bw_meter_timers[time_hash]; 2716 bw_meter_timers[time_hash] = x; 2717 x->bm_time_hash = time_hash; 2718 2719 continue; 2720 } 2721 2722 /* 2723 * Test if we should deliver an upcall 2724 */ 2725 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2726 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2727 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2728 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2729 /* Prepare an upcall for delivery */ 2730 bw_meter_prepare_upcall(x, &now); 2731 } 2732 2733 /* 2734 * Reschedule for next processing 2735 */ 2736 schedule_bw_meter(x, &now); 2737 } 2738 } 2739 splx(s); 2740 2741 /* Send all upcalls that are pending delivery */ 2742 bw_upcalls_send(); 2743 } 2744 2745 /* 2746 * A periodic function for sending all upcalls that are pending delivery 2747 */ 2748 static void 2749 expire_bw_upcalls_send(void *unused) 2750 { 2751 bw_upcalls_send(); 2752 2753 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2754 expire_bw_upcalls_send, NULL); 2755 } 2756 2757 /* 2758 * A periodic function for periodic scanning of the multicast forwarding 2759 * table for processing all "<=" bw_meter entries. 2760 */ 2761 static void 2762 expire_bw_meter_process(void *unused) 2763 { 2764 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2765 bw_meter_process(); 2766 2767 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 2768 expire_bw_meter_process, NULL); 2769 } 2770 2771 /* 2772 * End of bandwidth monitoring code 2773 */ 2774 2775 #ifdef PIM 2776 /* 2777 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2778 * 2779 */ 2780 static int 2781 pim_register_send(struct ip *ip, struct vif *vifp, 2782 struct mbuf *m, struct mfc *rt) 2783 { 2784 struct mbuf *mb_copy, *mm; 2785 2786 if (mrtdebug & DEBUG_PIM) 2787 log(LOG_DEBUG, "pim_register_send: "); 2788 2789 mb_copy = pim_register_prepare(ip, m); 2790 if (mb_copy == NULL) 2791 return ENOBUFS; 2792 2793 /* 2794 * Send all the fragments. Note that the mbuf for each fragment 2795 * is freed by the sending machinery. 2796 */ 2797 for (mm = mb_copy; mm; mm = mb_copy) { 2798 mb_copy = mm->m_nextpkt; 2799 mm->m_nextpkt = 0; 2800 mm = m_pullup(mm, sizeof(struct ip)); 2801 if (mm != NULL) { 2802 ip = mtod(mm, struct ip *); 2803 if ((mrt_api_config & MRT_MFC_RP) && 2804 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2805 pim_register_send_rp(ip, vifp, mm, rt); 2806 } else { 2807 pim_register_send_upcall(ip, vifp, mm, rt); 2808 } 2809 } 2810 } 2811 2812 return 0; 2813 } 2814 2815 /* 2816 * Return a copy of the data packet that is ready for PIM Register 2817 * encapsulation. 2818 * XXX: Note that in the returned copy the IP header is a valid one. 2819 */ 2820 static struct mbuf * 2821 pim_register_prepare(struct ip *ip, struct mbuf *m) 2822 { 2823 struct mbuf *mb_copy = NULL; 2824 int mtu; 2825 2826 /* Take care of delayed checksums */ 2827 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2828 in_delayed_cksum(m); 2829 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2830 } 2831 2832 /* 2833 * Copy the old packet & pullup its IP header into the 2834 * new mbuf so we can modify it. 2835 */ 2836 mb_copy = m_copypacket(m, MB_DONTWAIT); 2837 if (mb_copy == NULL) 2838 return NULL; 2839 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2840 if (mb_copy == NULL) 2841 return NULL; 2842 2843 /* take care of the TTL */ 2844 ip = mtod(mb_copy, struct ip *); 2845 --ip->ip_ttl; 2846 2847 /* Compute the MTU after the PIM Register encapsulation */ 2848 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2849 2850 if (ip->ip_len <= mtu) { 2851 /* Turn the IP header into a valid one */ 2852 ip->ip_len = htons(ip->ip_len); 2853 ip->ip_off = htons(ip->ip_off); 2854 ip->ip_sum = 0; 2855 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2856 } else { 2857 /* Fragment the packet */ 2858 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2859 m_freem(mb_copy); 2860 return NULL; 2861 } 2862 } 2863 return mb_copy; 2864 } 2865 2866 /* 2867 * Send an upcall with the data packet to the user-level process. 2868 */ 2869 static int 2870 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2871 struct mbuf *mb_copy, struct mfc *rt) 2872 { 2873 struct mbuf *mb_first; 2874 int len = ntohs(ip->ip_len); 2875 struct igmpmsg *im; 2876 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2877 2878 /* 2879 * Add a new mbuf with an upcall header 2880 */ 2881 MGETHDR(mb_first, MB_DONTWAIT, MT_HEADER); 2882 if (mb_first == NULL) { 2883 m_freem(mb_copy); 2884 return ENOBUFS; 2885 } 2886 mb_first->m_data += max_linkhdr; 2887 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2888 mb_first->m_len = sizeof(struct igmpmsg); 2889 mb_first->m_next = mb_copy; 2890 2891 /* Send message to routing daemon */ 2892 im = mtod(mb_first, struct igmpmsg *); 2893 im->im_msgtype = IGMPMSG_WHOLEPKT; 2894 im->im_mbz = 0; 2895 im->im_vif = vifp - viftable; 2896 im->im_src = ip->ip_src; 2897 im->im_dst = ip->ip_dst; 2898 2899 k_igmpsrc.sin_addr = ip->ip_src; 2900 2901 mrtstat.mrts_upcalls++; 2902 2903 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2904 if (mrtdebug & DEBUG_PIM) 2905 log(LOG_WARNING, 2906 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2907 ++mrtstat.mrts_upq_sockfull; 2908 return ENOBUFS; 2909 } 2910 2911 /* Keep statistics */ 2912 pimstat.pims_snd_registers_msgs++; 2913 pimstat.pims_snd_registers_bytes += len; 2914 2915 return 0; 2916 } 2917 2918 /* 2919 * Encapsulate the data packet in PIM Register message and send it to the RP. 2920 */ 2921 static int 2922 pim_register_send_rp(struct ip *ip, struct vif *vifp, 2923 struct mbuf *mb_copy, struct mfc *rt) 2924 { 2925 struct mbuf *mb_first; 2926 struct ip *ip_outer; 2927 struct pim_encap_pimhdr *pimhdr; 2928 int len = ntohs(ip->ip_len); 2929 vifi_t vifi = rt->mfc_parent; 2930 2931 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2932 m_freem(mb_copy); 2933 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2934 } 2935 2936 /* 2937 * Add a new mbuf with the encapsulating header 2938 */ 2939 MGETHDR(mb_first, MB_DONTWAIT, MT_HEADER); 2940 if (mb_first == NULL) { 2941 m_freem(mb_copy); 2942 return ENOBUFS; 2943 } 2944 mb_first->m_data += max_linkhdr; 2945 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2946 mb_first->m_next = mb_copy; 2947 2948 mb_first->m_pkthdr.len = len + mb_first->m_len; 2949 2950 /* 2951 * Fill in the encapsulating IP and PIM header 2952 */ 2953 ip_outer = mtod(mb_first, struct ip *); 2954 *ip_outer = pim_encap_iphdr; 2955 #ifdef RANDOM_IP_ID 2956 ip_outer->ip_id = ip_randomid(); 2957 #else 2958 ip_outer->ip_id = htons(ip_id++); 2959 #endif 2960 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2961 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2962 ip_outer->ip_dst = rt->mfc_rp; 2963 /* 2964 * Copy the inner header TOS to the outer header, and take care of the 2965 * IP_DF bit. 2966 */ 2967 ip_outer->ip_tos = ip->ip_tos; 2968 if (ntohs(ip->ip_off) & IP_DF) 2969 ip_outer->ip_off |= IP_DF; 2970 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2971 + sizeof(pim_encap_iphdr)); 2972 *pimhdr = pim_encap_pimhdr; 2973 /* If the iif crosses a border, set the Border-bit */ 2974 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2975 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2976 2977 mb_first->m_data += sizeof(pim_encap_iphdr); 2978 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2979 mb_first->m_data -= sizeof(pim_encap_iphdr); 2980 2981 if (vifp->v_rate_limit == 0) 2982 tbf_send_packet(vifp, mb_first); 2983 else 2984 tbf_control(vifp, mb_first, ip, ip_outer->ip_len); 2985 2986 /* Keep statistics */ 2987 pimstat.pims_snd_registers_msgs++; 2988 pimstat.pims_snd_registers_bytes += len; 2989 2990 return 0; 2991 } 2992 2993 /* 2994 * PIM-SMv2 and PIM-DM messages processing. 2995 * Receives and verifies the PIM control messages, and passes them 2996 * up to the listening socket, using rip_input(). 2997 * The only message with special processing is the PIM_REGISTER message 2998 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2999 * is passed to if_simloop(). 3000 */ 3001 void 3002 pim_input(struct mbuf *m, ...) 3003 { 3004 int off, proto; 3005 struct ip *ip = mtod(m, struct ip *); 3006 struct pim *pim; 3007 int minlen; 3008 int datalen = ip->ip_len; 3009 int ip_tos; 3010 int iphlen; 3011 __va_list ap; 3012 3013 __va_start(ap, m); 3014 off = __va_arg(ap, int); 3015 proto = __va_arg(ap, int); 3016 __va_end(ap); 3017 3018 iphlen = off; 3019 3020 /* Keep statistics */ 3021 pimstat.pims_rcv_total_msgs++; 3022 pimstat.pims_rcv_total_bytes += datalen; 3023 3024 /* 3025 * Validate lengths 3026 */ 3027 if (datalen < PIM_MINLEN) { 3028 pimstat.pims_rcv_tooshort++; 3029 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3030 datalen, (u_long)ip->ip_src.s_addr); 3031 m_freem(m); 3032 return; 3033 } 3034 3035 /* 3036 * If the packet is at least as big as a REGISTER, go agead 3037 * and grab the PIM REGISTER header size, to avoid another 3038 * possible m_pullup() later. 3039 * 3040 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3041 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3042 */ 3043 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3044 /* 3045 * Get the IP and PIM headers in contiguous memory, and 3046 * possibly the PIM REGISTER header. 3047 */ 3048 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3049 (m = m_pullup(m, minlen)) == 0) { 3050 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3051 return; 3052 } 3053 /* m_pullup() may have given us a new mbuf so reset ip. */ 3054 ip = mtod(m, struct ip *); 3055 ip_tos = ip->ip_tos; 3056 3057 /* adjust mbuf to point to the PIM header */ 3058 m->m_data += iphlen; 3059 m->m_len -= iphlen; 3060 pim = mtod(m, struct pim *); 3061 3062 /* 3063 * Validate checksum. If PIM REGISTER, exclude the data packet. 3064 * 3065 * XXX: some older PIMv2 implementations don't make this distinction, 3066 * so for compatibility reason perform the checksum over part of the 3067 * message, and if error, then over the whole message. 3068 */ 3069 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3070 /* do nothing, checksum okay */ 3071 } else if (in_cksum(m, datalen)) { 3072 pimstat.pims_rcv_badsum++; 3073 if (mrtdebug & DEBUG_PIM) 3074 log(LOG_DEBUG, "pim_input: invalid checksum"); 3075 m_freem(m); 3076 return; 3077 } 3078 3079 /* PIM version check */ 3080 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3081 pimstat.pims_rcv_badversion++; 3082 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3083 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3084 m_freem(m); 3085 return; 3086 } 3087 3088 /* restore mbuf back to the outer IP */ 3089 m->m_data -= iphlen; 3090 m->m_len += iphlen; 3091 3092 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3093 /* 3094 * Since this is a REGISTER, we'll make a copy of the register 3095 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3096 * routing daemon. 3097 */ 3098 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3099 struct mbuf *mcp; 3100 struct ip *encap_ip; 3101 u_int32_t *reghdr; 3102 3103 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3104 if (mrtdebug & DEBUG_PIM) 3105 log(LOG_DEBUG, 3106 "pim_input: register vif not set: %d\n", reg_vif_num); 3107 m_freem(m); 3108 return; 3109 } 3110 3111 /* 3112 * Validate length 3113 */ 3114 if (datalen < PIM_REG_MINLEN) { 3115 pimstat.pims_rcv_tooshort++; 3116 pimstat.pims_rcv_badregisters++; 3117 log(LOG_ERR, 3118 "pim_input: register packet size too small %d from %lx\n", 3119 datalen, (u_long)ip->ip_src.s_addr); 3120 m_freem(m); 3121 return; 3122 } 3123 3124 reghdr = (u_int32_t *)(pim + 1); 3125 encap_ip = (struct ip *)(reghdr + 1); 3126 3127 if (mrtdebug & DEBUG_PIM) { 3128 log(LOG_DEBUG, 3129 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3130 (u_long)ntohl(encap_ip->ip_src.s_addr), 3131 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3132 ntohs(encap_ip->ip_len)); 3133 } 3134 3135 /* verify the version number of the inner packet */ 3136 if (encap_ip->ip_v != IPVERSION) { 3137 pimstat.pims_rcv_badregisters++; 3138 if (mrtdebug & DEBUG_PIM) { 3139 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3140 "of the inner packet\n", encap_ip->ip_v); 3141 } 3142 m_freem(m); 3143 return; 3144 } 3145 3146 /* verify the inner packet is destined to a mcast group */ 3147 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 3148 pimstat.pims_rcv_badregisters++; 3149 if (mrtdebug & DEBUG_PIM) 3150 log(LOG_DEBUG, 3151 "pim_input: inner packet of register is not " 3152 "multicast %lx\n", 3153 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3154 m_freem(m); 3155 return; 3156 } 3157 3158 /* If a NULL_REGISTER, pass it to the daemon */ 3159 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3160 goto pim_input_to_daemon; 3161 3162 /* 3163 * Copy the TOS from the outer IP header to the inner IP header. 3164 */ 3165 if (encap_ip->ip_tos != ip_tos) { 3166 /* Outer TOS -> inner TOS */ 3167 encap_ip->ip_tos = ip_tos; 3168 /* Recompute the inner header checksum. Sigh... */ 3169 3170 /* adjust mbuf to point to the inner IP header */ 3171 m->m_data += (iphlen + PIM_MINLEN); 3172 m->m_len -= (iphlen + PIM_MINLEN); 3173 3174 encap_ip->ip_sum = 0; 3175 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3176 3177 /* restore mbuf to point back to the outer IP header */ 3178 m->m_data -= (iphlen + PIM_MINLEN); 3179 m->m_len += (iphlen + PIM_MINLEN); 3180 } 3181 3182 /* 3183 * Decapsulate the inner IP packet and loopback to forward it 3184 * as a normal multicast packet. Also, make a copy of the 3185 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3186 * to pass to the daemon later, so it can take the appropriate 3187 * actions (e.g., send back PIM_REGISTER_STOP). 3188 * XXX: here m->m_data points to the outer IP header. 3189 */ 3190 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3191 if (mcp == NULL) { 3192 log(LOG_ERR, 3193 "pim_input: pim register: could not copy register head\n"); 3194 m_freem(m); 3195 return; 3196 } 3197 3198 /* Keep statistics */ 3199 /* XXX: registers_bytes include only the encap. mcast pkt */ 3200 pimstat.pims_rcv_registers_msgs++; 3201 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3202 3203 /* 3204 * forward the inner ip packet; point m_data at the inner ip. 3205 */ 3206 m_adj(m, iphlen + PIM_MINLEN); 3207 3208 if (mrtdebug & DEBUG_PIM) { 3209 log(LOG_DEBUG, 3210 "pim_input: forwarding decapsulated register: " 3211 "src %lx, dst %lx, vif %d\n", 3212 (u_long)ntohl(encap_ip->ip_src.s_addr), 3213 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3214 reg_vif_num); 3215 } 3216 if_simloop(viftable[reg_vif_num].v_ifp, m, dst.sin_family, 0); 3217 3218 /* prepare the register head to send to the mrouting daemon */ 3219 m = mcp; 3220 } 3221 3222 pim_input_to_daemon: 3223 /* 3224 * Pass the PIM message up to the daemon; if it is a Register message, 3225 * pass the 'head' only up to the daemon. This includes the 3226 * outer IP header, PIM header, PIM-Register header and the 3227 * inner IP header. 3228 * XXX: the outer IP header pkt size of a Register is not adjust to 3229 * reflect the fact that the inner multicast data is truncated. 3230 */ 3231 rip_input(m, iphlen, proto); 3232 3233 return; 3234 } 3235 #endif /* PIM */ 3236 3237 static int 3238 ip_mroute_modevent(module_t mod, int type, void *unused) 3239 { 3240 int s; 3241 3242 switch (type) { 3243 case MOD_LOAD: 3244 s = splnet(); 3245 /* XXX Protect against multiple loading */ 3246 ip_mcast_src = X_ip_mcast_src; 3247 ip_mforward = X_ip_mforward; 3248 ip_mrouter_done = X_ip_mrouter_done; 3249 ip_mrouter_get = X_ip_mrouter_get; 3250 ip_mrouter_set = X_ip_mrouter_set; 3251 ip_rsvp_force_done = X_ip_rsvp_force_done; 3252 ip_rsvp_vif = X_ip_rsvp_vif; 3253 ipip_input = X_ipip_input; 3254 legal_vif_num = X_legal_vif_num; 3255 mrt_ioctl = X_mrt_ioctl; 3256 rsvp_input_p = X_rsvp_input; 3257 splx(s); 3258 break; 3259 3260 case MOD_UNLOAD: 3261 if (ip_mrouter) 3262 return EINVAL; 3263 3264 s = splnet(); 3265 ip_mcast_src = NULL; 3266 ip_mforward = NULL; 3267 ip_mrouter_done = NULL; 3268 ip_mrouter_get = NULL; 3269 ip_mrouter_set = NULL; 3270 ip_rsvp_force_done = NULL; 3271 ip_rsvp_vif = NULL; 3272 ipip_input = NULL; 3273 legal_vif_num = NULL; 3274 mrt_ioctl = NULL; 3275 rsvp_input_p = NULL; 3276 splx(s); 3277 break; 3278 } 3279 return 0; 3280 } 3281 3282 static moduledata_t ip_mroutemod = { 3283 "ip_mroute", 3284 ip_mroute_modevent, 3285 0 3286 }; 3287 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3288