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 */ 22 23 #include "opt_mrouting.h" 24 25 #ifdef PIM 26 #define _PIM_VT 1 27 #endif 28 29 #include <sys/param.h> 30 #include <sys/kernel.h> 31 #include <sys/malloc.h> 32 #include <sys/mbuf.h> 33 #include <sys/protosw.h> 34 #include <sys/socket.h> 35 #include <sys/socketvar.h> 36 #include <sys/sockio.h> 37 #include <sys/sysctl.h> 38 #include <sys/syslog.h> 39 #include <sys/systm.h> 40 #include <sys/thread2.h> 41 #include <sys/time.h> 42 #include <sys/in_cksum.h> 43 44 #include <machine/stdarg.h> 45 46 #include <net/if.h> 47 #include <net/netisr.h> 48 #include <net/route.h> 49 #include <netinet/in.h> 50 #include <netinet/igmp.h> 51 #include <netinet/in_systm.h> 52 #include <netinet/in_var.h> 53 #include <netinet/ip.h> 54 #include "ip_mroute.h" 55 #include <netinet/ip_var.h> 56 #ifdef PIM 57 #include <netinet/pim.h> 58 #include <netinet/pim_var.h> 59 #endif 60 #ifdef ALTQ 61 #include <netinet/in_pcb.h> 62 #endif 63 #include <netinet/udp.h> 64 65 /* 66 * Control debugging code for rsvp and multicast routing code. 67 * Can only set them with the debugger. 68 */ 69 static u_int rsvpdebug; /* non-zero enables debugging */ 70 71 static u_int mrtdebug; /* any set of the flags below */ 72 73 #define DEBUG_MFC 0x02 74 #define DEBUG_FORWARD 0x04 75 #define DEBUG_EXPIRE 0x08 76 #define DEBUG_XMIT 0x10 77 #define DEBUG_PIM 0x20 78 79 #define VIFI_INVALID ((vifi_t) -1) 80 81 #define M_HASCL(m) ((m)->m_flags & M_EXT) 82 83 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 84 85 static struct mrtstat mrtstat; 86 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 87 &mrtstat, mrtstat, 88 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 89 90 static struct mfc *mfctable[MFCTBLSIZ]; 91 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 92 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 93 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 94 95 static struct vif viftable[MAXVIFS]; 96 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 97 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 98 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 99 100 static u_char nexpire[MFCTBLSIZ]; 101 102 struct lwkt_token mroute_token = LWKT_TOKEN_INITIALIZER(mroute_token); 103 104 105 static struct callout expire_upcalls_ch; 106 static struct callout tbf_reprocess_q_ch; 107 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 108 #define UPCALL_EXPIRE 6 /* number of timeouts */ 109 110 /* 111 * Define the token bucket filter structures 112 * tbftable -> each vif has one of these for storing info 113 */ 114 115 static struct tbf tbftable[MAXVIFS]; 116 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 117 118 /* 119 * 'Interfaces' associated with decapsulator (so we can tell 120 * packets that went through it from ones that get reflected 121 * by a broken gateway). These interfaces are never linked into 122 * the system ifnet list & no routes point to them. I.e., packets 123 * can't be sent this way. They only exist as a placeholder for 124 * multicast source verification. 125 */ 126 static struct ifnet multicast_decap_if[MAXVIFS]; 127 128 #define ENCAP_TTL 64 129 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 130 131 /* prototype IP hdr for encapsulated packets */ 132 static struct ip multicast_encap_iphdr = { 133 #if BYTE_ORDER == LITTLE_ENDIAN 134 sizeof(struct ip) >> 2, IPVERSION, 135 #else 136 IPVERSION, sizeof(struct ip) >> 2, 137 #endif 138 0, /* tos */ 139 sizeof(struct ip), /* total length */ 140 0, /* id */ 141 0, /* frag offset */ 142 ENCAP_TTL, ENCAP_PROTO, 143 0, /* checksum */ 144 }; 145 146 /* 147 * Bandwidth meter variables and constants 148 */ 149 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 150 /* 151 * Pending timeouts are stored in a hash table, the key being the 152 * expiration time. Periodically, the entries are analysed and processed. 153 */ 154 #define BW_METER_BUCKETS 1024 155 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 156 static struct callout bw_meter_ch; 157 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 158 159 /* 160 * Pending upcalls are stored in a vector which is flushed when 161 * full, or periodically 162 */ 163 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 164 static u_int bw_upcalls_n; /* # of pending upcalls */ 165 static struct callout bw_upcalls_ch; 166 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 167 168 #ifdef PIM 169 static struct pimstat pimstat; 170 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 171 &pimstat, pimstat, 172 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 173 174 /* 175 * Note: the PIM Register encapsulation adds the following in front of a 176 * data packet: 177 * 178 * struct pim_encap_hdr { 179 * struct ip ip; 180 * struct pim_encap_pimhdr pim; 181 * } 182 * 183 */ 184 185 struct pim_encap_pimhdr { 186 struct pim pim; 187 uint32_t flags; 188 }; 189 190 static struct ip pim_encap_iphdr = { 191 #if BYTE_ORDER == LITTLE_ENDIAN 192 sizeof(struct ip) >> 2, 193 IPVERSION, 194 #else 195 IPVERSION, 196 sizeof(struct ip) >> 2, 197 #endif 198 0, /* tos */ 199 sizeof(struct ip), /* total length */ 200 0, /* id */ 201 0, /* frag offset */ 202 ENCAP_TTL, 203 IPPROTO_PIM, 204 0, /* checksum */ 205 }; 206 207 static struct pim_encap_pimhdr pim_encap_pimhdr = { 208 { 209 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 210 0, /* reserved */ 211 0, /* checksum */ 212 }, 213 0 /* flags */ 214 }; 215 216 static struct ifnet multicast_register_if; 217 static vifi_t reg_vif_num = VIFI_INVALID; 218 #endif /* PIM */ 219 220 /* 221 * Private variables. 222 */ 223 static vifi_t numvifs; 224 static int have_encap_tunnel; 225 226 /* 227 * one-back cache used by ipip_input to locate a tunnel's vif 228 * given a datagram's src ip address. 229 */ 230 static u_long last_encap_src; 231 static struct vif *last_encap_vif; 232 233 static u_long X_ip_mcast_src(int vifi); 234 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 235 struct mbuf *m, struct ip_moptions *imo); 236 static int X_ip_mrouter_done(void); 237 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 238 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 239 static int X_legal_vif_num(int vif); 240 static int X_mrt_ioctl(u_long cmd, caddr_t data); 241 242 static int get_sg_cnt(struct sioc_sg_req *); 243 static int get_vif_cnt(struct sioc_vif_req *); 244 static int ip_mrouter_init(struct socket *, int); 245 static int add_vif(struct vifctl *); 246 static int del_vif(vifi_t); 247 static int add_mfc(struct mfcctl2 *); 248 static int del_mfc(struct mfcctl2 *); 249 static int set_api_config(uint32_t *); /* chose API capabilities */ 250 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 251 static int set_assert(int); 252 static void expire_upcalls(void *); 253 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 254 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 255 static void encap_send(struct ip *, struct vif *, struct mbuf *); 256 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 257 static void tbf_queue(struct vif *, struct mbuf *); 258 static void tbf_process_q(struct vif *); 259 static void tbf_reprocess_q(void *); 260 static int tbf_dq_sel(struct vif *, struct ip *); 261 static void tbf_send_packet(struct vif *, struct mbuf *); 262 static void tbf_update_tokens(struct vif *); 263 static int priority(struct vif *, struct ip *); 264 265 /* 266 * Bandwidth monitoring 267 */ 268 static void free_bw_list(struct bw_meter *list); 269 static int add_bw_upcall(struct bw_upcall *); 270 static int del_bw_upcall(struct bw_upcall *); 271 static void bw_meter_receive_packet(struct bw_meter *x, int plen, 272 struct timeval *nowp); 273 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 274 static void bw_upcalls_send(void); 275 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 276 static void unschedule_bw_meter(struct bw_meter *x); 277 static void bw_meter_process(void); 278 static void expire_bw_upcalls_send(void *); 279 static void expire_bw_meter_process(void *); 280 281 #ifdef PIM 282 static int pim_register_send(struct ip *, struct vif *, 283 struct mbuf *, struct mfc *); 284 static int pim_register_send_rp(struct ip *, struct vif *, 285 struct mbuf *, struct mfc *); 286 static int pim_register_send_upcall(struct ip *, struct vif *, 287 struct mbuf *, struct mfc *); 288 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 289 #endif 290 291 /* 292 * whether or not special PIM assert processing is enabled. 293 */ 294 static int pim_assert; 295 /* 296 * Rate limit for assert notification messages, in usec 297 */ 298 #define ASSERT_MSG_TIME 3000000 299 300 /* 301 * Kernel multicast routing API capabilities and setup. 302 * If more API capabilities are added to the kernel, they should be 303 * recorded in `mrt_api_support'. 304 */ 305 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 306 MRT_MFC_FLAGS_BORDER_VIF | 307 MRT_MFC_RP | 308 MRT_MFC_BW_UPCALL); 309 static uint32_t mrt_api_config = 0; 310 311 /* 312 * Hash function for a source, group entry 313 */ 314 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 315 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 316 317 /* 318 * Find a route for a given origin IP address and Multicast group address 319 * Type of service parameter to be added in the future!!! 320 * Statistics are updated by the caller if needed 321 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 322 */ 323 static struct mfc * 324 mfc_find(in_addr_t o, in_addr_t g) 325 { 326 struct mfc *rt; 327 328 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 329 if ((rt->mfc_origin.s_addr == o) && 330 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 331 break; 332 return rt; 333 } 334 335 /* 336 * Macros to compute elapsed time efficiently 337 * Borrowed from Van Jacobson's scheduling code 338 */ 339 #define TV_DELTA(a, b, delta) { \ 340 int xxs; \ 341 delta = (a).tv_usec - (b).tv_usec; \ 342 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 343 switch (xxs) { \ 344 case 2: \ 345 delta += 1000000; \ 346 /* FALLTHROUGH */ \ 347 case 1: \ 348 delta += 1000000; \ 349 break; \ 350 default: \ 351 delta += (1000000 * xxs); \ 352 } \ 353 } \ 354 } 355 356 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 357 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 358 359 /* 360 * Handle MRT setsockopt commands to modify the multicast routing tables. 361 */ 362 static int 363 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 364 { 365 int error, optval; 366 vifi_t vifi; 367 struct vifctl vifc; 368 struct mfcctl2 mfc; 369 struct bw_upcall bw_upcall; 370 uint32_t i; 371 372 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 373 return EPERM; 374 375 error = 0; 376 switch (sopt->sopt_name) { 377 case MRT_INIT: 378 error = soopt_to_kbuf(sopt, &optval, sizeof optval, sizeof optval); 379 if (error) 380 break; 381 error = ip_mrouter_init(so, optval); 382 break; 383 384 case MRT_DONE: 385 error = ip_mrouter_done(); 386 break; 387 388 case MRT_ADD_VIF: 389 error = soopt_to_kbuf(sopt, &vifc, sizeof vifc, sizeof vifc); 390 if (error) 391 break; 392 error = add_vif(&vifc); 393 break; 394 395 case MRT_DEL_VIF: 396 error = soopt_to_kbuf(sopt, &vifi, sizeof vifi, sizeof vifi); 397 if (error) 398 break; 399 error = del_vif(vifi); 400 break; 401 402 case MRT_ADD_MFC: 403 case MRT_DEL_MFC: 404 /* 405 * select data size depending on API version. 406 */ 407 if (sopt->sopt_name == MRT_ADD_MFC && 408 mrt_api_config & MRT_API_FLAGS_ALL) { 409 error = soopt_to_kbuf(sopt, &mfc, sizeof(struct mfcctl2), 410 sizeof(struct mfcctl2)); 411 } else { 412 error = soopt_to_kbuf(sopt, &mfc, sizeof(struct mfcctl), 413 sizeof(struct mfcctl)); 414 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 415 sizeof(mfc) - sizeof(struct mfcctl)); 416 } 417 if (error) 418 break; 419 if (sopt->sopt_name == MRT_ADD_MFC) 420 error = add_mfc(&mfc); 421 else 422 error = del_mfc(&mfc); 423 break; 424 425 case MRT_ASSERT: 426 error = soopt_to_kbuf(sopt, &optval, sizeof optval, sizeof optval); 427 if (error) 428 break; 429 set_assert(optval); 430 break; 431 432 case MRT_API_CONFIG: 433 error = soopt_to_kbuf(sopt, &i, sizeof i, sizeof i); 434 if (!error) 435 error = set_api_config(&i); 436 if (!error) 437 soopt_from_kbuf(sopt, &i, sizeof i); 438 break; 439 440 case MRT_ADD_BW_UPCALL: 441 case MRT_DEL_BW_UPCALL: 442 error = soopt_to_kbuf(sopt, &bw_upcall, sizeof bw_upcall, 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 error = 0; 468 switch (sopt->sopt_name) { 469 case MRT_VERSION: 470 soopt_from_kbuf(sopt, &version, sizeof version); 471 break; 472 473 case MRT_ASSERT: 474 soopt_from_kbuf(sopt, &pim_assert, sizeof pim_assert); 475 break; 476 477 case MRT_API_SUPPORT: 478 soopt_from_kbuf(sopt, &mrt_api_support, sizeof mrt_api_support); 479 break; 480 481 case MRT_API_CONFIG: 482 soopt_from_kbuf(sopt, &mrt_api_config, sizeof mrt_api_config); 483 break; 484 485 default: 486 error = EOPNOTSUPP; 487 break; 488 } 489 return error; 490 } 491 492 /* 493 * Handle ioctl commands to obtain information from the cache 494 */ 495 static int 496 X_mrt_ioctl(u_long cmd, caddr_t data) 497 { 498 int error = 0; 499 500 switch (cmd) { 501 case SIOCGETVIFCNT: 502 error = get_vif_cnt((struct sioc_vif_req *)data); 503 break; 504 505 case SIOCGETSGCNT: 506 error = get_sg_cnt((struct sioc_sg_req *)data); 507 break; 508 509 default: 510 error = EINVAL; 511 break; 512 } 513 return error; 514 } 515 516 /* 517 * returns the packet, byte, rpf-failure count for the source group provided 518 */ 519 static int 520 get_sg_cnt(struct sioc_sg_req *req) 521 { 522 struct mfc *rt; 523 524 lwkt_gettoken(&mroute_token); 525 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 526 if (rt == NULL) { 527 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 528 lwkt_reltoken(&mroute_token); 529 return EADDRNOTAVAIL; 530 } 531 req->pktcnt = rt->mfc_pkt_cnt; 532 req->bytecnt = rt->mfc_byte_cnt; 533 req->wrong_if = rt->mfc_wrong_if; 534 lwkt_reltoken(&mroute_token); 535 return 0; 536 } 537 538 /* 539 * returns the input and output packet and byte counts on the vif provided 540 */ 541 static int 542 get_vif_cnt(struct sioc_vif_req *req) 543 { 544 vifi_t vifi = req->vifi; 545 546 if (vifi >= numvifs) 547 return EINVAL; 548 549 req->icount = viftable[vifi].v_pkt_in; 550 req->ocount = viftable[vifi].v_pkt_out; 551 req->ibytes = viftable[vifi].v_bytes_in; 552 req->obytes = viftable[vifi].v_bytes_out; 553 554 return 0; 555 } 556 557 /* 558 * Enable multicast routing 559 */ 560 static int 561 ip_mrouter_init(struct socket *so, int version) 562 { 563 if (mrtdebug) 564 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 565 so->so_type, so->so_proto->pr_protocol); 566 567 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 568 return EOPNOTSUPP; 569 570 if (version != 1) 571 return ENOPROTOOPT; 572 573 if (ip_mrouter != NULL) 574 return EADDRINUSE; 575 576 ip_mrouter = so; 577 578 bzero((caddr_t)mfctable, sizeof(mfctable)); 579 bzero((caddr_t)nexpire, sizeof(nexpire)); 580 581 pim_assert = 0; 582 bw_upcalls_n = 0; 583 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 584 585 callout_init(&expire_upcalls_ch); 586 callout_init(&bw_upcalls_ch); 587 callout_init(&bw_meter_ch); 588 callout_init(&tbf_reprocess_q_ch); 589 590 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 591 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 592 expire_bw_upcalls_send, NULL); 593 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 594 595 mrt_api_config = 0; 596 597 if (mrtdebug) 598 log(LOG_DEBUG, "ip_mrouter_init\n"); 599 600 return 0; 601 } 602 603 /* 604 * Disable multicast routing 605 */ 606 static int 607 X_ip_mrouter_done(void) 608 { 609 vifi_t vifi; 610 int i; 611 struct ifnet *ifp; 612 struct ifreq ifr; 613 struct mfc *rt; 614 struct rtdetq *rte; 615 616 lwkt_gettoken(&mroute_token); 617 618 /* 619 * For each phyint in use, disable promiscuous reception of all IP 620 * multicasts. 621 */ 622 for (vifi = 0; vifi < numvifs; vifi++) { 623 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 624 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 625 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 626 627 so->sin_len = sizeof(struct sockaddr_in); 628 so->sin_family = AF_INET; 629 so->sin_addr.s_addr = INADDR_ANY; 630 ifp = viftable[vifi].v_ifp; 631 if_allmulti(ifp, 0); 632 } 633 } 634 bzero((caddr_t)tbftable, sizeof(tbftable)); 635 bzero((caddr_t)viftable, sizeof(viftable)); 636 numvifs = 0; 637 pim_assert = 0; 638 639 callout_stop(&expire_upcalls_ch); 640 641 mrt_api_config = 0; 642 bw_upcalls_n = 0; 643 callout_stop(&bw_upcalls_ch); 644 callout_stop(&bw_meter_ch); 645 callout_stop(&tbf_reprocess_q_ch); 646 647 /* 648 * Free all multicast forwarding cache entries. 649 */ 650 for (i = 0; i < MFCTBLSIZ; i++) { 651 for (rt = mfctable[i]; rt != NULL; ) { 652 struct mfc *nr = rt->mfc_next; 653 654 for (rte = rt->mfc_stall; rte != NULL; ) { 655 struct rtdetq *n = rte->next; 656 657 m_freem(rte->m); 658 kfree(rte, M_MRTABLE); 659 rte = n; 660 } 661 free_bw_list(rt->mfc_bw_meter); 662 kfree(rt, M_MRTABLE); 663 rt = nr; 664 } 665 } 666 667 bzero((caddr_t)mfctable, sizeof(mfctable)); 668 669 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 670 671 /* 672 * Reset de-encapsulation cache 673 */ 674 last_encap_src = INADDR_ANY; 675 last_encap_vif = NULL; 676 #ifdef PIM 677 reg_vif_num = VIFI_INVALID; 678 #endif 679 have_encap_tunnel = 0; 680 681 ip_mrouter = NULL; 682 683 lwkt_reltoken(&mroute_token); 684 685 if (mrtdebug) 686 log(LOG_DEBUG, "ip_mrouter_done\n"); 687 688 return 0; 689 } 690 691 /* 692 * Set PIM assert processing global 693 */ 694 static int 695 set_assert(int i) 696 { 697 if ((i != 1) && (i != 0)) 698 return EINVAL; 699 700 pim_assert = i; 701 702 return 0; 703 } 704 705 /* 706 * Configure API capabilities 707 */ 708 static int 709 set_api_config(uint32_t *apival) 710 { 711 int i; 712 713 /* 714 * We can set the API capabilities only if it is the first operation 715 * after MRT_INIT. I.e.: 716 * - there are no vifs installed 717 * - pim_assert is not enabled 718 * - the MFC table is empty 719 */ 720 if (numvifs > 0) { 721 *apival = 0; 722 return EPERM; 723 } 724 if (pim_assert) { 725 *apival = 0; 726 return EPERM; 727 } 728 for (i = 0; i < MFCTBLSIZ; i++) { 729 if (mfctable[i] != NULL) { 730 *apival = 0; 731 return EPERM; 732 } 733 } 734 735 mrt_api_config = *apival & mrt_api_support; 736 *apival = mrt_api_config; 737 738 return 0; 739 } 740 741 /* 742 * Add a vif to the vif table 743 */ 744 static int 745 add_vif(struct vifctl *vifcp) 746 { 747 struct vif *vifp = viftable + vifcp->vifc_vifi; 748 struct sockaddr_in sin = {sizeof sin, AF_INET}; 749 struct ifaddr *ifa; 750 struct ifnet *ifp; 751 int error, i; 752 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 753 754 if (vifcp->vifc_vifi >= MAXVIFS) 755 return EINVAL; 756 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) 757 return EADDRINUSE; 758 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) 759 return EADDRNOTAVAIL; 760 761 /* Find the interface with an address in AF_INET family */ 762 #ifdef PIM 763 if (vifcp->vifc_flags & VIFF_REGISTER) { 764 /* 765 * XXX: Because VIFF_REGISTER does not really need a valid 766 * local interface (e.g. it could be 127.0.0.2), we don't 767 * check its address. 768 */ 769 ifp = NULL; 770 } else 771 #endif 772 { 773 sin.sin_addr = vifcp->vifc_lcl_addr; 774 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 775 if (ifa == NULL) 776 return EADDRNOTAVAIL; 777 ifp = ifa->ifa_ifp; 778 } 779 780 if (vifcp->vifc_flags & VIFF_TUNNEL) { 781 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 782 /* 783 * An encapsulating tunnel is wanted. Tell ipip_input() to 784 * start paying attention to encapsulated packets. 785 */ 786 if (have_encap_tunnel == 0) { 787 have_encap_tunnel = 1; 788 for (i = 0; i < MAXVIFS; i++) { 789 if_initname(&multicast_decap_if[i], "mdecap", i); 790 } 791 } 792 /* 793 * Set interface to fake encapsulator interface 794 */ 795 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 796 /* 797 * Prepare cached route entry 798 */ 799 bzero(&vifp->v_route, sizeof(vifp->v_route)); 800 } else { 801 log(LOG_ERR, "source routed tunnels not supported\n"); 802 return EOPNOTSUPP; 803 } 804 #ifdef PIM 805 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 806 ifp = &multicast_register_if; 807 if (mrtdebug) 808 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 809 (void *)&multicast_register_if); 810 if (reg_vif_num == VIFI_INVALID) { 811 if_initname(&multicast_register_if, "register_vif", 0); 812 multicast_register_if.if_flags = IFF_LOOPBACK; 813 bzero(&vifp->v_route, sizeof(vifp->v_route)); 814 reg_vif_num = vifcp->vifc_vifi; 815 } 816 #endif 817 } else { /* Make sure the interface supports multicast */ 818 if ((ifp->if_flags & IFF_MULTICAST) == 0) 819 return EOPNOTSUPP; 820 821 /* Enable promiscuous reception of all IP multicasts from the if */ 822 lwkt_gettoken(&mroute_token); 823 error = if_allmulti(ifp, 1); 824 lwkt_reltoken(&mroute_token); 825 if (error) 826 return error; 827 } 828 829 lwkt_gettoken(&mroute_token); 830 /* define parameters for the tbf structure */ 831 vifp->v_tbf = v_tbf; 832 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 833 vifp->v_tbf->tbf_n_tok = 0; 834 vifp->v_tbf->tbf_q_len = 0; 835 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 836 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 837 838 vifp->v_flags = vifcp->vifc_flags; 839 vifp->v_threshold = vifcp->vifc_threshold; 840 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 841 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 842 vifp->v_ifp = ifp; 843 /* scaling up here allows division by 1024 in critical code */ 844 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 845 vifp->v_rsvp_on = 0; 846 vifp->v_rsvpd = NULL; 847 /* initialize per vif pkt counters */ 848 vifp->v_pkt_in = 0; 849 vifp->v_pkt_out = 0; 850 vifp->v_bytes_in = 0; 851 vifp->v_bytes_out = 0; 852 853 /* Adjust numvifs up if the vifi is higher than numvifs */ 854 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 855 856 lwkt_reltoken(&mroute_token); 857 858 if (mrtdebug) 859 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 860 vifcp->vifc_vifi, 861 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 862 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 863 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 864 vifcp->vifc_threshold, 865 vifcp->vifc_rate_limit); 866 867 return 0; 868 } 869 870 /* 871 * Delete a vif from the vif table 872 */ 873 static int 874 del_vif(vifi_t vifi) 875 { 876 struct vif *vifp; 877 878 if (vifi >= numvifs) 879 return EINVAL; 880 vifp = &viftable[vifi]; 881 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) 882 return EADDRNOTAVAIL; 883 884 lwkt_gettoken(&mroute_token); 885 886 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 887 if_allmulti(vifp->v_ifp, 0); 888 889 if (vifp == last_encap_vif) { 890 last_encap_vif = NULL; 891 last_encap_src = INADDR_ANY; 892 } 893 894 /* 895 * Free packets queued at the interface 896 */ 897 while (vifp->v_tbf->tbf_q) { 898 struct mbuf *m = vifp->v_tbf->tbf_q; 899 900 vifp->v_tbf->tbf_q = m->m_nextpkt; 901 m_freem(m); 902 } 903 904 #ifdef PIM 905 if (vifp->v_flags & VIFF_REGISTER) 906 reg_vif_num = VIFI_INVALID; 907 #endif 908 909 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 910 bzero((caddr_t)vifp, sizeof (*vifp)); 911 912 if (mrtdebug) 913 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 914 915 /* Adjust numvifs down */ 916 for (vifi = numvifs; vifi > 0; vifi--) 917 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 918 break; 919 numvifs = vifi; 920 921 lwkt_reltoken(&mroute_token); 922 923 return 0; 924 } 925 926 /* 927 * update an mfc entry without resetting counters and S,G addresses. 928 */ 929 static void 930 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 931 { 932 int i; 933 934 rt->mfc_parent = mfccp->mfcc_parent; 935 for (i = 0; i < numvifs; i++) { 936 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 937 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 938 MRT_MFC_FLAGS_ALL; 939 } 940 /* set the RP address */ 941 if (mrt_api_config & MRT_MFC_RP) 942 rt->mfc_rp = mfccp->mfcc_rp; 943 else 944 rt->mfc_rp.s_addr = INADDR_ANY; 945 } 946 947 /* 948 * fully initialize an mfc entry from the parameter. 949 */ 950 static void 951 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 952 { 953 rt->mfc_origin = mfccp->mfcc_origin; 954 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 955 956 update_mfc_params(rt, mfccp); 957 958 /* initialize pkt counters per src-grp */ 959 rt->mfc_pkt_cnt = 0; 960 rt->mfc_byte_cnt = 0; 961 rt->mfc_wrong_if = 0; 962 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 963 } 964 965 966 /* 967 * Add an mfc entry 968 */ 969 static int 970 add_mfc(struct mfcctl2 *mfccp) 971 { 972 struct mfc *rt; 973 u_long hash; 974 struct rtdetq *rte; 975 u_short nstl; 976 977 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 978 979 /* If an entry already exists, just update the fields */ 980 if (rt) { 981 if (mrtdebug & DEBUG_MFC) 982 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 983 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 984 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 985 mfccp->mfcc_parent); 986 987 lwkt_gettoken(&mroute_token); 988 update_mfc_params(rt, mfccp); 989 lwkt_reltoken(&mroute_token); 990 return 0; 991 } 992 993 /* 994 * Find the entry for which the upcall was made and update 995 */ 996 lwkt_gettoken(&mroute_token); 997 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 998 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 999 1000 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1001 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1002 (rt->mfc_stall != NULL)) { 1003 1004 if (nstl++) 1005 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1006 "multiple kernel entries", 1007 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1008 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1009 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1010 1011 if (mrtdebug & DEBUG_MFC) 1012 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1013 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1014 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1015 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1016 1017 init_mfc_params(rt, mfccp); 1018 1019 rt->mfc_expire = 0; /* Don't clean this guy up */ 1020 nexpire[hash]--; 1021 1022 /* free packets Qed at the end of this entry */ 1023 for (rte = rt->mfc_stall; rte != NULL; ) { 1024 struct rtdetq *n = rte->next; 1025 1026 ip_mdq(rte->m, rte->ifp, rt, -1); 1027 m_freem(rte->m); 1028 kfree(rte, M_MRTABLE); 1029 rte = n; 1030 } 1031 rt->mfc_stall = NULL; 1032 } 1033 } 1034 1035 /* 1036 * It is possible that an entry is being inserted without an upcall 1037 */ 1038 if (nstl == 0) { 1039 if (mrtdebug & DEBUG_MFC) 1040 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1041 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1042 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1043 mfccp->mfcc_parent); 1044 1045 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1046 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1047 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1048 init_mfc_params(rt, mfccp); 1049 if (rt->mfc_expire) 1050 nexpire[hash]--; 1051 rt->mfc_expire = 0; 1052 break; /* XXX */ 1053 } 1054 } 1055 if (rt == NULL) { /* no upcall, so make a new entry */ 1056 rt = kmalloc(sizeof(*rt), M_MRTABLE, M_INTWAIT | M_NULLOK); 1057 if (rt == NULL) { 1058 lwkt_reltoken(&mroute_token); 1059 return ENOBUFS; 1060 } 1061 1062 init_mfc_params(rt, mfccp); 1063 rt->mfc_expire = 0; 1064 rt->mfc_stall = NULL; 1065 1066 rt->mfc_bw_meter = NULL; 1067 /* insert new entry at head of hash chain */ 1068 rt->mfc_next = mfctable[hash]; 1069 mfctable[hash] = rt; 1070 } 1071 } 1072 lwkt_reltoken(&mroute_token); 1073 return 0; 1074 } 1075 1076 /* 1077 * Delete an mfc entry 1078 */ 1079 static int 1080 del_mfc(struct mfcctl2 *mfccp) 1081 { 1082 struct in_addr origin; 1083 struct in_addr mcastgrp; 1084 struct mfc *rt; 1085 struct mfc **nptr; 1086 u_long hash; 1087 struct bw_meter *list; 1088 1089 origin = mfccp->mfcc_origin; 1090 mcastgrp = mfccp->mfcc_mcastgrp; 1091 1092 if (mrtdebug & DEBUG_MFC) 1093 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1094 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1095 1096 lwkt_gettoken(&mroute_token); 1097 1098 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1099 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1100 if (origin.s_addr == rt->mfc_origin.s_addr && 1101 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1102 rt->mfc_stall == NULL) 1103 break; 1104 if (rt == NULL) { 1105 lwkt_reltoken(&mroute_token); 1106 return EADDRNOTAVAIL; 1107 } 1108 1109 *nptr = rt->mfc_next; 1110 1111 /* 1112 * free the bw_meter entries 1113 */ 1114 list = rt->mfc_bw_meter; 1115 rt->mfc_bw_meter = NULL; 1116 lwkt_reltoken(&mroute_token); 1117 1118 kfree(rt, M_MRTABLE); 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 (ssb_appendaddr(&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 vifi_t vifi; 1159 1160 if (mrtdebug & DEBUG_FORWARD) 1161 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1162 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1163 (void *)ifp); 1164 1165 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1166 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1167 /* 1168 * Packet arrived via a physical interface or 1169 * an encapsulated tunnel or a register_vif. 1170 */ 1171 } else { 1172 /* 1173 * Packet arrived through a source-route tunnel. 1174 * Source-route tunnels are no longer supported. 1175 */ 1176 static time_t last_log; 1177 if (last_log != time_uptime) { 1178 last_log = time_uptime; 1179 log(LOG_ERR, 1180 "ip_mforward: received source-routed packet from %lx\n", 1181 (u_long)ntohl(ip->ip_src.s_addr)); 1182 } 1183 return 1; 1184 } 1185 1186 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1187 if (ip->ip_ttl < 255) 1188 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1189 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1190 struct vif *vifp = viftable + vifi; 1191 1192 kprintf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1193 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1194 vifi, 1195 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1196 vifp->v_ifp->if_xname); 1197 } 1198 return ip_mdq(m, ifp, NULL, vifi); 1199 } 1200 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1201 kprintf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1202 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1203 if (!imo) 1204 kprintf("In fact, no options were specified at all\n"); 1205 } 1206 1207 /* 1208 * Don't forward a packet with time-to-live of zero or one, 1209 * or a packet destined to a local-only group. 1210 */ 1211 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1212 return 0; 1213 1214 /* 1215 * Determine forwarding vifs from the forwarding cache table 1216 */ 1217 lwkt_gettoken(&mroute_token); 1218 ++mrtstat.mrts_mfc_lookups; 1219 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1220 1221 /* Entry exists, so forward if necessary */ 1222 if (rt != NULL) { 1223 int ipres = ip_mdq(m, ifp, rt, -1); 1224 lwkt_reltoken(&mroute_token); 1225 return ipres; 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 = kmalloc((sizeof *rte), M_MRTABLE, M_INTWAIT | M_NULLOK); 1251 if (rte == NULL) { 1252 lwkt_reltoken(&mroute_token); 1253 return ENOBUFS; 1254 } 1255 1256 mb0 = m_copypacket(m, M_NOWAIT); 1257 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1258 mb0 = m_pullup(mb0, hlen); 1259 if (mb0 == NULL) { 1260 kfree(rte, M_MRTABLE); 1261 lwkt_reltoken(&mroute_token); 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 = kmalloc(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 kfree(rt, M_MRTABLE); 1317 fail: 1318 kfree(rte, M_MRTABLE); 1319 m_freem(mb0); 1320 lwkt_reltoken(&mroute_token); 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 kfree(rte, M_MRTABLE); 1361 m_freem(mb0); 1362 lwkt_reltoken(&mroute_token); 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 lwkt_reltoken(&mroute_token); 1375 return 0; 1376 } 1377 } 1378 1379 /* 1380 * Clean up the cache entry if upcall is not serviced 1381 */ 1382 static void 1383 expire_upcalls(void *unused) 1384 { 1385 struct rtdetq *rte; 1386 struct mfc *mfc, **nptr; 1387 int i; 1388 1389 lwkt_gettoken(&mroute_token); 1390 for (i = 0; i < MFCTBLSIZ; i++) { 1391 if (nexpire[i] == 0) 1392 continue; 1393 nptr = &mfctable[i]; 1394 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1395 /* 1396 * Skip real cache entries 1397 * Make sure it wasn't marked to not expire (shouldn't happen) 1398 * If it expires now 1399 */ 1400 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1401 --mfc->mfc_expire == 0) { 1402 if (mrtdebug & DEBUG_EXPIRE) 1403 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1404 (u_long)ntohl(mfc->mfc_origin.s_addr), 1405 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1406 /* 1407 * drop all the packets 1408 * free the mbuf with the pkt, if, timing info 1409 */ 1410 for (rte = mfc->mfc_stall; rte; ) { 1411 struct rtdetq *n = rte->next; 1412 1413 m_freem(rte->m); 1414 kfree(rte, M_MRTABLE); 1415 rte = n; 1416 } 1417 ++mrtstat.mrts_cache_cleanups; 1418 nexpire[i]--; 1419 1420 /* 1421 * free the bw_meter entries 1422 */ 1423 while (mfc->mfc_bw_meter != NULL) { 1424 struct bw_meter *x = mfc->mfc_bw_meter; 1425 1426 mfc->mfc_bw_meter = x->bm_mfc_next; 1427 kfree(x, M_BWMETER); 1428 } 1429 1430 *nptr = mfc->mfc_next; 1431 kfree(mfc, M_MRTABLE); 1432 } else { 1433 nptr = &mfc->mfc_next; 1434 } 1435 } 1436 } 1437 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1438 lwkt_reltoken(&mroute_token); 1439 } 1440 1441 /* 1442 * Packet forwarding routine once entry in the cache is made 1443 */ 1444 static int 1445 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1446 { 1447 struct ip *ip = mtod(m, struct ip *); 1448 vifi_t vifi; 1449 int plen = ip->ip_len; 1450 1451 /* 1452 * Macro to send packet on vif. Since RSVP packets don't get counted on 1453 * input, they shouldn't get counted on output, so statistics keeping is 1454 * separate. 1455 */ 1456 #define MC_SEND(ip,vifp,m) { \ 1457 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1458 encap_send((ip), (vifp), (m)); \ 1459 else \ 1460 phyint_send((ip), (vifp), (m)); \ 1461 } 1462 1463 /* 1464 * If xmt_vif is not -1, send on only the requested vif. 1465 * 1466 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1467 */ 1468 if (xmt_vif < numvifs) { 1469 #ifdef PIM 1470 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1471 pim_register_send(ip, viftable + xmt_vif, m, rt); 1472 else 1473 #endif 1474 MC_SEND(ip, viftable + xmt_vif, m); 1475 return 1; 1476 } 1477 1478 /* 1479 * Don't forward if it didn't arrive from the parent vif for its origin. 1480 */ 1481 vifi = rt->mfc_parent; 1482 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1483 /* came in the wrong interface */ 1484 if (mrtdebug & DEBUG_FORWARD) 1485 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1486 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1487 ++mrtstat.mrts_wrong_if; 1488 ++rt->mfc_wrong_if; 1489 /* 1490 * If we are doing PIM assert processing, send a message 1491 * to the routing daemon. 1492 * 1493 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1494 * can complete the SPT switch, regardless of the type 1495 * of the iif (broadcast media, GRE tunnel, etc). 1496 */ 1497 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1498 struct timeval now; 1499 u_long delta; 1500 1501 #ifdef PIM 1502 if (ifp == &multicast_register_if) 1503 pimstat.pims_rcv_registers_wrongiif++; 1504 #endif 1505 1506 /* Get vifi for the incoming packet */ 1507 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1508 ; 1509 if (vifi >= numvifs) 1510 return 0; /* The iif is not found: ignore the packet. */ 1511 1512 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1513 return 0; /* WRONGVIF disabled: ignore the packet */ 1514 1515 GET_TIME(now); 1516 1517 TV_DELTA(rt->mfc_last_assert, now, delta); 1518 1519 if (delta > ASSERT_MSG_TIME) { 1520 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1521 struct igmpmsg *im; 1522 int hlen = ip->ip_hl << 2; 1523 struct mbuf *mm = m_copy(m, 0, hlen); 1524 1525 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1526 mm = m_pullup(mm, hlen); 1527 if (mm == NULL) 1528 return ENOBUFS; 1529 1530 rt->mfc_last_assert = now; 1531 1532 im = mtod(mm, struct igmpmsg *); 1533 im->im_msgtype = IGMPMSG_WRONGVIF; 1534 im->im_mbz = 0; 1535 im->im_vif = vifi; 1536 1537 mrtstat.mrts_upcalls++; 1538 1539 k_igmpsrc.sin_addr = im->im_src; 1540 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1541 log(LOG_WARNING, 1542 "ip_mforward: ip_mrouter socket queue full\n"); 1543 ++mrtstat.mrts_upq_sockfull; 1544 return ENOBUFS; 1545 } 1546 } 1547 } 1548 return 0; 1549 } 1550 1551 /* If I sourced this packet, it counts as output, else it was input. */ 1552 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1553 viftable[vifi].v_pkt_out++; 1554 viftable[vifi].v_bytes_out += plen; 1555 } else { 1556 viftable[vifi].v_pkt_in++; 1557 viftable[vifi].v_bytes_in += plen; 1558 } 1559 rt->mfc_pkt_cnt++; 1560 rt->mfc_byte_cnt += plen; 1561 1562 /* 1563 * For each vif, decide if a copy of the packet should be forwarded. 1564 * Forward if: 1565 * - the ttl exceeds the vif's threshold 1566 * - there are group members downstream on interface 1567 */ 1568 for (vifi = 0; vifi < numvifs; vifi++) 1569 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1570 viftable[vifi].v_pkt_out++; 1571 viftable[vifi].v_bytes_out += plen; 1572 #ifdef PIM 1573 if (viftable[vifi].v_flags & VIFF_REGISTER) 1574 pim_register_send(ip, viftable + vifi, m, rt); 1575 else 1576 #endif 1577 MC_SEND(ip, viftable+vifi, m); 1578 } 1579 1580 /* 1581 * Perform upcall-related bw measuring. 1582 */ 1583 if (rt->mfc_bw_meter != NULL) { 1584 struct bw_meter *x; 1585 struct timeval now; 1586 1587 GET_TIME(now); 1588 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1589 bw_meter_receive_packet(x, plen, &now); 1590 } 1591 1592 return 0; 1593 } 1594 1595 /* 1596 * check if a vif number is legal/ok. This is used by ip_output. 1597 */ 1598 static int 1599 X_legal_vif_num(int vif) 1600 { 1601 return (vif >= 0 && vif < numvifs); 1602 } 1603 1604 /* 1605 * Return the local address used by this vif 1606 */ 1607 static u_long 1608 X_ip_mcast_src(int vifi) 1609 { 1610 if (vifi >= 0 && vifi < numvifs) 1611 return viftable[vifi].v_lcl_addr.s_addr; 1612 else 1613 return INADDR_ANY; 1614 } 1615 1616 static void 1617 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1618 { 1619 struct mbuf *mb_copy; 1620 int hlen = ip->ip_hl << 2; 1621 1622 /* 1623 * Make a new reference to the packet; make sure that 1624 * the IP header is actually copied, not just referenced, 1625 * so that ip_output() only scribbles on the copy. 1626 */ 1627 mb_copy = m_copypacket(m, M_NOWAIT); 1628 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1629 mb_copy = m_pullup(mb_copy, hlen); 1630 if (mb_copy == NULL) 1631 return; 1632 1633 if (vifp->v_rate_limit == 0) 1634 tbf_send_packet(vifp, mb_copy); 1635 else 1636 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1637 } 1638 1639 static void 1640 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1641 { 1642 struct mbuf *mb_copy; 1643 struct ip *ip_copy; 1644 int i, len = ip->ip_len; 1645 1646 /* Take care of delayed checksums */ 1647 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1648 in_delayed_cksum(m); 1649 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1650 } 1651 1652 /* 1653 * copy the old packet & pullup its IP header into the 1654 * new mbuf so we can modify it. Try to fill the new 1655 * mbuf since if we don't the ethernet driver will. 1656 */ 1657 MGETHDR(mb_copy, M_NOWAIT, MT_HEADER); 1658 if (mb_copy == NULL) 1659 return; 1660 mb_copy->m_data += max_linkhdr; 1661 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1662 1663 if ((mb_copy->m_next = m_copypacket(m, M_NOWAIT)) == NULL) { 1664 m_freem(mb_copy); 1665 return; 1666 } 1667 i = MHLEN - M_LEADINGSPACE(mb_copy); 1668 if (i > len) 1669 i = len; 1670 mb_copy = m_pullup(mb_copy, i); 1671 if (mb_copy == NULL) 1672 return; 1673 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1674 1675 /* 1676 * fill in the encapsulating IP header. 1677 */ 1678 ip_copy = mtod(mb_copy, struct ip *); 1679 *ip_copy = multicast_encap_iphdr; 1680 ip_copy->ip_id = ip_newid(); 1681 ip_copy->ip_len += len; 1682 ip_copy->ip_src = vifp->v_lcl_addr; 1683 ip_copy->ip_dst = vifp->v_rmt_addr; 1684 1685 /* 1686 * turn the encapsulated IP header back into a valid one. 1687 */ 1688 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1689 --ip->ip_ttl; 1690 ip->ip_len = htons(ip->ip_len); 1691 ip->ip_off = htons(ip->ip_off); 1692 ip->ip_sum = 0; 1693 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1694 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1695 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1696 1697 if (vifp->v_rate_limit == 0) 1698 tbf_send_packet(vifp, mb_copy); 1699 else 1700 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1701 } 1702 1703 /* 1704 * De-encapsulate a packet and feed it back through ip input (this 1705 * routine is called whenever IP gets a packet with proto type 1706 * ENCAP_PROTO and a local destination address). 1707 * 1708 * This is similar to mroute_encapcheck() + mroute_encap_input() in -current. 1709 */ 1710 static int 1711 X_ipip_input(struct mbuf **mp, int *offp, int proto) 1712 { 1713 struct mbuf *m = *mp; 1714 struct ip *ip = mtod(m, struct ip *); 1715 int hlen = ip->ip_hl << 2; 1716 1717 if (!have_encap_tunnel) { 1718 rip_input(mp, offp, proto); 1719 return(IPPROTO_DONE); 1720 } 1721 *mp = NULL; 1722 1723 /* 1724 * dump the packet if it's not to a multicast destination or if 1725 * we don't have an encapsulating tunnel with the source. 1726 * Note: This code assumes that the remote site IP address 1727 * uniquely identifies the tunnel (i.e., that this site has 1728 * at most one tunnel with the remote site). 1729 */ 1730 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) { 1731 ++mrtstat.mrts_bad_tunnel; 1732 m_freem(m); 1733 return(IPPROTO_DONE); 1734 } 1735 if (ip->ip_src.s_addr != last_encap_src) { 1736 struct vif *vifp = viftable; 1737 struct vif *vife = vifp + numvifs; 1738 1739 last_encap_src = ip->ip_src.s_addr; 1740 last_encap_vif = NULL; 1741 for ( ; vifp < vife; ++vifp) 1742 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1743 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1744 == VIFF_TUNNEL) 1745 last_encap_vif = vifp; 1746 break; 1747 } 1748 } 1749 if (last_encap_vif == NULL) { 1750 last_encap_src = INADDR_ANY; 1751 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1752 m_freem(m); 1753 if (mrtdebug) 1754 log(LOG_DEBUG, "ip_mforward: no tunnel with %lx\n", 1755 (u_long)ntohl(ip->ip_src.s_addr)); 1756 return(IPPROTO_DONE); 1757 } 1758 1759 if (hlen > sizeof(struct ip)) 1760 ip_stripoptions(m); 1761 m->m_data += sizeof(struct ip); 1762 m->m_len -= sizeof(struct ip); 1763 m->m_pkthdr.len -= sizeof(struct ip); 1764 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 1765 1766 netisr_queue(NETISR_IP, m); 1767 return(IPPROTO_DONE); 1768 } 1769 1770 /* 1771 * Token bucket filter module 1772 */ 1773 1774 static void 1775 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len) 1776 { 1777 struct tbf *t = vifp->v_tbf; 1778 1779 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */ 1780 mrtstat.mrts_pkt2large++; 1781 m_freem(m); 1782 return; 1783 } 1784 1785 tbf_update_tokens(vifp); 1786 1787 if (t->tbf_q_len == 0) { /* queue empty... */ 1788 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */ 1789 t->tbf_n_tok -= p_len; 1790 tbf_send_packet(vifp, m); 1791 } else { /* no, queue packet and try later */ 1792 tbf_queue(vifp, m); 1793 callout_reset(&tbf_reprocess_q_ch, TBF_REPROCESS, 1794 tbf_reprocess_q, vifp); 1795 } 1796 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1797 /* finite queue length, so queue pkts and process queue */ 1798 tbf_queue(vifp, m); 1799 tbf_process_q(vifp); 1800 } else { 1801 /* queue full, try to dq and queue and process */ 1802 if (!tbf_dq_sel(vifp, ip)) { 1803 mrtstat.mrts_q_overflow++; 1804 m_freem(m); 1805 } else { 1806 tbf_queue(vifp, m); 1807 tbf_process_q(vifp); 1808 } 1809 } 1810 } 1811 1812 /* 1813 * adds a packet to the queue at the interface 1814 */ 1815 static void 1816 tbf_queue(struct vif *vifp, struct mbuf *m) 1817 { 1818 struct tbf *t = vifp->v_tbf; 1819 1820 lwkt_gettoken(&mroute_token); 1821 1822 if (t->tbf_t == NULL) /* Queue was empty */ 1823 t->tbf_q = m; 1824 else /* Insert at tail */ 1825 t->tbf_t->m_nextpkt = m; 1826 1827 t->tbf_t = m; /* Set new tail pointer */ 1828 1829 #ifdef DIAGNOSTIC 1830 /* Make sure we didn't get fed a bogus mbuf */ 1831 if (m->m_nextpkt) 1832 panic("tbf_queue: m_nextpkt"); 1833 #endif 1834 m->m_nextpkt = NULL; 1835 1836 t->tbf_q_len++; 1837 1838 lwkt_reltoken(&mroute_token); 1839 } 1840 1841 /* 1842 * processes the queue at the interface 1843 */ 1844 static void 1845 tbf_process_q(struct vif *vifp) 1846 { 1847 struct tbf *t = vifp->v_tbf; 1848 1849 lwkt_gettoken(&mroute_token); 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 lwkt_reltoken(&mroute_token); 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 u_int p; 1895 struct mbuf *m, *last; 1896 struct mbuf **np; 1897 struct tbf *t = vifp->v_tbf; 1898 1899 lwkt_gettoken(&mroute_token); 1900 1901 p = priority(vifp, ip); 1902 1903 np = &t->tbf_q; 1904 last = NULL; 1905 while ((m = *np) != NULL) { 1906 if (p > priority(vifp, mtod(m, struct ip *))) { 1907 *np = m->m_nextpkt; 1908 /* If we're removing the last packet, fix the tail pointer */ 1909 if (m == t->tbf_t) 1910 t->tbf_t = last; 1911 m_freem(m); 1912 /* It's impossible for the queue to be empty, but check anyways. */ 1913 if (--t->tbf_q_len == 0) 1914 t->tbf_t = NULL; 1915 mrtstat.mrts_drop_sel++; 1916 lwkt_reltoken(&mroute_token); 1917 return 1; 1918 } 1919 np = &m->m_nextpkt; 1920 last = m; 1921 } 1922 lwkt_reltoken(&mroute_token); 1923 return 0; 1924 } 1925 1926 static void 1927 tbf_send_packet(struct vif *vifp, struct mbuf *m) 1928 { 1929 lwkt_gettoken(&mroute_token); 1930 1931 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */ 1932 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 1933 else { 1934 struct ip_moptions imo; 1935 int error; 1936 static struct route ro; /* XXX check this */ 1937 1938 imo.imo_multicast_ifp = vifp->v_ifp; 1939 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1940 imo.imo_multicast_loop = 1; 1941 imo.imo_multicast_vif = -1; 1942 1943 /* 1944 * Re-entrancy should not be a problem here, because 1945 * the packets that we send out and are looped back at us 1946 * should get rejected because they appear to come from 1947 * the loopback interface, thus preventing looping. 1948 */ 1949 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL); 1950 1951 if (mrtdebug & DEBUG_XMIT) 1952 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1953 (int)(vifp - viftable), error); 1954 } 1955 lwkt_reltoken(&mroute_token); 1956 } 1957 1958 /* determine the current time and then 1959 * the elapsed time (between the last time and time now) 1960 * in milliseconds & update the no. of tokens in the bucket 1961 */ 1962 static void 1963 tbf_update_tokens(struct vif *vifp) 1964 { 1965 struct timeval tp; 1966 u_long tm; 1967 struct tbf *t = vifp->v_tbf; 1968 1969 lwkt_gettoken(&mroute_token); 1970 1971 GET_TIME(tp); 1972 1973 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1974 1975 /* 1976 * This formula is actually 1977 * "time in seconds" * "bytes/second". 1978 * 1979 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1980 * 1981 * The (1000/1024) was introduced in add_vif to optimize 1982 * this divide into a shift. 1983 */ 1984 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1985 t->tbf_last_pkt_t = tp; 1986 1987 if (t->tbf_n_tok > MAX_BKT_SIZE) 1988 t->tbf_n_tok = MAX_BKT_SIZE; 1989 1990 lwkt_reltoken(&mroute_token); 1991 } 1992 1993 static int 1994 priority(struct vif *vifp, struct ip *ip) 1995 { 1996 int prio = 50; /* the lowest priority -- default case */ 1997 1998 /* temporary hack; may add general packet classifier some day */ 1999 2000 /* 2001 * The UDP port space is divided up into four priority ranges: 2002 * [0, 16384) : unclassified - lowest priority 2003 * [16384, 32768) : audio - highest priority 2004 * [32768, 49152) : whiteboard - medium priority 2005 * [49152, 65536) : video - low priority 2006 * 2007 * Everything else gets lowest priority. 2008 */ 2009 if (ip->ip_p == IPPROTO_UDP) { 2010 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2011 switch (ntohs(udp->uh_dport) & 0xc000) { 2012 case 0x4000: 2013 prio = 70; 2014 break; 2015 case 0x8000: 2016 prio = 60; 2017 break; 2018 case 0xc000: 2019 prio = 55; 2020 break; 2021 } 2022 } 2023 return prio; 2024 } 2025 2026 /* 2027 * End of token bucket filter modifications 2028 */ 2029 2030 static int 2031 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 2032 { 2033 int error, vifi; 2034 2035 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2036 return EOPNOTSUPP; 2037 2038 error = soopt_to_kbuf(sopt, &vifi, sizeof vifi, sizeof vifi); 2039 if (error) 2040 return error; 2041 2042 lwkt_gettoken(&mroute_token); 2043 2044 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 2045 lwkt_reltoken(&mroute_token); 2046 return EADDRNOTAVAIL; 2047 } 2048 2049 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 2050 /* Check if socket is available. */ 2051 if (viftable[vifi].v_rsvpd != NULL) { 2052 lwkt_reltoken(&mroute_token); 2053 return EADDRINUSE; 2054 } 2055 2056 viftable[vifi].v_rsvpd = so; 2057 /* This may seem silly, but we need to be sure we don't over-increment 2058 * the RSVP counter, in case something slips up. 2059 */ 2060 if (!viftable[vifi].v_rsvp_on) { 2061 viftable[vifi].v_rsvp_on = 1; 2062 rsvp_on++; 2063 } 2064 } else { /* must be VIF_OFF */ 2065 /* 2066 * XXX as an additional consistency check, one could make sure 2067 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 2068 * first parameter is pretty useless. 2069 */ 2070 viftable[vifi].v_rsvpd = NULL; 2071 /* 2072 * This may seem silly, but we need to be sure we don't over-decrement 2073 * the RSVP counter, in case something slips up. 2074 */ 2075 if (viftable[vifi].v_rsvp_on) { 2076 viftable[vifi].v_rsvp_on = 0; 2077 rsvp_on--; 2078 } 2079 } 2080 lwkt_reltoken(&mroute_token); 2081 return 0; 2082 } 2083 2084 static void 2085 X_ip_rsvp_force_done(struct socket *so) 2086 { 2087 int vifi; 2088 2089 /* Don't bother if it is not the right type of socket. */ 2090 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2091 return; 2092 2093 lwkt_gettoken(&mroute_token); 2094 2095 /* The socket may be attached to more than one vif...this 2096 * is perfectly legal. 2097 */ 2098 for (vifi = 0; vifi < numvifs; vifi++) { 2099 if (viftable[vifi].v_rsvpd == so) { 2100 viftable[vifi].v_rsvpd = NULL; 2101 /* This may seem silly, but we need to be sure we don't 2102 * over-decrement the RSVP counter, in case something slips up. 2103 */ 2104 if (viftable[vifi].v_rsvp_on) { 2105 viftable[vifi].v_rsvp_on = 0; 2106 rsvp_on--; 2107 } 2108 } 2109 } 2110 2111 lwkt_reltoken(&mroute_token); 2112 } 2113 2114 static int 2115 X_rsvp_input(struct mbuf **mp, int *offp, int proto) 2116 { 2117 int vifi; 2118 struct mbuf *m = *mp; 2119 struct ip *ip = mtod(m, struct ip *); 2120 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2121 struct ifnet *ifp; 2122 #ifdef ALTQ 2123 /* support IP_RECVIF used by rsvpd rel4.2a1 */ 2124 struct inpcb *inp; 2125 struct socket *so; 2126 struct mbuf *opts; 2127 #endif 2128 2129 *mp = NULL; 2130 2131 if (rsvpdebug) 2132 kprintf("rsvp_input: rsvp_on %d\n",rsvp_on); 2133 2134 /* Can still get packets with rsvp_on = 0 if there is a local member 2135 * of the group to which the RSVP packet is addressed. But in this 2136 * case we want to throw the packet away. 2137 */ 2138 if (!rsvp_on) { 2139 m_freem(m); 2140 return(IPPROTO_DONE); 2141 } 2142 2143 lwkt_gettoken(&mroute_token); 2144 2145 if (rsvpdebug) 2146 kprintf("rsvp_input: check vifs\n"); 2147 2148 #ifdef DIAGNOSTIC 2149 if (!(m->m_flags & M_PKTHDR)) 2150 panic("rsvp_input no hdr"); 2151 #endif 2152 2153 ifp = m->m_pkthdr.rcvif; 2154 /* Find which vif the packet arrived on. */ 2155 for (vifi = 0; vifi < numvifs; vifi++) 2156 if (viftable[vifi].v_ifp == ifp) 2157 break; 2158 2159 #ifdef ALTQ 2160 if (vifi == numvifs || (so = viftable[vifi].v_rsvpd) == NULL) { 2161 #else 2162 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2163 #endif 2164 /* 2165 * If the old-style non-vif-associated socket is set, 2166 * then use it. Otherwise, drop packet since there 2167 * is no specific socket for this vif. 2168 */ 2169 if (ip_rsvpd != NULL) { 2170 if (rsvpdebug) 2171 kprintf("rsvp_input: Sending packet up old-style socket\n"); 2172 *mp = m; 2173 rip_input(mp, offp, proto); /* xxx */ 2174 } else { 2175 if (rsvpdebug && vifi == numvifs) 2176 kprintf("rsvp_input: Can't find vif for packet.\n"); 2177 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2178 kprintf("rsvp_input: No socket defined for vif %d\n",vifi); 2179 m_freem(m); 2180 } 2181 lwkt_reltoken(&mroute_token); 2182 return(IPPROTO_DONE); 2183 } 2184 rsvp_src.sin_addr = ip->ip_src; 2185 2186 if (rsvpdebug && m) 2187 kprintf("rsvp_input: m->m_len = %d, ssb_space() = %ld\n", 2188 m->m_len,ssb_space(&(viftable[vifi].v_rsvpd->so_rcv))); 2189 2190 #ifdef ALTQ 2191 opts = NULL; 2192 inp = (struct inpcb *)so->so_pcb; 2193 if (inp->inp_flags & INP_CONTROLOPTS || 2194 inp->inp_socket->so_options & SO_TIMESTAMP) { 2195 ip_savecontrol(inp, &opts, ip, m); 2196 } 2197 if (ssb_appendaddr(&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 kprintf("rsvp_input: Failed to append to socket\n"); 2204 } 2205 else { 2206 sorwakeup(so); 2207 if (rsvpdebug) 2208 kprintf("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 kprintf("rsvp_input: Failed to append to socket\n"); 2214 } else { 2215 if (rsvpdebug) 2216 kprintf("rsvp_input: send packet up\n"); 2217 } 2218 #endif /* !ALTQ */ 2219 lwkt_reltoken(&mroute_token); 2220 return(IPPROTO_DONE); 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 lwkt_gettoken(&mroute_token); 2286 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2287 if (mfc == NULL) { 2288 lwkt_reltoken(&mroute_token); 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 lwkt_reltoken(&mroute_token); 2298 return 0; /* XXX Already installed */ 2299 } 2300 } 2301 lwkt_reltoken(&mroute_token); 2302 2303 /* Allocate the new bw_meter entry */ 2304 x = kmalloc(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 lwkt_gettoken(&mroute_token); 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 lwkt_reltoken(&mroute_token); 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 kfree(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 lwkt_gettoken(&mroute_token); 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 lwkt_reltoken(&mroute_token); 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 lwkt_reltoken(&mroute_token); 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 unschedule_bw_meter(x); 2392 lwkt_reltoken(&mroute_token); 2393 /* Free the bw_meter entry */ 2394 kfree(x, M_BWMETER); 2395 return 0; 2396 } else { 2397 lwkt_reltoken(&mroute_token); 2398 return EINVAL; 2399 } 2400 } 2401 /* NOTREACHED */ 2402 } 2403 2404 /* 2405 * Perform bandwidth measurement processing that may result in an upcall 2406 */ 2407 static void 2408 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2409 { 2410 struct timeval delta; 2411 2412 lwkt_gettoken(&mroute_token); 2413 delta = *nowp; 2414 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2415 2416 if (x->bm_flags & BW_METER_GEQ) { 2417 /* 2418 * Processing for ">=" type of bw_meter entry 2419 */ 2420 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2421 /* Reset the bw_meter entry */ 2422 x->bm_start_time = *nowp; 2423 x->bm_measured.b_packets = 0; 2424 x->bm_measured.b_bytes = 0; 2425 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2426 } 2427 2428 /* Record that a packet is received */ 2429 x->bm_measured.b_packets++; 2430 x->bm_measured.b_bytes += plen; 2431 2432 /* 2433 * Test if we should deliver an upcall 2434 */ 2435 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2436 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2437 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2438 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2439 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2440 /* Prepare an upcall for delivery */ 2441 bw_meter_prepare_upcall(x, nowp); 2442 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2443 } 2444 } 2445 } else if (x->bm_flags & BW_METER_LEQ) { 2446 /* 2447 * Processing for "<=" type of bw_meter entry 2448 */ 2449 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2450 /* 2451 * We are behind time with the multicast forwarding table 2452 * scanning for "<=" type of bw_meter entries, so test now 2453 * if we should deliver an upcall. 2454 */ 2455 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2456 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2457 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2458 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2459 /* Prepare an upcall for delivery */ 2460 bw_meter_prepare_upcall(x, nowp); 2461 } 2462 /* Reschedule the bw_meter entry */ 2463 unschedule_bw_meter(x); 2464 schedule_bw_meter(x, nowp); 2465 } 2466 2467 /* Record that a packet is received */ 2468 x->bm_measured.b_packets++; 2469 x->bm_measured.b_bytes += plen; 2470 2471 /* 2472 * Test if we should restart the measuring interval 2473 */ 2474 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2475 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2476 (x->bm_flags & BW_METER_UNIT_BYTES && 2477 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2478 /* Don't restart the measuring interval */ 2479 } else { 2480 /* Do restart the measuring interval */ 2481 /* 2482 * XXX: note that we don't unschedule and schedule, because this 2483 * might be too much overhead per packet. Instead, when we process 2484 * all entries for a given timer hash bin, we check whether it is 2485 * really a timeout. If not, we reschedule at that time. 2486 */ 2487 x->bm_start_time = *nowp; 2488 x->bm_measured.b_packets = 0; 2489 x->bm_measured.b_bytes = 0; 2490 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2491 } 2492 } 2493 lwkt_reltoken(&mroute_token); 2494 } 2495 2496 /* 2497 * Prepare a bandwidth-related upcall 2498 */ 2499 static void 2500 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2501 { 2502 struct timeval delta; 2503 struct bw_upcall *u; 2504 2505 lwkt_gettoken(&mroute_token); 2506 2507 /* 2508 * Compute the measured time interval 2509 */ 2510 delta = *nowp; 2511 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2512 2513 /* 2514 * If there are too many pending upcalls, deliver them now 2515 */ 2516 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2517 bw_upcalls_send(); 2518 2519 /* 2520 * Set the bw_upcall entry 2521 */ 2522 u = &bw_upcalls[bw_upcalls_n++]; 2523 u->bu_src = x->bm_mfc->mfc_origin; 2524 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2525 u->bu_threshold.b_time = x->bm_threshold.b_time; 2526 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2527 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2528 u->bu_measured.b_time = delta; 2529 u->bu_measured.b_packets = x->bm_measured.b_packets; 2530 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2531 u->bu_flags = 0; 2532 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2533 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2534 if (x->bm_flags & BW_METER_UNIT_BYTES) 2535 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2536 if (x->bm_flags & BW_METER_GEQ) 2537 u->bu_flags |= BW_UPCALL_GEQ; 2538 if (x->bm_flags & BW_METER_LEQ) 2539 u->bu_flags |= BW_UPCALL_LEQ; 2540 2541 lwkt_reltoken(&mroute_token); 2542 } 2543 2544 /* 2545 * Send the pending bandwidth-related upcalls 2546 */ 2547 static void 2548 bw_upcalls_send(void) 2549 { 2550 struct mbuf *m; 2551 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2552 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2553 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2554 0, /* unused2 */ 2555 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2556 0, /* im_mbz */ 2557 0, /* im_vif */ 2558 0, /* unused3 */ 2559 { 0 }, /* im_src */ 2560 { 0 } }; /* im_dst */ 2561 2562 if (bw_upcalls_n == 0) 2563 return; /* No pending upcalls */ 2564 2565 bw_upcalls_n = 0; 2566 2567 /* 2568 * Allocate a new mbuf, initialize it with the header and 2569 * the payload for the pending calls. 2570 */ 2571 MGETHDR(m, M_NOWAIT, MT_HEADER); 2572 if (m == NULL) { 2573 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2574 return; 2575 } 2576 2577 m->m_len = m->m_pkthdr.len = 0; 2578 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2579 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2580 2581 /* 2582 * Send the upcalls 2583 * XXX do we need to set the address in k_igmpsrc ? 2584 */ 2585 mrtstat.mrts_upcalls++; 2586 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2587 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2588 ++mrtstat.mrts_upq_sockfull; 2589 } 2590 } 2591 2592 /* 2593 * Compute the timeout hash value for the bw_meter entries 2594 */ 2595 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2596 do { \ 2597 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2598 \ 2599 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2600 (hash) = next_timeval.tv_sec; \ 2601 if (next_timeval.tv_usec) \ 2602 (hash)++; /* XXX: make sure we don't timeout early */ \ 2603 (hash) %= BW_METER_BUCKETS; \ 2604 } while (0) 2605 2606 /* 2607 * Schedule a timer to process periodically bw_meter entry of type "<=" 2608 * by linking the entry in the proper hash bucket. 2609 */ 2610 static void 2611 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2612 { 2613 int time_hash; 2614 2615 if (!(x->bm_flags & BW_METER_LEQ)) 2616 return; /* XXX: we schedule timers only for "<=" entries */ 2617 2618 /* 2619 * Reset the bw_meter entry 2620 */ 2621 lwkt_gettoken(&mroute_token); 2622 x->bm_start_time = *nowp; 2623 x->bm_measured.b_packets = 0; 2624 x->bm_measured.b_bytes = 0; 2625 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2626 2627 /* 2628 * Compute the timeout hash value and insert the entry 2629 */ 2630 BW_METER_TIMEHASH(x, time_hash); 2631 x->bm_time_next = bw_meter_timers[time_hash]; 2632 bw_meter_timers[time_hash] = x; 2633 x->bm_time_hash = time_hash; 2634 2635 lwkt_reltoken(&mroute_token); 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(void) 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 lwkt_gettoken(&mroute_token); 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 /* Send all upcalls that are pending delivery */ 2765 bw_upcalls_send(); 2766 lwkt_reltoken(&mroute_token); 2767 } 2768 2769 /* 2770 * A periodic function for sending all upcalls that are pending delivery 2771 */ 2772 static void 2773 expire_bw_upcalls_send(void *unused) 2774 { 2775 bw_upcalls_send(); 2776 2777 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2778 expire_bw_upcalls_send, NULL); 2779 } 2780 2781 /* 2782 * A periodic function for periodic scanning of the multicast forwarding 2783 * table for processing all "<=" bw_meter entries. 2784 */ 2785 static void 2786 expire_bw_meter_process(void *unused) 2787 { 2788 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2789 bw_meter_process(); 2790 2791 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 2792 expire_bw_meter_process, NULL); 2793 } 2794 2795 /* 2796 * End of bandwidth monitoring code 2797 */ 2798 2799 #ifdef PIM 2800 /* 2801 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2802 * 2803 */ 2804 static int 2805 pim_register_send(struct ip *ip, struct vif *vifp, 2806 struct mbuf *m, struct mfc *rt) 2807 { 2808 struct mbuf *mb_copy, *mm; 2809 2810 if (mrtdebug & DEBUG_PIM) 2811 log(LOG_DEBUG, "pim_register_send: "); 2812 2813 mb_copy = pim_register_prepare(ip, m); 2814 if (mb_copy == NULL) 2815 return ENOBUFS; 2816 2817 /* 2818 * Send all the fragments. Note that the mbuf for each fragment 2819 * is freed by the sending machinery. 2820 */ 2821 for (mm = mb_copy; mm; mm = mb_copy) { 2822 mb_copy = mm->m_nextpkt; 2823 mm->m_nextpkt = 0; 2824 mm = m_pullup(mm, sizeof(struct ip)); 2825 if (mm != NULL) { 2826 ip = mtod(mm, struct ip *); 2827 if ((mrt_api_config & MRT_MFC_RP) && 2828 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2829 pim_register_send_rp(ip, vifp, mm, rt); 2830 } else { 2831 pim_register_send_upcall(ip, vifp, mm, rt); 2832 } 2833 } 2834 } 2835 2836 return 0; 2837 } 2838 2839 /* 2840 * Return a copy of the data packet that is ready for PIM Register 2841 * encapsulation. 2842 * XXX: Note that in the returned copy the IP header is a valid one. 2843 */ 2844 static struct mbuf * 2845 pim_register_prepare(struct ip *ip, struct mbuf *m) 2846 { 2847 struct mbuf *mb_copy = NULL; 2848 int mtu; 2849 2850 /* Take care of delayed checksums */ 2851 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2852 in_delayed_cksum(m); 2853 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2854 } 2855 2856 /* 2857 * Copy the old packet & pullup its IP header into the 2858 * new mbuf so we can modify it. 2859 */ 2860 mb_copy = m_copypacket(m, M_NOWAIT); 2861 if (mb_copy == NULL) 2862 return NULL; 2863 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2864 if (mb_copy == NULL) 2865 return NULL; 2866 2867 /* take care of the TTL */ 2868 ip = mtod(mb_copy, struct ip *); 2869 --ip->ip_ttl; 2870 2871 /* Compute the MTU after the PIM Register encapsulation */ 2872 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2873 2874 if (ip->ip_len <= mtu) { 2875 /* Turn the IP header into a valid one */ 2876 ip->ip_len = htons(ip->ip_len); 2877 ip->ip_off = htons(ip->ip_off); 2878 ip->ip_sum = 0; 2879 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2880 } else { 2881 /* Fragment the packet */ 2882 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2883 m_freem(mb_copy); 2884 return NULL; 2885 } 2886 } 2887 return mb_copy; 2888 } 2889 2890 /* 2891 * Send an upcall with the data packet to the user-level process. 2892 */ 2893 static int 2894 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2895 struct mbuf *mb_copy, struct mfc *rt) 2896 { 2897 struct mbuf *mb_first; 2898 int len = ntohs(ip->ip_len); 2899 struct igmpmsg *im; 2900 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2901 2902 /* 2903 * Add a new mbuf with an upcall header 2904 */ 2905 MGETHDR(mb_first, M_NOWAIT, MT_HEADER); 2906 if (mb_first == NULL) { 2907 m_freem(mb_copy); 2908 return ENOBUFS; 2909 } 2910 mb_first->m_data += max_linkhdr; 2911 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2912 mb_first->m_len = sizeof(struct igmpmsg); 2913 mb_first->m_next = mb_copy; 2914 2915 /* Send message to routing daemon */ 2916 im = mtod(mb_first, struct igmpmsg *); 2917 im->im_msgtype = IGMPMSG_WHOLEPKT; 2918 im->im_mbz = 0; 2919 im->im_vif = vifp - viftable; 2920 im->im_src = ip->ip_src; 2921 im->im_dst = ip->ip_dst; 2922 2923 k_igmpsrc.sin_addr = ip->ip_src; 2924 2925 mrtstat.mrts_upcalls++; 2926 2927 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2928 if (mrtdebug & DEBUG_PIM) 2929 log(LOG_WARNING, 2930 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2931 ++mrtstat.mrts_upq_sockfull; 2932 return ENOBUFS; 2933 } 2934 2935 /* Keep statistics */ 2936 pimstat.pims_snd_registers_msgs++; 2937 pimstat.pims_snd_registers_bytes += len; 2938 2939 return 0; 2940 } 2941 2942 /* 2943 * Encapsulate the data packet in PIM Register message and send it to the RP. 2944 */ 2945 static int 2946 pim_register_send_rp(struct ip *ip, struct vif *vifp, 2947 struct mbuf *mb_copy, struct mfc *rt) 2948 { 2949 struct mbuf *mb_first; 2950 struct ip *ip_outer; 2951 struct pim_encap_pimhdr *pimhdr; 2952 int len = ntohs(ip->ip_len); 2953 vifi_t vifi = rt->mfc_parent; 2954 2955 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2956 m_freem(mb_copy); 2957 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2958 } 2959 2960 /* 2961 * Add a new mbuf with the encapsulating header 2962 */ 2963 MGETHDR(mb_first, M_NOWAIT, MT_HEADER); 2964 if (mb_first == NULL) { 2965 m_freem(mb_copy); 2966 return ENOBUFS; 2967 } 2968 mb_first->m_data += max_linkhdr; 2969 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2970 mb_first->m_next = mb_copy; 2971 2972 mb_first->m_pkthdr.len = len + mb_first->m_len; 2973 2974 /* 2975 * Fill in the encapsulating IP and PIM header 2976 */ 2977 ip_outer = mtod(mb_first, struct ip *); 2978 *ip_outer = pim_encap_iphdr; 2979 ip_outer->ip_id = ip_newid(); 2980 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2981 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2982 ip_outer->ip_dst = rt->mfc_rp; 2983 /* 2984 * Copy the inner header TOS to the outer header, and take care of the 2985 * IP_DF bit. 2986 */ 2987 ip_outer->ip_tos = ip->ip_tos; 2988 if (ntohs(ip->ip_off) & IP_DF) 2989 ip_outer->ip_off |= IP_DF; 2990 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2991 + sizeof(pim_encap_iphdr)); 2992 *pimhdr = pim_encap_pimhdr; 2993 /* If the iif crosses a border, set the Border-bit */ 2994 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2995 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2996 2997 mb_first->m_data += sizeof(pim_encap_iphdr); 2998 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2999 mb_first->m_data -= sizeof(pim_encap_iphdr); 3000 3001 if (vifp->v_rate_limit == 0) 3002 tbf_send_packet(vifp, mb_first); 3003 else 3004 tbf_control(vifp, mb_first, ip, ip_outer->ip_len); 3005 3006 /* Keep statistics */ 3007 pimstat.pims_snd_registers_msgs++; 3008 pimstat.pims_snd_registers_bytes += len; 3009 3010 return 0; 3011 } 3012 3013 /* 3014 * PIM-SMv2 and PIM-DM messages processing. 3015 * Receives and verifies the PIM control messages, and passes them 3016 * up to the listening socket, using rip_input(). 3017 * The only message with special processing is the PIM_REGISTER message 3018 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3019 * is passed to if_simloop(). 3020 */ 3021 int 3022 pim_input(struct mbuf **mp, int *offp, int proto) 3023 { 3024 struct mbuf *m = *mp; 3025 struct ip *ip = mtod(m, struct ip *); 3026 struct pim *pim; 3027 int minlen; 3028 int datalen = ip->ip_len; 3029 int ip_tos; 3030 int iphlen; 3031 3032 iphlen = *offp; 3033 *mp = NULL; 3034 3035 /* Keep statistics */ 3036 pimstat.pims_rcv_total_msgs++; 3037 pimstat.pims_rcv_total_bytes += datalen; 3038 3039 /* 3040 * Validate lengths 3041 */ 3042 if (datalen < PIM_MINLEN) { 3043 pimstat.pims_rcv_tooshort++; 3044 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3045 datalen, (u_long)ip->ip_src.s_addr); 3046 m_freem(m); 3047 return(IPPROTO_DONE); 3048 } 3049 3050 /* 3051 * If the packet is at least as big as a REGISTER, go agead 3052 * and grab the PIM REGISTER header size, to avoid another 3053 * possible m_pullup() later. 3054 * 3055 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3056 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3057 */ 3058 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3059 /* 3060 * Get the IP and PIM headers in contiguous memory, and 3061 * possibly the PIM REGISTER header. 3062 */ 3063 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3064 (m = m_pullup(m, minlen)) == NULL) { 3065 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3066 return(IPPROTO_DONE); 3067 } 3068 /* m_pullup() may have given us a new mbuf so reset ip. */ 3069 ip = mtod(m, struct ip *); 3070 ip_tos = ip->ip_tos; 3071 3072 /* adjust mbuf to point to the PIM header */ 3073 m->m_data += iphlen; 3074 m->m_len -= iphlen; 3075 pim = mtod(m, struct pim *); 3076 3077 /* 3078 * Validate checksum. If PIM REGISTER, exclude the data packet. 3079 * 3080 * XXX: some older PIMv2 implementations don't make this distinction, 3081 * so for compatibility reason perform the checksum over part of the 3082 * message, and if error, then over the whole message. 3083 */ 3084 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3085 /* do nothing, checksum okay */ 3086 } else if (in_cksum(m, datalen)) { 3087 pimstat.pims_rcv_badsum++; 3088 if (mrtdebug & DEBUG_PIM) 3089 log(LOG_DEBUG, "pim_input: invalid checksum"); 3090 m_freem(m); 3091 return(IPPROTO_DONE); 3092 } 3093 3094 /* PIM version check */ 3095 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3096 pimstat.pims_rcv_badversion++; 3097 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3098 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3099 m_freem(m); 3100 return(IPPROTO_DONE); 3101 } 3102 3103 /* restore mbuf back to the outer IP */ 3104 m->m_data -= iphlen; 3105 m->m_len += iphlen; 3106 3107 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3108 /* 3109 * Since this is a REGISTER, we'll make a copy of the register 3110 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3111 * routing daemon. 3112 */ 3113 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3114 struct mbuf *mcp; 3115 struct ip *encap_ip; 3116 u_int32_t *reghdr; 3117 3118 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3119 if (mrtdebug & DEBUG_PIM) 3120 log(LOG_DEBUG, 3121 "pim_input: register vif not set: %d\n", reg_vif_num); 3122 m_freem(m); 3123 return(IPPROTO_DONE); 3124 } 3125 3126 /* 3127 * Validate length 3128 */ 3129 if (datalen < PIM_REG_MINLEN) { 3130 pimstat.pims_rcv_tooshort++; 3131 pimstat.pims_rcv_badregisters++; 3132 log(LOG_ERR, 3133 "pim_input: register packet size too small %d from %lx\n", 3134 datalen, (u_long)ip->ip_src.s_addr); 3135 m_freem(m); 3136 return(IPPROTO_DONE); 3137 } 3138 3139 reghdr = (u_int32_t *)(pim + 1); 3140 encap_ip = (struct ip *)(reghdr + 1); 3141 3142 if (mrtdebug & DEBUG_PIM) { 3143 log(LOG_DEBUG, 3144 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3145 (u_long)ntohl(encap_ip->ip_src.s_addr), 3146 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3147 ntohs(encap_ip->ip_len)); 3148 } 3149 3150 /* verify the version number of the inner packet */ 3151 if (encap_ip->ip_v != IPVERSION) { 3152 pimstat.pims_rcv_badregisters++; 3153 if (mrtdebug & DEBUG_PIM) { 3154 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3155 "of the inner packet\n", encap_ip->ip_v); 3156 } 3157 m_freem(m); 3158 return(IPPROTO_DONE); 3159 } 3160 3161 /* verify the inner packet is destined to a mcast group */ 3162 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 3163 pimstat.pims_rcv_badregisters++; 3164 if (mrtdebug & DEBUG_PIM) 3165 log(LOG_DEBUG, 3166 "pim_input: inner packet of register is not " 3167 "multicast %lx\n", 3168 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3169 m_freem(m); 3170 return(IPPROTO_DONE); 3171 } 3172 3173 /* If a NULL_REGISTER, pass it to the daemon */ 3174 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3175 goto pim_input_to_daemon; 3176 3177 /* 3178 * Copy the TOS from the outer IP header to the inner IP header. 3179 */ 3180 if (encap_ip->ip_tos != ip_tos) { 3181 /* Outer TOS -> inner TOS */ 3182 encap_ip->ip_tos = ip_tos; 3183 /* Recompute the inner header checksum. Sigh... */ 3184 3185 /* adjust mbuf to point to the inner IP header */ 3186 m->m_data += (iphlen + PIM_MINLEN); 3187 m->m_len -= (iphlen + PIM_MINLEN); 3188 3189 encap_ip->ip_sum = 0; 3190 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3191 3192 /* restore mbuf to point back to the outer IP header */ 3193 m->m_data -= (iphlen + PIM_MINLEN); 3194 m->m_len += (iphlen + PIM_MINLEN); 3195 } 3196 3197 /* 3198 * Decapsulate the inner IP packet and loopback to forward it 3199 * as a normal multicast packet. Also, make a copy of the 3200 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3201 * to pass to the daemon later, so it can take the appropriate 3202 * actions (e.g., send back PIM_REGISTER_STOP). 3203 * XXX: here m->m_data points to the outer IP header. 3204 */ 3205 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3206 if (mcp == NULL) { 3207 log(LOG_ERR, 3208 "pim_input: pim register: could not copy register head\n"); 3209 m_freem(m); 3210 return(IPPROTO_DONE); 3211 } 3212 3213 /* Keep statistics */ 3214 /* XXX: registers_bytes include only the encap. mcast pkt */ 3215 pimstat.pims_rcv_registers_msgs++; 3216 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3217 3218 /* 3219 * forward the inner ip packet; point m_data at the inner ip. 3220 */ 3221 m_adj(m, iphlen + PIM_MINLEN); 3222 3223 if (mrtdebug & DEBUG_PIM) { 3224 log(LOG_DEBUG, 3225 "pim_input: forwarding decapsulated register: " 3226 "src %lx, dst %lx, vif %d\n", 3227 (u_long)ntohl(encap_ip->ip_src.s_addr), 3228 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3229 reg_vif_num); 3230 } 3231 if_simloop(viftable[reg_vif_num].v_ifp, m, dst.sin_family, 0); 3232 3233 /* prepare the register head to send to the mrouting daemon */ 3234 m = mcp; 3235 } 3236 3237 pim_input_to_daemon: 3238 /* 3239 * Pass the PIM message up to the daemon; if it is a Register message, 3240 * pass the 'head' only up to the daemon. This includes the 3241 * outer IP header, PIM header, PIM-Register header and the 3242 * inner IP header. 3243 * XXX: the outer IP header pkt size of a Register is not adjust to 3244 * reflect the fact that the inner multicast data is truncated. 3245 */ 3246 *mp = m; 3247 *offp = iphlen; 3248 rip_input(mp, offp, proto); 3249 return(IPPROTO_DONE); 3250 } 3251 #endif /* PIM */ 3252 3253 static int 3254 ip_mroute_modevent(module_t mod, int type, void *unused) 3255 { 3256 switch (type) { 3257 case MOD_LOAD: 3258 lwkt_gettoken(&mroute_token); 3259 /* XXX Protect against multiple loading */ 3260 ip_mcast_src = X_ip_mcast_src; 3261 ip_mforward = X_ip_mforward; 3262 ip_mrouter_done = X_ip_mrouter_done; 3263 ip_mrouter_get = X_ip_mrouter_get; 3264 ip_mrouter_set = X_ip_mrouter_set; 3265 ip_rsvp_force_done = X_ip_rsvp_force_done; 3266 ip_rsvp_vif = X_ip_rsvp_vif; 3267 ipip_input = X_ipip_input; 3268 legal_vif_num = X_legal_vif_num; 3269 mrt_ioctl = X_mrt_ioctl; 3270 rsvp_input_p = X_rsvp_input; 3271 lwkt_reltoken(&mroute_token); 3272 break; 3273 3274 case MOD_UNLOAD: 3275 if (ip_mrouter) 3276 return EINVAL; 3277 3278 lwkt_gettoken(&mroute_token); 3279 ip_mcast_src = NULL; 3280 ip_mforward = NULL; 3281 ip_mrouter_done = NULL; 3282 ip_mrouter_get = NULL; 3283 ip_mrouter_set = NULL; 3284 ip_rsvp_force_done = NULL; 3285 ip_rsvp_vif = NULL; 3286 ipip_input = NULL; 3287 legal_vif_num = NULL; 3288 mrt_ioctl = NULL; 3289 rsvp_input_p = NULL; 3290 lwkt_reltoken(&mroute_token); 3291 break; 3292 } 3293 return 0; 3294 } 3295 3296 static moduledata_t ip_mroutemod = { 3297 "ip_mroute", 3298 ip_mroute_modevent, 3299 0 3300 }; 3301 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3302