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 * $FreeBSD$ 21 */ 22 23 #include "opt_mac.h" 24 #include "opt_mrouting.h" 25 #include "opt_random_ip_id.h" 26 27 #ifdef PIM 28 #define _PIM_VT 1 29 #endif 30 31 #include <sys/param.h> 32 #include <sys/kernel.h> 33 #include <sys/lock.h> 34 #include <sys/mac.h> 35 #include <sys/malloc.h> 36 #include <sys/mbuf.h> 37 #include <sys/protosw.h> 38 #include <sys/signalvar.h> 39 #include <sys/socket.h> 40 #include <sys/socketvar.h> 41 #include <sys/sockio.h> 42 #include <sys/sx.h> 43 #include <sys/sysctl.h> 44 #include <sys/syslog.h> 45 #include <sys/systm.h> 46 #include <sys/time.h> 47 #include <net/if.h> 48 #include <net/netisr.h> 49 #include <net/route.h> 50 #include <netinet/in.h> 51 #include <netinet/igmp.h> 52 #include <netinet/in_systm.h> 53 #include <netinet/in_var.h> 54 #include <netinet/ip.h> 55 #include <netinet/ip_encap.h> 56 #include <netinet/ip_mroute.h> 57 #include <netinet/ip_var.h> 58 #ifdef PIM 59 #include <netinet/pim.h> 60 #include <netinet/pim_var.h> 61 #endif 62 #include <netinet/udp.h> 63 #include <machine/in_cksum.h> 64 65 /* 66 * Control debugging code for rsvp and multicast routing code. 67 * Can only set them with the debugger. 68 */ 69 static u_int rsvpdebug; /* non-zero enables debugging */ 70 71 static u_int mrtdebug; /* any set of the flags below */ 72 #define DEBUG_MFC 0x02 73 #define DEBUG_FORWARD 0x04 74 #define DEBUG_EXPIRE 0x08 75 #define DEBUG_XMIT 0x10 76 #define DEBUG_PIM 0x20 77 78 #define VIFI_INVALID ((vifi_t) -1) 79 80 #define M_HASCL(m) ((m)->m_flags & M_EXT) 81 82 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 83 84 /* 85 * Locking. We use two locks: one for the virtual interface table and 86 * one for the forwarding table. These locks may be nested in which case 87 * the VIF lock must always be taken first. Note that each lock is used 88 * to cover not only the specific data structure but also related data 89 * structures. It may be better to add more fine-grained locking later; 90 * it's not clear how performance-critical this code is. 91 */ 92 93 static struct mrtstat mrtstat; 94 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 95 &mrtstat, mrtstat, 96 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 97 98 static struct mfc *mfctable[MFCTBLSIZ]; 99 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 100 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 101 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 102 103 static struct mtx mfc_mtx; 104 #define MFC_LOCK() mtx_lock(&mfc_mtx) 105 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 106 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED) 107 #define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF) 108 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 109 110 static struct vif viftable[MAXVIFS]; 111 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 112 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 113 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 114 115 static struct mtx vif_mtx; 116 #define VIF_LOCK() mtx_lock(&vif_mtx) 117 #define VIF_UNLOCK() mtx_unlock(&vif_mtx) 118 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 119 #define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF) 120 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 121 122 static u_char nexpire[MFCTBLSIZ]; 123 124 static struct callout expire_upcalls_ch; 125 126 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 127 #define UPCALL_EXPIRE 6 /* number of timeouts */ 128 129 /* 130 * Define the token bucket filter structures 131 * tbftable -> each vif has one of these for storing info 132 */ 133 134 static struct tbf tbftable[MAXVIFS]; 135 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 136 137 /* 138 * 'Interfaces' associated with decapsulator (so we can tell 139 * packets that went through it from ones that get reflected 140 * by a broken gateway). These interfaces are never linked into 141 * the system ifnet list & no routes point to them. I.e., packets 142 * can't be sent this way. They only exist as a placeholder for 143 * multicast source verification. 144 */ 145 static struct ifnet multicast_decap_if[MAXVIFS]; 146 147 #define ENCAP_TTL 64 148 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 149 150 /* prototype IP hdr for encapsulated packets */ 151 static struct ip multicast_encap_iphdr = { 152 #if BYTE_ORDER == LITTLE_ENDIAN 153 sizeof(struct ip) >> 2, IPVERSION, 154 #else 155 IPVERSION, sizeof(struct ip) >> 2, 156 #endif 157 0, /* tos */ 158 sizeof(struct ip), /* total length */ 159 0, /* id */ 160 0, /* frag offset */ 161 ENCAP_TTL, ENCAP_PROTO, 162 0, /* checksum */ 163 }; 164 165 /* 166 * Bandwidth meter variables and constants 167 */ 168 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 169 /* 170 * Pending timeouts are stored in a hash table, the key being the 171 * expiration time. Periodically, the entries are analysed and processed. 172 */ 173 #define BW_METER_BUCKETS 1024 174 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 175 static struct callout bw_meter_ch; 176 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 177 178 /* 179 * Pending upcalls are stored in a vector which is flushed when 180 * full, or periodically 181 */ 182 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 183 static u_int bw_upcalls_n; /* # of pending upcalls */ 184 static struct callout bw_upcalls_ch; 185 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 186 187 #ifdef PIM 188 static struct pimstat pimstat; 189 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 190 &pimstat, pimstat, 191 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 192 193 /* 194 * Note: the PIM Register encapsulation adds the following in front of a 195 * data packet: 196 * 197 * struct pim_encap_hdr { 198 * struct ip ip; 199 * struct pim_encap_pimhdr pim; 200 * } 201 * 202 */ 203 204 struct pim_encap_pimhdr { 205 struct pim pim; 206 uint32_t flags; 207 }; 208 209 static struct ip pim_encap_iphdr = { 210 #if BYTE_ORDER == LITTLE_ENDIAN 211 sizeof(struct ip) >> 2, 212 IPVERSION, 213 #else 214 IPVERSION, 215 sizeof(struct ip) >> 2, 216 #endif 217 0, /* tos */ 218 sizeof(struct ip), /* total length */ 219 0, /* id */ 220 0, /* frag offset */ 221 ENCAP_TTL, 222 IPPROTO_PIM, 223 0, /* checksum */ 224 }; 225 226 static struct pim_encap_pimhdr pim_encap_pimhdr = { 227 { 228 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 229 0, /* reserved */ 230 0, /* checksum */ 231 }, 232 0 /* flags */ 233 }; 234 235 static struct ifnet multicast_register_if; 236 static vifi_t reg_vif_num = VIFI_INVALID; 237 #endif /* PIM */ 238 239 /* 240 * Private variables. 241 */ 242 static vifi_t numvifs; 243 static const struct encaptab *encap_cookie; 244 245 /* 246 * one-back cache used by mroute_encapcheck to locate a tunnel's vif 247 * given a datagram's src ip address. 248 */ 249 static u_long last_encap_src; 250 static struct vif *last_encap_vif; 251 252 /* 253 * Callout for queue processing. 254 */ 255 static struct callout tbf_reprocess_ch; 256 257 static u_long X_ip_mcast_src(int vifi); 258 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 259 struct mbuf *m, struct ip_moptions *imo); 260 static int X_ip_mrouter_done(void); 261 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 262 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 263 static int X_legal_vif_num(int vif); 264 static int X_mrt_ioctl(int cmd, caddr_t data); 265 266 static int get_sg_cnt(struct sioc_sg_req *); 267 static int get_vif_cnt(struct sioc_vif_req *); 268 static int ip_mrouter_init(struct socket *, int); 269 static int add_vif(struct vifctl *); 270 static int del_vif(vifi_t); 271 static int add_mfc(struct mfcctl2 *); 272 static int del_mfc(struct mfcctl2 *); 273 static int set_api_config(uint32_t *); /* chose API capabilities */ 274 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 275 static int set_assert(int); 276 static void expire_upcalls(void *); 277 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 278 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 279 static void encap_send(struct ip *, struct vif *, struct mbuf *); 280 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 281 static void tbf_queue(struct vif *, struct mbuf *); 282 static void tbf_process_q(struct vif *); 283 static void tbf_reprocess_q(void *); 284 static int tbf_dq_sel(struct vif *, struct ip *); 285 static void tbf_send_packet(struct vif *, struct mbuf *); 286 static void tbf_update_tokens(struct vif *); 287 static int priority(struct vif *, struct ip *); 288 289 /* 290 * Bandwidth monitoring 291 */ 292 static void free_bw_list(struct bw_meter *list); 293 static int add_bw_upcall(struct bw_upcall *); 294 static int del_bw_upcall(struct bw_upcall *); 295 static void bw_meter_receive_packet(struct bw_meter *x, int plen, 296 struct timeval *nowp); 297 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 298 static void bw_upcalls_send(void); 299 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 300 static void unschedule_bw_meter(struct bw_meter *x); 301 static void bw_meter_process(void); 302 static void expire_bw_upcalls_send(void *); 303 static void expire_bw_meter_process(void *); 304 305 #ifdef PIM 306 static int pim_register_send(struct ip *, struct vif *, 307 struct mbuf *, struct mfc *); 308 static int pim_register_send_rp(struct ip *, struct vif *, 309 struct mbuf *, struct mfc *); 310 static int pim_register_send_upcall(struct ip *, struct vif *, 311 struct mbuf *, struct mfc *); 312 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 313 #endif 314 315 /* 316 * whether or not special PIM assert processing is enabled. 317 */ 318 static int pim_assert; 319 /* 320 * Rate limit for assert notification messages, in usec 321 */ 322 #define ASSERT_MSG_TIME 3000000 323 324 /* 325 * Kernel multicast routing API capabilities and setup. 326 * If more API capabilities are added to the kernel, they should be 327 * recorded in `mrt_api_support'. 328 */ 329 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 330 MRT_MFC_FLAGS_BORDER_VIF | 331 MRT_MFC_RP | 332 MRT_MFC_BW_UPCALL); 333 static uint32_t mrt_api_config = 0; 334 335 /* 336 * Hash function for a source, group entry 337 */ 338 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 339 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 340 341 /* 342 * Find a route for a given origin IP address and Multicast group address 343 * Type of service parameter to be added in the future!!! 344 * Statistics are updated by the caller if needed 345 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 346 */ 347 static struct mfc * 348 mfc_find(in_addr_t o, in_addr_t g) 349 { 350 struct mfc *rt; 351 352 MFC_LOCK_ASSERT(); 353 354 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 355 if ((rt->mfc_origin.s_addr == o) && 356 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 357 break; 358 return rt; 359 } 360 361 /* 362 * Macros to compute elapsed time efficiently 363 * Borrowed from Van Jacobson's scheduling code 364 */ 365 #define TV_DELTA(a, b, delta) { \ 366 int xxs; \ 367 delta = (a).tv_usec - (b).tv_usec; \ 368 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 369 switch (xxs) { \ 370 case 2: \ 371 delta += 1000000; \ 372 /* FALLTHROUGH */ \ 373 case 1: \ 374 delta += 1000000; \ 375 break; \ 376 default: \ 377 delta += (1000000 * xxs); \ 378 } \ 379 } \ 380 } 381 382 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 383 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 384 385 /* 386 * Handle MRT setsockopt commands to modify the multicast routing tables. 387 */ 388 static int 389 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 390 { 391 int error, optval; 392 vifi_t vifi; 393 struct vifctl vifc; 394 struct mfcctl2 mfc; 395 struct bw_upcall bw_upcall; 396 uint32_t i; 397 398 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 399 return EPERM; 400 401 error = 0; 402 switch (sopt->sopt_name) { 403 case MRT_INIT: 404 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 405 if (error) 406 break; 407 error = ip_mrouter_init(so, optval); 408 break; 409 410 case MRT_DONE: 411 error = ip_mrouter_done(); 412 break; 413 414 case MRT_ADD_VIF: 415 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 416 if (error) 417 break; 418 error = add_vif(&vifc); 419 break; 420 421 case MRT_DEL_VIF: 422 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 423 if (error) 424 break; 425 error = del_vif(vifi); 426 break; 427 428 case MRT_ADD_MFC: 429 case MRT_DEL_MFC: 430 /* 431 * select data size depending on API version. 432 */ 433 if (sopt->sopt_name == MRT_ADD_MFC && 434 mrt_api_config & MRT_API_FLAGS_ALL) { 435 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 436 sizeof(struct mfcctl2)); 437 } else { 438 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 439 sizeof(struct mfcctl)); 440 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 441 sizeof(mfc) - sizeof(struct mfcctl)); 442 } 443 if (error) 444 break; 445 if (sopt->sopt_name == MRT_ADD_MFC) 446 error = add_mfc(&mfc); 447 else 448 error = del_mfc(&mfc); 449 break; 450 451 case MRT_ASSERT: 452 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 453 if (error) 454 break; 455 set_assert(optval); 456 break; 457 458 case MRT_API_CONFIG: 459 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 460 if (!error) 461 error = set_api_config(&i); 462 if (!error) 463 error = sooptcopyout(sopt, &i, sizeof i); 464 break; 465 466 case MRT_ADD_BW_UPCALL: 467 case MRT_DEL_BW_UPCALL: 468 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 469 sizeof bw_upcall); 470 if (error) 471 break; 472 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 473 error = add_bw_upcall(&bw_upcall); 474 else 475 error = del_bw_upcall(&bw_upcall); 476 break; 477 478 default: 479 error = EOPNOTSUPP; 480 break; 481 } 482 return error; 483 } 484 485 /* 486 * Handle MRT getsockopt commands 487 */ 488 static int 489 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 490 { 491 int error; 492 static int version = 0x0305; /* !!! why is this here? XXX */ 493 494 switch (sopt->sopt_name) { 495 case MRT_VERSION: 496 error = sooptcopyout(sopt, &version, sizeof version); 497 break; 498 499 case MRT_ASSERT: 500 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 501 break; 502 503 case MRT_API_SUPPORT: 504 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 505 break; 506 507 case MRT_API_CONFIG: 508 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 509 break; 510 511 default: 512 error = EOPNOTSUPP; 513 break; 514 } 515 return error; 516 } 517 518 /* 519 * Handle ioctl commands to obtain information from the cache 520 */ 521 static int 522 X_mrt_ioctl(int cmd, caddr_t data) 523 { 524 int error = 0; 525 526 switch (cmd) { 527 case (SIOCGETVIFCNT): 528 error = get_vif_cnt((struct sioc_vif_req *)data); 529 break; 530 531 case (SIOCGETSGCNT): 532 error = get_sg_cnt((struct sioc_sg_req *)data); 533 break; 534 535 default: 536 error = EINVAL; 537 break; 538 } 539 return error; 540 } 541 542 /* 543 * returns the packet, byte, rpf-failure count for the source group provided 544 */ 545 static int 546 get_sg_cnt(struct sioc_sg_req *req) 547 { 548 struct mfc *rt; 549 550 MFC_LOCK(); 551 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 552 if (rt == NULL) { 553 MFC_UNLOCK(); 554 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 555 return EADDRNOTAVAIL; 556 } 557 req->pktcnt = rt->mfc_pkt_cnt; 558 req->bytecnt = rt->mfc_byte_cnt; 559 req->wrong_if = rt->mfc_wrong_if; 560 MFC_UNLOCK(); 561 return 0; 562 } 563 564 /* 565 * returns the input and output packet and byte counts on the vif provided 566 */ 567 static int 568 get_vif_cnt(struct sioc_vif_req *req) 569 { 570 vifi_t vifi = req->vifi; 571 572 VIF_LOCK(); 573 if (vifi >= numvifs) { 574 VIF_UNLOCK(); 575 return EINVAL; 576 } 577 578 req->icount = viftable[vifi].v_pkt_in; 579 req->ocount = viftable[vifi].v_pkt_out; 580 req->ibytes = viftable[vifi].v_bytes_in; 581 req->obytes = viftable[vifi].v_bytes_out; 582 VIF_UNLOCK(); 583 584 return 0; 585 } 586 587 static void 588 ip_mrouter_reset(void) 589 { 590 bzero((caddr_t)mfctable, sizeof(mfctable)); 591 MFC_LOCK_INIT(); 592 VIF_LOCK_INIT(); 593 bzero((caddr_t)nexpire, sizeof(nexpire)); 594 595 pim_assert = 0; 596 mrt_api_config = 0; 597 598 callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE); 599 600 bw_upcalls_n = 0; 601 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 602 callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE); 603 callout_init(&bw_meter_ch, CALLOUT_MPSAFE); 604 605 callout_init(&tbf_reprocess_ch, CALLOUT_MPSAFE); 606 } 607 608 /* 609 * Enable multicast routing 610 */ 611 static int 612 ip_mrouter_init(struct socket *so, int version) 613 { 614 if (mrtdebug) 615 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 616 so->so_type, so->so_proto->pr_protocol); 617 618 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 619 return EOPNOTSUPP; 620 621 if (version != 1) 622 return ENOPROTOOPT; 623 624 if (ip_mrouter != NULL) 625 return EADDRINUSE; 626 627 ip_mrouter_reset(); 628 629 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 630 631 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 632 expire_bw_upcalls_send, NULL); 633 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 634 635 ip_mrouter = so; 636 637 if (mrtdebug) 638 log(LOG_DEBUG, "ip_mrouter_init\n"); 639 640 return 0; 641 } 642 643 /* 644 * Disable multicast routing 645 */ 646 static int 647 X_ip_mrouter_done(void) 648 { 649 vifi_t vifi; 650 int i; 651 struct ifnet *ifp; 652 struct ifreq ifr; 653 struct mfc *rt; 654 struct rtdetq *rte; 655 656 /* 657 * Detach/disable hooks to the reset of the system. 658 */ 659 ip_mrouter = NULL; 660 mrt_api_config = 0; 661 662 VIF_LOCK(); 663 if (encap_cookie) { 664 const struct encaptab *c = encap_cookie; 665 encap_cookie = NULL; 666 encap_detach(c); 667 } 668 VIF_UNLOCK(); 669 670 callout_stop(&tbf_reprocess_ch); 671 672 VIF_LOCK(); 673 /* 674 * For each phyint in use, disable promiscuous reception of all IP 675 * multicasts. 676 */ 677 for (vifi = 0; vifi < numvifs; vifi++) { 678 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 679 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 680 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 681 682 so->sin_len = sizeof(struct sockaddr_in); 683 so->sin_family = AF_INET; 684 so->sin_addr.s_addr = INADDR_ANY; 685 ifp = viftable[vifi].v_ifp; 686 if_allmulti(ifp, 0); 687 } 688 } 689 bzero((caddr_t)tbftable, sizeof(tbftable)); 690 bzero((caddr_t)viftable, sizeof(viftable)); 691 numvifs = 0; 692 pim_assert = 0; 693 VIF_LOCK_DESTROY(); 694 695 /* 696 * Free all multicast forwarding cache entries. 697 */ 698 callout_stop(&expire_upcalls_ch); 699 callout_stop(&bw_upcalls_ch); 700 callout_stop(&bw_meter_ch); 701 702 MFC_LOCK(); 703 for (i = 0; i < MFCTBLSIZ; i++) { 704 for (rt = mfctable[i]; rt != NULL; ) { 705 struct mfc *nr = rt->mfc_next; 706 707 for (rte = rt->mfc_stall; rte != NULL; ) { 708 struct rtdetq *n = rte->next; 709 710 m_freem(rte->m); 711 free(rte, M_MRTABLE); 712 rte = n; 713 } 714 free_bw_list(rt->mfc_bw_meter); 715 free(rt, M_MRTABLE); 716 rt = nr; 717 } 718 } 719 bzero((caddr_t)mfctable, sizeof(mfctable)); 720 bw_upcalls_n = 0; 721 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 722 MFC_LOCK_DESTROY(); 723 724 /* 725 * Reset de-encapsulation cache 726 */ 727 last_encap_src = INADDR_ANY; 728 last_encap_vif = NULL; 729 #ifdef PIM 730 reg_vif_num = VIFI_INVALID; 731 #endif 732 733 if (mrtdebug) 734 log(LOG_DEBUG, "ip_mrouter_done\n"); 735 736 return 0; 737 } 738 739 /* 740 * Set PIM assert processing global 741 */ 742 static int 743 set_assert(int i) 744 { 745 if ((i != 1) && (i != 0)) 746 return EINVAL; 747 748 pim_assert = i; 749 750 return 0; 751 } 752 753 /* 754 * Configure API capabilities 755 */ 756 int 757 set_api_config(uint32_t *apival) 758 { 759 int i; 760 761 /* 762 * We can set the API capabilities only if it is the first operation 763 * after MRT_INIT. I.e.: 764 * - there are no vifs installed 765 * - pim_assert is not enabled 766 * - the MFC table is empty 767 */ 768 if (numvifs > 0) { 769 *apival = 0; 770 return EPERM; 771 } 772 if (pim_assert) { 773 *apival = 0; 774 return EPERM; 775 } 776 for (i = 0; i < MFCTBLSIZ; i++) { 777 if (mfctable[i] != NULL) { 778 *apival = 0; 779 return EPERM; 780 } 781 } 782 783 mrt_api_config = *apival & mrt_api_support; 784 *apival = mrt_api_config; 785 786 return 0; 787 } 788 789 /* 790 * Decide if a packet is from a tunnelled peer. 791 * Return 0 if not, 64 if so. XXX yuck.. 64 ??? 792 */ 793 static int 794 mroute_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 795 { 796 struct ip *ip = mtod(m, struct ip *); 797 int hlen = ip->ip_hl << 2; 798 799 /* 800 * don't claim the packet if it's not to a multicast destination or if 801 * we don't have an encapsulating tunnel with the source. 802 * Note: This code assumes that the remote site IP address 803 * uniquely identifies the tunnel (i.e., that this site has 804 * at most one tunnel with the remote site). 805 */ 806 if (!IN_MULTICAST(ntohl(((struct ip *)((char *)ip+hlen))->ip_dst.s_addr))) 807 return 0; 808 if (ip->ip_src.s_addr != last_encap_src) { 809 struct vif *vifp = viftable; 810 struct vif *vife = vifp + numvifs; 811 812 last_encap_src = ip->ip_src.s_addr; 813 last_encap_vif = NULL; 814 for ( ; vifp < vife; ++vifp) 815 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 816 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) == VIFF_TUNNEL) 817 last_encap_vif = vifp; 818 break; 819 } 820 } 821 if (last_encap_vif == NULL) { 822 last_encap_src = INADDR_ANY; 823 return 0; 824 } 825 return 64; 826 } 827 828 /* 829 * De-encapsulate a packet and feed it back through ip input (this 830 * routine is called whenever IP gets a packet that mroute_encap_func() 831 * claimed). 832 */ 833 static void 834 mroute_encap_input(struct mbuf *m, int off) 835 { 836 struct ip *ip = mtod(m, struct ip *); 837 int hlen = ip->ip_hl << 2; 838 839 if (hlen > sizeof(struct ip)) 840 ip_stripoptions(m, (struct mbuf *) 0); 841 m->m_data += sizeof(struct ip); 842 m->m_len -= sizeof(struct ip); 843 m->m_pkthdr.len -= sizeof(struct ip); 844 845 m->m_pkthdr.rcvif = last_encap_vif->v_ifp; 846 847 netisr_queue(NETISR_IP, m); 848 /* 849 * normally we would need a "schednetisr(NETISR_IP)" 850 * here but we were called by ip_input and it is going 851 * to loop back & try to dequeue the packet we just 852 * queued as soon as we return so we avoid the 853 * unnecessary software interrrupt. 854 * 855 * XXX 856 * This no longer holds - we may have direct-dispatched the packet, 857 * or there may be a queue processing limit. 858 */ 859 } 860 861 extern struct domain inetdomain; 862 static struct protosw mroute_encap_protosw = 863 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 864 mroute_encap_input, 0, 0, rip_ctloutput, 865 0, 866 0, 0, 0, 0, 867 &rip_usrreqs 868 }; 869 870 /* 871 * Add a vif to the vif table 872 */ 873 static int 874 add_vif(struct vifctl *vifcp) 875 { 876 struct vif *vifp = viftable + vifcp->vifc_vifi; 877 struct sockaddr_in sin = {sizeof sin, AF_INET}; 878 struct ifaddr *ifa; 879 struct ifnet *ifp; 880 int error; 881 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 882 883 VIF_LOCK(); 884 if (vifcp->vifc_vifi >= MAXVIFS) { 885 VIF_UNLOCK(); 886 return EINVAL; 887 } 888 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) { 889 VIF_UNLOCK(); 890 return EADDRINUSE; 891 } 892 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) { 893 VIF_UNLOCK(); 894 return EADDRNOTAVAIL; 895 } 896 897 /* Find the interface with an address in AF_INET family */ 898 #ifdef PIM 899 if (vifcp->vifc_flags & VIFF_REGISTER) { 900 /* 901 * XXX: Because VIFF_REGISTER does not really need a valid 902 * local interface (e.g. it could be 127.0.0.2), we don't 903 * check its address. 904 */ 905 ifp = NULL; 906 } else 907 #endif 908 { 909 sin.sin_addr = vifcp->vifc_lcl_addr; 910 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 911 if (ifa == NULL) { 912 VIF_UNLOCK(); 913 return EADDRNOTAVAIL; 914 } 915 ifp = ifa->ifa_ifp; 916 } 917 918 if (vifcp->vifc_flags & VIFF_TUNNEL) { 919 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 920 /* 921 * An encapsulating tunnel is wanted. Tell 922 * mroute_encap_input() to start paying attention 923 * to encapsulated packets. 924 */ 925 if (encap_cookie == NULL) { 926 int i; 927 928 encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 929 mroute_encapcheck, 930 (struct protosw *)&mroute_encap_protosw, NULL); 931 932 if (encap_cookie == NULL) { 933 printf("ip_mroute: unable to attach encap\n"); 934 VIF_UNLOCK(); 935 return EIO; /* XXX */ 936 } 937 for (i = 0; i < MAXVIFS; ++i) { 938 if_initname(&multicast_decap_if[i], "mdecap", i); 939 } 940 } 941 /* 942 * Set interface to fake encapsulator interface 943 */ 944 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 945 /* 946 * Prepare cached route entry 947 */ 948 bzero(&vifp->v_route, sizeof(vifp->v_route)); 949 } else { 950 log(LOG_ERR, "source routed tunnels not supported\n"); 951 VIF_UNLOCK(); 952 return EOPNOTSUPP; 953 } 954 #ifdef PIM 955 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 956 ifp = &multicast_register_if; 957 if (mrtdebug) 958 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 959 (void *)&multicast_register_if); 960 if (reg_vif_num == VIFI_INVALID) { 961 if_initname(&multicast_register_if, "register_vif", 0); 962 multicast_register_if.if_flags = IFF_LOOPBACK; 963 bzero(&vifp->v_route, sizeof(vifp->v_route)); 964 reg_vif_num = vifcp->vifc_vifi; 965 } 966 #endif 967 } else { /* Make sure the interface supports multicast */ 968 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 969 VIF_UNLOCK(); 970 return EOPNOTSUPP; 971 } 972 973 /* Enable promiscuous reception of all IP multicasts from the if */ 974 error = if_allmulti(ifp, 1); 975 if (error) { 976 VIF_UNLOCK(); 977 return error; 978 } 979 } 980 981 /* define parameters for the tbf structure */ 982 vifp->v_tbf = v_tbf; 983 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 984 vifp->v_tbf->tbf_n_tok = 0; 985 vifp->v_tbf->tbf_q_len = 0; 986 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 987 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 988 989 vifp->v_flags = vifcp->vifc_flags; 990 vifp->v_threshold = vifcp->vifc_threshold; 991 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 992 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 993 vifp->v_ifp = ifp; 994 /* scaling up here allows division by 1024 in critical code */ 995 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 996 vifp->v_rsvp_on = 0; 997 vifp->v_rsvpd = NULL; 998 /* initialize per vif pkt counters */ 999 vifp->v_pkt_in = 0; 1000 vifp->v_pkt_out = 0; 1001 vifp->v_bytes_in = 0; 1002 vifp->v_bytes_out = 0; 1003 1004 /* Adjust numvifs up if the vifi is higher than numvifs */ 1005 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 1006 1007 VIF_UNLOCK(); 1008 1009 if (mrtdebug) 1010 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x, rate %d\n", 1011 vifcp->vifc_vifi, 1012 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 1013 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 1014 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 1015 vifcp->vifc_threshold, 1016 vifcp->vifc_rate_limit); 1017 1018 return 0; 1019 } 1020 1021 /* 1022 * Delete a vif from the vif table 1023 */ 1024 static int 1025 del_vif(vifi_t vifi) 1026 { 1027 struct vif *vifp; 1028 1029 VIF_LOCK(); 1030 1031 if (vifi >= numvifs) { 1032 VIF_UNLOCK(); 1033 return EINVAL; 1034 } 1035 vifp = &viftable[vifi]; 1036 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) { 1037 VIF_UNLOCK(); 1038 return EADDRNOTAVAIL; 1039 } 1040 1041 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 1042 if_allmulti(vifp->v_ifp, 0); 1043 1044 if (vifp == last_encap_vif) { 1045 last_encap_vif = NULL; 1046 last_encap_src = INADDR_ANY; 1047 } 1048 1049 /* 1050 * Free packets queued at the interface 1051 */ 1052 while (vifp->v_tbf->tbf_q) { 1053 struct mbuf *m = vifp->v_tbf->tbf_q; 1054 1055 vifp->v_tbf->tbf_q = m->m_act; 1056 m_freem(m); 1057 } 1058 1059 #ifdef PIM 1060 if (vifp->v_flags & VIFF_REGISTER) 1061 reg_vif_num = VIFI_INVALID; 1062 #endif 1063 1064 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 1065 bzero((caddr_t)vifp, sizeof (*vifp)); 1066 1067 if (mrtdebug) 1068 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1069 1070 /* Adjust numvifs down */ 1071 for (vifi = numvifs; vifi > 0; vifi--) 1072 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1073 break; 1074 numvifs = vifi; 1075 1076 VIF_UNLOCK(); 1077 1078 return 0; 1079 } 1080 1081 /* 1082 * update an mfc entry without resetting counters and S,G addresses. 1083 */ 1084 static void 1085 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1086 { 1087 int i; 1088 1089 rt->mfc_parent = mfccp->mfcc_parent; 1090 for (i = 0; i < numvifs; i++) { 1091 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1092 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1093 MRT_MFC_FLAGS_ALL; 1094 } 1095 /* set the RP address */ 1096 if (mrt_api_config & MRT_MFC_RP) 1097 rt->mfc_rp = mfccp->mfcc_rp; 1098 else 1099 rt->mfc_rp.s_addr = INADDR_ANY; 1100 } 1101 1102 /* 1103 * fully initialize an mfc entry from the parameter. 1104 */ 1105 static void 1106 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1107 { 1108 rt->mfc_origin = mfccp->mfcc_origin; 1109 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1110 1111 update_mfc_params(rt, mfccp); 1112 1113 /* initialize pkt counters per src-grp */ 1114 rt->mfc_pkt_cnt = 0; 1115 rt->mfc_byte_cnt = 0; 1116 rt->mfc_wrong_if = 0; 1117 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1118 } 1119 1120 1121 /* 1122 * Add an mfc entry 1123 */ 1124 static int 1125 add_mfc(struct mfcctl2 *mfccp) 1126 { 1127 struct mfc *rt; 1128 u_long hash; 1129 struct rtdetq *rte; 1130 u_short nstl; 1131 1132 VIF_LOCK(); 1133 MFC_LOCK(); 1134 1135 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1136 1137 /* If an entry already exists, just update the fields */ 1138 if (rt) { 1139 if (mrtdebug & DEBUG_MFC) 1140 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1141 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1142 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1143 mfccp->mfcc_parent); 1144 1145 update_mfc_params(rt, mfccp); 1146 MFC_UNLOCK(); 1147 VIF_UNLOCK(); 1148 return 0; 1149 } 1150 1151 /* 1152 * Find the entry for which the upcall was made and update 1153 */ 1154 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1155 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1156 1157 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1158 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1159 (rt->mfc_stall != NULL)) { 1160 1161 if (nstl++) 1162 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1163 "multiple kernel entries", 1164 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1165 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1166 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1167 1168 if (mrtdebug & DEBUG_MFC) 1169 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1170 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1171 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1172 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1173 1174 init_mfc_params(rt, mfccp); 1175 1176 rt->mfc_expire = 0; /* Don't clean this guy up */ 1177 nexpire[hash]--; 1178 1179 /* free packets Qed at the end of this entry */ 1180 for (rte = rt->mfc_stall; rte != NULL; ) { 1181 struct rtdetq *n = rte->next; 1182 1183 ip_mdq(rte->m, rte->ifp, rt, -1); 1184 m_freem(rte->m); 1185 free(rte, M_MRTABLE); 1186 rte = n; 1187 } 1188 rt->mfc_stall = NULL; 1189 } 1190 } 1191 1192 /* 1193 * It is possible that an entry is being inserted without an upcall 1194 */ 1195 if (nstl == 0) { 1196 if (mrtdebug & DEBUG_MFC) 1197 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1198 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1199 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1200 mfccp->mfcc_parent); 1201 1202 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1203 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1204 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1205 init_mfc_params(rt, mfccp); 1206 if (rt->mfc_expire) 1207 nexpire[hash]--; 1208 rt->mfc_expire = 0; 1209 break; /* XXX */ 1210 } 1211 } 1212 if (rt == NULL) { /* no upcall, so make a new entry */ 1213 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1214 if (rt == NULL) { 1215 MFC_UNLOCK(); 1216 VIF_UNLOCK(); 1217 return ENOBUFS; 1218 } 1219 1220 init_mfc_params(rt, mfccp); 1221 rt->mfc_expire = 0; 1222 rt->mfc_stall = NULL; 1223 1224 rt->mfc_bw_meter = NULL; 1225 /* insert new entry at head of hash chain */ 1226 rt->mfc_next = mfctable[hash]; 1227 mfctable[hash] = rt; 1228 } 1229 } 1230 MFC_UNLOCK(); 1231 VIF_UNLOCK(); 1232 return 0; 1233 } 1234 1235 /* 1236 * Delete an mfc entry 1237 */ 1238 static int 1239 del_mfc(struct mfcctl2 *mfccp) 1240 { 1241 struct in_addr origin; 1242 struct in_addr mcastgrp; 1243 struct mfc *rt; 1244 struct mfc **nptr; 1245 u_long hash; 1246 struct bw_meter *list; 1247 1248 origin = mfccp->mfcc_origin; 1249 mcastgrp = mfccp->mfcc_mcastgrp; 1250 1251 if (mrtdebug & DEBUG_MFC) 1252 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1253 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1254 1255 MFC_LOCK(); 1256 1257 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1258 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1259 if (origin.s_addr == rt->mfc_origin.s_addr && 1260 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1261 rt->mfc_stall == NULL) 1262 break; 1263 if (rt == NULL) { 1264 MFC_UNLOCK(); 1265 return EADDRNOTAVAIL; 1266 } 1267 1268 *nptr = rt->mfc_next; 1269 1270 /* 1271 * free the bw_meter entries 1272 */ 1273 list = rt->mfc_bw_meter; 1274 rt->mfc_bw_meter = NULL; 1275 1276 free(rt, M_MRTABLE); 1277 1278 free_bw_list(list); 1279 1280 MFC_UNLOCK(); 1281 1282 return 0; 1283 } 1284 1285 /* 1286 * Send a message to mrouted on the multicast routing socket 1287 */ 1288 static int 1289 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1290 { 1291 if (s) { 1292 if (sbappendaddr(&s->so_rcv, (struct sockaddr *)src, mm, NULL) != 0) { 1293 sorwakeup(s); 1294 return 0; 1295 } 1296 } 1297 m_freem(mm); 1298 return -1; 1299 } 1300 1301 /* 1302 * IP multicast forwarding function. This function assumes that the packet 1303 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1304 * pointed to by "ifp", and the packet is to be relayed to other networks 1305 * that have members of the packet's destination IP multicast group. 1306 * 1307 * The packet is returned unscathed to the caller, unless it is 1308 * erroneous, in which case a non-zero return value tells the caller to 1309 * discard it. 1310 */ 1311 1312 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1313 1314 static int 1315 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1316 struct ip_moptions *imo) 1317 { 1318 struct mfc *rt; 1319 int error; 1320 vifi_t vifi; 1321 1322 if (mrtdebug & DEBUG_FORWARD) 1323 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1324 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1325 (void *)ifp); 1326 1327 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1328 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1329 /* 1330 * Packet arrived via a physical interface or 1331 * an encapsulated tunnel or a register_vif. 1332 */ 1333 } else { 1334 /* 1335 * Packet arrived through a source-route tunnel. 1336 * Source-route tunnels are no longer supported. 1337 */ 1338 static int last_log; 1339 if (last_log != time_second) { 1340 last_log = time_second; 1341 log(LOG_ERR, 1342 "ip_mforward: received source-routed packet from %lx\n", 1343 (u_long)ntohl(ip->ip_src.s_addr)); 1344 } 1345 return 1; 1346 } 1347 1348 VIF_LOCK(); 1349 MFC_LOCK(); 1350 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1351 if (ip->ip_ttl < 255) 1352 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1353 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1354 struct vif *vifp = viftable + vifi; 1355 1356 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1357 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1358 vifi, 1359 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1360 vifp->v_ifp->if_xname); 1361 } 1362 error = ip_mdq(m, ifp, NULL, vifi); 1363 MFC_UNLOCK(); 1364 VIF_UNLOCK(); 1365 return error; 1366 } 1367 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1368 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1369 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1370 if (!imo) 1371 printf("In fact, no options were specified at all\n"); 1372 } 1373 1374 /* 1375 * Don't forward a packet with time-to-live of zero or one, 1376 * or a packet destined to a local-only group. 1377 */ 1378 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) { 1379 MFC_UNLOCK(); 1380 VIF_UNLOCK(); 1381 return 0; 1382 } 1383 1384 /* 1385 * Determine forwarding vifs from the forwarding cache table 1386 */ 1387 ++mrtstat.mrts_mfc_lookups; 1388 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1389 1390 /* Entry exists, so forward if necessary */ 1391 if (rt != NULL) { 1392 error = ip_mdq(m, ifp, rt, -1); 1393 MFC_UNLOCK(); 1394 VIF_UNLOCK(); 1395 return error; 1396 } else { 1397 /* 1398 * If we don't have a route for packet's origin, 1399 * Make a copy of the packet & send message to routing daemon 1400 */ 1401 1402 struct mbuf *mb0; 1403 struct rtdetq *rte; 1404 u_long hash; 1405 int hlen = ip->ip_hl << 2; 1406 1407 ++mrtstat.mrts_mfc_misses; 1408 1409 mrtstat.mrts_no_route++; 1410 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1411 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1412 (u_long)ntohl(ip->ip_src.s_addr), 1413 (u_long)ntohl(ip->ip_dst.s_addr)); 1414 1415 /* 1416 * Allocate mbufs early so that we don't do extra work if we are 1417 * just going to fail anyway. Make sure to pullup the header so 1418 * that other people can't step on it. 1419 */ 1420 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1421 if (rte == NULL) { 1422 MFC_UNLOCK(); 1423 VIF_UNLOCK(); 1424 return ENOBUFS; 1425 } 1426 mb0 = m_copypacket(m, M_DONTWAIT); 1427 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1428 mb0 = m_pullup(mb0, hlen); 1429 if (mb0 == NULL) { 1430 free(rte, M_MRTABLE); 1431 MFC_UNLOCK(); 1432 VIF_UNLOCK(); 1433 return ENOBUFS; 1434 } 1435 1436 /* is there an upcall waiting for this flow ? */ 1437 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1438 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1439 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1440 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1441 (rt->mfc_stall != NULL)) 1442 break; 1443 } 1444 1445 if (rt == NULL) { 1446 int i; 1447 struct igmpmsg *im; 1448 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1449 struct mbuf *mm; 1450 1451 /* 1452 * Locate the vifi for the incoming interface for this packet. 1453 * If none found, drop packet. 1454 */ 1455 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1456 ; 1457 if (vifi >= numvifs) /* vif not found, drop packet */ 1458 goto non_fatal; 1459 1460 /* no upcall, so make a new entry */ 1461 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1462 if (rt == NULL) 1463 goto fail; 1464 /* Make a copy of the header to send to the user level process */ 1465 mm = m_copy(mb0, 0, hlen); 1466 if (mm == NULL) 1467 goto fail1; 1468 1469 /* 1470 * Send message to routing daemon to install 1471 * a route into the kernel table 1472 */ 1473 1474 im = mtod(mm, struct igmpmsg *); 1475 im->im_msgtype = IGMPMSG_NOCACHE; 1476 im->im_mbz = 0; 1477 im->im_vif = vifi; 1478 1479 mrtstat.mrts_upcalls++; 1480 1481 k_igmpsrc.sin_addr = ip->ip_src; 1482 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1483 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1484 ++mrtstat.mrts_upq_sockfull; 1485 fail1: 1486 free(rt, M_MRTABLE); 1487 fail: 1488 free(rte, M_MRTABLE); 1489 m_freem(mb0); 1490 MFC_UNLOCK(); 1491 VIF_UNLOCK(); 1492 return ENOBUFS; 1493 } 1494 1495 /* insert new entry at head of hash chain */ 1496 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1497 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1498 rt->mfc_expire = UPCALL_EXPIRE; 1499 nexpire[hash]++; 1500 for (i = 0; i < numvifs; i++) { 1501 rt->mfc_ttls[i] = 0; 1502 rt->mfc_flags[i] = 0; 1503 } 1504 rt->mfc_parent = -1; 1505 1506 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1507 1508 rt->mfc_bw_meter = NULL; 1509 1510 /* link into table */ 1511 rt->mfc_next = mfctable[hash]; 1512 mfctable[hash] = rt; 1513 rt->mfc_stall = rte; 1514 1515 } else { 1516 /* determine if q has overflowed */ 1517 int npkts = 0; 1518 struct rtdetq **p; 1519 1520 /* 1521 * XXX ouch! we need to append to the list, but we 1522 * only have a pointer to the front, so we have to 1523 * scan the entire list every time. 1524 */ 1525 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1526 npkts++; 1527 1528 if (npkts > MAX_UPQ) { 1529 mrtstat.mrts_upq_ovflw++; 1530 non_fatal: 1531 free(rte, M_MRTABLE); 1532 m_freem(mb0); 1533 MFC_UNLOCK(); 1534 VIF_UNLOCK(); 1535 return 0; 1536 } 1537 1538 /* Add this entry to the end of the queue */ 1539 *p = rte; 1540 } 1541 1542 rte->m = mb0; 1543 rte->ifp = ifp; 1544 rte->next = NULL; 1545 1546 MFC_UNLOCK(); 1547 VIF_UNLOCK(); 1548 1549 return 0; 1550 } 1551 } 1552 1553 /* 1554 * Clean up the cache entry if upcall is not serviced 1555 */ 1556 static void 1557 expire_upcalls(void *unused) 1558 { 1559 struct rtdetq *rte; 1560 struct mfc *mfc, **nptr; 1561 int i; 1562 1563 MFC_LOCK(); 1564 for (i = 0; i < MFCTBLSIZ; i++) { 1565 if (nexpire[i] == 0) 1566 continue; 1567 nptr = &mfctable[i]; 1568 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1569 /* 1570 * Skip real cache entries 1571 * Make sure it wasn't marked to not expire (shouldn't happen) 1572 * If it expires now 1573 */ 1574 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1575 --mfc->mfc_expire == 0) { 1576 if (mrtdebug & DEBUG_EXPIRE) 1577 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1578 (u_long)ntohl(mfc->mfc_origin.s_addr), 1579 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1580 /* 1581 * drop all the packets 1582 * free the mbuf with the pkt, if, timing info 1583 */ 1584 for (rte = mfc->mfc_stall; rte; ) { 1585 struct rtdetq *n = rte->next; 1586 1587 m_freem(rte->m); 1588 free(rte, M_MRTABLE); 1589 rte = n; 1590 } 1591 ++mrtstat.mrts_cache_cleanups; 1592 nexpire[i]--; 1593 1594 /* 1595 * free the bw_meter entries 1596 */ 1597 while (mfc->mfc_bw_meter != NULL) { 1598 struct bw_meter *x = mfc->mfc_bw_meter; 1599 1600 mfc->mfc_bw_meter = x->bm_mfc_next; 1601 free(x, M_BWMETER); 1602 } 1603 1604 *nptr = mfc->mfc_next; 1605 free(mfc, M_MRTABLE); 1606 } else { 1607 nptr = &mfc->mfc_next; 1608 } 1609 } 1610 } 1611 MFC_UNLOCK(); 1612 1613 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1614 } 1615 1616 /* 1617 * Packet forwarding routine once entry in the cache is made 1618 */ 1619 static int 1620 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1621 { 1622 struct ip *ip = mtod(m, struct ip *); 1623 vifi_t vifi; 1624 int plen = ip->ip_len; 1625 1626 VIF_LOCK_ASSERT(); 1627 /* 1628 * Macro to send packet on vif. Since RSVP packets don't get counted on 1629 * input, they shouldn't get counted on output, so statistics keeping is 1630 * separate. 1631 */ 1632 #define MC_SEND(ip,vifp,m) { \ 1633 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1634 encap_send((ip), (vifp), (m)); \ 1635 else \ 1636 phyint_send((ip), (vifp), (m)); \ 1637 } 1638 1639 /* 1640 * If xmt_vif is not -1, send on only the requested vif. 1641 * 1642 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1643 */ 1644 if (xmt_vif < numvifs) { 1645 #ifdef PIM 1646 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1647 pim_register_send(ip, viftable + xmt_vif, m, rt); 1648 else 1649 #endif 1650 MC_SEND(ip, viftable + xmt_vif, m); 1651 return 1; 1652 } 1653 1654 /* 1655 * Don't forward if it didn't arrive from the parent vif for its origin. 1656 */ 1657 vifi = rt->mfc_parent; 1658 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1659 /* came in the wrong interface */ 1660 if (mrtdebug & DEBUG_FORWARD) 1661 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1662 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1663 ++mrtstat.mrts_wrong_if; 1664 ++rt->mfc_wrong_if; 1665 /* 1666 * If we are doing PIM assert processing, send a message 1667 * to the routing daemon. 1668 * 1669 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1670 * can complete the SPT switch, regardless of the type 1671 * of the iif (broadcast media, GRE tunnel, etc). 1672 */ 1673 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1674 struct timeval now; 1675 u_long delta; 1676 1677 #ifdef PIM 1678 if (ifp == &multicast_register_if) 1679 pimstat.pims_rcv_registers_wrongiif++; 1680 #endif 1681 1682 /* Get vifi for the incoming packet */ 1683 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1684 ; 1685 if (vifi >= numvifs) 1686 return 0; /* The iif is not found: ignore the packet. */ 1687 1688 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1689 return 0; /* WRONGVIF disabled: ignore the packet */ 1690 1691 GET_TIME(now); 1692 1693 TV_DELTA(rt->mfc_last_assert, now, delta); 1694 1695 if (delta > ASSERT_MSG_TIME) { 1696 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1697 struct igmpmsg *im; 1698 int hlen = ip->ip_hl << 2; 1699 struct mbuf *mm = m_copy(m, 0, hlen); 1700 1701 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1702 mm = m_pullup(mm, hlen); 1703 if (mm == NULL) 1704 return ENOBUFS; 1705 1706 rt->mfc_last_assert = now; 1707 1708 im = mtod(mm, struct igmpmsg *); 1709 im->im_msgtype = IGMPMSG_WRONGVIF; 1710 im->im_mbz = 0; 1711 im->im_vif = vifi; 1712 1713 mrtstat.mrts_upcalls++; 1714 1715 k_igmpsrc.sin_addr = im->im_src; 1716 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1717 log(LOG_WARNING, 1718 "ip_mforward: ip_mrouter socket queue full\n"); 1719 ++mrtstat.mrts_upq_sockfull; 1720 return ENOBUFS; 1721 } 1722 } 1723 } 1724 return 0; 1725 } 1726 1727 /* If I sourced this packet, it counts as output, else it was input. */ 1728 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1729 viftable[vifi].v_pkt_out++; 1730 viftable[vifi].v_bytes_out += plen; 1731 } else { 1732 viftable[vifi].v_pkt_in++; 1733 viftable[vifi].v_bytes_in += plen; 1734 } 1735 rt->mfc_pkt_cnt++; 1736 rt->mfc_byte_cnt += plen; 1737 1738 /* 1739 * For each vif, decide if a copy of the packet should be forwarded. 1740 * Forward if: 1741 * - the ttl exceeds the vif's threshold 1742 * - there are group members downstream on interface 1743 */ 1744 for (vifi = 0; vifi < numvifs; vifi++) 1745 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1746 viftable[vifi].v_pkt_out++; 1747 viftable[vifi].v_bytes_out += plen; 1748 #ifdef PIM 1749 if (viftable[vifi].v_flags & VIFF_REGISTER) 1750 pim_register_send(ip, viftable + vifi, m, rt); 1751 else 1752 #endif 1753 MC_SEND(ip, viftable+vifi, m); 1754 } 1755 1756 /* 1757 * Perform upcall-related bw measuring. 1758 */ 1759 if (rt->mfc_bw_meter != NULL) { 1760 struct bw_meter *x; 1761 struct timeval now; 1762 1763 GET_TIME(now); 1764 MFC_LOCK_ASSERT(); 1765 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1766 bw_meter_receive_packet(x, plen, &now); 1767 } 1768 1769 return 0; 1770 } 1771 1772 /* 1773 * check if a vif number is legal/ok. This is used by ip_output. 1774 */ 1775 static int 1776 X_legal_vif_num(int vif) 1777 { 1778 /* XXX unlocked, matter? */ 1779 return (vif >= 0 && vif < numvifs); 1780 } 1781 1782 /* 1783 * Return the local address used by this vif 1784 */ 1785 static u_long 1786 X_ip_mcast_src(int vifi) 1787 { 1788 /* XXX unlocked, matter? */ 1789 if (vifi >= 0 && vifi < numvifs) 1790 return viftable[vifi].v_lcl_addr.s_addr; 1791 else 1792 return INADDR_ANY; 1793 } 1794 1795 static void 1796 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1797 { 1798 struct mbuf *mb_copy; 1799 int hlen = ip->ip_hl << 2; 1800 1801 VIF_LOCK_ASSERT(); 1802 1803 /* 1804 * Make a new reference to the packet; make sure that 1805 * the IP header is actually copied, not just referenced, 1806 * so that ip_output() only scribbles on the copy. 1807 */ 1808 mb_copy = m_copypacket(m, M_DONTWAIT); 1809 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1810 mb_copy = m_pullup(mb_copy, hlen); 1811 if (mb_copy == NULL) 1812 return; 1813 1814 if (vifp->v_rate_limit == 0) 1815 tbf_send_packet(vifp, mb_copy); 1816 else 1817 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1818 } 1819 1820 static void 1821 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1822 { 1823 struct mbuf *mb_copy; 1824 struct ip *ip_copy; 1825 int i, len = ip->ip_len; 1826 1827 VIF_LOCK_ASSERT(); 1828 1829 /* Take care of delayed checksums */ 1830 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 1831 in_delayed_cksum(m); 1832 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 1833 } 1834 1835 /* 1836 * copy the old packet & pullup its IP header into the 1837 * new mbuf so we can modify it. Try to fill the new 1838 * mbuf since if we don't the ethernet driver will. 1839 */ 1840 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1841 if (mb_copy == NULL) 1842 return; 1843 #ifdef MAC 1844 mac_create_mbuf_multicast_encap(m, vifp->v_ifp, mb_copy); 1845 #endif 1846 mb_copy->m_data += max_linkhdr; 1847 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1848 1849 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) { 1850 m_freem(mb_copy); 1851 return; 1852 } 1853 i = MHLEN - M_LEADINGSPACE(mb_copy); 1854 if (i > len) 1855 i = len; 1856 mb_copy = m_pullup(mb_copy, i); 1857 if (mb_copy == NULL) 1858 return; 1859 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1860 1861 /* 1862 * fill in the encapsulating IP header. 1863 */ 1864 ip_copy = mtod(mb_copy, struct ip *); 1865 *ip_copy = multicast_encap_iphdr; 1866 #ifdef RANDOM_IP_ID 1867 ip_copy->ip_id = ip_randomid(); 1868 #else 1869 ip_copy->ip_id = htons(ip_id++); 1870 #endif 1871 ip_copy->ip_len += len; 1872 ip_copy->ip_src = vifp->v_lcl_addr; 1873 ip_copy->ip_dst = vifp->v_rmt_addr; 1874 1875 /* 1876 * turn the encapsulated IP header back into a valid one. 1877 */ 1878 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1879 --ip->ip_ttl; 1880 ip->ip_len = htons(ip->ip_len); 1881 ip->ip_off = htons(ip->ip_off); 1882 ip->ip_sum = 0; 1883 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1884 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1885 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1886 1887 if (vifp->v_rate_limit == 0) 1888 tbf_send_packet(vifp, mb_copy); 1889 else 1890 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1891 } 1892 1893 /* 1894 * Token bucket filter module 1895 */ 1896 1897 static void 1898 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_long p_len) 1899 { 1900 struct tbf *t = vifp->v_tbf; 1901 1902 VIF_LOCK_ASSERT(); 1903 1904 if (p_len > MAX_BKT_SIZE) { /* drop if packet is too large */ 1905 mrtstat.mrts_pkt2large++; 1906 m_freem(m); 1907 return; 1908 } 1909 1910 tbf_update_tokens(vifp); 1911 1912 if (t->tbf_q_len == 0) { /* queue empty... */ 1913 if (p_len <= t->tbf_n_tok) { /* send packet if enough tokens */ 1914 t->tbf_n_tok -= p_len; 1915 tbf_send_packet(vifp, m); 1916 } else { /* no, queue packet and try later */ 1917 tbf_queue(vifp, m); 1918 callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, 1919 tbf_reprocess_q, vifp); 1920 } 1921 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1922 /* finite queue length, so queue pkts and process queue */ 1923 tbf_queue(vifp, m); 1924 tbf_process_q(vifp); 1925 } else { 1926 /* queue full, try to dq and queue and process */ 1927 if (!tbf_dq_sel(vifp, ip)) { 1928 mrtstat.mrts_q_overflow++; 1929 m_freem(m); 1930 } else { 1931 tbf_queue(vifp, m); 1932 tbf_process_q(vifp); 1933 } 1934 } 1935 } 1936 1937 /* 1938 * adds a packet to the queue at the interface 1939 */ 1940 static void 1941 tbf_queue(struct vif *vifp, struct mbuf *m) 1942 { 1943 struct tbf *t = vifp->v_tbf; 1944 1945 VIF_LOCK_ASSERT(); 1946 1947 if (t->tbf_t == NULL) /* Queue was empty */ 1948 t->tbf_q = m; 1949 else /* Insert at tail */ 1950 t->tbf_t->m_act = m; 1951 1952 t->tbf_t = m; /* Set new tail pointer */ 1953 1954 #ifdef DIAGNOSTIC 1955 /* Make sure we didn't get fed a bogus mbuf */ 1956 if (m->m_act) 1957 panic("tbf_queue: m_act"); 1958 #endif 1959 m->m_act = NULL; 1960 1961 t->tbf_q_len++; 1962 } 1963 1964 /* 1965 * processes the queue at the interface 1966 */ 1967 static void 1968 tbf_process_q(struct vif *vifp) 1969 { 1970 struct tbf *t = vifp->v_tbf; 1971 1972 VIF_LOCK_ASSERT(); 1973 1974 /* loop through the queue at the interface and send as many packets 1975 * as possible 1976 */ 1977 while (t->tbf_q_len > 0) { 1978 struct mbuf *m = t->tbf_q; 1979 int len = mtod(m, struct ip *)->ip_len; 1980 1981 /* determine if the packet can be sent */ 1982 if (len > t->tbf_n_tok) /* not enough tokens, we are done */ 1983 break; 1984 /* ok, reduce no of tokens, dequeue and send the packet. */ 1985 t->tbf_n_tok -= len; 1986 1987 t->tbf_q = m->m_act; 1988 if (--t->tbf_q_len == 0) 1989 t->tbf_t = NULL; 1990 1991 m->m_act = NULL; 1992 tbf_send_packet(vifp, m); 1993 } 1994 } 1995 1996 static void 1997 tbf_reprocess_q(void *xvifp) 1998 { 1999 struct vif *vifp = xvifp; 2000 2001 if (ip_mrouter == NULL) 2002 return; 2003 VIF_LOCK(); 2004 tbf_update_tokens(vifp); 2005 tbf_process_q(vifp); 2006 if (vifp->v_tbf->tbf_q_len) 2007 callout_reset(&tbf_reprocess_ch, TBF_REPROCESS, tbf_reprocess_q, vifp); 2008 VIF_UNLOCK(); 2009 } 2010 2011 /* function that will selectively discard a member of the queue 2012 * based on the precedence value and the priority 2013 */ 2014 static int 2015 tbf_dq_sel(struct vif *vifp, struct ip *ip) 2016 { 2017 u_int p; 2018 struct mbuf *m, *last; 2019 struct mbuf **np; 2020 struct tbf *t = vifp->v_tbf; 2021 2022 VIF_LOCK_ASSERT(); 2023 2024 p = priority(vifp, ip); 2025 2026 np = &t->tbf_q; 2027 last = NULL; 2028 while ((m = *np) != NULL) { 2029 if (p > priority(vifp, mtod(m, struct ip *))) { 2030 *np = m->m_act; 2031 /* If we're removing the last packet, fix the tail pointer */ 2032 if (m == t->tbf_t) 2033 t->tbf_t = last; 2034 m_freem(m); 2035 /* It's impossible for the queue to be empty, but check anyways. */ 2036 if (--t->tbf_q_len == 0) 2037 t->tbf_t = NULL; 2038 mrtstat.mrts_drop_sel++; 2039 return 1; 2040 } 2041 np = &m->m_act; 2042 last = m; 2043 } 2044 return 0; 2045 } 2046 2047 static void 2048 tbf_send_packet(struct vif *vifp, struct mbuf *m) 2049 { 2050 VIF_LOCK_ASSERT(); 2051 2052 if (vifp->v_flags & VIFF_TUNNEL) /* If tunnel options */ 2053 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 2054 else { 2055 struct ip_moptions imo; 2056 int error; 2057 static struct route ro; /* XXX check this */ 2058 2059 imo.imo_multicast_ifp = vifp->v_ifp; 2060 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 2061 imo.imo_multicast_loop = 1; 2062 imo.imo_multicast_vif = -1; 2063 2064 /* 2065 * Re-entrancy should not be a problem here, because 2066 * the packets that we send out and are looped back at us 2067 * should get rejected because they appear to come from 2068 * the loopback interface, thus preventing looping. 2069 */ 2070 error = ip_output(m, NULL, &ro, IP_FORWARDING, &imo, NULL); 2071 2072 if (mrtdebug & DEBUG_XMIT) 2073 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 2074 (int)(vifp - viftable), error); 2075 } 2076 } 2077 2078 /* determine the current time and then 2079 * the elapsed time (between the last time and time now) 2080 * in milliseconds & update the no. of tokens in the bucket 2081 */ 2082 static void 2083 tbf_update_tokens(struct vif *vifp) 2084 { 2085 struct timeval tp; 2086 u_long tm; 2087 struct tbf *t = vifp->v_tbf; 2088 2089 VIF_LOCK_ASSERT(); 2090 2091 GET_TIME(tp); 2092 2093 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 2094 2095 /* 2096 * This formula is actually 2097 * "time in seconds" * "bytes/second". 2098 * 2099 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2100 * 2101 * The (1000/1024) was introduced in add_vif to optimize 2102 * this divide into a shift. 2103 */ 2104 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 2105 t->tbf_last_pkt_t = tp; 2106 2107 if (t->tbf_n_tok > MAX_BKT_SIZE) 2108 t->tbf_n_tok = MAX_BKT_SIZE; 2109 } 2110 2111 static int 2112 priority(struct vif *vifp, struct ip *ip) 2113 { 2114 int prio = 50; /* the lowest priority -- default case */ 2115 2116 /* temporary hack; may add general packet classifier some day */ 2117 2118 /* 2119 * The UDP port space is divided up into four priority ranges: 2120 * [0, 16384) : unclassified - lowest priority 2121 * [16384, 32768) : audio - highest priority 2122 * [32768, 49152) : whiteboard - medium priority 2123 * [49152, 65536) : video - low priority 2124 * 2125 * Everything else gets lowest priority. 2126 */ 2127 if (ip->ip_p == IPPROTO_UDP) { 2128 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2129 switch (ntohs(udp->uh_dport) & 0xc000) { 2130 case 0x4000: 2131 prio = 70; 2132 break; 2133 case 0x8000: 2134 prio = 60; 2135 break; 2136 case 0xc000: 2137 prio = 55; 2138 break; 2139 } 2140 } 2141 return prio; 2142 } 2143 2144 /* 2145 * End of token bucket filter modifications 2146 */ 2147 2148 static int 2149 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 2150 { 2151 int error, vifi; 2152 2153 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2154 return EOPNOTSUPP; 2155 2156 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 2157 if (error) 2158 return error; 2159 2160 VIF_LOCK(); 2161 2162 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 2163 VIF_UNLOCK(); 2164 return EADDRNOTAVAIL; 2165 } 2166 2167 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 2168 /* Check if socket is available. */ 2169 if (viftable[vifi].v_rsvpd != NULL) { 2170 VIF_UNLOCK(); 2171 return EADDRINUSE; 2172 } 2173 2174 viftable[vifi].v_rsvpd = so; 2175 /* This may seem silly, but we need to be sure we don't over-increment 2176 * the RSVP counter, in case something slips up. 2177 */ 2178 if (!viftable[vifi].v_rsvp_on) { 2179 viftable[vifi].v_rsvp_on = 1; 2180 rsvp_on++; 2181 } 2182 } else { /* must be VIF_OFF */ 2183 /* 2184 * XXX as an additional consistency check, one could make sure 2185 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 2186 * first parameter is pretty useless. 2187 */ 2188 viftable[vifi].v_rsvpd = NULL; 2189 /* 2190 * This may seem silly, but we need to be sure we don't over-decrement 2191 * the RSVP counter, in case something slips up. 2192 */ 2193 if (viftable[vifi].v_rsvp_on) { 2194 viftable[vifi].v_rsvp_on = 0; 2195 rsvp_on--; 2196 } 2197 } 2198 VIF_UNLOCK(); 2199 return 0; 2200 } 2201 2202 static void 2203 X_ip_rsvp_force_done(struct socket *so) 2204 { 2205 int vifi; 2206 2207 /* Don't bother if it is not the right type of socket. */ 2208 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2209 return; 2210 2211 VIF_LOCK(); 2212 2213 /* The socket may be attached to more than one vif...this 2214 * is perfectly legal. 2215 */ 2216 for (vifi = 0; vifi < numvifs; vifi++) { 2217 if (viftable[vifi].v_rsvpd == so) { 2218 viftable[vifi].v_rsvpd = NULL; 2219 /* This may seem silly, but we need to be sure we don't 2220 * over-decrement the RSVP counter, in case something slips up. 2221 */ 2222 if (viftable[vifi].v_rsvp_on) { 2223 viftable[vifi].v_rsvp_on = 0; 2224 rsvp_on--; 2225 } 2226 } 2227 } 2228 2229 VIF_UNLOCK(); 2230 } 2231 2232 static void 2233 X_rsvp_input(struct mbuf *m, int off) 2234 { 2235 int vifi; 2236 struct ip *ip = mtod(m, struct ip *); 2237 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2238 struct ifnet *ifp; 2239 2240 if (rsvpdebug) 2241 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2242 2243 /* Can still get packets with rsvp_on = 0 if there is a local member 2244 * of the group to which the RSVP packet is addressed. But in this 2245 * case we want to throw the packet away. 2246 */ 2247 if (!rsvp_on) { 2248 m_freem(m); 2249 return; 2250 } 2251 2252 if (rsvpdebug) 2253 printf("rsvp_input: check vifs\n"); 2254 2255 #ifdef DIAGNOSTIC 2256 M_ASSERTPKTHDR(m); 2257 #endif 2258 2259 ifp = m->m_pkthdr.rcvif; 2260 2261 VIF_LOCK(); 2262 /* Find which vif the packet arrived on. */ 2263 for (vifi = 0; vifi < numvifs; vifi++) 2264 if (viftable[vifi].v_ifp == ifp) 2265 break; 2266 2267 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 2268 /* 2269 * Drop the lock here to avoid holding it across rip_input. 2270 * This could make rsvpdebug printfs wrong. If you care, 2271 * record the state of stuff before dropping the lock. 2272 */ 2273 VIF_UNLOCK(); 2274 /* 2275 * If the old-style non-vif-associated socket is set, 2276 * then use it. Otherwise, drop packet since there 2277 * is no specific socket for this vif. 2278 */ 2279 if (ip_rsvpd != NULL) { 2280 if (rsvpdebug) 2281 printf("rsvp_input: Sending packet up old-style socket\n"); 2282 rip_input(m, off); /* xxx */ 2283 } else { 2284 if (rsvpdebug && vifi == numvifs) 2285 printf("rsvp_input: Can't find vif for packet.\n"); 2286 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 2287 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2288 m_freem(m); 2289 } 2290 return; 2291 } 2292 rsvp_src.sin_addr = ip->ip_src; 2293 2294 if (rsvpdebug && m) 2295 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2296 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2297 2298 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 2299 if (rsvpdebug) 2300 printf("rsvp_input: Failed to append to socket\n"); 2301 } else { 2302 if (rsvpdebug) 2303 printf("rsvp_input: send packet up\n"); 2304 } 2305 VIF_UNLOCK(); 2306 } 2307 2308 /* 2309 * Code for bandwidth monitors 2310 */ 2311 2312 /* 2313 * Define common interface for timeval-related methods 2314 */ 2315 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 2316 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 2317 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 2318 2319 static uint32_t 2320 compute_bw_meter_flags(struct bw_upcall *req) 2321 { 2322 uint32_t flags = 0; 2323 2324 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2325 flags |= BW_METER_UNIT_PACKETS; 2326 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2327 flags |= BW_METER_UNIT_BYTES; 2328 if (req->bu_flags & BW_UPCALL_GEQ) 2329 flags |= BW_METER_GEQ; 2330 if (req->bu_flags & BW_UPCALL_LEQ) 2331 flags |= BW_METER_LEQ; 2332 2333 return flags; 2334 } 2335 2336 /* 2337 * Add a bw_meter entry 2338 */ 2339 static int 2340 add_bw_upcall(struct bw_upcall *req) 2341 { 2342 struct mfc *mfc; 2343 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2344 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2345 struct timeval now; 2346 struct bw_meter *x; 2347 uint32_t flags; 2348 2349 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2350 return EOPNOTSUPP; 2351 2352 /* Test if the flags are valid */ 2353 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2354 return EINVAL; 2355 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2356 return EINVAL; 2357 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2358 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2359 return EINVAL; 2360 2361 /* Test if the threshold time interval is valid */ 2362 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2363 return EINVAL; 2364 2365 flags = compute_bw_meter_flags(req); 2366 2367 /* 2368 * Find if we have already same bw_meter entry 2369 */ 2370 MFC_LOCK(); 2371 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2372 if (mfc == NULL) { 2373 MFC_UNLOCK(); 2374 return EADDRNOTAVAIL; 2375 } 2376 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2377 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2378 &req->bu_threshold.b_time, ==)) && 2379 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2380 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2381 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2382 MFC_UNLOCK(); 2383 return 0; /* XXX Already installed */ 2384 } 2385 } 2386 2387 /* Allocate the new bw_meter entry */ 2388 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2389 if (x == NULL) { 2390 MFC_UNLOCK(); 2391 return ENOBUFS; 2392 } 2393 2394 /* Set the new bw_meter entry */ 2395 x->bm_threshold.b_time = req->bu_threshold.b_time; 2396 GET_TIME(now); 2397 x->bm_start_time = now; 2398 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2399 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2400 x->bm_measured.b_packets = 0; 2401 x->bm_measured.b_bytes = 0; 2402 x->bm_flags = flags; 2403 x->bm_time_next = NULL; 2404 x->bm_time_hash = BW_METER_BUCKETS; 2405 2406 /* Add the new bw_meter entry to the front of entries for this MFC */ 2407 x->bm_mfc = mfc; 2408 x->bm_mfc_next = mfc->mfc_bw_meter; 2409 mfc->mfc_bw_meter = x; 2410 schedule_bw_meter(x, &now); 2411 MFC_UNLOCK(); 2412 2413 return 0; 2414 } 2415 2416 static void 2417 free_bw_list(struct bw_meter *list) 2418 { 2419 while (list != NULL) { 2420 struct bw_meter *x = list; 2421 2422 list = list->bm_mfc_next; 2423 unschedule_bw_meter(x); 2424 free(x, M_BWMETER); 2425 } 2426 } 2427 2428 /* 2429 * Delete one or multiple bw_meter entries 2430 */ 2431 static int 2432 del_bw_upcall(struct bw_upcall *req) 2433 { 2434 struct mfc *mfc; 2435 struct bw_meter *x; 2436 2437 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2438 return EOPNOTSUPP; 2439 2440 MFC_LOCK(); 2441 /* Find the corresponding MFC entry */ 2442 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2443 if (mfc == NULL) { 2444 MFC_UNLOCK(); 2445 return EADDRNOTAVAIL; 2446 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2447 /* 2448 * Delete all bw_meter entries for this mfc 2449 */ 2450 struct bw_meter *list; 2451 2452 list = mfc->mfc_bw_meter; 2453 mfc->mfc_bw_meter = NULL; 2454 free_bw_list(list); 2455 MFC_UNLOCK(); 2456 return 0; 2457 } else { /* Delete a single bw_meter entry */ 2458 struct bw_meter *prev; 2459 uint32_t flags = 0; 2460 2461 flags = compute_bw_meter_flags(req); 2462 2463 /* Find the bw_meter entry to delete */ 2464 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2465 x = x->bm_mfc_next) { 2466 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2467 &req->bu_threshold.b_time, ==)) && 2468 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2469 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2470 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2471 break; 2472 } 2473 if (x != NULL) { /* Delete entry from the list for this MFC */ 2474 if (prev != NULL) 2475 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2476 else 2477 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2478 2479 unschedule_bw_meter(x); 2480 MFC_UNLOCK(); 2481 /* Free the bw_meter entry */ 2482 free(x, M_BWMETER); 2483 return 0; 2484 } else { 2485 MFC_UNLOCK(); 2486 return EINVAL; 2487 } 2488 } 2489 /* NOTREACHED */ 2490 } 2491 2492 /* 2493 * Perform bandwidth measurement processing that may result in an upcall 2494 */ 2495 static void 2496 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2497 { 2498 struct timeval delta; 2499 2500 MFC_LOCK_ASSERT(); 2501 2502 delta = *nowp; 2503 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2504 2505 if (x->bm_flags & BW_METER_GEQ) { 2506 /* 2507 * Processing for ">=" type of bw_meter entry 2508 */ 2509 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2510 /* Reset the bw_meter entry */ 2511 x->bm_start_time = *nowp; 2512 x->bm_measured.b_packets = 0; 2513 x->bm_measured.b_bytes = 0; 2514 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2515 } 2516 2517 /* Record that a packet is received */ 2518 x->bm_measured.b_packets++; 2519 x->bm_measured.b_bytes += plen; 2520 2521 /* 2522 * Test if we should deliver an upcall 2523 */ 2524 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2525 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2526 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2527 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2528 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2529 /* Prepare an upcall for delivery */ 2530 bw_meter_prepare_upcall(x, nowp); 2531 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2532 } 2533 } 2534 } else if (x->bm_flags & BW_METER_LEQ) { 2535 /* 2536 * Processing for "<=" type of bw_meter entry 2537 */ 2538 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2539 /* 2540 * We are behind time with the multicast forwarding table 2541 * scanning for "<=" type of bw_meter entries, so test now 2542 * if we should deliver an upcall. 2543 */ 2544 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2545 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2546 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2547 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2548 /* Prepare an upcall for delivery */ 2549 bw_meter_prepare_upcall(x, nowp); 2550 } 2551 /* Reschedule the bw_meter entry */ 2552 unschedule_bw_meter(x); 2553 schedule_bw_meter(x, nowp); 2554 } 2555 2556 /* Record that a packet is received */ 2557 x->bm_measured.b_packets++; 2558 x->bm_measured.b_bytes += plen; 2559 2560 /* 2561 * Test if we should restart the measuring interval 2562 */ 2563 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2564 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2565 (x->bm_flags & BW_METER_UNIT_BYTES && 2566 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2567 /* Don't restart the measuring interval */ 2568 } else { 2569 /* Do restart the measuring interval */ 2570 /* 2571 * XXX: note that we don't unschedule and schedule, because this 2572 * might be too much overhead per packet. Instead, when we process 2573 * all entries for a given timer hash bin, we check whether it is 2574 * really a timeout. If not, we reschedule at that time. 2575 */ 2576 x->bm_start_time = *nowp; 2577 x->bm_measured.b_packets = 0; 2578 x->bm_measured.b_bytes = 0; 2579 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2580 } 2581 } 2582 } 2583 2584 /* 2585 * Prepare a bandwidth-related upcall 2586 */ 2587 static void 2588 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2589 { 2590 struct timeval delta; 2591 struct bw_upcall *u; 2592 2593 MFC_LOCK_ASSERT(); 2594 2595 /* 2596 * Compute the measured time interval 2597 */ 2598 delta = *nowp; 2599 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2600 2601 /* 2602 * If there are too many pending upcalls, deliver them now 2603 */ 2604 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2605 bw_upcalls_send(); 2606 2607 /* 2608 * Set the bw_upcall entry 2609 */ 2610 u = &bw_upcalls[bw_upcalls_n++]; 2611 u->bu_src = x->bm_mfc->mfc_origin; 2612 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2613 u->bu_threshold.b_time = x->bm_threshold.b_time; 2614 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2615 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2616 u->bu_measured.b_time = delta; 2617 u->bu_measured.b_packets = x->bm_measured.b_packets; 2618 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2619 u->bu_flags = 0; 2620 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2621 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2622 if (x->bm_flags & BW_METER_UNIT_BYTES) 2623 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2624 if (x->bm_flags & BW_METER_GEQ) 2625 u->bu_flags |= BW_UPCALL_GEQ; 2626 if (x->bm_flags & BW_METER_LEQ) 2627 u->bu_flags |= BW_UPCALL_LEQ; 2628 } 2629 2630 /* 2631 * Send the pending bandwidth-related upcalls 2632 */ 2633 static void 2634 bw_upcalls_send(void) 2635 { 2636 struct mbuf *m; 2637 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2638 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2639 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2640 0, /* unused2 */ 2641 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2642 0, /* im_mbz */ 2643 0, /* im_vif */ 2644 0, /* unused3 */ 2645 { 0 }, /* im_src */ 2646 { 0 } }; /* im_dst */ 2647 2648 MFC_LOCK_ASSERT(); 2649 2650 if (bw_upcalls_n == 0) 2651 return; /* No pending upcalls */ 2652 2653 bw_upcalls_n = 0; 2654 2655 /* 2656 * Allocate a new mbuf, initialize it with the header and 2657 * the payload for the pending calls. 2658 */ 2659 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2660 if (m == NULL) { 2661 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2662 return; 2663 } 2664 2665 m->m_len = m->m_pkthdr.len = 0; 2666 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2667 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2668 2669 /* 2670 * Send the upcalls 2671 * XXX do we need to set the address in k_igmpsrc ? 2672 */ 2673 mrtstat.mrts_upcalls++; 2674 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2675 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2676 ++mrtstat.mrts_upq_sockfull; 2677 } 2678 } 2679 2680 /* 2681 * Compute the timeout hash value for the bw_meter entries 2682 */ 2683 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2684 do { \ 2685 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2686 \ 2687 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2688 (hash) = next_timeval.tv_sec; \ 2689 if (next_timeval.tv_usec) \ 2690 (hash)++; /* XXX: make sure we don't timeout early */ \ 2691 (hash) %= BW_METER_BUCKETS; \ 2692 } while (0) 2693 2694 /* 2695 * Schedule a timer to process periodically bw_meter entry of type "<=" 2696 * by linking the entry in the proper hash bucket. 2697 */ 2698 static void 2699 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2700 { 2701 int time_hash; 2702 2703 MFC_LOCK_ASSERT(); 2704 2705 if (!(x->bm_flags & BW_METER_LEQ)) 2706 return; /* XXX: we schedule timers only for "<=" entries */ 2707 2708 /* 2709 * Reset the bw_meter entry 2710 */ 2711 x->bm_start_time = *nowp; 2712 x->bm_measured.b_packets = 0; 2713 x->bm_measured.b_bytes = 0; 2714 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2715 2716 /* 2717 * Compute the timeout hash value and insert the entry 2718 */ 2719 BW_METER_TIMEHASH(x, time_hash); 2720 x->bm_time_next = bw_meter_timers[time_hash]; 2721 bw_meter_timers[time_hash] = x; 2722 x->bm_time_hash = time_hash; 2723 } 2724 2725 /* 2726 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2727 * by removing the entry from the proper hash bucket. 2728 */ 2729 static void 2730 unschedule_bw_meter(struct bw_meter *x) 2731 { 2732 int time_hash; 2733 struct bw_meter *prev, *tmp; 2734 2735 MFC_LOCK_ASSERT(); 2736 2737 if (!(x->bm_flags & BW_METER_LEQ)) 2738 return; /* XXX: we schedule timers only for "<=" entries */ 2739 2740 /* 2741 * Compute the timeout hash value and delete the entry 2742 */ 2743 time_hash = x->bm_time_hash; 2744 if (time_hash >= BW_METER_BUCKETS) 2745 return; /* Entry was not scheduled */ 2746 2747 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2748 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2749 if (tmp == x) 2750 break; 2751 2752 if (tmp == NULL) 2753 panic("unschedule_bw_meter: bw_meter entry not found"); 2754 2755 if (prev != NULL) 2756 prev->bm_time_next = x->bm_time_next; 2757 else 2758 bw_meter_timers[time_hash] = x->bm_time_next; 2759 2760 x->bm_time_next = NULL; 2761 x->bm_time_hash = BW_METER_BUCKETS; 2762 } 2763 2764 2765 /* 2766 * Process all "<=" type of bw_meter that should be processed now, 2767 * and for each entry prepare an upcall if necessary. Each processed 2768 * entry is rescheduled again for the (periodic) processing. 2769 * 2770 * This is run periodically (once per second normally). On each round, 2771 * all the potentially matching entries are in the hash slot that we are 2772 * looking at. 2773 */ 2774 static void 2775 bw_meter_process() 2776 { 2777 static uint32_t last_tv_sec; /* last time we processed this */ 2778 2779 uint32_t loops; 2780 int i; 2781 struct timeval now, process_endtime; 2782 2783 GET_TIME(now); 2784 if (last_tv_sec == now.tv_sec) 2785 return; /* nothing to do */ 2786 2787 loops = now.tv_sec - last_tv_sec; 2788 last_tv_sec = now.tv_sec; 2789 if (loops > BW_METER_BUCKETS) 2790 loops = BW_METER_BUCKETS; 2791 2792 MFC_LOCK(); 2793 /* 2794 * Process all bins of bw_meter entries from the one after the last 2795 * processed to the current one. On entry, i points to the last bucket 2796 * visited, so we need to increment i at the beginning of the loop. 2797 */ 2798 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2799 struct bw_meter *x, *tmp_list; 2800 2801 if (++i >= BW_METER_BUCKETS) 2802 i = 0; 2803 2804 /* Disconnect the list of bw_meter entries from the bin */ 2805 tmp_list = bw_meter_timers[i]; 2806 bw_meter_timers[i] = NULL; 2807 2808 /* Process the list of bw_meter entries */ 2809 while (tmp_list != NULL) { 2810 x = tmp_list; 2811 tmp_list = tmp_list->bm_time_next; 2812 2813 /* Test if the time interval is over */ 2814 process_endtime = x->bm_start_time; 2815 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2816 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2817 /* Not yet: reschedule, but don't reset */ 2818 int time_hash; 2819 2820 BW_METER_TIMEHASH(x, time_hash); 2821 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2822 /* 2823 * XXX: somehow the bin processing is a bit ahead of time. 2824 * Put the entry in the next bin. 2825 */ 2826 if (++time_hash >= BW_METER_BUCKETS) 2827 time_hash = 0; 2828 } 2829 x->bm_time_next = bw_meter_timers[time_hash]; 2830 bw_meter_timers[time_hash] = x; 2831 x->bm_time_hash = time_hash; 2832 2833 continue; 2834 } 2835 2836 /* 2837 * Test if we should deliver an upcall 2838 */ 2839 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2840 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2841 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2842 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2843 /* Prepare an upcall for delivery */ 2844 bw_meter_prepare_upcall(x, &now); 2845 } 2846 2847 /* 2848 * Reschedule for next processing 2849 */ 2850 schedule_bw_meter(x, &now); 2851 } 2852 } 2853 2854 /* Send all upcalls that are pending delivery */ 2855 bw_upcalls_send(); 2856 2857 MFC_UNLOCK(); 2858 } 2859 2860 /* 2861 * A periodic function for sending all upcalls that are pending delivery 2862 */ 2863 static void 2864 expire_bw_upcalls_send(void *unused) 2865 { 2866 MFC_LOCK(); 2867 bw_upcalls_send(); 2868 MFC_UNLOCK(); 2869 2870 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2871 expire_bw_upcalls_send, NULL); 2872 } 2873 2874 /* 2875 * A periodic function for periodic scanning of the multicast forwarding 2876 * table for processing all "<=" bw_meter entries. 2877 */ 2878 static void 2879 expire_bw_meter_process(void *unused) 2880 { 2881 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2882 bw_meter_process(); 2883 2884 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2885 } 2886 2887 /* 2888 * End of bandwidth monitoring code 2889 */ 2890 2891 #ifdef PIM 2892 /* 2893 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2894 * 2895 */ 2896 static int 2897 pim_register_send(struct ip *ip, struct vif *vifp, 2898 struct mbuf *m, struct mfc *rt) 2899 { 2900 struct mbuf *mb_copy, *mm; 2901 2902 if (mrtdebug & DEBUG_PIM) 2903 log(LOG_DEBUG, "pim_register_send: "); 2904 2905 mb_copy = pim_register_prepare(ip, m); 2906 if (mb_copy == NULL) 2907 return ENOBUFS; 2908 2909 /* 2910 * Send all the fragments. Note that the mbuf for each fragment 2911 * is freed by the sending machinery. 2912 */ 2913 for (mm = mb_copy; mm; mm = mb_copy) { 2914 mb_copy = mm->m_nextpkt; 2915 mm->m_nextpkt = 0; 2916 mm = m_pullup(mm, sizeof(struct ip)); 2917 if (mm != NULL) { 2918 ip = mtod(mm, struct ip *); 2919 if ((mrt_api_config & MRT_MFC_RP) && 2920 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2921 pim_register_send_rp(ip, vifp, mm, rt); 2922 } else { 2923 pim_register_send_upcall(ip, vifp, mm, rt); 2924 } 2925 } 2926 } 2927 2928 return 0; 2929 } 2930 2931 /* 2932 * Return a copy of the data packet that is ready for PIM Register 2933 * encapsulation. 2934 * XXX: Note that in the returned copy the IP header is a valid one. 2935 */ 2936 static struct mbuf * 2937 pim_register_prepare(struct ip *ip, struct mbuf *m) 2938 { 2939 struct mbuf *mb_copy = NULL; 2940 int mtu; 2941 2942 /* Take care of delayed checksums */ 2943 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2944 in_delayed_cksum(m); 2945 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2946 } 2947 2948 /* 2949 * Copy the old packet & pullup its IP header into the 2950 * new mbuf so we can modify it. 2951 */ 2952 mb_copy = m_copypacket(m, M_DONTWAIT); 2953 if (mb_copy == NULL) 2954 return NULL; 2955 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2956 if (mb_copy == NULL) 2957 return NULL; 2958 2959 /* take care of the TTL */ 2960 ip = mtod(mb_copy, struct ip *); 2961 --ip->ip_ttl; 2962 2963 /* Compute the MTU after the PIM Register encapsulation */ 2964 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2965 2966 if (ip->ip_len <= mtu) { 2967 /* Turn the IP header into a valid one */ 2968 ip->ip_len = htons(ip->ip_len); 2969 ip->ip_off = htons(ip->ip_off); 2970 ip->ip_sum = 0; 2971 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2972 } else { 2973 /* Fragment the packet */ 2974 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2975 m_freem(mb_copy); 2976 return NULL; 2977 } 2978 } 2979 return mb_copy; 2980 } 2981 2982 /* 2983 * Send an upcall with the data packet to the user-level process. 2984 */ 2985 static int 2986 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2987 struct mbuf *mb_copy, struct mfc *rt) 2988 { 2989 struct mbuf *mb_first; 2990 int len = ntohs(ip->ip_len); 2991 struct igmpmsg *im; 2992 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2993 2994 VIF_LOCK_ASSERT(); 2995 2996 /* 2997 * Add a new mbuf with an upcall header 2998 */ 2999 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3000 if (mb_first == NULL) { 3001 m_freem(mb_copy); 3002 return ENOBUFS; 3003 } 3004 mb_first->m_data += max_linkhdr; 3005 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3006 mb_first->m_len = sizeof(struct igmpmsg); 3007 mb_first->m_next = mb_copy; 3008 3009 /* Send message to routing daemon */ 3010 im = mtod(mb_first, struct igmpmsg *); 3011 im->im_msgtype = IGMPMSG_WHOLEPKT; 3012 im->im_mbz = 0; 3013 im->im_vif = vifp - viftable; 3014 im->im_src = ip->ip_src; 3015 im->im_dst = ip->ip_dst; 3016 3017 k_igmpsrc.sin_addr = ip->ip_src; 3018 3019 mrtstat.mrts_upcalls++; 3020 3021 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3022 if (mrtdebug & DEBUG_PIM) 3023 log(LOG_WARNING, 3024 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3025 ++mrtstat.mrts_upq_sockfull; 3026 return ENOBUFS; 3027 } 3028 3029 /* Keep statistics */ 3030 pimstat.pims_snd_registers_msgs++; 3031 pimstat.pims_snd_registers_bytes += len; 3032 3033 return 0; 3034 } 3035 3036 /* 3037 * Encapsulate the data packet in PIM Register message and send it to the RP. 3038 */ 3039 static int 3040 pim_register_send_rp(struct ip *ip, struct vif *vifp, 3041 struct mbuf *mb_copy, struct mfc *rt) 3042 { 3043 struct mbuf *mb_first; 3044 struct ip *ip_outer; 3045 struct pim_encap_pimhdr *pimhdr; 3046 int len = ntohs(ip->ip_len); 3047 vifi_t vifi = rt->mfc_parent; 3048 3049 VIF_LOCK_ASSERT(); 3050 3051 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 3052 m_freem(mb_copy); 3053 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3054 } 3055 3056 /* 3057 * Add a new mbuf with the encapsulating header 3058 */ 3059 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3060 if (mb_first == NULL) { 3061 m_freem(mb_copy); 3062 return ENOBUFS; 3063 } 3064 mb_first->m_data += max_linkhdr; 3065 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3066 mb_first->m_next = mb_copy; 3067 3068 mb_first->m_pkthdr.len = len + mb_first->m_len; 3069 3070 /* 3071 * Fill in the encapsulating IP and PIM header 3072 */ 3073 ip_outer = mtod(mb_first, struct ip *); 3074 *ip_outer = pim_encap_iphdr; 3075 #ifdef RANDOM_IP_ID 3076 ip_outer->ip_id = ip_randomid(); 3077 #else 3078 ip_outer->ip_id = htons(ip_id++); 3079 #endif 3080 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3081 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3082 ip_outer->ip_dst = rt->mfc_rp; 3083 /* 3084 * Copy the inner header TOS to the outer header, and take care of the 3085 * IP_DF bit. 3086 */ 3087 ip_outer->ip_tos = ip->ip_tos; 3088 if (ntohs(ip->ip_off) & IP_DF) 3089 ip_outer->ip_off |= IP_DF; 3090 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 3091 + sizeof(pim_encap_iphdr)); 3092 *pimhdr = pim_encap_pimhdr; 3093 /* If the iif crosses a border, set the Border-bit */ 3094 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3095 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3096 3097 mb_first->m_data += sizeof(pim_encap_iphdr); 3098 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3099 mb_first->m_data -= sizeof(pim_encap_iphdr); 3100 3101 if (vifp->v_rate_limit == 0) 3102 tbf_send_packet(vifp, mb_first); 3103 else 3104 tbf_control(vifp, mb_first, ip, ip_outer->ip_len); 3105 3106 /* Keep statistics */ 3107 pimstat.pims_snd_registers_msgs++; 3108 pimstat.pims_snd_registers_bytes += len; 3109 3110 return 0; 3111 } 3112 3113 /* 3114 * PIM-SMv2 and PIM-DM messages processing. 3115 * Receives and verifies the PIM control messages, and passes them 3116 * up to the listening socket, using rip_input(). 3117 * The only message with special processing is the PIM_REGISTER message 3118 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3119 * is passed to if_simloop(). 3120 */ 3121 void 3122 pim_input(struct mbuf *m, int off) 3123 { 3124 struct ip *ip = mtod(m, struct ip *); 3125 struct pim *pim; 3126 int minlen; 3127 int datalen = ip->ip_len; 3128 int ip_tos; 3129 int iphlen = off; 3130 3131 /* Keep statistics */ 3132 pimstat.pims_rcv_total_msgs++; 3133 pimstat.pims_rcv_total_bytes += datalen; 3134 3135 /* 3136 * Validate lengths 3137 */ 3138 if (datalen < PIM_MINLEN) { 3139 pimstat.pims_rcv_tooshort++; 3140 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3141 datalen, (u_long)ip->ip_src.s_addr); 3142 m_freem(m); 3143 return; 3144 } 3145 3146 /* 3147 * If the packet is at least as big as a REGISTER, go agead 3148 * and grab the PIM REGISTER header size, to avoid another 3149 * possible m_pullup() later. 3150 * 3151 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3152 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3153 */ 3154 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3155 /* 3156 * Get the IP and PIM headers in contiguous memory, and 3157 * possibly the PIM REGISTER header. 3158 */ 3159 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3160 (m = m_pullup(m, minlen)) == 0) { 3161 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3162 return; 3163 } 3164 /* m_pullup() may have given us a new mbuf so reset ip. */ 3165 ip = mtod(m, struct ip *); 3166 ip_tos = ip->ip_tos; 3167 3168 /* adjust mbuf to point to the PIM header */ 3169 m->m_data += iphlen; 3170 m->m_len -= iphlen; 3171 pim = mtod(m, struct pim *); 3172 3173 /* 3174 * Validate checksum. If PIM REGISTER, exclude the data packet. 3175 * 3176 * XXX: some older PIMv2 implementations don't make this distinction, 3177 * so for compatibility reason perform the checksum over part of the 3178 * message, and if error, then over the whole message. 3179 */ 3180 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3181 /* do nothing, checksum okay */ 3182 } else if (in_cksum(m, datalen)) { 3183 pimstat.pims_rcv_badsum++; 3184 if (mrtdebug & DEBUG_PIM) 3185 log(LOG_DEBUG, "pim_input: invalid checksum"); 3186 m_freem(m); 3187 return; 3188 } 3189 3190 /* PIM version check */ 3191 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3192 pimstat.pims_rcv_badversion++; 3193 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3194 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3195 m_freem(m); 3196 return; 3197 } 3198 3199 /* restore mbuf back to the outer IP */ 3200 m->m_data -= iphlen; 3201 m->m_len += iphlen; 3202 3203 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3204 /* 3205 * Since this is a REGISTER, we'll make a copy of the register 3206 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3207 * routing daemon. 3208 */ 3209 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3210 struct mbuf *mcp; 3211 struct ip *encap_ip; 3212 u_int32_t *reghdr; 3213 struct ifnet *vifp; 3214 3215 VIF_LOCK(); 3216 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3217 VIF_UNLOCK(); 3218 if (mrtdebug & DEBUG_PIM) 3219 log(LOG_DEBUG, 3220 "pim_input: register vif not set: %d\n", reg_vif_num); 3221 m_freem(m); 3222 return; 3223 } 3224 /* XXX need refcnt? */ 3225 vifp = viftable[reg_vif_num].v_ifp; 3226 VIF_UNLOCK(); 3227 3228 /* 3229 * Validate length 3230 */ 3231 if (datalen < PIM_REG_MINLEN) { 3232 pimstat.pims_rcv_tooshort++; 3233 pimstat.pims_rcv_badregisters++; 3234 log(LOG_ERR, 3235 "pim_input: register packet size too small %d from %lx\n", 3236 datalen, (u_long)ip->ip_src.s_addr); 3237 m_freem(m); 3238 return; 3239 } 3240 3241 reghdr = (u_int32_t *)(pim + 1); 3242 encap_ip = (struct ip *)(reghdr + 1); 3243 3244 if (mrtdebug & DEBUG_PIM) { 3245 log(LOG_DEBUG, 3246 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3247 (u_long)ntohl(encap_ip->ip_src.s_addr), 3248 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3249 ntohs(encap_ip->ip_len)); 3250 } 3251 3252 /* verify the version number of the inner packet */ 3253 if (encap_ip->ip_v != IPVERSION) { 3254 pimstat.pims_rcv_badregisters++; 3255 if (mrtdebug & DEBUG_PIM) { 3256 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3257 "of the inner packet\n", encap_ip->ip_v); 3258 } 3259 m_freem(m); 3260 return; 3261 } 3262 3263 /* verify the inner packet is destined to a mcast group */ 3264 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 3265 pimstat.pims_rcv_badregisters++; 3266 if (mrtdebug & DEBUG_PIM) 3267 log(LOG_DEBUG, 3268 "pim_input: inner packet of register is not " 3269 "multicast %lx\n", 3270 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3271 m_freem(m); 3272 return; 3273 } 3274 3275 /* 3276 * Copy the TOS from the outer IP header to the inner IP header. 3277 */ 3278 if (encap_ip->ip_tos != ip_tos) { 3279 /* Outer TOS -> inner TOS */ 3280 encap_ip->ip_tos = ip_tos; 3281 /* Recompute the inner header checksum. Sigh... */ 3282 3283 /* adjust mbuf to point to the inner IP header */ 3284 m->m_data += (iphlen + PIM_MINLEN); 3285 m->m_len -= (iphlen + PIM_MINLEN); 3286 3287 encap_ip->ip_sum = 0; 3288 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3289 3290 /* restore mbuf to point back to the outer IP header */ 3291 m->m_data -= (iphlen + PIM_MINLEN); 3292 m->m_len += (iphlen + PIM_MINLEN); 3293 } 3294 3295 /* If a NULL_REGISTER, pass it to the daemon */ 3296 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3297 goto pim_input_to_daemon; 3298 3299 /* 3300 * Decapsulate the inner IP packet and loopback to forward it 3301 * as a normal multicast packet. Also, make a copy of the 3302 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3303 * to pass to the daemon later, so it can take the appropriate 3304 * actions (e.g., send back PIM_REGISTER_STOP). 3305 * XXX: here m->m_data points to the outer IP header. 3306 */ 3307 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3308 if (mcp == NULL) { 3309 log(LOG_ERR, 3310 "pim_input: pim register: could not copy register head\n"); 3311 m_freem(m); 3312 return; 3313 } 3314 3315 /* Keep statistics */ 3316 /* XXX: registers_bytes include only the encap. mcast pkt */ 3317 pimstat.pims_rcv_registers_msgs++; 3318 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3319 3320 /* 3321 * forward the inner ip packet; point m_data at the inner ip. 3322 */ 3323 m_adj(m, iphlen + PIM_MINLEN); 3324 3325 if (mrtdebug & DEBUG_PIM) { 3326 log(LOG_DEBUG, 3327 "pim_input: forwarding decapsulated register: " 3328 "src %lx, dst %lx, vif %d\n", 3329 (u_long)ntohl(encap_ip->ip_src.s_addr), 3330 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3331 reg_vif_num); 3332 } 3333 /* NB: vifp was collected above; can it change on us? */ 3334 if_simloop(vifp, m, dst.sin_family, 0); 3335 3336 /* prepare the register head to send to the mrouting daemon */ 3337 m = mcp; 3338 } 3339 3340 pim_input_to_daemon: 3341 /* 3342 * Pass the PIM message up to the daemon; if it is a Register message, 3343 * pass the 'head' only up to the daemon. This includes the 3344 * outer IP header, PIM header, PIM-Register header and the 3345 * inner IP header. 3346 * XXX: the outer IP header pkt size of a Register is not adjust to 3347 * reflect the fact that the inner multicast data is truncated. 3348 */ 3349 rip_input(m, iphlen); 3350 3351 return; 3352 } 3353 #endif /* PIM */ 3354 3355 static int 3356 ip_mroute_modevent(module_t mod, int type, void *unused) 3357 { 3358 int s; 3359 3360 switch (type) { 3361 case MOD_LOAD: 3362 s = splnet(); 3363 ip_mrouter_reset(); 3364 /* XXX synchronize setup */ 3365 ip_mcast_src = X_ip_mcast_src; 3366 ip_mforward = X_ip_mforward; 3367 ip_mrouter_done = X_ip_mrouter_done; 3368 ip_mrouter_get = X_ip_mrouter_get; 3369 ip_mrouter_set = X_ip_mrouter_set; 3370 ip_rsvp_force_done = X_ip_rsvp_force_done; 3371 ip_rsvp_vif = X_ip_rsvp_vif; 3372 legal_vif_num = X_legal_vif_num; 3373 mrt_ioctl = X_mrt_ioctl; 3374 rsvp_input_p = X_rsvp_input; 3375 break; 3376 3377 case MOD_UNLOAD: 3378 /* 3379 * Typically module unload happens after the user-level 3380 * process has shutdown the kernel services (the check 3381 * below insures someone can't just yank the module out 3382 * from under a running process). But if the module is 3383 * just loaded and then unloaded w/o starting up a user 3384 * process we still need to cleanup. 3385 */ 3386 if (ip_mrouter) 3387 return EINVAL; 3388 3389 X_ip_mrouter_done(); 3390 ip_mcast_src = NULL; 3391 ip_mforward = NULL; 3392 ip_mrouter_done = NULL; 3393 ip_mrouter_get = NULL; 3394 ip_mrouter_set = NULL; 3395 ip_rsvp_force_done = NULL; 3396 ip_rsvp_vif = NULL; 3397 legal_vif_num = NULL; 3398 mrt_ioctl = NULL; 3399 rsvp_input_p = NULL; 3400 break; 3401 } 3402 return 0; 3403 } 3404 3405 static moduledata_t ip_mroutemod = { 3406 "ip_mroute", 3407 ip_mroute_modevent, 3408 0 3409 }; 3410 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3411