1 /*- 2 * Copyright (c) 1989 Stephen Deering 3 * Copyright (c) 1992, 1993 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Stephen Deering of Stanford University. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 34 */ 35 36 /* 37 * IP multicast forwarding procedures 38 * 39 * Written by David Waitzman, BBN Labs, August 1988. 40 * Modified by Steve Deering, Stanford, February 1989. 41 * Modified by Mark J. Steiglitz, Stanford, May, 1991 42 * Modified by Van Jacobson, LBL, January 1993 43 * Modified by Ajit Thyagarajan, PARC, August 1993 44 * Modified by Bill Fenner, PARC, April 1995 45 * Modified by Ahmed Helmy, SGI, June 1996 46 * Modified by George Edmond Eddy (Rusty), ISI, February 1998 47 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000 48 * Modified by Hitoshi Asaeda, WIDE, August 2000 49 * Modified by Pavlin Radoslavov, ICSI, October 2002 50 * 51 * MROUTING Revision: 3.5 52 * and PIM-SMv2 and PIM-DM support, advanced API support, 53 * bandwidth metering and signaling 54 */ 55 56 #include <sys/cdefs.h> 57 __FBSDID("$FreeBSD$"); 58 59 #include "opt_inet.h" 60 #include "opt_inet6.h" 61 #include "opt_mac.h" 62 #include "opt_mrouting.h" 63 64 #define _PIM_VT 1 65 66 #include <sys/param.h> 67 #include <sys/kernel.h> 68 #include <sys/lock.h> 69 #include <sys/malloc.h> 70 #include <sys/mbuf.h> 71 #include <sys/module.h> 72 #include <sys/priv.h> 73 #include <sys/protosw.h> 74 #include <sys/signalvar.h> 75 #include <sys/socket.h> 76 #include <sys/socketvar.h> 77 #include <sys/sockio.h> 78 #include <sys/sx.h> 79 #include <sys/sysctl.h> 80 #include <sys/syslog.h> 81 #include <sys/systm.h> 82 #include <sys/time.h> 83 #include <sys/vimage.h> 84 #include <net/if.h> 85 #include <net/netisr.h> 86 #include <net/route.h> 87 #include <netinet/in.h> 88 #include <netinet/igmp.h> 89 #include <netinet/in_systm.h> 90 #include <netinet/in_var.h> 91 #include <netinet/ip.h> 92 #include <netinet/ip_encap.h> 93 #include <netinet/ip_mroute.h> 94 #include <netinet/ip_var.h> 95 #include <netinet/ip_options.h> 96 #include <netinet/pim.h> 97 #include <netinet/pim_var.h> 98 #include <netinet/udp.h> 99 #include <netinet/vinet.h> 100 101 #ifdef INET6 102 #include <netinet/ip6.h> 103 #include <netinet6/in6_var.h> 104 #include <netinet6/ip6_mroute.h> 105 #include <netinet6/ip6_var.h> 106 #endif 107 #include <machine/in_cksum.h> 108 109 #include <security/mac/mac_framework.h> 110 111 /* 112 * Control debugging code for rsvp and multicast routing code. 113 * Can only set them with the debugger. 114 */ 115 static u_int rsvpdebug; /* non-zero enables debugging */ 116 117 static u_int mrtdebug; /* any set of the flags below */ 118 #define DEBUG_MFC 0x02 119 #define DEBUG_FORWARD 0x04 120 #define DEBUG_EXPIRE 0x08 121 #define DEBUG_XMIT 0x10 122 #define DEBUG_PIM 0x20 123 124 #define VIFI_INVALID ((vifi_t) -1) 125 126 #define M_HASCL(m) ((m)->m_flags & M_EXT) 127 128 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 129 130 /* 131 * Locking. We use two locks: one for the virtual interface table and 132 * one for the forwarding table. These locks may be nested in which case 133 * the VIF lock must always be taken first. Note that each lock is used 134 * to cover not only the specific data structure but also related data 135 * structures. It may be better to add more fine-grained locking later; 136 * it's not clear how performance-critical this code is. 137 * 138 * XXX: This module could particularly benefit from being cleaned 139 * up to use the <sys/queue.h> macros. 140 * 141 */ 142 143 static struct mrtstat mrtstat; 144 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 145 &mrtstat, mrtstat, 146 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 147 148 static struct mfc *mfctable[MFCTBLSIZ]; 149 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 150 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 151 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 152 153 static struct mtx mrouter_mtx; 154 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx) 155 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx) 156 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED) 157 #define MROUTER_LOCK_INIT() \ 158 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF) 159 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx) 160 161 static struct mtx mfc_mtx; 162 #define MFC_LOCK() mtx_lock(&mfc_mtx) 163 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 164 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED) 165 #define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF) 166 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 167 168 static struct vif viftable[MAXVIFS]; 169 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 170 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 171 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 172 173 static struct mtx vif_mtx; 174 #define VIF_LOCK() mtx_lock(&vif_mtx) 175 #define VIF_UNLOCK() mtx_unlock(&vif_mtx) 176 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 177 #define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF) 178 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 179 180 static u_char nexpire[MFCTBLSIZ]; 181 182 static eventhandler_tag if_detach_event_tag = NULL; 183 184 static struct callout expire_upcalls_ch; 185 186 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 187 #define UPCALL_EXPIRE 6 /* number of timeouts */ 188 189 #define ENCAP_TTL 64 190 191 /* 192 * Bandwidth meter variables and constants 193 */ 194 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 195 /* 196 * Pending timeouts are stored in a hash table, the key being the 197 * expiration time. Periodically, the entries are analysed and processed. 198 */ 199 #define BW_METER_BUCKETS 1024 200 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 201 static struct callout bw_meter_ch; 202 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 203 204 /* 205 * Pending upcalls are stored in a vector which is flushed when 206 * full, or periodically 207 */ 208 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 209 static u_int bw_upcalls_n; /* # of pending upcalls */ 210 static struct callout bw_upcalls_ch; 211 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 212 213 static struct pimstat pimstat; 214 215 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); 216 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 217 &pimstat, pimstat, 218 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 219 220 static u_long pim_squelch_wholepkt = 0; 221 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, 222 &pim_squelch_wholepkt, 0, 223 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); 224 225 extern struct domain inetdomain; 226 struct protosw in_pim_protosw = { 227 .pr_type = SOCK_RAW, 228 .pr_domain = &inetdomain, 229 .pr_protocol = IPPROTO_PIM, 230 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, 231 .pr_input = pim_input, 232 .pr_output = (pr_output_t*)rip_output, 233 .pr_ctloutput = rip_ctloutput, 234 .pr_usrreqs = &rip_usrreqs 235 }; 236 static const struct encaptab *pim_encap_cookie; 237 238 #ifdef INET6 239 /* ip6_mroute.c glue */ 240 extern struct in6_protosw in6_pim_protosw; 241 static const struct encaptab *pim6_encap_cookie; 242 243 extern int X_ip6_mrouter_set(struct socket *, struct sockopt *); 244 extern int X_ip6_mrouter_get(struct socket *, struct sockopt *); 245 extern int X_ip6_mrouter_done(void); 246 extern int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *); 247 extern int X_mrt6_ioctl(int, caddr_t); 248 #endif 249 250 static int pim_encapcheck(const struct mbuf *, int, int, void *); 251 252 /* 253 * Note: the PIM Register encapsulation adds the following in front of a 254 * data packet: 255 * 256 * struct pim_encap_hdr { 257 * struct ip ip; 258 * struct pim_encap_pimhdr pim; 259 * } 260 * 261 */ 262 263 struct pim_encap_pimhdr { 264 struct pim pim; 265 uint32_t flags; 266 }; 267 268 static struct ip pim_encap_iphdr = { 269 #if BYTE_ORDER == LITTLE_ENDIAN 270 sizeof(struct ip) >> 2, 271 IPVERSION, 272 #else 273 IPVERSION, 274 sizeof(struct ip) >> 2, 275 #endif 276 0, /* tos */ 277 sizeof(struct ip), /* total length */ 278 0, /* id */ 279 0, /* frag offset */ 280 ENCAP_TTL, 281 IPPROTO_PIM, 282 0, /* checksum */ 283 }; 284 285 static struct pim_encap_pimhdr pim_encap_pimhdr = { 286 { 287 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 288 0, /* reserved */ 289 0, /* checksum */ 290 }, 291 0 /* flags */ 292 }; 293 294 static struct ifnet multicast_register_if; 295 static vifi_t reg_vif_num = VIFI_INVALID; 296 297 /* 298 * Private variables. 299 */ 300 static vifi_t numvifs; 301 302 static u_long X_ip_mcast_src(int vifi); 303 static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 304 struct mbuf *m, struct ip_moptions *imo); 305 static int X_ip_mrouter_done(void); 306 static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 307 static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 308 static int X_legal_vif_num(int vif); 309 static int X_mrt_ioctl(int cmd, caddr_t data, int fibnum); 310 311 static int get_sg_cnt(struct sioc_sg_req *); 312 static int get_vif_cnt(struct sioc_vif_req *); 313 static void if_detached_event(void *arg __unused, struct ifnet *); 314 static int ip_mrouter_init(struct socket *, int); 315 static int add_vif(struct vifctl *); 316 static int del_vif_locked(vifi_t); 317 static int del_vif(vifi_t); 318 static int add_mfc(struct mfcctl2 *); 319 static int del_mfc(struct mfcctl2 *); 320 static int set_api_config(uint32_t *); /* chose API capabilities */ 321 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 322 static int set_assert(int); 323 static void expire_upcalls(void *); 324 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 325 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 326 static void send_packet(struct vif *, struct mbuf *); 327 328 /* 329 * Bandwidth monitoring 330 */ 331 static void free_bw_list(struct bw_meter *list); 332 static int add_bw_upcall(struct bw_upcall *); 333 static int del_bw_upcall(struct bw_upcall *); 334 static void bw_meter_receive_packet(struct bw_meter *x, int plen, 335 struct timeval *nowp); 336 static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 337 static void bw_upcalls_send(void); 338 static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 339 static void unschedule_bw_meter(struct bw_meter *x); 340 static void bw_meter_process(void); 341 static void expire_bw_upcalls_send(void *); 342 static void expire_bw_meter_process(void *); 343 344 static int pim_register_send(struct ip *, struct vif *, 345 struct mbuf *, struct mfc *); 346 static int pim_register_send_rp(struct ip *, struct vif *, 347 struct mbuf *, struct mfc *); 348 static int pim_register_send_upcall(struct ip *, struct vif *, 349 struct mbuf *, struct mfc *); 350 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 351 352 /* 353 * whether or not special PIM assert processing is enabled. 354 */ 355 static int pim_assert; 356 /* 357 * Rate limit for assert notification messages, in usec 358 */ 359 #define ASSERT_MSG_TIME 3000000 360 361 /* 362 * Kernel multicast routing API capabilities and setup. 363 * If more API capabilities are added to the kernel, they should be 364 * recorded in `mrt_api_support'. 365 */ 366 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 367 MRT_MFC_FLAGS_BORDER_VIF | 368 MRT_MFC_RP | 369 MRT_MFC_BW_UPCALL); 370 static uint32_t mrt_api_config = 0; 371 372 /* 373 * Hash function for a source, group entry 374 */ 375 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 376 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 377 378 /* 379 * Find a route for a given origin IP address and Multicast group address 380 * Statistics are updated by the caller if needed 381 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 382 */ 383 static struct mfc * 384 mfc_find(in_addr_t o, in_addr_t g) 385 { 386 struct mfc *rt; 387 388 MFC_LOCK_ASSERT(); 389 390 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 391 if ((rt->mfc_origin.s_addr == o) && 392 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 393 break; 394 return rt; 395 } 396 397 /* 398 * Macros to compute elapsed time efficiently 399 * Borrowed from Van Jacobson's scheduling code 400 */ 401 #define TV_DELTA(a, b, delta) { \ 402 int xxs; \ 403 delta = (a).tv_usec - (b).tv_usec; \ 404 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 405 switch (xxs) { \ 406 case 2: \ 407 delta += 1000000; \ 408 /* FALLTHROUGH */ \ 409 case 1: \ 410 delta += 1000000; \ 411 break; \ 412 default: \ 413 delta += (1000000 * xxs); \ 414 } \ 415 } \ 416 } 417 418 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 419 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 420 421 /* 422 * Handle MRT setsockopt commands to modify the multicast routing tables. 423 */ 424 static int 425 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 426 { 427 INIT_VNET_INET(curvnet); 428 int error, optval; 429 vifi_t vifi; 430 struct vifctl vifc; 431 struct mfcctl2 mfc; 432 struct bw_upcall bw_upcall; 433 uint32_t i; 434 435 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT) 436 return EPERM; 437 438 error = 0; 439 switch (sopt->sopt_name) { 440 case MRT_INIT: 441 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 442 if (error) 443 break; 444 error = ip_mrouter_init(so, optval); 445 break; 446 447 case MRT_DONE: 448 error = ip_mrouter_done(); 449 break; 450 451 case MRT_ADD_VIF: 452 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 453 if (error) 454 break; 455 error = add_vif(&vifc); 456 break; 457 458 case MRT_DEL_VIF: 459 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 460 if (error) 461 break; 462 error = del_vif(vifi); 463 break; 464 465 case MRT_ADD_MFC: 466 case MRT_DEL_MFC: 467 /* 468 * select data size depending on API version. 469 */ 470 if (sopt->sopt_name == MRT_ADD_MFC && 471 mrt_api_config & MRT_API_FLAGS_ALL) { 472 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 473 sizeof(struct mfcctl2)); 474 } else { 475 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 476 sizeof(struct mfcctl)); 477 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 478 sizeof(mfc) - sizeof(struct mfcctl)); 479 } 480 if (error) 481 break; 482 if (sopt->sopt_name == MRT_ADD_MFC) 483 error = add_mfc(&mfc); 484 else 485 error = del_mfc(&mfc); 486 break; 487 488 case MRT_ASSERT: 489 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 490 if (error) 491 break; 492 set_assert(optval); 493 break; 494 495 case MRT_API_CONFIG: 496 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 497 if (!error) 498 error = set_api_config(&i); 499 if (!error) 500 error = sooptcopyout(sopt, &i, sizeof i); 501 break; 502 503 case MRT_ADD_BW_UPCALL: 504 case MRT_DEL_BW_UPCALL: 505 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 506 sizeof bw_upcall); 507 if (error) 508 break; 509 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 510 error = add_bw_upcall(&bw_upcall); 511 else 512 error = del_bw_upcall(&bw_upcall); 513 break; 514 515 default: 516 error = EOPNOTSUPP; 517 break; 518 } 519 return error; 520 } 521 522 /* 523 * Handle MRT getsockopt commands 524 */ 525 static int 526 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 527 { 528 int error; 529 static int version = 0x0305; /* !!! why is this here? XXX */ 530 531 switch (sopt->sopt_name) { 532 case MRT_VERSION: 533 error = sooptcopyout(sopt, &version, sizeof version); 534 break; 535 536 case MRT_ASSERT: 537 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 538 break; 539 540 case MRT_API_SUPPORT: 541 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 542 break; 543 544 case MRT_API_CONFIG: 545 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 546 break; 547 548 default: 549 error = EOPNOTSUPP; 550 break; 551 } 552 return error; 553 } 554 555 /* 556 * Handle ioctl commands to obtain information from the cache 557 */ 558 static int 559 X_mrt_ioctl(int cmd, caddr_t data, int fibnum) 560 { 561 int error = 0; 562 563 /* 564 * Currently the only function calling this ioctl routine is rtioctl(). 565 * Typically, only root can create the raw socket in order to execute 566 * this ioctl method, however the request might be coming from a prison 567 */ 568 error = priv_check(curthread, PRIV_NETINET_MROUTE); 569 if (error) 570 return (error); 571 switch (cmd) { 572 case (SIOCGETVIFCNT): 573 error = get_vif_cnt((struct sioc_vif_req *)data); 574 break; 575 576 case (SIOCGETSGCNT): 577 error = get_sg_cnt((struct sioc_sg_req *)data); 578 break; 579 580 default: 581 error = EINVAL; 582 break; 583 } 584 return error; 585 } 586 587 /* 588 * returns the packet, byte, rpf-failure count for the source group provided 589 */ 590 static int 591 get_sg_cnt(struct sioc_sg_req *req) 592 { 593 struct mfc *rt; 594 595 MFC_LOCK(); 596 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 597 if (rt == NULL) { 598 MFC_UNLOCK(); 599 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 600 return EADDRNOTAVAIL; 601 } 602 req->pktcnt = rt->mfc_pkt_cnt; 603 req->bytecnt = rt->mfc_byte_cnt; 604 req->wrong_if = rt->mfc_wrong_if; 605 MFC_UNLOCK(); 606 return 0; 607 } 608 609 /* 610 * returns the input and output packet and byte counts on the vif provided 611 */ 612 static int 613 get_vif_cnt(struct sioc_vif_req *req) 614 { 615 vifi_t vifi = req->vifi; 616 617 VIF_LOCK(); 618 if (vifi >= numvifs) { 619 VIF_UNLOCK(); 620 return EINVAL; 621 } 622 623 req->icount = viftable[vifi].v_pkt_in; 624 req->ocount = viftable[vifi].v_pkt_out; 625 req->ibytes = viftable[vifi].v_bytes_in; 626 req->obytes = viftable[vifi].v_bytes_out; 627 VIF_UNLOCK(); 628 629 return 0; 630 } 631 632 static void 633 ip_mrouter_reset(void) 634 { 635 bzero((caddr_t)mfctable, sizeof(mfctable)); 636 bzero((caddr_t)nexpire, sizeof(nexpire)); 637 638 pim_assert = 0; 639 mrt_api_config = 0; 640 641 callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE); 642 643 bw_upcalls_n = 0; 644 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 645 callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE); 646 callout_init(&bw_meter_ch, CALLOUT_MPSAFE); 647 } 648 649 static void 650 if_detached_event(void *arg __unused, struct ifnet *ifp) 651 { 652 INIT_VNET_INET(curvnet); 653 vifi_t vifi; 654 int i; 655 struct mfc *mfc; 656 struct mfc *nmfc; 657 struct mfc **ppmfc; /* Pointer to previous node's next-pointer */ 658 struct rtdetq *pq; 659 struct rtdetq *npq; 660 661 MROUTER_LOCK(); 662 if (V_ip_mrouter == NULL) { 663 MROUTER_UNLOCK(); 664 } 665 666 /* 667 * Tear down multicast forwarder state associated with this ifnet. 668 * 1. Walk the vif list, matching vifs against this ifnet. 669 * 2. Walk the multicast forwarding cache (mfc) looking for 670 * inner matches with this vif's index. 671 * 3. Free any pending mbufs for this mfc. 672 * 4. Free the associated mfc entry and state associated with this vif. 673 * Be very careful about unlinking from a singly-linked list whose 674 * "head node" is a pointer in a simple array. 675 * 5. Free vif state. This should disable ALLMULTI on the interface. 676 */ 677 VIF_LOCK(); 678 MFC_LOCK(); 679 for (vifi = 0; vifi < numvifs; vifi++) { 680 if (viftable[vifi].v_ifp != ifp) 681 continue; 682 for (i = 0; i < MFCTBLSIZ; i++) { 683 ppmfc = &mfctable[i]; 684 for (mfc = mfctable[i]; mfc != NULL; ) { 685 nmfc = mfc->mfc_next; 686 if (mfc->mfc_parent == vifi) { 687 for (pq = mfc->mfc_stall; pq != NULL; ) { 688 npq = pq->next; 689 m_freem(pq->m); 690 free(pq, M_MRTABLE); 691 pq = npq; 692 } 693 free_bw_list(mfc->mfc_bw_meter); 694 free(mfc, M_MRTABLE); 695 *ppmfc = nmfc; 696 } else { 697 ppmfc = &mfc->mfc_next; 698 } 699 mfc = nmfc; 700 } 701 } 702 del_vif_locked(vifi); 703 } 704 MFC_UNLOCK(); 705 VIF_UNLOCK(); 706 707 MROUTER_UNLOCK(); 708 } 709 710 /* 711 * Enable multicast routing 712 */ 713 static int 714 ip_mrouter_init(struct socket *so, int version) 715 { 716 INIT_VNET_INET(curvnet); 717 718 if (mrtdebug) 719 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 720 so->so_type, so->so_proto->pr_protocol); 721 722 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 723 return EOPNOTSUPP; 724 725 if (version != 1) 726 return ENOPROTOOPT; 727 728 MROUTER_LOCK(); 729 730 if (V_ip_mrouter != NULL) { 731 MROUTER_UNLOCK(); 732 return EADDRINUSE; 733 } 734 735 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 736 if_detached_event, NULL, EVENTHANDLER_PRI_ANY); 737 if (if_detach_event_tag == NULL) { 738 MROUTER_UNLOCK(); 739 return (ENOMEM); 740 } 741 742 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 743 744 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 745 expire_bw_upcalls_send, NULL); 746 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 747 748 V_ip_mrouter = so; 749 750 MROUTER_UNLOCK(); 751 752 if (mrtdebug) 753 log(LOG_DEBUG, "ip_mrouter_init\n"); 754 755 return 0; 756 } 757 758 /* 759 * Disable multicast routing 760 */ 761 static int 762 X_ip_mrouter_done(void) 763 { 764 INIT_VNET_INET(curvnet); 765 vifi_t vifi; 766 int i; 767 struct ifnet *ifp; 768 struct ifreq ifr; 769 struct mfc *rt; 770 struct rtdetq *rte; 771 772 MROUTER_LOCK(); 773 774 if (V_ip_mrouter == NULL) { 775 MROUTER_UNLOCK(); 776 return EINVAL; 777 } 778 779 /* 780 * Detach/disable hooks to the reset of the system. 781 */ 782 V_ip_mrouter = NULL; 783 mrt_api_config = 0; 784 785 VIF_LOCK(); 786 /* 787 * For each phyint in use, disable promiscuous reception of all IP 788 * multicasts. 789 */ 790 for (vifi = 0; vifi < numvifs; vifi++) { 791 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 792 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 793 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 794 795 so->sin_len = sizeof(struct sockaddr_in); 796 so->sin_family = AF_INET; 797 so->sin_addr.s_addr = INADDR_ANY; 798 ifp = viftable[vifi].v_ifp; 799 if_allmulti(ifp, 0); 800 } 801 } 802 bzero((caddr_t)viftable, sizeof(viftable)); 803 numvifs = 0; 804 pim_assert = 0; 805 VIF_UNLOCK(); 806 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 807 808 /* 809 * Free all multicast forwarding cache entries. 810 */ 811 callout_stop(&expire_upcalls_ch); 812 callout_stop(&bw_upcalls_ch); 813 callout_stop(&bw_meter_ch); 814 815 MFC_LOCK(); 816 for (i = 0; i < MFCTBLSIZ; i++) { 817 for (rt = mfctable[i]; rt != NULL; ) { 818 struct mfc *nr = rt->mfc_next; 819 820 for (rte = rt->mfc_stall; rte != NULL; ) { 821 struct rtdetq *n = rte->next; 822 823 m_freem(rte->m); 824 free(rte, M_MRTABLE); 825 rte = n; 826 } 827 free_bw_list(rt->mfc_bw_meter); 828 free(rt, M_MRTABLE); 829 rt = nr; 830 } 831 } 832 bzero((caddr_t)mfctable, sizeof(mfctable)); 833 bzero((caddr_t)nexpire, sizeof(nexpire)); 834 bw_upcalls_n = 0; 835 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 836 MFC_UNLOCK(); 837 838 reg_vif_num = VIFI_INVALID; 839 840 MROUTER_UNLOCK(); 841 842 if (mrtdebug) 843 log(LOG_DEBUG, "ip_mrouter_done\n"); 844 845 return 0; 846 } 847 848 /* 849 * Set PIM assert processing global 850 */ 851 static int 852 set_assert(int i) 853 { 854 if ((i != 1) && (i != 0)) 855 return EINVAL; 856 857 pim_assert = i; 858 859 return 0; 860 } 861 862 /* 863 * Configure API capabilities 864 */ 865 int 866 set_api_config(uint32_t *apival) 867 { 868 int i; 869 870 /* 871 * We can set the API capabilities only if it is the first operation 872 * after MRT_INIT. I.e.: 873 * - there are no vifs installed 874 * - pim_assert is not enabled 875 * - the MFC table is empty 876 */ 877 if (numvifs > 0) { 878 *apival = 0; 879 return EPERM; 880 } 881 if (pim_assert) { 882 *apival = 0; 883 return EPERM; 884 } 885 for (i = 0; i < MFCTBLSIZ; i++) { 886 if (mfctable[i] != NULL) { 887 *apival = 0; 888 return EPERM; 889 } 890 } 891 892 mrt_api_config = *apival & mrt_api_support; 893 *apival = mrt_api_config; 894 895 return 0; 896 } 897 898 /* 899 * Add a vif to the vif table 900 */ 901 static int 902 add_vif(struct vifctl *vifcp) 903 { 904 struct vif *vifp = viftable + vifcp->vifc_vifi; 905 struct sockaddr_in sin = {sizeof sin, AF_INET}; 906 struct ifaddr *ifa; 907 struct ifnet *ifp; 908 int error; 909 910 VIF_LOCK(); 911 if (vifcp->vifc_vifi >= MAXVIFS) { 912 VIF_UNLOCK(); 913 return EINVAL; 914 } 915 /* rate limiting is no longer supported by this code */ 916 if (vifcp->vifc_rate_limit != 0) { 917 log(LOG_ERR, "rate limiting is no longer supported\n"); 918 VIF_UNLOCK(); 919 return EINVAL; 920 } 921 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) { 922 VIF_UNLOCK(); 923 return EADDRINUSE; 924 } 925 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) { 926 VIF_UNLOCK(); 927 return EADDRNOTAVAIL; 928 } 929 930 /* Find the interface with an address in AF_INET family */ 931 if (vifcp->vifc_flags & VIFF_REGISTER) { 932 /* 933 * XXX: Because VIFF_REGISTER does not really need a valid 934 * local interface (e.g. it could be 127.0.0.2), we don't 935 * check its address. 936 */ 937 ifp = NULL; 938 } else { 939 sin.sin_addr = vifcp->vifc_lcl_addr; 940 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 941 if (ifa == NULL) { 942 VIF_UNLOCK(); 943 return EADDRNOTAVAIL; 944 } 945 ifp = ifa->ifa_ifp; 946 } 947 948 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { 949 log(LOG_ERR, "tunnels are no longer supported\n"); 950 VIF_UNLOCK(); 951 return EOPNOTSUPP; 952 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 953 ifp = &multicast_register_if; 954 if (mrtdebug) 955 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 956 (void *)&multicast_register_if); 957 if (reg_vif_num == VIFI_INVALID) { 958 if_initname(&multicast_register_if, "register_vif", 0); 959 multicast_register_if.if_flags = IFF_LOOPBACK; 960 reg_vif_num = vifcp->vifc_vifi; 961 } 962 } else { /* Make sure the interface supports multicast */ 963 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 964 VIF_UNLOCK(); 965 return EOPNOTSUPP; 966 } 967 968 /* Enable promiscuous reception of all IP multicasts from the if */ 969 error = if_allmulti(ifp, 1); 970 if (error) { 971 VIF_UNLOCK(); 972 return error; 973 } 974 } 975 976 vifp->v_flags = vifcp->vifc_flags; 977 vifp->v_threshold = vifcp->vifc_threshold; 978 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 979 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 980 vifp->v_ifp = ifp; 981 vifp->v_rsvp_on = 0; 982 vifp->v_rsvpd = NULL; 983 /* initialize per vif pkt counters */ 984 vifp->v_pkt_in = 0; 985 vifp->v_pkt_out = 0; 986 vifp->v_bytes_in = 0; 987 vifp->v_bytes_out = 0; 988 bzero(&vifp->v_route, sizeof(vifp->v_route)); 989 990 /* Adjust numvifs up if the vifi is higher than numvifs */ 991 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 992 993 VIF_UNLOCK(); 994 995 if (mrtdebug) 996 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x\n", 997 vifcp->vifc_vifi, 998 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 999 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 1000 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 1001 vifcp->vifc_threshold); 1002 1003 return 0; 1004 } 1005 1006 /* 1007 * Delete a vif from the vif table 1008 */ 1009 static int 1010 del_vif_locked(vifi_t vifi) 1011 { 1012 struct vif *vifp; 1013 1014 VIF_LOCK_ASSERT(); 1015 1016 if (vifi >= numvifs) { 1017 return EINVAL; 1018 } 1019 vifp = &viftable[vifi]; 1020 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) { 1021 return EADDRNOTAVAIL; 1022 } 1023 1024 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 1025 if_allmulti(vifp->v_ifp, 0); 1026 1027 if (vifp->v_flags & VIFF_REGISTER) 1028 reg_vif_num = VIFI_INVALID; 1029 1030 bzero((caddr_t)vifp, sizeof (*vifp)); 1031 1032 if (mrtdebug) 1033 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1034 1035 /* Adjust numvifs down */ 1036 for (vifi = numvifs; vifi > 0; vifi--) 1037 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1038 break; 1039 numvifs = vifi; 1040 1041 return 0; 1042 } 1043 1044 static int 1045 del_vif(vifi_t vifi) 1046 { 1047 int cc; 1048 1049 VIF_LOCK(); 1050 cc = del_vif_locked(vifi); 1051 VIF_UNLOCK(); 1052 1053 return cc; 1054 } 1055 1056 /* 1057 * update an mfc entry without resetting counters and S,G addresses. 1058 */ 1059 static void 1060 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1061 { 1062 int i; 1063 1064 rt->mfc_parent = mfccp->mfcc_parent; 1065 for (i = 0; i < numvifs; i++) { 1066 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1067 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1068 MRT_MFC_FLAGS_ALL; 1069 } 1070 /* set the RP address */ 1071 if (mrt_api_config & MRT_MFC_RP) 1072 rt->mfc_rp = mfccp->mfcc_rp; 1073 else 1074 rt->mfc_rp.s_addr = INADDR_ANY; 1075 } 1076 1077 /* 1078 * fully initialize an mfc entry from the parameter. 1079 */ 1080 static void 1081 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1082 { 1083 rt->mfc_origin = mfccp->mfcc_origin; 1084 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1085 1086 update_mfc_params(rt, mfccp); 1087 1088 /* initialize pkt counters per src-grp */ 1089 rt->mfc_pkt_cnt = 0; 1090 rt->mfc_byte_cnt = 0; 1091 rt->mfc_wrong_if = 0; 1092 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1093 } 1094 1095 1096 /* 1097 * Add an mfc entry 1098 */ 1099 static int 1100 add_mfc(struct mfcctl2 *mfccp) 1101 { 1102 struct mfc *rt; 1103 u_long hash; 1104 struct rtdetq *rte; 1105 u_short nstl; 1106 1107 VIF_LOCK(); 1108 MFC_LOCK(); 1109 1110 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1111 1112 /* If an entry already exists, just update the fields */ 1113 if (rt) { 1114 if (mrtdebug & DEBUG_MFC) 1115 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1116 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1117 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1118 mfccp->mfcc_parent); 1119 1120 update_mfc_params(rt, mfccp); 1121 MFC_UNLOCK(); 1122 VIF_UNLOCK(); 1123 return 0; 1124 } 1125 1126 /* 1127 * Find the entry for which the upcall was made and update 1128 */ 1129 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1130 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1131 1132 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1133 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1134 (rt->mfc_stall != NULL)) { 1135 1136 if (nstl++) 1137 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1138 "multiple kernel entries", 1139 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1140 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1141 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1142 1143 if (mrtdebug & DEBUG_MFC) 1144 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1145 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1146 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1147 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1148 1149 init_mfc_params(rt, mfccp); 1150 1151 rt->mfc_expire = 0; /* Don't clean this guy up */ 1152 nexpire[hash]--; 1153 1154 /* free packets Qed at the end of this entry */ 1155 for (rte = rt->mfc_stall; rte != NULL; ) { 1156 struct rtdetq *n = rte->next; 1157 1158 ip_mdq(rte->m, rte->ifp, rt, -1); 1159 m_freem(rte->m); 1160 free(rte, M_MRTABLE); 1161 rte = n; 1162 } 1163 rt->mfc_stall = NULL; 1164 } 1165 } 1166 1167 /* 1168 * It is possible that an entry is being inserted without an upcall 1169 */ 1170 if (nstl == 0) { 1171 if (mrtdebug & DEBUG_MFC) 1172 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1173 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1174 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1175 mfccp->mfcc_parent); 1176 1177 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1178 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1179 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1180 init_mfc_params(rt, mfccp); 1181 if (rt->mfc_expire) 1182 nexpire[hash]--; 1183 rt->mfc_expire = 0; 1184 break; /* XXX */ 1185 } 1186 } 1187 if (rt == NULL) { /* no upcall, so make a new entry */ 1188 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1189 if (rt == NULL) { 1190 MFC_UNLOCK(); 1191 VIF_UNLOCK(); 1192 return ENOBUFS; 1193 } 1194 1195 init_mfc_params(rt, mfccp); 1196 rt->mfc_expire = 0; 1197 rt->mfc_stall = NULL; 1198 1199 rt->mfc_bw_meter = NULL; 1200 /* insert new entry at head of hash chain */ 1201 rt->mfc_next = mfctable[hash]; 1202 mfctable[hash] = rt; 1203 } 1204 } 1205 MFC_UNLOCK(); 1206 VIF_UNLOCK(); 1207 return 0; 1208 } 1209 1210 /* 1211 * Delete an mfc entry 1212 */ 1213 static int 1214 del_mfc(struct mfcctl2 *mfccp) 1215 { 1216 struct in_addr origin; 1217 struct in_addr mcastgrp; 1218 struct mfc *rt; 1219 struct mfc **nptr; 1220 u_long hash; 1221 struct bw_meter *list; 1222 1223 origin = mfccp->mfcc_origin; 1224 mcastgrp = mfccp->mfcc_mcastgrp; 1225 1226 if (mrtdebug & DEBUG_MFC) 1227 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1228 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1229 1230 MFC_LOCK(); 1231 1232 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1233 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1234 if (origin.s_addr == rt->mfc_origin.s_addr && 1235 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1236 rt->mfc_stall == NULL) 1237 break; 1238 if (rt == NULL) { 1239 MFC_UNLOCK(); 1240 return EADDRNOTAVAIL; 1241 } 1242 1243 *nptr = rt->mfc_next; 1244 1245 /* 1246 * free the bw_meter entries 1247 */ 1248 list = rt->mfc_bw_meter; 1249 rt->mfc_bw_meter = NULL; 1250 1251 free(rt, M_MRTABLE); 1252 1253 free_bw_list(list); 1254 1255 MFC_UNLOCK(); 1256 1257 return 0; 1258 } 1259 1260 /* 1261 * Send a message to the routing daemon on the multicast routing socket 1262 */ 1263 static int 1264 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1265 { 1266 if (s) { 1267 SOCKBUF_LOCK(&s->so_rcv); 1268 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, 1269 NULL) != 0) { 1270 sorwakeup_locked(s); 1271 return 0; 1272 } 1273 SOCKBUF_UNLOCK(&s->so_rcv); 1274 } 1275 m_freem(mm); 1276 return -1; 1277 } 1278 1279 /* 1280 * IP multicast forwarding function. This function assumes that the packet 1281 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1282 * pointed to by "ifp", and the packet is to be relayed to other networks 1283 * that have members of the packet's destination IP multicast group. 1284 * 1285 * The packet is returned unscathed to the caller, unless it is 1286 * erroneous, in which case a non-zero return value tells the caller to 1287 * discard it. 1288 */ 1289 1290 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1291 1292 static int 1293 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1294 struct ip_moptions *imo) 1295 { 1296 INIT_VNET_INET(curvnet); 1297 struct mfc *rt; 1298 int error; 1299 vifi_t vifi; 1300 1301 if (mrtdebug & DEBUG_FORWARD) 1302 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1303 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1304 (void *)ifp); 1305 1306 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1307 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1308 /* 1309 * Packet arrived via a physical interface or 1310 * an encapsulated tunnel or a register_vif. 1311 */ 1312 } else { 1313 /* 1314 * Packet arrived through a source-route tunnel. 1315 * Source-route tunnels are no longer supported. 1316 */ 1317 static int last_log; 1318 if (last_log != time_uptime) { 1319 last_log = time_uptime; 1320 log(LOG_ERR, 1321 "ip_mforward: received source-routed packet from %lx\n", 1322 (u_long)ntohl(ip->ip_src.s_addr)); 1323 } 1324 return 1; 1325 } 1326 1327 VIF_LOCK(); 1328 MFC_LOCK(); 1329 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1330 if (ip->ip_ttl < MAXTTL) 1331 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1332 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1333 struct vif *vifp = viftable + vifi; 1334 1335 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1336 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1337 vifi, 1338 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1339 vifp->v_ifp->if_xname); 1340 } 1341 error = ip_mdq(m, ifp, NULL, vifi); 1342 MFC_UNLOCK(); 1343 VIF_UNLOCK(); 1344 return error; 1345 } 1346 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1347 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1348 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1349 if (!imo) 1350 printf("In fact, no options were specified at all\n"); 1351 } 1352 1353 /* 1354 * Don't forward a packet with time-to-live of zero or one, 1355 * or a packet destined to a local-only group. 1356 */ 1357 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) { 1358 MFC_UNLOCK(); 1359 VIF_UNLOCK(); 1360 return 0; 1361 } 1362 1363 /* 1364 * Determine forwarding vifs from the forwarding cache table 1365 */ 1366 ++mrtstat.mrts_mfc_lookups; 1367 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1368 1369 /* Entry exists, so forward if necessary */ 1370 if (rt != NULL) { 1371 error = ip_mdq(m, ifp, rt, -1); 1372 MFC_UNLOCK(); 1373 VIF_UNLOCK(); 1374 return error; 1375 } else { 1376 /* 1377 * If we don't have a route for packet's origin, 1378 * Make a copy of the packet & send message to routing daemon 1379 */ 1380 1381 struct mbuf *mb0; 1382 struct rtdetq *rte; 1383 u_long hash; 1384 int hlen = ip->ip_hl << 2; 1385 1386 ++mrtstat.mrts_mfc_misses; 1387 1388 mrtstat.mrts_no_route++; 1389 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1390 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1391 (u_long)ntohl(ip->ip_src.s_addr), 1392 (u_long)ntohl(ip->ip_dst.s_addr)); 1393 1394 /* 1395 * Allocate mbufs early so that we don't do extra work if we are 1396 * just going to fail anyway. Make sure to pullup the header so 1397 * that other people can't step on it. 1398 */ 1399 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1400 if (rte == NULL) { 1401 MFC_UNLOCK(); 1402 VIF_UNLOCK(); 1403 return ENOBUFS; 1404 } 1405 mb0 = m_copypacket(m, M_DONTWAIT); 1406 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1407 mb0 = m_pullup(mb0, hlen); 1408 if (mb0 == NULL) { 1409 free(rte, M_MRTABLE); 1410 MFC_UNLOCK(); 1411 VIF_UNLOCK(); 1412 return ENOBUFS; 1413 } 1414 1415 /* is there an upcall waiting for this flow ? */ 1416 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1417 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1418 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1419 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1420 (rt->mfc_stall != NULL)) 1421 break; 1422 } 1423 1424 if (rt == NULL) { 1425 int i; 1426 struct igmpmsg *im; 1427 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1428 struct mbuf *mm; 1429 1430 /* 1431 * Locate the vifi for the incoming interface for this packet. 1432 * If none found, drop packet. 1433 */ 1434 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1435 ; 1436 if (vifi >= numvifs) /* vif not found, drop packet */ 1437 goto non_fatal; 1438 1439 /* no upcall, so make a new entry */ 1440 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1441 if (rt == NULL) 1442 goto fail; 1443 /* Make a copy of the header to send to the user level process */ 1444 mm = m_copy(mb0, 0, hlen); 1445 if (mm == NULL) 1446 goto fail1; 1447 1448 /* 1449 * Send message to routing daemon to install 1450 * a route into the kernel table 1451 */ 1452 1453 im = mtod(mm, struct igmpmsg *); 1454 im->im_msgtype = IGMPMSG_NOCACHE; 1455 im->im_mbz = 0; 1456 im->im_vif = vifi; 1457 1458 mrtstat.mrts_upcalls++; 1459 1460 k_igmpsrc.sin_addr = ip->ip_src; 1461 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { 1462 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1463 ++mrtstat.mrts_upq_sockfull; 1464 fail1: 1465 free(rt, M_MRTABLE); 1466 fail: 1467 free(rte, M_MRTABLE); 1468 m_freem(mb0); 1469 MFC_UNLOCK(); 1470 VIF_UNLOCK(); 1471 return ENOBUFS; 1472 } 1473 1474 /* insert new entry at head of hash chain */ 1475 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1476 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1477 rt->mfc_expire = UPCALL_EXPIRE; 1478 nexpire[hash]++; 1479 for (i = 0; i < numvifs; i++) { 1480 rt->mfc_ttls[i] = 0; 1481 rt->mfc_flags[i] = 0; 1482 } 1483 rt->mfc_parent = -1; 1484 1485 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1486 1487 rt->mfc_bw_meter = NULL; 1488 1489 /* link into table */ 1490 rt->mfc_next = mfctable[hash]; 1491 mfctable[hash] = rt; 1492 rt->mfc_stall = rte; 1493 1494 } else { 1495 /* determine if q has overflowed */ 1496 int npkts = 0; 1497 struct rtdetq **p; 1498 1499 /* 1500 * XXX ouch! we need to append to the list, but we 1501 * only have a pointer to the front, so we have to 1502 * scan the entire list every time. 1503 */ 1504 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1505 npkts++; 1506 1507 if (npkts > MAX_UPQ) { 1508 mrtstat.mrts_upq_ovflw++; 1509 non_fatal: 1510 free(rte, M_MRTABLE); 1511 m_freem(mb0); 1512 MFC_UNLOCK(); 1513 VIF_UNLOCK(); 1514 return 0; 1515 } 1516 1517 /* Add this entry to the end of the queue */ 1518 *p = rte; 1519 } 1520 1521 rte->m = mb0; 1522 rte->ifp = ifp; 1523 rte->next = NULL; 1524 1525 MFC_UNLOCK(); 1526 VIF_UNLOCK(); 1527 1528 return 0; 1529 } 1530 } 1531 1532 /* 1533 * Clean up the cache entry if upcall is not serviced 1534 */ 1535 static void 1536 expire_upcalls(void *unused) 1537 { 1538 struct rtdetq *rte; 1539 struct mfc *mfc, **nptr; 1540 int i; 1541 1542 MFC_LOCK(); 1543 for (i = 0; i < MFCTBLSIZ; i++) { 1544 if (nexpire[i] == 0) 1545 continue; 1546 nptr = &mfctable[i]; 1547 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1548 /* 1549 * Skip real cache entries 1550 * Make sure it wasn't marked to not expire (shouldn't happen) 1551 * If it expires now 1552 */ 1553 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1554 --mfc->mfc_expire == 0) { 1555 if (mrtdebug & DEBUG_EXPIRE) 1556 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1557 (u_long)ntohl(mfc->mfc_origin.s_addr), 1558 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1559 /* 1560 * drop all the packets 1561 * free the mbuf with the pkt, if, timing info 1562 */ 1563 for (rte = mfc->mfc_stall; rte; ) { 1564 struct rtdetq *n = rte->next; 1565 1566 m_freem(rte->m); 1567 free(rte, M_MRTABLE); 1568 rte = n; 1569 } 1570 ++mrtstat.mrts_cache_cleanups; 1571 nexpire[i]--; 1572 1573 /* 1574 * free the bw_meter entries 1575 */ 1576 while (mfc->mfc_bw_meter != NULL) { 1577 struct bw_meter *x = mfc->mfc_bw_meter; 1578 1579 mfc->mfc_bw_meter = x->bm_mfc_next; 1580 free(x, M_BWMETER); 1581 } 1582 1583 *nptr = mfc->mfc_next; 1584 free(mfc, M_MRTABLE); 1585 } else { 1586 nptr = &mfc->mfc_next; 1587 } 1588 } 1589 } 1590 MFC_UNLOCK(); 1591 1592 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1593 } 1594 1595 /* 1596 * Packet forwarding routine once entry in the cache is made 1597 */ 1598 static int 1599 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1600 { 1601 INIT_VNET_INET(curvnet); 1602 struct ip *ip = mtod(m, struct ip *); 1603 vifi_t vifi; 1604 int plen = ip->ip_len; 1605 1606 VIF_LOCK_ASSERT(); 1607 1608 /* 1609 * If xmt_vif is not -1, send on only the requested vif. 1610 * 1611 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1612 */ 1613 if (xmt_vif < numvifs) { 1614 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1615 pim_register_send(ip, viftable + xmt_vif, m, rt); 1616 else 1617 phyint_send(ip, viftable + xmt_vif, m); 1618 return 1; 1619 } 1620 1621 /* 1622 * Don't forward if it didn't arrive from the parent vif for its origin. 1623 */ 1624 vifi = rt->mfc_parent; 1625 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1626 /* came in the wrong interface */ 1627 if (mrtdebug & DEBUG_FORWARD) 1628 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1629 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1630 ++mrtstat.mrts_wrong_if; 1631 ++rt->mfc_wrong_if; 1632 /* 1633 * If we are doing PIM assert processing, send a message 1634 * to the routing daemon. 1635 * 1636 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1637 * can complete the SPT switch, regardless of the type 1638 * of the iif (broadcast media, GRE tunnel, etc). 1639 */ 1640 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1641 struct timeval now; 1642 u_long delta; 1643 1644 if (ifp == &multicast_register_if) 1645 pimstat.pims_rcv_registers_wrongiif++; 1646 1647 /* Get vifi for the incoming packet */ 1648 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1649 ; 1650 if (vifi >= numvifs) 1651 return 0; /* The iif is not found: ignore the packet. */ 1652 1653 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1654 return 0; /* WRONGVIF disabled: ignore the packet */ 1655 1656 GET_TIME(now); 1657 1658 TV_DELTA(now, rt->mfc_last_assert, delta); 1659 1660 if (delta > ASSERT_MSG_TIME) { 1661 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1662 struct igmpmsg *im; 1663 int hlen = ip->ip_hl << 2; 1664 struct mbuf *mm = m_copy(m, 0, hlen); 1665 1666 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1667 mm = m_pullup(mm, hlen); 1668 if (mm == NULL) 1669 return ENOBUFS; 1670 1671 rt->mfc_last_assert = now; 1672 1673 im = mtod(mm, struct igmpmsg *); 1674 im->im_msgtype = IGMPMSG_WRONGVIF; 1675 im->im_mbz = 0; 1676 im->im_vif = vifi; 1677 1678 mrtstat.mrts_upcalls++; 1679 1680 k_igmpsrc.sin_addr = im->im_src; 1681 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { 1682 log(LOG_WARNING, 1683 "ip_mforward: ip_mrouter socket queue full\n"); 1684 ++mrtstat.mrts_upq_sockfull; 1685 return ENOBUFS; 1686 } 1687 } 1688 } 1689 return 0; 1690 } 1691 1692 /* If I sourced this packet, it counts as output, else it was input. */ 1693 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1694 viftable[vifi].v_pkt_out++; 1695 viftable[vifi].v_bytes_out += plen; 1696 } else { 1697 viftable[vifi].v_pkt_in++; 1698 viftable[vifi].v_bytes_in += plen; 1699 } 1700 rt->mfc_pkt_cnt++; 1701 rt->mfc_byte_cnt += plen; 1702 1703 /* 1704 * For each vif, decide if a copy of the packet should be forwarded. 1705 * Forward if: 1706 * - the ttl exceeds the vif's threshold 1707 * - there are group members downstream on interface 1708 */ 1709 for (vifi = 0; vifi < numvifs; vifi++) 1710 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1711 viftable[vifi].v_pkt_out++; 1712 viftable[vifi].v_bytes_out += plen; 1713 if (viftable[vifi].v_flags & VIFF_REGISTER) 1714 pim_register_send(ip, viftable + vifi, m, rt); 1715 else 1716 phyint_send(ip, viftable + vifi, m); 1717 } 1718 1719 /* 1720 * Perform upcall-related bw measuring. 1721 */ 1722 if (rt->mfc_bw_meter != NULL) { 1723 struct bw_meter *x; 1724 struct timeval now; 1725 1726 GET_TIME(now); 1727 MFC_LOCK_ASSERT(); 1728 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1729 bw_meter_receive_packet(x, plen, &now); 1730 } 1731 1732 return 0; 1733 } 1734 1735 /* 1736 * check if a vif number is legal/ok. This is used by ip_output. 1737 */ 1738 static int 1739 X_legal_vif_num(int vif) 1740 { 1741 /* XXX unlocked, matter? */ 1742 return (vif >= 0 && vif < numvifs); 1743 } 1744 1745 /* 1746 * Return the local address used by this vif 1747 */ 1748 static u_long 1749 X_ip_mcast_src(int vifi) 1750 { 1751 /* XXX unlocked, matter? */ 1752 if (vifi >= 0 && vifi < numvifs) 1753 return viftable[vifi].v_lcl_addr.s_addr; 1754 else 1755 return INADDR_ANY; 1756 } 1757 1758 static void 1759 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1760 { 1761 struct mbuf *mb_copy; 1762 int hlen = ip->ip_hl << 2; 1763 1764 VIF_LOCK_ASSERT(); 1765 1766 /* 1767 * Make a new reference to the packet; make sure that 1768 * the IP header is actually copied, not just referenced, 1769 * so that ip_output() only scribbles on the copy. 1770 */ 1771 mb_copy = m_copypacket(m, M_DONTWAIT); 1772 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1773 mb_copy = m_pullup(mb_copy, hlen); 1774 if (mb_copy == NULL) 1775 return; 1776 1777 send_packet(vifp, mb_copy); 1778 } 1779 1780 static void 1781 send_packet(struct vif *vifp, struct mbuf *m) 1782 { 1783 struct ip_moptions imo; 1784 struct in_multi *imm[2]; 1785 int error; 1786 1787 VIF_LOCK_ASSERT(); 1788 1789 imo.imo_multicast_ifp = vifp->v_ifp; 1790 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1791 imo.imo_multicast_loop = 1; 1792 imo.imo_multicast_vif = -1; 1793 imo.imo_num_memberships = 0; 1794 imo.imo_max_memberships = 2; 1795 imo.imo_membership = &imm[0]; 1796 1797 /* 1798 * Re-entrancy should not be a problem here, because 1799 * the packets that we send out and are looped back at us 1800 * should get rejected because they appear to come from 1801 * the loopback interface, thus preventing looping. 1802 */ 1803 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL); 1804 if (mrtdebug & DEBUG_XMIT) { 1805 log(LOG_DEBUG, "phyint_send on vif %td err %d\n", 1806 vifp - viftable, error); 1807 } 1808 } 1809 1810 static int 1811 X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 1812 { 1813 INIT_VNET_INET(curvnet); 1814 int error, vifi; 1815 1816 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1817 return EOPNOTSUPP; 1818 1819 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 1820 if (error) 1821 return error; 1822 1823 VIF_LOCK(); 1824 1825 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 1826 VIF_UNLOCK(); 1827 return EADDRNOTAVAIL; 1828 } 1829 1830 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 1831 /* Check if socket is available. */ 1832 if (viftable[vifi].v_rsvpd != NULL) { 1833 VIF_UNLOCK(); 1834 return EADDRINUSE; 1835 } 1836 1837 viftable[vifi].v_rsvpd = so; 1838 /* This may seem silly, but we need to be sure we don't over-increment 1839 * the RSVP counter, in case something slips up. 1840 */ 1841 if (!viftable[vifi].v_rsvp_on) { 1842 viftable[vifi].v_rsvp_on = 1; 1843 V_rsvp_on++; 1844 } 1845 } else { /* must be VIF_OFF */ 1846 /* 1847 * XXX as an additional consistency check, one could make sure 1848 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 1849 * first parameter is pretty useless. 1850 */ 1851 viftable[vifi].v_rsvpd = NULL; 1852 /* 1853 * This may seem silly, but we need to be sure we don't over-decrement 1854 * the RSVP counter, in case something slips up. 1855 */ 1856 if (viftable[vifi].v_rsvp_on) { 1857 viftable[vifi].v_rsvp_on = 0; 1858 V_rsvp_on--; 1859 } 1860 } 1861 VIF_UNLOCK(); 1862 return 0; 1863 } 1864 1865 static void 1866 X_ip_rsvp_force_done(struct socket *so) 1867 { 1868 INIT_VNET_INET(curvnet); 1869 int vifi; 1870 1871 /* Don't bother if it is not the right type of socket. */ 1872 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1873 return; 1874 1875 VIF_LOCK(); 1876 1877 /* The socket may be attached to more than one vif...this 1878 * is perfectly legal. 1879 */ 1880 for (vifi = 0; vifi < numvifs; vifi++) { 1881 if (viftable[vifi].v_rsvpd == so) { 1882 viftable[vifi].v_rsvpd = NULL; 1883 /* This may seem silly, but we need to be sure we don't 1884 * over-decrement the RSVP counter, in case something slips up. 1885 */ 1886 if (viftable[vifi].v_rsvp_on) { 1887 viftable[vifi].v_rsvp_on = 0; 1888 V_rsvp_on--; 1889 } 1890 } 1891 } 1892 1893 VIF_UNLOCK(); 1894 } 1895 1896 static void 1897 X_rsvp_input(struct mbuf *m, int off) 1898 { 1899 INIT_VNET_INET(curvnet); 1900 int vifi; 1901 struct ip *ip = mtod(m, struct ip *); 1902 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 1903 struct ifnet *ifp; 1904 1905 if (rsvpdebug) 1906 printf("rsvp_input: rsvp_on %d\n", V_rsvp_on); 1907 1908 /* Can still get packets with rsvp_on = 0 if there is a local member 1909 * of the group to which the RSVP packet is addressed. But in this 1910 * case we want to throw the packet away. 1911 */ 1912 if (!V_rsvp_on) { 1913 m_freem(m); 1914 return; 1915 } 1916 1917 if (rsvpdebug) 1918 printf("rsvp_input: check vifs\n"); 1919 1920 #ifdef DIAGNOSTIC 1921 M_ASSERTPKTHDR(m); 1922 #endif 1923 1924 ifp = m->m_pkthdr.rcvif; 1925 1926 VIF_LOCK(); 1927 /* Find which vif the packet arrived on. */ 1928 for (vifi = 0; vifi < numvifs; vifi++) 1929 if (viftable[vifi].v_ifp == ifp) 1930 break; 1931 1932 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 1933 /* 1934 * Drop the lock here to avoid holding it across rip_input. 1935 * This could make rsvpdebug printfs wrong. If you care, 1936 * record the state of stuff before dropping the lock. 1937 */ 1938 VIF_UNLOCK(); 1939 /* 1940 * If the old-style non-vif-associated socket is set, 1941 * then use it. Otherwise, drop packet since there 1942 * is no specific socket for this vif. 1943 */ 1944 if (V_ip_rsvpd != NULL) { 1945 if (rsvpdebug) 1946 printf("rsvp_input: Sending packet up old-style socket\n"); 1947 rip_input(m, off); /* xxx */ 1948 } else { 1949 if (rsvpdebug && vifi == numvifs) 1950 printf("rsvp_input: Can't find vif for packet.\n"); 1951 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 1952 printf("rsvp_input: No socket defined for vif %d\n",vifi); 1953 m_freem(m); 1954 } 1955 return; 1956 } 1957 rsvp_src.sin_addr = ip->ip_src; 1958 1959 if (rsvpdebug && m) 1960 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 1961 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 1962 1963 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 1964 if (rsvpdebug) 1965 printf("rsvp_input: Failed to append to socket\n"); 1966 } else { 1967 if (rsvpdebug) 1968 printf("rsvp_input: send packet up\n"); 1969 } 1970 VIF_UNLOCK(); 1971 } 1972 1973 /* 1974 * Code for bandwidth monitors 1975 */ 1976 1977 /* 1978 * Define common interface for timeval-related methods 1979 */ 1980 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 1981 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 1982 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 1983 1984 static uint32_t 1985 compute_bw_meter_flags(struct bw_upcall *req) 1986 { 1987 uint32_t flags = 0; 1988 1989 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 1990 flags |= BW_METER_UNIT_PACKETS; 1991 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 1992 flags |= BW_METER_UNIT_BYTES; 1993 if (req->bu_flags & BW_UPCALL_GEQ) 1994 flags |= BW_METER_GEQ; 1995 if (req->bu_flags & BW_UPCALL_LEQ) 1996 flags |= BW_METER_LEQ; 1997 1998 return flags; 1999 } 2000 2001 /* 2002 * Add a bw_meter entry 2003 */ 2004 static int 2005 add_bw_upcall(struct bw_upcall *req) 2006 { 2007 struct mfc *mfc; 2008 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2009 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2010 struct timeval now; 2011 struct bw_meter *x; 2012 uint32_t flags; 2013 2014 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2015 return EOPNOTSUPP; 2016 2017 /* Test if the flags are valid */ 2018 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2019 return EINVAL; 2020 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2021 return EINVAL; 2022 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2023 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2024 return EINVAL; 2025 2026 /* Test if the threshold time interval is valid */ 2027 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2028 return EINVAL; 2029 2030 flags = compute_bw_meter_flags(req); 2031 2032 /* 2033 * Find if we have already same bw_meter entry 2034 */ 2035 MFC_LOCK(); 2036 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2037 if (mfc == NULL) { 2038 MFC_UNLOCK(); 2039 return EADDRNOTAVAIL; 2040 } 2041 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2042 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2043 &req->bu_threshold.b_time, ==)) && 2044 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2045 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2046 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2047 MFC_UNLOCK(); 2048 return 0; /* XXX Already installed */ 2049 } 2050 } 2051 2052 /* Allocate the new bw_meter entry */ 2053 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2054 if (x == NULL) { 2055 MFC_UNLOCK(); 2056 return ENOBUFS; 2057 } 2058 2059 /* Set the new bw_meter entry */ 2060 x->bm_threshold.b_time = req->bu_threshold.b_time; 2061 GET_TIME(now); 2062 x->bm_start_time = now; 2063 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2064 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2065 x->bm_measured.b_packets = 0; 2066 x->bm_measured.b_bytes = 0; 2067 x->bm_flags = flags; 2068 x->bm_time_next = NULL; 2069 x->bm_time_hash = BW_METER_BUCKETS; 2070 2071 /* Add the new bw_meter entry to the front of entries for this MFC */ 2072 x->bm_mfc = mfc; 2073 x->bm_mfc_next = mfc->mfc_bw_meter; 2074 mfc->mfc_bw_meter = x; 2075 schedule_bw_meter(x, &now); 2076 MFC_UNLOCK(); 2077 2078 return 0; 2079 } 2080 2081 static void 2082 free_bw_list(struct bw_meter *list) 2083 { 2084 while (list != NULL) { 2085 struct bw_meter *x = list; 2086 2087 list = list->bm_mfc_next; 2088 unschedule_bw_meter(x); 2089 free(x, M_BWMETER); 2090 } 2091 } 2092 2093 /* 2094 * Delete one or multiple bw_meter entries 2095 */ 2096 static int 2097 del_bw_upcall(struct bw_upcall *req) 2098 { 2099 struct mfc *mfc; 2100 struct bw_meter *x; 2101 2102 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2103 return EOPNOTSUPP; 2104 2105 MFC_LOCK(); 2106 /* Find the corresponding MFC entry */ 2107 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2108 if (mfc == NULL) { 2109 MFC_UNLOCK(); 2110 return EADDRNOTAVAIL; 2111 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2112 /* 2113 * Delete all bw_meter entries for this mfc 2114 */ 2115 struct bw_meter *list; 2116 2117 list = mfc->mfc_bw_meter; 2118 mfc->mfc_bw_meter = NULL; 2119 free_bw_list(list); 2120 MFC_UNLOCK(); 2121 return 0; 2122 } else { /* Delete a single bw_meter entry */ 2123 struct bw_meter *prev; 2124 uint32_t flags = 0; 2125 2126 flags = compute_bw_meter_flags(req); 2127 2128 /* Find the bw_meter entry to delete */ 2129 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2130 prev = x, x = x->bm_mfc_next) { 2131 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2132 &req->bu_threshold.b_time, ==)) && 2133 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2134 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2135 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2136 break; 2137 } 2138 if (x != NULL) { /* Delete entry from the list for this MFC */ 2139 if (prev != NULL) 2140 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2141 else 2142 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2143 2144 unschedule_bw_meter(x); 2145 MFC_UNLOCK(); 2146 /* Free the bw_meter entry */ 2147 free(x, M_BWMETER); 2148 return 0; 2149 } else { 2150 MFC_UNLOCK(); 2151 return EINVAL; 2152 } 2153 } 2154 /* NOTREACHED */ 2155 } 2156 2157 /* 2158 * Perform bandwidth measurement processing that may result in an upcall 2159 */ 2160 static void 2161 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2162 { 2163 struct timeval delta; 2164 2165 MFC_LOCK_ASSERT(); 2166 2167 delta = *nowp; 2168 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2169 2170 if (x->bm_flags & BW_METER_GEQ) { 2171 /* 2172 * Processing for ">=" type of bw_meter entry 2173 */ 2174 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2175 /* Reset the bw_meter entry */ 2176 x->bm_start_time = *nowp; 2177 x->bm_measured.b_packets = 0; 2178 x->bm_measured.b_bytes = 0; 2179 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2180 } 2181 2182 /* Record that a packet is received */ 2183 x->bm_measured.b_packets++; 2184 x->bm_measured.b_bytes += plen; 2185 2186 /* 2187 * Test if we should deliver an upcall 2188 */ 2189 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2190 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2191 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2192 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2193 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2194 /* Prepare an upcall for delivery */ 2195 bw_meter_prepare_upcall(x, nowp); 2196 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2197 } 2198 } 2199 } else if (x->bm_flags & BW_METER_LEQ) { 2200 /* 2201 * Processing for "<=" type of bw_meter entry 2202 */ 2203 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2204 /* 2205 * We are behind time with the multicast forwarding table 2206 * scanning for "<=" type of bw_meter entries, so test now 2207 * if we should deliver an upcall. 2208 */ 2209 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2210 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2211 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2212 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2213 /* Prepare an upcall for delivery */ 2214 bw_meter_prepare_upcall(x, nowp); 2215 } 2216 /* Reschedule the bw_meter entry */ 2217 unschedule_bw_meter(x); 2218 schedule_bw_meter(x, nowp); 2219 } 2220 2221 /* Record that a packet is received */ 2222 x->bm_measured.b_packets++; 2223 x->bm_measured.b_bytes += plen; 2224 2225 /* 2226 * Test if we should restart the measuring interval 2227 */ 2228 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2229 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2230 (x->bm_flags & BW_METER_UNIT_BYTES && 2231 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2232 /* Don't restart the measuring interval */ 2233 } else { 2234 /* Do restart the measuring interval */ 2235 /* 2236 * XXX: note that we don't unschedule and schedule, because this 2237 * might be too much overhead per packet. Instead, when we process 2238 * all entries for a given timer hash bin, we check whether it is 2239 * really a timeout. If not, we reschedule at that time. 2240 */ 2241 x->bm_start_time = *nowp; 2242 x->bm_measured.b_packets = 0; 2243 x->bm_measured.b_bytes = 0; 2244 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2245 } 2246 } 2247 } 2248 2249 /* 2250 * Prepare a bandwidth-related upcall 2251 */ 2252 static void 2253 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2254 { 2255 struct timeval delta; 2256 struct bw_upcall *u; 2257 2258 MFC_LOCK_ASSERT(); 2259 2260 /* 2261 * Compute the measured time interval 2262 */ 2263 delta = *nowp; 2264 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2265 2266 /* 2267 * If there are too many pending upcalls, deliver them now 2268 */ 2269 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2270 bw_upcalls_send(); 2271 2272 /* 2273 * Set the bw_upcall entry 2274 */ 2275 u = &bw_upcalls[bw_upcalls_n++]; 2276 u->bu_src = x->bm_mfc->mfc_origin; 2277 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2278 u->bu_threshold.b_time = x->bm_threshold.b_time; 2279 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2280 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2281 u->bu_measured.b_time = delta; 2282 u->bu_measured.b_packets = x->bm_measured.b_packets; 2283 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2284 u->bu_flags = 0; 2285 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2286 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2287 if (x->bm_flags & BW_METER_UNIT_BYTES) 2288 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2289 if (x->bm_flags & BW_METER_GEQ) 2290 u->bu_flags |= BW_UPCALL_GEQ; 2291 if (x->bm_flags & BW_METER_LEQ) 2292 u->bu_flags |= BW_UPCALL_LEQ; 2293 } 2294 2295 /* 2296 * Send the pending bandwidth-related upcalls 2297 */ 2298 static void 2299 bw_upcalls_send(void) 2300 { 2301 INIT_VNET_INET(curvnet); 2302 struct mbuf *m; 2303 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2304 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2305 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2306 0, /* unused2 */ 2307 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2308 0, /* im_mbz */ 2309 0, /* im_vif */ 2310 0, /* unused3 */ 2311 { 0 }, /* im_src */ 2312 { 0 } }; /* im_dst */ 2313 2314 MFC_LOCK_ASSERT(); 2315 2316 if (bw_upcalls_n == 0) 2317 return; /* No pending upcalls */ 2318 2319 bw_upcalls_n = 0; 2320 2321 /* 2322 * Allocate a new mbuf, initialize it with the header and 2323 * the payload for the pending calls. 2324 */ 2325 MGETHDR(m, M_DONTWAIT, MT_DATA); 2326 if (m == NULL) { 2327 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2328 return; 2329 } 2330 2331 m->m_len = m->m_pkthdr.len = 0; 2332 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2333 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2334 2335 /* 2336 * Send the upcalls 2337 * XXX do we need to set the address in k_igmpsrc ? 2338 */ 2339 mrtstat.mrts_upcalls++; 2340 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) { 2341 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2342 ++mrtstat.mrts_upq_sockfull; 2343 } 2344 } 2345 2346 /* 2347 * Compute the timeout hash value for the bw_meter entries 2348 */ 2349 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2350 do { \ 2351 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2352 \ 2353 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2354 (hash) = next_timeval.tv_sec; \ 2355 if (next_timeval.tv_usec) \ 2356 (hash)++; /* XXX: make sure we don't timeout early */ \ 2357 (hash) %= BW_METER_BUCKETS; \ 2358 } while (0) 2359 2360 /* 2361 * Schedule a timer to process periodically bw_meter entry of type "<=" 2362 * by linking the entry in the proper hash bucket. 2363 */ 2364 static void 2365 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2366 { 2367 int time_hash; 2368 2369 MFC_LOCK_ASSERT(); 2370 2371 if (!(x->bm_flags & BW_METER_LEQ)) 2372 return; /* XXX: we schedule timers only for "<=" entries */ 2373 2374 /* 2375 * Reset the bw_meter entry 2376 */ 2377 x->bm_start_time = *nowp; 2378 x->bm_measured.b_packets = 0; 2379 x->bm_measured.b_bytes = 0; 2380 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2381 2382 /* 2383 * Compute the timeout hash value and insert the entry 2384 */ 2385 BW_METER_TIMEHASH(x, time_hash); 2386 x->bm_time_next = bw_meter_timers[time_hash]; 2387 bw_meter_timers[time_hash] = x; 2388 x->bm_time_hash = time_hash; 2389 } 2390 2391 /* 2392 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2393 * by removing the entry from the proper hash bucket. 2394 */ 2395 static void 2396 unschedule_bw_meter(struct bw_meter *x) 2397 { 2398 int time_hash; 2399 struct bw_meter *prev, *tmp; 2400 2401 MFC_LOCK_ASSERT(); 2402 2403 if (!(x->bm_flags & BW_METER_LEQ)) 2404 return; /* XXX: we schedule timers only for "<=" entries */ 2405 2406 /* 2407 * Compute the timeout hash value and delete the entry 2408 */ 2409 time_hash = x->bm_time_hash; 2410 if (time_hash >= BW_METER_BUCKETS) 2411 return; /* Entry was not scheduled */ 2412 2413 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2414 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2415 if (tmp == x) 2416 break; 2417 2418 if (tmp == NULL) 2419 panic("unschedule_bw_meter: bw_meter entry not found"); 2420 2421 if (prev != NULL) 2422 prev->bm_time_next = x->bm_time_next; 2423 else 2424 bw_meter_timers[time_hash] = x->bm_time_next; 2425 2426 x->bm_time_next = NULL; 2427 x->bm_time_hash = BW_METER_BUCKETS; 2428 } 2429 2430 2431 /* 2432 * Process all "<=" type of bw_meter that should be processed now, 2433 * and for each entry prepare an upcall if necessary. Each processed 2434 * entry is rescheduled again for the (periodic) processing. 2435 * 2436 * This is run periodically (once per second normally). On each round, 2437 * all the potentially matching entries are in the hash slot that we are 2438 * looking at. 2439 */ 2440 static void 2441 bw_meter_process() 2442 { 2443 static uint32_t last_tv_sec; /* last time we processed this */ 2444 2445 uint32_t loops; 2446 int i; 2447 struct timeval now, process_endtime; 2448 2449 GET_TIME(now); 2450 if (last_tv_sec == now.tv_sec) 2451 return; /* nothing to do */ 2452 2453 loops = now.tv_sec - last_tv_sec; 2454 last_tv_sec = now.tv_sec; 2455 if (loops > BW_METER_BUCKETS) 2456 loops = BW_METER_BUCKETS; 2457 2458 MFC_LOCK(); 2459 /* 2460 * Process all bins of bw_meter entries from the one after the last 2461 * processed to the current one. On entry, i points to the last bucket 2462 * visited, so we need to increment i at the beginning of the loop. 2463 */ 2464 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2465 struct bw_meter *x, *tmp_list; 2466 2467 if (++i >= BW_METER_BUCKETS) 2468 i = 0; 2469 2470 /* Disconnect the list of bw_meter entries from the bin */ 2471 tmp_list = bw_meter_timers[i]; 2472 bw_meter_timers[i] = NULL; 2473 2474 /* Process the list of bw_meter entries */ 2475 while (tmp_list != NULL) { 2476 x = tmp_list; 2477 tmp_list = tmp_list->bm_time_next; 2478 2479 /* Test if the time interval is over */ 2480 process_endtime = x->bm_start_time; 2481 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2482 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2483 /* Not yet: reschedule, but don't reset */ 2484 int time_hash; 2485 2486 BW_METER_TIMEHASH(x, time_hash); 2487 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2488 /* 2489 * XXX: somehow the bin processing is a bit ahead of time. 2490 * Put the entry in the next bin. 2491 */ 2492 if (++time_hash >= BW_METER_BUCKETS) 2493 time_hash = 0; 2494 } 2495 x->bm_time_next = bw_meter_timers[time_hash]; 2496 bw_meter_timers[time_hash] = x; 2497 x->bm_time_hash = time_hash; 2498 2499 continue; 2500 } 2501 2502 /* 2503 * Test if we should deliver an upcall 2504 */ 2505 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2506 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2507 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2508 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2509 /* Prepare an upcall for delivery */ 2510 bw_meter_prepare_upcall(x, &now); 2511 } 2512 2513 /* 2514 * Reschedule for next processing 2515 */ 2516 schedule_bw_meter(x, &now); 2517 } 2518 } 2519 2520 /* Send all upcalls that are pending delivery */ 2521 bw_upcalls_send(); 2522 2523 MFC_UNLOCK(); 2524 } 2525 2526 /* 2527 * A periodic function for sending all upcalls that are pending delivery 2528 */ 2529 static void 2530 expire_bw_upcalls_send(void *unused) 2531 { 2532 MFC_LOCK(); 2533 bw_upcalls_send(); 2534 MFC_UNLOCK(); 2535 2536 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2537 expire_bw_upcalls_send, NULL); 2538 } 2539 2540 /* 2541 * A periodic function for periodic scanning of the multicast forwarding 2542 * table for processing all "<=" bw_meter entries. 2543 */ 2544 static void 2545 expire_bw_meter_process(void *unused) 2546 { 2547 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2548 bw_meter_process(); 2549 2550 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2551 } 2552 2553 /* 2554 * End of bandwidth monitoring code 2555 */ 2556 2557 /* 2558 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2559 * 2560 */ 2561 static int 2562 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m, 2563 struct mfc *rt) 2564 { 2565 struct mbuf *mb_copy, *mm; 2566 2567 if (mrtdebug & DEBUG_PIM) 2568 log(LOG_DEBUG, "pim_register_send: "); 2569 2570 /* 2571 * Do not send IGMP_WHOLEPKT notifications to userland, if the 2572 * rendezvous point was unspecified, and we were told not to. 2573 */ 2574 if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) && 2575 (rt->mfc_rp.s_addr == INADDR_ANY)) 2576 return 0; 2577 2578 mb_copy = pim_register_prepare(ip, m); 2579 if (mb_copy == NULL) 2580 return ENOBUFS; 2581 2582 /* 2583 * Send all the fragments. Note that the mbuf for each fragment 2584 * is freed by the sending machinery. 2585 */ 2586 for (mm = mb_copy; mm; mm = mb_copy) { 2587 mb_copy = mm->m_nextpkt; 2588 mm->m_nextpkt = 0; 2589 mm = m_pullup(mm, sizeof(struct ip)); 2590 if (mm != NULL) { 2591 ip = mtod(mm, struct ip *); 2592 if ((mrt_api_config & MRT_MFC_RP) && 2593 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2594 pim_register_send_rp(ip, vifp, mm, rt); 2595 } else { 2596 pim_register_send_upcall(ip, vifp, mm, rt); 2597 } 2598 } 2599 } 2600 2601 return 0; 2602 } 2603 2604 /* 2605 * Return a copy of the data packet that is ready for PIM Register 2606 * encapsulation. 2607 * XXX: Note that in the returned copy the IP header is a valid one. 2608 */ 2609 static struct mbuf * 2610 pim_register_prepare(struct ip *ip, struct mbuf *m) 2611 { 2612 struct mbuf *mb_copy = NULL; 2613 int mtu; 2614 2615 /* Take care of delayed checksums */ 2616 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2617 in_delayed_cksum(m); 2618 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2619 } 2620 2621 /* 2622 * Copy the old packet & pullup its IP header into the 2623 * new mbuf so we can modify it. 2624 */ 2625 mb_copy = m_copypacket(m, M_DONTWAIT); 2626 if (mb_copy == NULL) 2627 return NULL; 2628 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2629 if (mb_copy == NULL) 2630 return NULL; 2631 2632 /* take care of the TTL */ 2633 ip = mtod(mb_copy, struct ip *); 2634 --ip->ip_ttl; 2635 2636 /* Compute the MTU after the PIM Register encapsulation */ 2637 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2638 2639 if (ip->ip_len <= mtu) { 2640 /* Turn the IP header into a valid one */ 2641 ip->ip_len = htons(ip->ip_len); 2642 ip->ip_off = htons(ip->ip_off); 2643 ip->ip_sum = 0; 2644 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2645 } else { 2646 /* Fragment the packet */ 2647 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2648 m_freem(mb_copy); 2649 return NULL; 2650 } 2651 } 2652 return mb_copy; 2653 } 2654 2655 /* 2656 * Send an upcall with the data packet to the user-level process. 2657 */ 2658 static int 2659 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2660 struct mbuf *mb_copy, struct mfc *rt) 2661 { 2662 INIT_VNET_INET(curvnet); 2663 struct mbuf *mb_first; 2664 int len = ntohs(ip->ip_len); 2665 struct igmpmsg *im; 2666 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2667 2668 VIF_LOCK_ASSERT(); 2669 2670 /* 2671 * Add a new mbuf with an upcall header 2672 */ 2673 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2674 if (mb_first == NULL) { 2675 m_freem(mb_copy); 2676 return ENOBUFS; 2677 } 2678 mb_first->m_data += max_linkhdr; 2679 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2680 mb_first->m_len = sizeof(struct igmpmsg); 2681 mb_first->m_next = mb_copy; 2682 2683 /* Send message to routing daemon */ 2684 im = mtod(mb_first, struct igmpmsg *); 2685 im->im_msgtype = IGMPMSG_WHOLEPKT; 2686 im->im_mbz = 0; 2687 im->im_vif = vifp - viftable; 2688 im->im_src = ip->ip_src; 2689 im->im_dst = ip->ip_dst; 2690 2691 k_igmpsrc.sin_addr = ip->ip_src; 2692 2693 mrtstat.mrts_upcalls++; 2694 2695 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2696 if (mrtdebug & DEBUG_PIM) 2697 log(LOG_WARNING, 2698 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2699 ++mrtstat.mrts_upq_sockfull; 2700 return ENOBUFS; 2701 } 2702 2703 /* Keep statistics */ 2704 pimstat.pims_snd_registers_msgs++; 2705 pimstat.pims_snd_registers_bytes += len; 2706 2707 return 0; 2708 } 2709 2710 /* 2711 * Encapsulate the data packet in PIM Register message and send it to the RP. 2712 */ 2713 static int 2714 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy, 2715 struct mfc *rt) 2716 { 2717 INIT_VNET_INET(curvnet); 2718 struct mbuf *mb_first; 2719 struct ip *ip_outer; 2720 struct pim_encap_pimhdr *pimhdr; 2721 int len = ntohs(ip->ip_len); 2722 vifi_t vifi = rt->mfc_parent; 2723 2724 VIF_LOCK_ASSERT(); 2725 2726 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2727 m_freem(mb_copy); 2728 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2729 } 2730 2731 /* 2732 * Add a new mbuf with the encapsulating header 2733 */ 2734 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2735 if (mb_first == NULL) { 2736 m_freem(mb_copy); 2737 return ENOBUFS; 2738 } 2739 mb_first->m_data += max_linkhdr; 2740 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2741 mb_first->m_next = mb_copy; 2742 2743 mb_first->m_pkthdr.len = len + mb_first->m_len; 2744 2745 /* 2746 * Fill in the encapsulating IP and PIM header 2747 */ 2748 ip_outer = mtod(mb_first, struct ip *); 2749 *ip_outer = pim_encap_iphdr; 2750 ip_outer->ip_id = ip_newid(); 2751 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2752 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2753 ip_outer->ip_dst = rt->mfc_rp; 2754 /* 2755 * Copy the inner header TOS to the outer header, and take care of the 2756 * IP_DF bit. 2757 */ 2758 ip_outer->ip_tos = ip->ip_tos; 2759 if (ntohs(ip->ip_off) & IP_DF) 2760 ip_outer->ip_off |= IP_DF; 2761 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2762 + sizeof(pim_encap_iphdr)); 2763 *pimhdr = pim_encap_pimhdr; 2764 /* If the iif crosses a border, set the Border-bit */ 2765 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2766 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2767 2768 mb_first->m_data += sizeof(pim_encap_iphdr); 2769 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2770 mb_first->m_data -= sizeof(pim_encap_iphdr); 2771 2772 send_packet(vifp, mb_first); 2773 2774 /* Keep statistics */ 2775 pimstat.pims_snd_registers_msgs++; 2776 pimstat.pims_snd_registers_bytes += len; 2777 2778 return 0; 2779 } 2780 2781 /* 2782 * pim_encapcheck() is called by the encap[46]_input() path at runtime to 2783 * determine if a packet is for PIM; allowing PIM to be dynamically loaded 2784 * into the kernel. 2785 */ 2786 static int 2787 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 2788 { 2789 2790 #ifdef DIAGNOSTIC 2791 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); 2792 #endif 2793 if (proto != IPPROTO_PIM) 2794 return 0; /* not for us; reject the datagram. */ 2795 2796 return 64; /* claim the datagram. */ 2797 } 2798 2799 /* 2800 * PIM-SMv2 and PIM-DM messages processing. 2801 * Receives and verifies the PIM control messages, and passes them 2802 * up to the listening socket, using rip_input(). 2803 * The only message with special processing is the PIM_REGISTER message 2804 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2805 * is passed to if_simloop(). 2806 */ 2807 void 2808 pim_input(struct mbuf *m, int off) 2809 { 2810 struct ip *ip = mtod(m, struct ip *); 2811 struct pim *pim; 2812 int minlen; 2813 int datalen = ip->ip_len; 2814 int ip_tos; 2815 int iphlen = off; 2816 2817 /* Keep statistics */ 2818 pimstat.pims_rcv_total_msgs++; 2819 pimstat.pims_rcv_total_bytes += datalen; 2820 2821 /* 2822 * Validate lengths 2823 */ 2824 if (datalen < PIM_MINLEN) { 2825 pimstat.pims_rcv_tooshort++; 2826 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 2827 datalen, (u_long)ip->ip_src.s_addr); 2828 m_freem(m); 2829 return; 2830 } 2831 2832 /* 2833 * If the packet is at least as big as a REGISTER, go agead 2834 * and grab the PIM REGISTER header size, to avoid another 2835 * possible m_pullup() later. 2836 * 2837 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 2838 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 2839 */ 2840 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 2841 /* 2842 * Get the IP and PIM headers in contiguous memory, and 2843 * possibly the PIM REGISTER header. 2844 */ 2845 if ((m->m_flags & M_EXT || m->m_len < minlen) && 2846 (m = m_pullup(m, minlen)) == 0) { 2847 log(LOG_ERR, "pim_input: m_pullup failure\n"); 2848 return; 2849 } 2850 /* m_pullup() may have given us a new mbuf so reset ip. */ 2851 ip = mtod(m, struct ip *); 2852 ip_tos = ip->ip_tos; 2853 2854 /* adjust mbuf to point to the PIM header */ 2855 m->m_data += iphlen; 2856 m->m_len -= iphlen; 2857 pim = mtod(m, struct pim *); 2858 2859 /* 2860 * Validate checksum. If PIM REGISTER, exclude the data packet. 2861 * 2862 * XXX: some older PIMv2 implementations don't make this distinction, 2863 * so for compatibility reason perform the checksum over part of the 2864 * message, and if error, then over the whole message. 2865 */ 2866 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 2867 /* do nothing, checksum okay */ 2868 } else if (in_cksum(m, datalen)) { 2869 pimstat.pims_rcv_badsum++; 2870 if (mrtdebug & DEBUG_PIM) 2871 log(LOG_DEBUG, "pim_input: invalid checksum"); 2872 m_freem(m); 2873 return; 2874 } 2875 2876 /* PIM version check */ 2877 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 2878 pimstat.pims_rcv_badversion++; 2879 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 2880 PIM_VT_V(pim->pim_vt), PIM_VERSION); 2881 m_freem(m); 2882 return; 2883 } 2884 2885 /* restore mbuf back to the outer IP */ 2886 m->m_data -= iphlen; 2887 m->m_len += iphlen; 2888 2889 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 2890 /* 2891 * Since this is a REGISTER, we'll make a copy of the register 2892 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 2893 * routing daemon. 2894 */ 2895 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 2896 struct mbuf *mcp; 2897 struct ip *encap_ip; 2898 u_int32_t *reghdr; 2899 struct ifnet *vifp; 2900 2901 VIF_LOCK(); 2902 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 2903 VIF_UNLOCK(); 2904 if (mrtdebug & DEBUG_PIM) 2905 log(LOG_DEBUG, 2906 "pim_input: register vif not set: %d\n", reg_vif_num); 2907 m_freem(m); 2908 return; 2909 } 2910 /* XXX need refcnt? */ 2911 vifp = viftable[reg_vif_num].v_ifp; 2912 VIF_UNLOCK(); 2913 2914 /* 2915 * Validate length 2916 */ 2917 if (datalen < PIM_REG_MINLEN) { 2918 pimstat.pims_rcv_tooshort++; 2919 pimstat.pims_rcv_badregisters++; 2920 log(LOG_ERR, 2921 "pim_input: register packet size too small %d from %lx\n", 2922 datalen, (u_long)ip->ip_src.s_addr); 2923 m_freem(m); 2924 return; 2925 } 2926 2927 reghdr = (u_int32_t *)(pim + 1); 2928 encap_ip = (struct ip *)(reghdr + 1); 2929 2930 if (mrtdebug & DEBUG_PIM) { 2931 log(LOG_DEBUG, 2932 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 2933 (u_long)ntohl(encap_ip->ip_src.s_addr), 2934 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2935 ntohs(encap_ip->ip_len)); 2936 } 2937 2938 /* verify the version number of the inner packet */ 2939 if (encap_ip->ip_v != IPVERSION) { 2940 pimstat.pims_rcv_badregisters++; 2941 if (mrtdebug & DEBUG_PIM) { 2942 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 2943 "of the inner packet\n", encap_ip->ip_v); 2944 } 2945 m_freem(m); 2946 return; 2947 } 2948 2949 /* verify the inner packet is destined to a mcast group */ 2950 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 2951 pimstat.pims_rcv_badregisters++; 2952 if (mrtdebug & DEBUG_PIM) 2953 log(LOG_DEBUG, 2954 "pim_input: inner packet of register is not " 2955 "multicast %lx\n", 2956 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 2957 m_freem(m); 2958 return; 2959 } 2960 2961 /* If a NULL_REGISTER, pass it to the daemon */ 2962 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 2963 goto pim_input_to_daemon; 2964 2965 /* 2966 * Copy the TOS from the outer IP header to the inner IP header. 2967 */ 2968 if (encap_ip->ip_tos != ip_tos) { 2969 /* Outer TOS -> inner TOS */ 2970 encap_ip->ip_tos = ip_tos; 2971 /* Recompute the inner header checksum. Sigh... */ 2972 2973 /* adjust mbuf to point to the inner IP header */ 2974 m->m_data += (iphlen + PIM_MINLEN); 2975 m->m_len -= (iphlen + PIM_MINLEN); 2976 2977 encap_ip->ip_sum = 0; 2978 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 2979 2980 /* restore mbuf to point back to the outer IP header */ 2981 m->m_data -= (iphlen + PIM_MINLEN); 2982 m->m_len += (iphlen + PIM_MINLEN); 2983 } 2984 2985 /* 2986 * Decapsulate the inner IP packet and loopback to forward it 2987 * as a normal multicast packet. Also, make a copy of the 2988 * outer_iphdr + pimhdr + reghdr + encap_iphdr 2989 * to pass to the daemon later, so it can take the appropriate 2990 * actions (e.g., send back PIM_REGISTER_STOP). 2991 * XXX: here m->m_data points to the outer IP header. 2992 */ 2993 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 2994 if (mcp == NULL) { 2995 log(LOG_ERR, 2996 "pim_input: pim register: could not copy register head\n"); 2997 m_freem(m); 2998 return; 2999 } 3000 3001 /* Keep statistics */ 3002 /* XXX: registers_bytes include only the encap. mcast pkt */ 3003 pimstat.pims_rcv_registers_msgs++; 3004 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3005 3006 /* 3007 * forward the inner ip packet; point m_data at the inner ip. 3008 */ 3009 m_adj(m, iphlen + PIM_MINLEN); 3010 3011 if (mrtdebug & DEBUG_PIM) { 3012 log(LOG_DEBUG, 3013 "pim_input: forwarding decapsulated register: " 3014 "src %lx, dst %lx, vif %d\n", 3015 (u_long)ntohl(encap_ip->ip_src.s_addr), 3016 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3017 reg_vif_num); 3018 } 3019 /* NB: vifp was collected above; can it change on us? */ 3020 if_simloop(vifp, m, dst.sin_family, 0); 3021 3022 /* prepare the register head to send to the mrouting daemon */ 3023 m = mcp; 3024 } 3025 3026 pim_input_to_daemon: 3027 /* 3028 * Pass the PIM message up to the daemon; if it is a Register message, 3029 * pass the 'head' only up to the daemon. This includes the 3030 * outer IP header, PIM header, PIM-Register header and the 3031 * inner IP header. 3032 * XXX: the outer IP header pkt size of a Register is not adjust to 3033 * reflect the fact that the inner multicast data is truncated. 3034 */ 3035 rip_input(m, iphlen); 3036 3037 return; 3038 } 3039 3040 /* 3041 * XXX: This is common code for dealing with initialization for both 3042 * the IPv4 and IPv6 multicast forwarding paths. It could do with cleanup. 3043 */ 3044 static int 3045 ip_mroute_modevent(module_t mod, int type, void *unused) 3046 { 3047 INIT_VNET_INET(curvnet); 3048 3049 switch (type) { 3050 case MOD_LOAD: 3051 MROUTER_LOCK_INIT(); 3052 MFC_LOCK_INIT(); 3053 VIF_LOCK_INIT(); 3054 ip_mrouter_reset(); 3055 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", 3056 &pim_squelch_wholepkt); 3057 3058 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM, 3059 pim_encapcheck, &in_pim_protosw, NULL); 3060 if (pim_encap_cookie == NULL) { 3061 printf("ip_mroute: unable to attach pim encap\n"); 3062 VIF_LOCK_DESTROY(); 3063 MFC_LOCK_DESTROY(); 3064 MROUTER_LOCK_DESTROY(); 3065 return (EINVAL); 3066 } 3067 3068 #ifdef INET6 3069 pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM, 3070 pim_encapcheck, (struct protosw *)&in6_pim_protosw, NULL); 3071 if (pim6_encap_cookie == NULL) { 3072 printf("ip_mroute: unable to attach pim6 encap\n"); 3073 if (pim_encap_cookie) { 3074 encap_detach(pim_encap_cookie); 3075 pim_encap_cookie = NULL; 3076 } 3077 VIF_LOCK_DESTROY(); 3078 MFC_LOCK_DESTROY(); 3079 MROUTER_LOCK_DESTROY(); 3080 return (EINVAL); 3081 } 3082 #endif 3083 3084 ip_mcast_src = X_ip_mcast_src; 3085 ip_mforward = X_ip_mforward; 3086 ip_mrouter_done = X_ip_mrouter_done; 3087 ip_mrouter_get = X_ip_mrouter_get; 3088 ip_mrouter_set = X_ip_mrouter_set; 3089 3090 #ifdef INET6 3091 ip6_mforward = X_ip6_mforward; 3092 ip6_mrouter_done = X_ip6_mrouter_done; 3093 ip6_mrouter_get = X_ip6_mrouter_get; 3094 ip6_mrouter_set = X_ip6_mrouter_set; 3095 mrt6_ioctl = X_mrt6_ioctl; 3096 #endif 3097 3098 ip_rsvp_force_done = X_ip_rsvp_force_done; 3099 ip_rsvp_vif = X_ip_rsvp_vif; 3100 3101 legal_vif_num = X_legal_vif_num; 3102 mrt_ioctl = X_mrt_ioctl; 3103 rsvp_input_p = X_rsvp_input; 3104 break; 3105 3106 case MOD_UNLOAD: 3107 /* 3108 * Typically module unload happens after the user-level 3109 * process has shutdown the kernel services (the check 3110 * below insures someone can't just yank the module out 3111 * from under a running process). But if the module is 3112 * just loaded and then unloaded w/o starting up a user 3113 * process we still need to cleanup. 3114 */ 3115 if (V_ip_mrouter 3116 #ifdef INET6 3117 || ip6_mrouter 3118 #endif 3119 ) 3120 return EINVAL; 3121 3122 #ifdef INET6 3123 if (pim6_encap_cookie) { 3124 encap_detach(pim6_encap_cookie); 3125 pim6_encap_cookie = NULL; 3126 } 3127 X_ip6_mrouter_done(); 3128 ip6_mforward = NULL; 3129 ip6_mrouter_done = NULL; 3130 ip6_mrouter_get = NULL; 3131 ip6_mrouter_set = NULL; 3132 mrt6_ioctl = NULL; 3133 #endif 3134 3135 if (pim_encap_cookie) { 3136 encap_detach(pim_encap_cookie); 3137 pim_encap_cookie = NULL; 3138 } 3139 X_ip_mrouter_done(); 3140 ip_mcast_src = NULL; 3141 ip_mforward = NULL; 3142 ip_mrouter_done = NULL; 3143 ip_mrouter_get = NULL; 3144 ip_mrouter_set = NULL; 3145 3146 ip_rsvp_force_done = NULL; 3147 ip_rsvp_vif = NULL; 3148 3149 legal_vif_num = NULL; 3150 mrt_ioctl = NULL; 3151 rsvp_input_p = NULL; 3152 3153 VIF_LOCK_DESTROY(); 3154 MFC_LOCK_DESTROY(); 3155 MROUTER_LOCK_DESTROY(); 3156 break; 3157 3158 default: 3159 return EOPNOTSUPP; 3160 } 3161 return 0; 3162 } 3163 3164 static moduledata_t ip_mroutemod = { 3165 "ip_mroute", 3166 ip_mroute_modevent, 3167 0 3168 }; 3169 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3170