ip_mroute.c revision 166972
1157873Simp/*- 2157873Simp * Copyright (c) 1989 Stephen Deering 3157873Simp * Copyright (c) 1992, 1993 4157873Simp * The Regents of the University of California. All rights reserved. 5157873Simp * 6157873Simp * This code is derived from software contributed to Berkeley by 7157873Simp * Stephen Deering of Stanford University. 8157873Simp * 9157873Simp * Redistribution and use in source and binary forms, with or without 10157873Simp * modification, are permitted provided that the following conditions 11157873Simp * are met: 12157873Simp * 1. Redistributions of source code must retain the above copyright 13157873Simp * notice, this list of conditions and the following disclaimer. 14157873Simp * 2. Redistributions in binary form must reproduce the above copyright 15157873Simp * notice, this list of conditions and the following disclaimer in the 16157873Simp * documentation and/or other materials provided with the distribution. 17157873Simp * 4. Neither the name of the University nor the names of its contributors 18157873Simp * may be used to endorse or promote products derived from this software 19157873Simp * without specific prior written permission. 20157873Simp * 21157873Simp * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22157873Simp * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23157873Simp * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24157873Simp * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25157873Simp * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26157873Simp * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27157873Simp * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28157873Simp * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29157873Simp * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30157873Simp * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31157873Simp * SUCH DAMAGE. 32157873Simp * 33157873Simp * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 34157873Simp */ 35157873Simp 36157873Simp/* 37157873Simp * IP multicast forwarding procedures 38157873Simp * 39157873Simp * Written by David Waitzman, BBN Labs, August 1988. 40157873Simp * Modified by Steve Deering, Stanford, February 1989. 41157873Simp * Modified by Mark J. Steiglitz, Stanford, May, 1991 42161190Simp * Modified by Van Jacobson, LBL, January 1993 43157873Simp * Modified by Ajit Thyagarajan, PARC, August 1993 44157873Simp * Modified by Bill Fenner, PARC, April 1995 45157873Simp * Modified by Ahmed Helmy, SGI, June 1996 46157873Simp * Modified by George Edmond Eddy (Rusty), ISI, February 1998 47157873Simp * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000 48157873Simp * Modified by Hitoshi Asaeda, WIDE, August 2000 49157873Simp * Modified by Pavlin Radoslavov, ICSI, October 2002 50157873Simp * 51157873Simp * MROUTING Revision: 3.5 52157873Simp * and PIM-SMv2 and PIM-DM support, advanced API support, 53157873Simp * bandwidth metering and signaling 54157873Simp * 55157873Simp * $FreeBSD: head/sys/netinet/ip_mroute.c 166972 2007-02-25 14:22:03Z bms $ 56157873Simp */ 57161190Simp 58157873Simp#include "opt_inet.h" 59161190Simp#include "opt_inet6.h" 60157873Simp#include "opt_mac.h" 61161190Simp#include "opt_mrouting.h" 62157873Simp 63157873Simp#define _PIM_VT 1 64157873Simp 65157873Simp#include <sys/param.h> 66157873Simp#include <sys/kernel.h> 67161190Simp#include <sys/lock.h> 68157873Simp#include <sys/malloc.h> 69157873Simp#include <sys/mbuf.h> 70157873Simp#include <sys/module.h> 71157873Simp#include <sys/priv.h> 72157873Simp#include <sys/protosw.h> 73157873Simp#include <sys/signalvar.h> 74157873Simp#include <sys/socket.h> 75157873Simp#include <sys/socketvar.h> 76157873Simp#include <sys/sockio.h> 77157873Simp#include <sys/sx.h> 78157873Simp#include <sys/sysctl.h> 79157873Simp#include <sys/syslog.h> 80157873Simp#include <sys/systm.h> 81161190Simp#include <sys/time.h> 82161190Simp#include <net/if.h> 83161190Simp#include <net/netisr.h> 84161190Simp#include <net/route.h> 85157873Simp#include <netinet/in.h> 86157873Simp#include <netinet/igmp.h> 87157873Simp#include <netinet/in_systm.h> 88157873Simp#include <netinet/in_var.h> 89157873Simp#include <netinet/ip.h> 90157873Simp#include <netinet/ip_encap.h> 91157873Simp#include <netinet/ip_mroute.h> 92157873Simp#include <netinet/ip_var.h> 93157873Simp#include <netinet/ip_options.h> 94157873Simp#include <netinet/pim.h> 95157873Simp#include <netinet/pim_var.h> 96157873Simp#include <netinet/udp.h> 97157873Simp#ifdef INET6 98157873Simp#include <netinet/ip6.h> 99157873Simp#include <netinet6/in6_var.h> 100157873Simp#include <netinet6/ip6_mroute.h> 101157873Simp#include <netinet6/ip6_var.h> 102157873Simp#endif 103157873Simp#include <machine/in_cksum.h> 104157873Simp 105157873Simp#include <security/mac/mac_framework.h> 106157873Simp 107157873Simp/* 108157873Simp * Control debugging code for rsvp and multicast routing code. 109157873Simp * Can only set them with the debugger. 110157873Simp */ 111157873Simpstatic u_int rsvpdebug; /* non-zero enables debugging */ 112157873Simp 113157873Simpstatic u_int mrtdebug; /* any set of the flags below */ 114157873Simp#define DEBUG_MFC 0x02 115157873Simp#define DEBUG_FORWARD 0x04 116157873Simp#define DEBUG_EXPIRE 0x08 117157873Simp#define DEBUG_XMIT 0x10 118157873Simp#define DEBUG_PIM 0x20 119157873Simp 120157873Simp#define VIFI_INVALID ((vifi_t) -1) 121157873Simp 122157873Simp#define M_HASCL(m) ((m)->m_flags & M_EXT) 123157873Simp 124157873Simpstatic MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 125157873Simp 126157873Simp/* 127157873Simp * Locking. We use two locks: one for the virtual interface table and 128157873Simp * one for the forwarding table. These locks may be nested in which case 129 * the VIF lock must always be taken first. Note that each lock is used 130 * to cover not only the specific data structure but also related data 131 * structures. It may be better to add more fine-grained locking later; 132 * it's not clear how performance-critical this code is. 133 * 134 * XXX: This module could particularly benefit from being cleaned 135 * up to use the <sys/queue.h> macros. 136 * 137 */ 138 139static struct mrtstat mrtstat; 140SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 141 &mrtstat, mrtstat, 142 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 143 144static struct mfc *mfctable[MFCTBLSIZ]; 145SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 146 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 147 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 148 149static struct mtx mfc_mtx; 150#define MFC_LOCK() mtx_lock(&mfc_mtx) 151#define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 152#define MFC_LOCK_ASSERT() do { \ 153 mtx_assert(&mfc_mtx, MA_OWNED); \ 154 NET_ASSERT_GIANT(); \ 155} while (0) 156#define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF) 157#define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 158 159static struct vif viftable[MAXVIFS]; 160SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 161 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 162 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 163 164static struct mtx vif_mtx; 165#define VIF_LOCK() mtx_lock(&vif_mtx) 166#define VIF_UNLOCK() mtx_unlock(&vif_mtx) 167#define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 168#define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF) 169#define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 170 171static u_char nexpire[MFCTBLSIZ]; 172 173static eventhandler_tag if_detach_event_tag = NULL; 174 175static struct callout expire_upcalls_ch; 176 177#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 178#define UPCALL_EXPIRE 6 /* number of timeouts */ 179 180#define ENCAP_TTL 64 181 182/* 183 * Bandwidth meter variables and constants 184 */ 185static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 186/* 187 * Pending timeouts are stored in a hash table, the key being the 188 * expiration time. Periodically, the entries are analysed and processed. 189 */ 190#define BW_METER_BUCKETS 1024 191static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 192static struct callout bw_meter_ch; 193#define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 194 195/* 196 * Pending upcalls are stored in a vector which is flushed when 197 * full, or periodically 198 */ 199static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 200static u_int bw_upcalls_n; /* # of pending upcalls */ 201static struct callout bw_upcalls_ch; 202#define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 203 204static struct pimstat pimstat; 205 206SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); 207SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 208 &pimstat, pimstat, 209 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 210 211static u_long pim_squelch_wholepkt = 0; 212SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, 213 &pim_squelch_wholepkt, 0, 214 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); 215 216extern struct domain inetdomain; 217struct protosw in_pim_protosw = { 218 .pr_type = SOCK_RAW, 219 .pr_domain = &inetdomain, 220 .pr_protocol = IPPROTO_PIM, 221 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, 222 .pr_input = pim_input, 223 .pr_output = (pr_output_t*)rip_output, 224 .pr_ctloutput = rip_ctloutput, 225 .pr_usrreqs = &rip_usrreqs 226}; 227static const struct encaptab *pim_encap_cookie; 228 229#ifdef INET6 230/* ip6_mroute.c glue */ 231extern struct in6_protosw in6_pim_protosw; 232static const struct encaptab *pim6_encap_cookie; 233 234extern int X_ip6_mrouter_set(struct socket *, struct sockopt *); 235extern int X_ip6_mrouter_get(struct socket *, struct sockopt *); 236extern int X_ip6_mrouter_done(void); 237extern int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *); 238extern int X_mrt6_ioctl(int, caddr_t); 239#endif 240 241static int pim_encapcheck(const struct mbuf *, int, int, void *); 242 243/* 244 * Note: the PIM Register encapsulation adds the following in front of a 245 * data packet: 246 * 247 * struct pim_encap_hdr { 248 * struct ip ip; 249 * struct pim_encap_pimhdr pim; 250 * } 251 * 252 */ 253 254struct pim_encap_pimhdr { 255 struct pim pim; 256 uint32_t flags; 257}; 258 259static struct ip pim_encap_iphdr = { 260#if BYTE_ORDER == LITTLE_ENDIAN 261 sizeof(struct ip) >> 2, 262 IPVERSION, 263#else 264 IPVERSION, 265 sizeof(struct ip) >> 2, 266#endif 267 0, /* tos */ 268 sizeof(struct ip), /* total length */ 269 0, /* id */ 270 0, /* frag offset */ 271 ENCAP_TTL, 272 IPPROTO_PIM, 273 0, /* checksum */ 274}; 275 276static struct pim_encap_pimhdr pim_encap_pimhdr = { 277 { 278 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 279 0, /* reserved */ 280 0, /* checksum */ 281 }, 282 0 /* flags */ 283}; 284 285static struct ifnet multicast_register_if; 286static vifi_t reg_vif_num = VIFI_INVALID; 287 288/* 289 * Private variables. 290 */ 291static vifi_t numvifs; 292 293static u_long X_ip_mcast_src(int vifi); 294static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 295 struct mbuf *m, struct ip_moptions *imo); 296static int X_ip_mrouter_done(void); 297static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 298static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 299static int X_legal_vif_num(int vif); 300static int X_mrt_ioctl(int cmd, caddr_t data); 301 302static int get_sg_cnt(struct sioc_sg_req *); 303static int get_vif_cnt(struct sioc_vif_req *); 304static void if_detached_event(void *arg __unused, struct ifnet *); 305static int ip_mrouter_init(struct socket *, int); 306static int add_vif(struct vifctl *); 307static int del_vif_locked(vifi_t); 308static int del_vif(vifi_t); 309static int add_mfc(struct mfcctl2 *); 310static int del_mfc(struct mfcctl2 *); 311static int set_api_config(uint32_t *); /* chose API capabilities */ 312static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 313static int set_assert(int); 314static void expire_upcalls(void *); 315static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 316static void phyint_send(struct ip *, struct vif *, struct mbuf *); 317static void send_packet(struct vif *, struct mbuf *); 318 319/* 320 * Bandwidth monitoring 321 */ 322static void free_bw_list(struct bw_meter *list); 323static int add_bw_upcall(struct bw_upcall *); 324static int del_bw_upcall(struct bw_upcall *); 325static void bw_meter_receive_packet(struct bw_meter *x, int plen, 326 struct timeval *nowp); 327static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 328static void bw_upcalls_send(void); 329static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 330static void unschedule_bw_meter(struct bw_meter *x); 331static void bw_meter_process(void); 332static void expire_bw_upcalls_send(void *); 333static void expire_bw_meter_process(void *); 334 335static int pim_register_send(struct ip *, struct vif *, 336 struct mbuf *, struct mfc *); 337static int pim_register_send_rp(struct ip *, struct vif *, 338 struct mbuf *, struct mfc *); 339static int pim_register_send_upcall(struct ip *, struct vif *, 340 struct mbuf *, struct mfc *); 341static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 342 343/* 344 * whether or not special PIM assert processing is enabled. 345 */ 346static int pim_assert; 347/* 348 * Rate limit for assert notification messages, in usec 349 */ 350#define ASSERT_MSG_TIME 3000000 351 352/* 353 * Kernel multicast routing API capabilities and setup. 354 * If more API capabilities are added to the kernel, they should be 355 * recorded in `mrt_api_support'. 356 */ 357static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 358 MRT_MFC_FLAGS_BORDER_VIF | 359 MRT_MFC_RP | 360 MRT_MFC_BW_UPCALL); 361static uint32_t mrt_api_config = 0; 362 363/* 364 * Hash function for a source, group entry 365 */ 366#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 367 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 368 369/* 370 * Find a route for a given origin IP address and Multicast group address 371 * Type of service parameter to be added in the future!!! 372 * Statistics are updated by the caller if needed 373 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 374 */ 375static struct mfc * 376mfc_find(in_addr_t o, in_addr_t g) 377{ 378 struct mfc *rt; 379 380 MFC_LOCK_ASSERT(); 381 382 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 383 if ((rt->mfc_origin.s_addr == o) && 384 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 385 break; 386 return rt; 387} 388 389/* 390 * Macros to compute elapsed time efficiently 391 * Borrowed from Van Jacobson's scheduling code 392 */ 393#define TV_DELTA(a, b, delta) { \ 394 int xxs; \ 395 delta = (a).tv_usec - (b).tv_usec; \ 396 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 397 switch (xxs) { \ 398 case 2: \ 399 delta += 1000000; \ 400 /* FALLTHROUGH */ \ 401 case 1: \ 402 delta += 1000000; \ 403 break; \ 404 default: \ 405 delta += (1000000 * xxs); \ 406 } \ 407 } \ 408} 409 410#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 411 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 412 413/* 414 * Handle MRT setsockopt commands to modify the multicast routing tables. 415 */ 416static int 417X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 418{ 419 int error, optval; 420 vifi_t vifi; 421 struct vifctl vifc; 422 struct mfcctl2 mfc; 423 struct bw_upcall bw_upcall; 424 uint32_t i; 425 426 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 427 return EPERM; 428 429 error = 0; 430 switch (sopt->sopt_name) { 431 case MRT_INIT: 432 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 433 if (error) 434 break; 435 error = ip_mrouter_init(so, optval); 436 break; 437 438 case MRT_DONE: 439 error = ip_mrouter_done(); 440 break; 441 442 case MRT_ADD_VIF: 443 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 444 if (error) 445 break; 446 error = add_vif(&vifc); 447 break; 448 449 case MRT_DEL_VIF: 450 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 451 if (error) 452 break; 453 error = del_vif(vifi); 454 break; 455 456 case MRT_ADD_MFC: 457 case MRT_DEL_MFC: 458 /* 459 * select data size depending on API version. 460 */ 461 if (sopt->sopt_name == MRT_ADD_MFC && 462 mrt_api_config & MRT_API_FLAGS_ALL) { 463 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 464 sizeof(struct mfcctl2)); 465 } else { 466 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 467 sizeof(struct mfcctl)); 468 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 469 sizeof(mfc) - sizeof(struct mfcctl)); 470 } 471 if (error) 472 break; 473 if (sopt->sopt_name == MRT_ADD_MFC) 474 error = add_mfc(&mfc); 475 else 476 error = del_mfc(&mfc); 477 break; 478 479 case MRT_ASSERT: 480 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 481 if (error) 482 break; 483 set_assert(optval); 484 break; 485 486 case MRT_API_CONFIG: 487 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 488 if (!error) 489 error = set_api_config(&i); 490 if (!error) 491 error = sooptcopyout(sopt, &i, sizeof i); 492 break; 493 494 case MRT_ADD_BW_UPCALL: 495 case MRT_DEL_BW_UPCALL: 496 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 497 sizeof bw_upcall); 498 if (error) 499 break; 500 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 501 error = add_bw_upcall(&bw_upcall); 502 else 503 error = del_bw_upcall(&bw_upcall); 504 break; 505 506 default: 507 error = EOPNOTSUPP; 508 break; 509 } 510 return error; 511} 512 513/* 514 * Handle MRT getsockopt commands 515 */ 516static int 517X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 518{ 519 int error; 520 static int version = 0x0305; /* !!! why is this here? XXX */ 521 522 switch (sopt->sopt_name) { 523 case MRT_VERSION: 524 error = sooptcopyout(sopt, &version, sizeof version); 525 break; 526 527 case MRT_ASSERT: 528 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 529 break; 530 531 case MRT_API_SUPPORT: 532 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 533 break; 534 535 case MRT_API_CONFIG: 536 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 537 break; 538 539 default: 540 error = EOPNOTSUPP; 541 break; 542 } 543 return error; 544} 545 546/* 547 * Handle ioctl commands to obtain information from the cache 548 */ 549static int 550X_mrt_ioctl(int cmd, caddr_t data) 551{ 552 int error = 0; 553 554 /* 555 * Currently the only function calling this ioctl routine is rtioctl(). 556 * Typically, only root can create the raw socket in order to execute 557 * this ioctl method, however the request might be coming from a prison 558 */ 559 error = priv_check(curthread, PRIV_NETINET_MROUTE); 560 if (error) 561 return (error); 562 switch (cmd) { 563 case (SIOCGETVIFCNT): 564 error = get_vif_cnt((struct sioc_vif_req *)data); 565 break; 566 567 case (SIOCGETSGCNT): 568 error = get_sg_cnt((struct sioc_sg_req *)data); 569 break; 570 571 default: 572 error = EINVAL; 573 break; 574 } 575 return error; 576} 577 578/* 579 * returns the packet, byte, rpf-failure count for the source group provided 580 */ 581static int 582get_sg_cnt(struct sioc_sg_req *req) 583{ 584 struct mfc *rt; 585 586 MFC_LOCK(); 587 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 588 if (rt == NULL) { 589 MFC_UNLOCK(); 590 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 591 return EADDRNOTAVAIL; 592 } 593 req->pktcnt = rt->mfc_pkt_cnt; 594 req->bytecnt = rt->mfc_byte_cnt; 595 req->wrong_if = rt->mfc_wrong_if; 596 MFC_UNLOCK(); 597 return 0; 598} 599 600/* 601 * returns the input and output packet and byte counts on the vif provided 602 */ 603static int 604get_vif_cnt(struct sioc_vif_req *req) 605{ 606 vifi_t vifi = req->vifi; 607 608 VIF_LOCK(); 609 if (vifi >= numvifs) { 610 VIF_UNLOCK(); 611 return EINVAL; 612 } 613 614 req->icount = viftable[vifi].v_pkt_in; 615 req->ocount = viftable[vifi].v_pkt_out; 616 req->ibytes = viftable[vifi].v_bytes_in; 617 req->obytes = viftable[vifi].v_bytes_out; 618 VIF_UNLOCK(); 619 620 return 0; 621} 622 623static void 624ip_mrouter_reset(void) 625{ 626 bzero((caddr_t)mfctable, sizeof(mfctable)); 627 bzero((caddr_t)nexpire, sizeof(nexpire)); 628 629 pim_assert = 0; 630 mrt_api_config = 0; 631 632 callout_init(&expire_upcalls_ch, NET_CALLOUT_MPSAFE); 633 634 bw_upcalls_n = 0; 635 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 636 callout_init(&bw_upcalls_ch, NET_CALLOUT_MPSAFE); 637 callout_init(&bw_meter_ch, NET_CALLOUT_MPSAFE); 638} 639 640static struct mtx mrouter_mtx; 641 642static void 643if_detached_event(void *arg __unused, struct ifnet *ifp) 644{ 645 vifi_t vifi; 646 int i; 647 struct mfc *mfc; 648 struct mfc *nmfc; 649 struct mfc **ppmfc; /* Pointer to previous node's next-pointer */ 650 struct rtdetq *pq; 651 struct rtdetq *npq; 652 653 mtx_lock(&mrouter_mtx); 654 if (ip_mrouter == NULL) { 655 mtx_unlock(&mrouter_mtx); 656 } 657 658 /* 659 * Tear down multicast forwarder state associated with this ifnet. 660 * 1. Walk the vif list, matching vifs against this ifnet. 661 * 2. Walk the multicast forwarding cache (mfc) looking for 662 * inner matches with this vif's index. 663 * 3. Free any pending mbufs for this mfc. 664 * 4. Free the associated mfc entry and state associated with this vif. 665 * Be very careful about unlinking from a singly-linked list whose 666 * "head node" is a pointer in a simple array. 667 * 5. Free vif state. This should disable ALLMULTI on the interface. 668 */ 669 VIF_LOCK(); 670 MFC_LOCK(); 671 for (vifi = 0; vifi < numvifs; vifi++) { 672 if (viftable[vifi].v_ifp != ifp) 673 continue; 674 for (i = 0; i < MFCTBLSIZ; i++) { 675 ppmfc = &mfctable[i]; 676 for (mfc = mfctable[i]; mfc != NULL; ) { 677 nmfc = mfc->mfc_next; 678 if (mfc->mfc_parent == vifi) { 679 for (pq = mfc->mfc_stall; pq != NULL; ) { 680 npq = pq->next; 681 m_freem(pq->m); 682 free(pq, M_MRTABLE); 683 pq = npq; 684 } 685 free_bw_list(mfc->mfc_bw_meter); 686 free(mfc, M_MRTABLE); 687 *ppmfc = nmfc; 688 } else { 689 ppmfc = &mfc->mfc_next; 690 } 691 mfc = nmfc; 692 } 693 } 694 del_vif_locked(vifi); 695 } 696 MFC_UNLOCK(); 697 VIF_UNLOCK(); 698 699 mtx_unlock(&mrouter_mtx); 700} 701 702/* 703 * Enable multicast routing 704 */ 705static int 706ip_mrouter_init(struct socket *so, int version) 707{ 708 if (mrtdebug) 709 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 710 so->so_type, so->so_proto->pr_protocol); 711 712 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 713 return EOPNOTSUPP; 714 715 if (version != 1) 716 return ENOPROTOOPT; 717 718 mtx_lock(&mrouter_mtx); 719 720 if (ip_mrouter != NULL) { 721 mtx_unlock(&mrouter_mtx); 722 return EADDRINUSE; 723 } 724 725 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 726 if_detached_event, NULL, EVENTHANDLER_PRI_ANY); 727 if (if_detach_event_tag == NULL) { 728 mtx_unlock(&mrouter_mtx); 729 return (ENOMEM); 730 } 731 732 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 733 734 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 735 expire_bw_upcalls_send, NULL); 736 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 737 738 ip_mrouter = so; 739 740 mtx_unlock(&mrouter_mtx); 741 742 if (mrtdebug) 743 log(LOG_DEBUG, "ip_mrouter_init\n"); 744 745 return 0; 746} 747 748/* 749 * Disable multicast routing 750 */ 751static int 752X_ip_mrouter_done(void) 753{ 754 vifi_t vifi; 755 int i; 756 struct ifnet *ifp; 757 struct ifreq ifr; 758 struct mfc *rt; 759 struct rtdetq *rte; 760 761 mtx_lock(&mrouter_mtx); 762 763 if (ip_mrouter == NULL) { 764 mtx_unlock(&mrouter_mtx); 765 return EINVAL; 766 } 767 768 /* 769 * Detach/disable hooks to the reset of the system. 770 */ 771 ip_mrouter = NULL; 772 mrt_api_config = 0; 773 774 VIF_LOCK(); 775 /* 776 * For each phyint in use, disable promiscuous reception of all IP 777 * multicasts. 778 */ 779 for (vifi = 0; vifi < numvifs; vifi++) { 780 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 781 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 782 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 783 784 so->sin_len = sizeof(struct sockaddr_in); 785 so->sin_family = AF_INET; 786 so->sin_addr.s_addr = INADDR_ANY; 787 ifp = viftable[vifi].v_ifp; 788 if_allmulti(ifp, 0); 789 } 790 } 791 bzero((caddr_t)viftable, sizeof(viftable)); 792 numvifs = 0; 793 pim_assert = 0; 794 VIF_UNLOCK(); 795 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 796 797 /* 798 * Free all multicast forwarding cache entries. 799 */ 800 callout_stop(&expire_upcalls_ch); 801 callout_stop(&bw_upcalls_ch); 802 callout_stop(&bw_meter_ch); 803 804 MFC_LOCK(); 805 for (i = 0; i < MFCTBLSIZ; i++) { 806 for (rt = mfctable[i]; rt != NULL; ) { 807 struct mfc *nr = rt->mfc_next; 808 809 for (rte = rt->mfc_stall; rte != NULL; ) { 810 struct rtdetq *n = rte->next; 811 812 m_freem(rte->m); 813 free(rte, M_MRTABLE); 814 rte = n; 815 } 816 free_bw_list(rt->mfc_bw_meter); 817 free(rt, M_MRTABLE); 818 rt = nr; 819 } 820 } 821 bzero((caddr_t)mfctable, sizeof(mfctable)); 822 bzero((caddr_t)nexpire, sizeof(nexpire)); 823 bw_upcalls_n = 0; 824 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 825 MFC_UNLOCK(); 826 827 reg_vif_num = VIFI_INVALID; 828 829 mtx_unlock(&mrouter_mtx); 830 831 if (mrtdebug) 832 log(LOG_DEBUG, "ip_mrouter_done\n"); 833 834 return 0; 835} 836 837/* 838 * Set PIM assert processing global 839 */ 840static int 841set_assert(int i) 842{ 843 if ((i != 1) && (i != 0)) 844 return EINVAL; 845 846 pim_assert = i; 847 848 return 0; 849} 850 851/* 852 * Configure API capabilities 853 */ 854int 855set_api_config(uint32_t *apival) 856{ 857 int i; 858 859 /* 860 * We can set the API capabilities only if it is the first operation 861 * after MRT_INIT. I.e.: 862 * - there are no vifs installed 863 * - pim_assert is not enabled 864 * - the MFC table is empty 865 */ 866 if (numvifs > 0) { 867 *apival = 0; 868 return EPERM; 869 } 870 if (pim_assert) { 871 *apival = 0; 872 return EPERM; 873 } 874 for (i = 0; i < MFCTBLSIZ; i++) { 875 if (mfctable[i] != NULL) { 876 *apival = 0; 877 return EPERM; 878 } 879 } 880 881 mrt_api_config = *apival & mrt_api_support; 882 *apival = mrt_api_config; 883 884 return 0; 885} 886 887/* 888 * Add a vif to the vif table 889 */ 890static int 891add_vif(struct vifctl *vifcp) 892{ 893 struct vif *vifp = viftable + vifcp->vifc_vifi; 894 struct sockaddr_in sin = {sizeof sin, AF_INET}; 895 struct ifaddr *ifa; 896 struct ifnet *ifp; 897 int error; 898 899 VIF_LOCK(); 900 if (vifcp->vifc_vifi >= MAXVIFS) { 901 VIF_UNLOCK(); 902 return EINVAL; 903 } 904 /* rate limiting is no longer supported by this code */ 905 if (vifcp->vifc_rate_limit != 0) { 906 log(LOG_ERR, "rate limiting is no longer supported\n"); 907 VIF_UNLOCK(); 908 return EINVAL; 909 } 910 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) { 911 VIF_UNLOCK(); 912 return EADDRINUSE; 913 } 914 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) { 915 VIF_UNLOCK(); 916 return EADDRNOTAVAIL; 917 } 918 919 /* Find the interface with an address in AF_INET family */ 920 if (vifcp->vifc_flags & VIFF_REGISTER) { 921 /* 922 * XXX: Because VIFF_REGISTER does not really need a valid 923 * local interface (e.g. it could be 127.0.0.2), we don't 924 * check its address. 925 */ 926 ifp = NULL; 927 } else { 928 sin.sin_addr = vifcp->vifc_lcl_addr; 929 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 930 if (ifa == NULL) { 931 VIF_UNLOCK(); 932 return EADDRNOTAVAIL; 933 } 934 ifp = ifa->ifa_ifp; 935 } 936 937 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { 938 log(LOG_ERR, "tunnels are no longer supported\n"); 939 VIF_UNLOCK(); 940 return EOPNOTSUPP; 941 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 942 ifp = &multicast_register_if; 943 if (mrtdebug) 944 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 945 (void *)&multicast_register_if); 946 if (reg_vif_num == VIFI_INVALID) { 947 if_initname(&multicast_register_if, "register_vif", 0); 948 multicast_register_if.if_flags = IFF_LOOPBACK; 949 reg_vif_num = vifcp->vifc_vifi; 950 } 951 } else { /* Make sure the interface supports multicast */ 952 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 953 VIF_UNLOCK(); 954 return EOPNOTSUPP; 955 } 956 957 /* Enable promiscuous reception of all IP multicasts from the if */ 958 error = if_allmulti(ifp, 1); 959 if (error) { 960 VIF_UNLOCK(); 961 return error; 962 } 963 } 964 965 vifp->v_flags = vifcp->vifc_flags; 966 vifp->v_threshold = vifcp->vifc_threshold; 967 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 968 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 969 vifp->v_ifp = ifp; 970 vifp->v_rsvp_on = 0; 971 vifp->v_rsvpd = NULL; 972 /* initialize per vif pkt counters */ 973 vifp->v_pkt_in = 0; 974 vifp->v_pkt_out = 0; 975 vifp->v_bytes_in = 0; 976 vifp->v_bytes_out = 0; 977 bzero(&vifp->v_route, sizeof(vifp->v_route)); 978 979 /* Adjust numvifs up if the vifi is higher than numvifs */ 980 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 981 982 VIF_UNLOCK(); 983 984 if (mrtdebug) 985 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x\n", 986 vifcp->vifc_vifi, 987 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 988 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 989 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 990 vifcp->vifc_threshold); 991 992 return 0; 993} 994 995/* 996 * Delete a vif from the vif table 997 */ 998static int 999del_vif_locked(vifi_t vifi) 1000{ 1001 struct vif *vifp; 1002 1003 VIF_LOCK_ASSERT(); 1004 1005 if (vifi >= numvifs) { 1006 return EINVAL; 1007 } 1008 vifp = &viftable[vifi]; 1009 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) { 1010 return EADDRNOTAVAIL; 1011 } 1012 1013 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 1014 if_allmulti(vifp->v_ifp, 0); 1015 1016 if (vifp->v_flags & VIFF_REGISTER) 1017 reg_vif_num = VIFI_INVALID; 1018 1019 bzero((caddr_t)vifp, sizeof (*vifp)); 1020 1021 if (mrtdebug) 1022 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1023 1024 /* Adjust numvifs down */ 1025 for (vifi = numvifs; vifi > 0; vifi--) 1026 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1027 break; 1028 numvifs = vifi; 1029 1030 return 0; 1031} 1032 1033static int 1034del_vif(vifi_t vifi) 1035{ 1036 int cc; 1037 1038 VIF_LOCK(); 1039 cc = del_vif_locked(vifi); 1040 VIF_UNLOCK(); 1041 1042 return cc; 1043} 1044 1045/* 1046 * update an mfc entry without resetting counters and S,G addresses. 1047 */ 1048static void 1049update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1050{ 1051 int i; 1052 1053 rt->mfc_parent = mfccp->mfcc_parent; 1054 for (i = 0; i < numvifs; i++) { 1055 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1056 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1057 MRT_MFC_FLAGS_ALL; 1058 } 1059 /* set the RP address */ 1060 if (mrt_api_config & MRT_MFC_RP) 1061 rt->mfc_rp = mfccp->mfcc_rp; 1062 else 1063 rt->mfc_rp.s_addr = INADDR_ANY; 1064} 1065 1066/* 1067 * fully initialize an mfc entry from the parameter. 1068 */ 1069static void 1070init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1071{ 1072 rt->mfc_origin = mfccp->mfcc_origin; 1073 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1074 1075 update_mfc_params(rt, mfccp); 1076 1077 /* initialize pkt counters per src-grp */ 1078 rt->mfc_pkt_cnt = 0; 1079 rt->mfc_byte_cnt = 0; 1080 rt->mfc_wrong_if = 0; 1081 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1082} 1083 1084 1085/* 1086 * Add an mfc entry 1087 */ 1088static int 1089add_mfc(struct mfcctl2 *mfccp) 1090{ 1091 struct mfc *rt; 1092 u_long hash; 1093 struct rtdetq *rte; 1094 u_short nstl; 1095 1096 VIF_LOCK(); 1097 MFC_LOCK(); 1098 1099 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1100 1101 /* If an entry already exists, just update the fields */ 1102 if (rt) { 1103 if (mrtdebug & DEBUG_MFC) 1104 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1105 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1106 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1107 mfccp->mfcc_parent); 1108 1109 update_mfc_params(rt, mfccp); 1110 MFC_UNLOCK(); 1111 VIF_UNLOCK(); 1112 return 0; 1113 } 1114 1115 /* 1116 * Find the entry for which the upcall was made and update 1117 */ 1118 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1119 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1120 1121 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1122 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1123 (rt->mfc_stall != NULL)) { 1124 1125 if (nstl++) 1126 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1127 "multiple kernel entries", 1128 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1129 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1130 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1131 1132 if (mrtdebug & DEBUG_MFC) 1133 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1134 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1135 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1136 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1137 1138 init_mfc_params(rt, mfccp); 1139 1140 rt->mfc_expire = 0; /* Don't clean this guy up */ 1141 nexpire[hash]--; 1142 1143 /* free packets Qed at the end of this entry */ 1144 for (rte = rt->mfc_stall; rte != NULL; ) { 1145 struct rtdetq *n = rte->next; 1146 1147 ip_mdq(rte->m, rte->ifp, rt, -1); 1148 m_freem(rte->m); 1149 free(rte, M_MRTABLE); 1150 rte = n; 1151 } 1152 rt->mfc_stall = NULL; 1153 } 1154 } 1155 1156 /* 1157 * It is possible that an entry is being inserted without an upcall 1158 */ 1159 if (nstl == 0) { 1160 if (mrtdebug & DEBUG_MFC) 1161 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1162 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1163 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1164 mfccp->mfcc_parent); 1165 1166 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1167 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1168 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1169 init_mfc_params(rt, mfccp); 1170 if (rt->mfc_expire) 1171 nexpire[hash]--; 1172 rt->mfc_expire = 0; 1173 break; /* XXX */ 1174 } 1175 } 1176 if (rt == NULL) { /* no upcall, so make a new entry */ 1177 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1178 if (rt == NULL) { 1179 MFC_UNLOCK(); 1180 VIF_UNLOCK(); 1181 return ENOBUFS; 1182 } 1183 1184 init_mfc_params(rt, mfccp); 1185 rt->mfc_expire = 0; 1186 rt->mfc_stall = NULL; 1187 1188 rt->mfc_bw_meter = NULL; 1189 /* insert new entry at head of hash chain */ 1190 rt->mfc_next = mfctable[hash]; 1191 mfctable[hash] = rt; 1192 } 1193 } 1194 MFC_UNLOCK(); 1195 VIF_UNLOCK(); 1196 return 0; 1197} 1198 1199/* 1200 * Delete an mfc entry 1201 */ 1202static int 1203del_mfc(struct mfcctl2 *mfccp) 1204{ 1205 struct in_addr origin; 1206 struct in_addr mcastgrp; 1207 struct mfc *rt; 1208 struct mfc **nptr; 1209 u_long hash; 1210 struct bw_meter *list; 1211 1212 origin = mfccp->mfcc_origin; 1213 mcastgrp = mfccp->mfcc_mcastgrp; 1214 1215 if (mrtdebug & DEBUG_MFC) 1216 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1217 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1218 1219 MFC_LOCK(); 1220 1221 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1222 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1223 if (origin.s_addr == rt->mfc_origin.s_addr && 1224 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1225 rt->mfc_stall == NULL) 1226 break; 1227 if (rt == NULL) { 1228 MFC_UNLOCK(); 1229 return EADDRNOTAVAIL; 1230 } 1231 1232 *nptr = rt->mfc_next; 1233 1234 /* 1235 * free the bw_meter entries 1236 */ 1237 list = rt->mfc_bw_meter; 1238 rt->mfc_bw_meter = NULL; 1239 1240 free(rt, M_MRTABLE); 1241 1242 free_bw_list(list); 1243 1244 MFC_UNLOCK(); 1245 1246 return 0; 1247} 1248 1249/* 1250 * Send a message to the routing daemon on the multicast routing socket 1251 */ 1252static int 1253socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1254{ 1255 if (s) { 1256 SOCKBUF_LOCK(&s->so_rcv); 1257 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, 1258 NULL) != 0) { 1259 sorwakeup_locked(s); 1260 return 0; 1261 } 1262 SOCKBUF_UNLOCK(&s->so_rcv); 1263 } 1264 m_freem(mm); 1265 return -1; 1266} 1267 1268/* 1269 * IP multicast forwarding function. This function assumes that the packet 1270 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1271 * pointed to by "ifp", and the packet is to be relayed to other networks 1272 * that have members of the packet's destination IP multicast group. 1273 * 1274 * The packet is returned unscathed to the caller, unless it is 1275 * erroneous, in which case a non-zero return value tells the caller to 1276 * discard it. 1277 */ 1278 1279#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1280 1281static int 1282X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1283 struct ip_moptions *imo) 1284{ 1285 struct mfc *rt; 1286 int error; 1287 vifi_t vifi; 1288 1289 if (mrtdebug & DEBUG_FORWARD) 1290 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1291 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1292 (void *)ifp); 1293 1294 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1295 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1296 /* 1297 * Packet arrived via a physical interface or 1298 * an encapsulated tunnel or a register_vif. 1299 */ 1300 } else { 1301 /* 1302 * Packet arrived through a source-route tunnel. 1303 * Source-route tunnels are no longer supported. 1304 */ 1305 static int last_log; 1306 if (last_log != time_uptime) { 1307 last_log = time_uptime; 1308 log(LOG_ERR, 1309 "ip_mforward: received source-routed packet from %lx\n", 1310 (u_long)ntohl(ip->ip_src.s_addr)); 1311 } 1312 return 1; 1313 } 1314 1315 VIF_LOCK(); 1316 MFC_LOCK(); 1317 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1318 if (ip->ip_ttl < MAXTTL) 1319 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1320 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1321 struct vif *vifp = viftable + vifi; 1322 1323 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1324 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1325 vifi, 1326 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1327 vifp->v_ifp->if_xname); 1328 } 1329 error = ip_mdq(m, ifp, NULL, vifi); 1330 MFC_UNLOCK(); 1331 VIF_UNLOCK(); 1332 return error; 1333 } 1334 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1335 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1336 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1337 if (!imo) 1338 printf("In fact, no options were specified at all\n"); 1339 } 1340 1341 /* 1342 * Don't forward a packet with time-to-live of zero or one, 1343 * or a packet destined to a local-only group. 1344 */ 1345 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) { 1346 MFC_UNLOCK(); 1347 VIF_UNLOCK(); 1348 return 0; 1349 } 1350 1351 /* 1352 * Determine forwarding vifs from the forwarding cache table 1353 */ 1354 ++mrtstat.mrts_mfc_lookups; 1355 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1356 1357 /* Entry exists, so forward if necessary */ 1358 if (rt != NULL) { 1359 error = ip_mdq(m, ifp, rt, -1); 1360 MFC_UNLOCK(); 1361 VIF_UNLOCK(); 1362 return error; 1363 } else { 1364 /* 1365 * If we don't have a route for packet's origin, 1366 * Make a copy of the packet & send message to routing daemon 1367 */ 1368 1369 struct mbuf *mb0; 1370 struct rtdetq *rte; 1371 u_long hash; 1372 int hlen = ip->ip_hl << 2; 1373 1374 ++mrtstat.mrts_mfc_misses; 1375 1376 mrtstat.mrts_no_route++; 1377 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1378 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1379 (u_long)ntohl(ip->ip_src.s_addr), 1380 (u_long)ntohl(ip->ip_dst.s_addr)); 1381 1382 /* 1383 * Allocate mbufs early so that we don't do extra work if we are 1384 * just going to fail anyway. Make sure to pullup the header so 1385 * that other people can't step on it. 1386 */ 1387 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1388 if (rte == NULL) { 1389 MFC_UNLOCK(); 1390 VIF_UNLOCK(); 1391 return ENOBUFS; 1392 } 1393 mb0 = m_copypacket(m, M_DONTWAIT); 1394 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1395 mb0 = m_pullup(mb0, hlen); 1396 if (mb0 == NULL) { 1397 free(rte, M_MRTABLE); 1398 MFC_UNLOCK(); 1399 VIF_UNLOCK(); 1400 return ENOBUFS; 1401 } 1402 1403 /* is there an upcall waiting for this flow ? */ 1404 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1405 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1406 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1407 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1408 (rt->mfc_stall != NULL)) 1409 break; 1410 } 1411 1412 if (rt == NULL) { 1413 int i; 1414 struct igmpmsg *im; 1415 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1416 struct mbuf *mm; 1417 1418 /* 1419 * Locate the vifi for the incoming interface for this packet. 1420 * If none found, drop packet. 1421 */ 1422 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1423 ; 1424 if (vifi >= numvifs) /* vif not found, drop packet */ 1425 goto non_fatal; 1426 1427 /* no upcall, so make a new entry */ 1428 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1429 if (rt == NULL) 1430 goto fail; 1431 /* Make a copy of the header to send to the user level process */ 1432 mm = m_copy(mb0, 0, hlen); 1433 if (mm == NULL) 1434 goto fail1; 1435 1436 /* 1437 * Send message to routing daemon to install 1438 * a route into the kernel table 1439 */ 1440 1441 im = mtod(mm, struct igmpmsg *); 1442 im->im_msgtype = IGMPMSG_NOCACHE; 1443 im->im_mbz = 0; 1444 im->im_vif = vifi; 1445 1446 mrtstat.mrts_upcalls++; 1447 1448 k_igmpsrc.sin_addr = ip->ip_src; 1449 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1450 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1451 ++mrtstat.mrts_upq_sockfull; 1452fail1: 1453 free(rt, M_MRTABLE); 1454fail: 1455 free(rte, M_MRTABLE); 1456 m_freem(mb0); 1457 MFC_UNLOCK(); 1458 VIF_UNLOCK(); 1459 return ENOBUFS; 1460 } 1461 1462 /* insert new entry at head of hash chain */ 1463 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1464 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1465 rt->mfc_expire = UPCALL_EXPIRE; 1466 nexpire[hash]++; 1467 for (i = 0; i < numvifs; i++) { 1468 rt->mfc_ttls[i] = 0; 1469 rt->mfc_flags[i] = 0; 1470 } 1471 rt->mfc_parent = -1; 1472 1473 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1474 1475 rt->mfc_bw_meter = NULL; 1476 1477 /* link into table */ 1478 rt->mfc_next = mfctable[hash]; 1479 mfctable[hash] = rt; 1480 rt->mfc_stall = rte; 1481 1482 } else { 1483 /* determine if q has overflowed */ 1484 int npkts = 0; 1485 struct rtdetq **p; 1486 1487 /* 1488 * XXX ouch! we need to append to the list, but we 1489 * only have a pointer to the front, so we have to 1490 * scan the entire list every time. 1491 */ 1492 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1493 npkts++; 1494 1495 if (npkts > MAX_UPQ) { 1496 mrtstat.mrts_upq_ovflw++; 1497non_fatal: 1498 free(rte, M_MRTABLE); 1499 m_freem(mb0); 1500 MFC_UNLOCK(); 1501 VIF_UNLOCK(); 1502 return 0; 1503 } 1504 1505 /* Add this entry to the end of the queue */ 1506 *p = rte; 1507 } 1508 1509 rte->m = mb0; 1510 rte->ifp = ifp; 1511 rte->next = NULL; 1512 1513 MFC_UNLOCK(); 1514 VIF_UNLOCK(); 1515 1516 return 0; 1517 } 1518} 1519 1520/* 1521 * Clean up the cache entry if upcall is not serviced 1522 */ 1523static void 1524expire_upcalls(void *unused) 1525{ 1526 struct rtdetq *rte; 1527 struct mfc *mfc, **nptr; 1528 int i; 1529 1530 MFC_LOCK(); 1531 for (i = 0; i < MFCTBLSIZ; i++) { 1532 if (nexpire[i] == 0) 1533 continue; 1534 nptr = &mfctable[i]; 1535 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1536 /* 1537 * Skip real cache entries 1538 * Make sure it wasn't marked to not expire (shouldn't happen) 1539 * If it expires now 1540 */ 1541 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1542 --mfc->mfc_expire == 0) { 1543 if (mrtdebug & DEBUG_EXPIRE) 1544 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1545 (u_long)ntohl(mfc->mfc_origin.s_addr), 1546 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1547 /* 1548 * drop all the packets 1549 * free the mbuf with the pkt, if, timing info 1550 */ 1551 for (rte = mfc->mfc_stall; rte; ) { 1552 struct rtdetq *n = rte->next; 1553 1554 m_freem(rte->m); 1555 free(rte, M_MRTABLE); 1556 rte = n; 1557 } 1558 ++mrtstat.mrts_cache_cleanups; 1559 nexpire[i]--; 1560 1561 /* 1562 * free the bw_meter entries 1563 */ 1564 while (mfc->mfc_bw_meter != NULL) { 1565 struct bw_meter *x = mfc->mfc_bw_meter; 1566 1567 mfc->mfc_bw_meter = x->bm_mfc_next; 1568 free(x, M_BWMETER); 1569 } 1570 1571 *nptr = mfc->mfc_next; 1572 free(mfc, M_MRTABLE); 1573 } else { 1574 nptr = &mfc->mfc_next; 1575 } 1576 } 1577 } 1578 MFC_UNLOCK(); 1579 1580 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1581} 1582 1583/* 1584 * Packet forwarding routine once entry in the cache is made 1585 */ 1586static int 1587ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1588{ 1589 struct ip *ip = mtod(m, struct ip *); 1590 vifi_t vifi; 1591 int plen = ip->ip_len; 1592 1593 VIF_LOCK_ASSERT(); 1594 1595 /* 1596 * If xmt_vif is not -1, send on only the requested vif. 1597 * 1598 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1599 */ 1600 if (xmt_vif < numvifs) { 1601 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1602 pim_register_send(ip, viftable + xmt_vif, m, rt); 1603 else 1604 phyint_send(ip, viftable + xmt_vif, m); 1605 return 1; 1606 } 1607 1608 /* 1609 * Don't forward if it didn't arrive from the parent vif for its origin. 1610 */ 1611 vifi = rt->mfc_parent; 1612 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1613 /* came in the wrong interface */ 1614 if (mrtdebug & DEBUG_FORWARD) 1615 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1616 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1617 ++mrtstat.mrts_wrong_if; 1618 ++rt->mfc_wrong_if; 1619 /* 1620 * If we are doing PIM assert processing, send a message 1621 * to the routing daemon. 1622 * 1623 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1624 * can complete the SPT switch, regardless of the type 1625 * of the iif (broadcast media, GRE tunnel, etc). 1626 */ 1627 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1628 struct timeval now; 1629 u_long delta; 1630 1631 if (ifp == &multicast_register_if) 1632 pimstat.pims_rcv_registers_wrongiif++; 1633 1634 /* Get vifi for the incoming packet */ 1635 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1636 ; 1637 if (vifi >= numvifs) 1638 return 0; /* The iif is not found: ignore the packet. */ 1639 1640 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1641 return 0; /* WRONGVIF disabled: ignore the packet */ 1642 1643 GET_TIME(now); 1644 1645 TV_DELTA(now, rt->mfc_last_assert, delta); 1646 1647 if (delta > ASSERT_MSG_TIME) { 1648 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1649 struct igmpmsg *im; 1650 int hlen = ip->ip_hl << 2; 1651 struct mbuf *mm = m_copy(m, 0, hlen); 1652 1653 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1654 mm = m_pullup(mm, hlen); 1655 if (mm == NULL) 1656 return ENOBUFS; 1657 1658 rt->mfc_last_assert = now; 1659 1660 im = mtod(mm, struct igmpmsg *); 1661 im->im_msgtype = IGMPMSG_WRONGVIF; 1662 im->im_mbz = 0; 1663 im->im_vif = vifi; 1664 1665 mrtstat.mrts_upcalls++; 1666 1667 k_igmpsrc.sin_addr = im->im_src; 1668 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1669 log(LOG_WARNING, 1670 "ip_mforward: ip_mrouter socket queue full\n"); 1671 ++mrtstat.mrts_upq_sockfull; 1672 return ENOBUFS; 1673 } 1674 } 1675 } 1676 return 0; 1677 } 1678 1679 /* If I sourced this packet, it counts as output, else it was input. */ 1680 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1681 viftable[vifi].v_pkt_out++; 1682 viftable[vifi].v_bytes_out += plen; 1683 } else { 1684 viftable[vifi].v_pkt_in++; 1685 viftable[vifi].v_bytes_in += plen; 1686 } 1687 rt->mfc_pkt_cnt++; 1688 rt->mfc_byte_cnt += plen; 1689 1690 /* 1691 * For each vif, decide if a copy of the packet should be forwarded. 1692 * Forward if: 1693 * - the ttl exceeds the vif's threshold 1694 * - there are group members downstream on interface 1695 */ 1696 for (vifi = 0; vifi < numvifs; vifi++) 1697 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1698 viftable[vifi].v_pkt_out++; 1699 viftable[vifi].v_bytes_out += plen; 1700 if (viftable[vifi].v_flags & VIFF_REGISTER) 1701 pim_register_send(ip, viftable + vifi, m, rt); 1702 else 1703 phyint_send(ip, viftable + vifi, m); 1704 } 1705 1706 /* 1707 * Perform upcall-related bw measuring. 1708 */ 1709 if (rt->mfc_bw_meter != NULL) { 1710 struct bw_meter *x; 1711 struct timeval now; 1712 1713 GET_TIME(now); 1714 MFC_LOCK_ASSERT(); 1715 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1716 bw_meter_receive_packet(x, plen, &now); 1717 } 1718 1719 return 0; 1720} 1721 1722/* 1723 * check if a vif number is legal/ok. This is used by ip_output. 1724 */ 1725static int 1726X_legal_vif_num(int vif) 1727{ 1728 /* XXX unlocked, matter? */ 1729 return (vif >= 0 && vif < numvifs); 1730} 1731 1732/* 1733 * Return the local address used by this vif 1734 */ 1735static u_long 1736X_ip_mcast_src(int vifi) 1737{ 1738 /* XXX unlocked, matter? */ 1739 if (vifi >= 0 && vifi < numvifs) 1740 return viftable[vifi].v_lcl_addr.s_addr; 1741 else 1742 return INADDR_ANY; 1743} 1744 1745static void 1746phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1747{ 1748 struct mbuf *mb_copy; 1749 int hlen = ip->ip_hl << 2; 1750 1751 VIF_LOCK_ASSERT(); 1752 1753 /* 1754 * Make a new reference to the packet; make sure that 1755 * the IP header is actually copied, not just referenced, 1756 * so that ip_output() only scribbles on the copy. 1757 */ 1758 mb_copy = m_copypacket(m, M_DONTWAIT); 1759 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1760 mb_copy = m_pullup(mb_copy, hlen); 1761 if (mb_copy == NULL) 1762 return; 1763 1764 send_packet(vifp, mb_copy); 1765} 1766 1767static void 1768send_packet(struct vif *vifp, struct mbuf *m) 1769{ 1770 struct ip_moptions imo; 1771 struct in_multi *imm[2]; 1772 int error; 1773 1774 VIF_LOCK_ASSERT(); 1775 1776 imo.imo_multicast_ifp = vifp->v_ifp; 1777 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1778 imo.imo_multicast_loop = 1; 1779 imo.imo_multicast_vif = -1; 1780 imo.imo_num_memberships = 0; 1781 imo.imo_max_memberships = 2; 1782 imo.imo_membership = &imm[0]; 1783 1784 /* 1785 * Re-entrancy should not be a problem here, because 1786 * the packets that we send out and are looped back at us 1787 * should get rejected because they appear to come from 1788 * the loopback interface, thus preventing looping. 1789 */ 1790 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL); 1791 if (mrtdebug & DEBUG_XMIT) { 1792 log(LOG_DEBUG, "phyint_send on vif %td err %d\n", 1793 vifp - viftable, error); 1794 } 1795} 1796 1797static int 1798X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 1799{ 1800 int error, vifi; 1801 1802 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1803 return EOPNOTSUPP; 1804 1805 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 1806 if (error) 1807 return error; 1808 1809 VIF_LOCK(); 1810 1811 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 1812 VIF_UNLOCK(); 1813 return EADDRNOTAVAIL; 1814 } 1815 1816 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 1817 /* Check if socket is available. */ 1818 if (viftable[vifi].v_rsvpd != NULL) { 1819 VIF_UNLOCK(); 1820 return EADDRINUSE; 1821 } 1822 1823 viftable[vifi].v_rsvpd = so; 1824 /* This may seem silly, but we need to be sure we don't over-increment 1825 * the RSVP counter, in case something slips up. 1826 */ 1827 if (!viftable[vifi].v_rsvp_on) { 1828 viftable[vifi].v_rsvp_on = 1; 1829 rsvp_on++; 1830 } 1831 } else { /* must be VIF_OFF */ 1832 /* 1833 * XXX as an additional consistency check, one could make sure 1834 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 1835 * first parameter is pretty useless. 1836 */ 1837 viftable[vifi].v_rsvpd = NULL; 1838 /* 1839 * This may seem silly, but we need to be sure we don't over-decrement 1840 * the RSVP counter, in case something slips up. 1841 */ 1842 if (viftable[vifi].v_rsvp_on) { 1843 viftable[vifi].v_rsvp_on = 0; 1844 rsvp_on--; 1845 } 1846 } 1847 VIF_UNLOCK(); 1848 return 0; 1849} 1850 1851static void 1852X_ip_rsvp_force_done(struct socket *so) 1853{ 1854 int vifi; 1855 1856 /* Don't bother if it is not the right type of socket. */ 1857 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1858 return; 1859 1860 VIF_LOCK(); 1861 1862 /* The socket may be attached to more than one vif...this 1863 * is perfectly legal. 1864 */ 1865 for (vifi = 0; vifi < numvifs; vifi++) { 1866 if (viftable[vifi].v_rsvpd == so) { 1867 viftable[vifi].v_rsvpd = NULL; 1868 /* This may seem silly, but we need to be sure we don't 1869 * over-decrement the RSVP counter, in case something slips up. 1870 */ 1871 if (viftable[vifi].v_rsvp_on) { 1872 viftable[vifi].v_rsvp_on = 0; 1873 rsvp_on--; 1874 } 1875 } 1876 } 1877 1878 VIF_UNLOCK(); 1879} 1880 1881static void 1882X_rsvp_input(struct mbuf *m, int off) 1883{ 1884 int vifi; 1885 struct ip *ip = mtod(m, struct ip *); 1886 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 1887 struct ifnet *ifp; 1888 1889 if (rsvpdebug) 1890 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 1891 1892 /* Can still get packets with rsvp_on = 0 if there is a local member 1893 * of the group to which the RSVP packet is addressed. But in this 1894 * case we want to throw the packet away. 1895 */ 1896 if (!rsvp_on) { 1897 m_freem(m); 1898 return; 1899 } 1900 1901 if (rsvpdebug) 1902 printf("rsvp_input: check vifs\n"); 1903 1904#ifdef DIAGNOSTIC 1905 M_ASSERTPKTHDR(m); 1906#endif 1907 1908 ifp = m->m_pkthdr.rcvif; 1909 1910 VIF_LOCK(); 1911 /* Find which vif the packet arrived on. */ 1912 for (vifi = 0; vifi < numvifs; vifi++) 1913 if (viftable[vifi].v_ifp == ifp) 1914 break; 1915 1916 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 1917 /* 1918 * Drop the lock here to avoid holding it across rip_input. 1919 * This could make rsvpdebug printfs wrong. If you care, 1920 * record the state of stuff before dropping the lock. 1921 */ 1922 VIF_UNLOCK(); 1923 /* 1924 * If the old-style non-vif-associated socket is set, 1925 * then use it. Otherwise, drop packet since there 1926 * is no specific socket for this vif. 1927 */ 1928 if (ip_rsvpd != NULL) { 1929 if (rsvpdebug) 1930 printf("rsvp_input: Sending packet up old-style socket\n"); 1931 rip_input(m, off); /* xxx */ 1932 } else { 1933 if (rsvpdebug && vifi == numvifs) 1934 printf("rsvp_input: Can't find vif for packet.\n"); 1935 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 1936 printf("rsvp_input: No socket defined for vif %d\n",vifi); 1937 m_freem(m); 1938 } 1939 return; 1940 } 1941 rsvp_src.sin_addr = ip->ip_src; 1942 1943 if (rsvpdebug && m) 1944 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 1945 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 1946 1947 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 1948 if (rsvpdebug) 1949 printf("rsvp_input: Failed to append to socket\n"); 1950 } else { 1951 if (rsvpdebug) 1952 printf("rsvp_input: send packet up\n"); 1953 } 1954 VIF_UNLOCK(); 1955} 1956 1957/* 1958 * Code for bandwidth monitors 1959 */ 1960 1961/* 1962 * Define common interface for timeval-related methods 1963 */ 1964#define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 1965#define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 1966#define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 1967 1968static uint32_t 1969compute_bw_meter_flags(struct bw_upcall *req) 1970{ 1971 uint32_t flags = 0; 1972 1973 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 1974 flags |= BW_METER_UNIT_PACKETS; 1975 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 1976 flags |= BW_METER_UNIT_BYTES; 1977 if (req->bu_flags & BW_UPCALL_GEQ) 1978 flags |= BW_METER_GEQ; 1979 if (req->bu_flags & BW_UPCALL_LEQ) 1980 flags |= BW_METER_LEQ; 1981 1982 return flags; 1983} 1984 1985/* 1986 * Add a bw_meter entry 1987 */ 1988static int 1989add_bw_upcall(struct bw_upcall *req) 1990{ 1991 struct mfc *mfc; 1992 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 1993 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 1994 struct timeval now; 1995 struct bw_meter *x; 1996 uint32_t flags; 1997 1998 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 1999 return EOPNOTSUPP; 2000 2001 /* Test if the flags are valid */ 2002 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2003 return EINVAL; 2004 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2005 return EINVAL; 2006 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2007 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2008 return EINVAL; 2009 2010 /* Test if the threshold time interval is valid */ 2011 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2012 return EINVAL; 2013 2014 flags = compute_bw_meter_flags(req); 2015 2016 /* 2017 * Find if we have already same bw_meter entry 2018 */ 2019 MFC_LOCK(); 2020 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2021 if (mfc == NULL) { 2022 MFC_UNLOCK(); 2023 return EADDRNOTAVAIL; 2024 } 2025 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2026 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2027 &req->bu_threshold.b_time, ==)) && 2028 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2029 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2030 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2031 MFC_UNLOCK(); 2032 return 0; /* XXX Already installed */ 2033 } 2034 } 2035 2036 /* Allocate the new bw_meter entry */ 2037 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2038 if (x == NULL) { 2039 MFC_UNLOCK(); 2040 return ENOBUFS; 2041 } 2042 2043 /* Set the new bw_meter entry */ 2044 x->bm_threshold.b_time = req->bu_threshold.b_time; 2045 GET_TIME(now); 2046 x->bm_start_time = now; 2047 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2048 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2049 x->bm_measured.b_packets = 0; 2050 x->bm_measured.b_bytes = 0; 2051 x->bm_flags = flags; 2052 x->bm_time_next = NULL; 2053 x->bm_time_hash = BW_METER_BUCKETS; 2054 2055 /* Add the new bw_meter entry to the front of entries for this MFC */ 2056 x->bm_mfc = mfc; 2057 x->bm_mfc_next = mfc->mfc_bw_meter; 2058 mfc->mfc_bw_meter = x; 2059 schedule_bw_meter(x, &now); 2060 MFC_UNLOCK(); 2061 2062 return 0; 2063} 2064 2065static void 2066free_bw_list(struct bw_meter *list) 2067{ 2068 while (list != NULL) { 2069 struct bw_meter *x = list; 2070 2071 list = list->bm_mfc_next; 2072 unschedule_bw_meter(x); 2073 free(x, M_BWMETER); 2074 } 2075} 2076 2077/* 2078 * Delete one or multiple bw_meter entries 2079 */ 2080static int 2081del_bw_upcall(struct bw_upcall *req) 2082{ 2083 struct mfc *mfc; 2084 struct bw_meter *x; 2085 2086 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2087 return EOPNOTSUPP; 2088 2089 MFC_LOCK(); 2090 /* Find the corresponding MFC entry */ 2091 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2092 if (mfc == NULL) { 2093 MFC_UNLOCK(); 2094 return EADDRNOTAVAIL; 2095 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2096 /* 2097 * Delete all bw_meter entries for this mfc 2098 */ 2099 struct bw_meter *list; 2100 2101 list = mfc->mfc_bw_meter; 2102 mfc->mfc_bw_meter = NULL; 2103 free_bw_list(list); 2104 MFC_UNLOCK(); 2105 return 0; 2106 } else { /* Delete a single bw_meter entry */ 2107 struct bw_meter *prev; 2108 uint32_t flags = 0; 2109 2110 flags = compute_bw_meter_flags(req); 2111 2112 /* Find the bw_meter entry to delete */ 2113 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2114 prev = x, x = x->bm_mfc_next) { 2115 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2116 &req->bu_threshold.b_time, ==)) && 2117 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2118 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2119 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2120 break; 2121 } 2122 if (x != NULL) { /* Delete entry from the list for this MFC */ 2123 if (prev != NULL) 2124 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2125 else 2126 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2127 2128 unschedule_bw_meter(x); 2129 MFC_UNLOCK(); 2130 /* Free the bw_meter entry */ 2131 free(x, M_BWMETER); 2132 return 0; 2133 } else { 2134 MFC_UNLOCK(); 2135 return EINVAL; 2136 } 2137 } 2138 /* NOTREACHED */ 2139} 2140 2141/* 2142 * Perform bandwidth measurement processing that may result in an upcall 2143 */ 2144static void 2145bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2146{ 2147 struct timeval delta; 2148 2149 MFC_LOCK_ASSERT(); 2150 2151 delta = *nowp; 2152 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2153 2154 if (x->bm_flags & BW_METER_GEQ) { 2155 /* 2156 * Processing for ">=" type of bw_meter entry 2157 */ 2158 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2159 /* Reset the bw_meter entry */ 2160 x->bm_start_time = *nowp; 2161 x->bm_measured.b_packets = 0; 2162 x->bm_measured.b_bytes = 0; 2163 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2164 } 2165 2166 /* Record that a packet is received */ 2167 x->bm_measured.b_packets++; 2168 x->bm_measured.b_bytes += plen; 2169 2170 /* 2171 * Test if we should deliver an upcall 2172 */ 2173 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2174 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2175 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2176 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2177 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2178 /* Prepare an upcall for delivery */ 2179 bw_meter_prepare_upcall(x, nowp); 2180 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2181 } 2182 } 2183 } else if (x->bm_flags & BW_METER_LEQ) { 2184 /* 2185 * Processing for "<=" type of bw_meter entry 2186 */ 2187 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2188 /* 2189 * We are behind time with the multicast forwarding table 2190 * scanning for "<=" type of bw_meter entries, so test now 2191 * if we should deliver an upcall. 2192 */ 2193 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2194 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2195 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2196 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2197 /* Prepare an upcall for delivery */ 2198 bw_meter_prepare_upcall(x, nowp); 2199 } 2200 /* Reschedule the bw_meter entry */ 2201 unschedule_bw_meter(x); 2202 schedule_bw_meter(x, nowp); 2203 } 2204 2205 /* Record that a packet is received */ 2206 x->bm_measured.b_packets++; 2207 x->bm_measured.b_bytes += plen; 2208 2209 /* 2210 * Test if we should restart the measuring interval 2211 */ 2212 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2213 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2214 (x->bm_flags & BW_METER_UNIT_BYTES && 2215 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2216 /* Don't restart the measuring interval */ 2217 } else { 2218 /* Do restart the measuring interval */ 2219 /* 2220 * XXX: note that we don't unschedule and schedule, because this 2221 * might be too much overhead per packet. Instead, when we process 2222 * all entries for a given timer hash bin, we check whether it is 2223 * really a timeout. If not, we reschedule at that time. 2224 */ 2225 x->bm_start_time = *nowp; 2226 x->bm_measured.b_packets = 0; 2227 x->bm_measured.b_bytes = 0; 2228 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2229 } 2230 } 2231} 2232 2233/* 2234 * Prepare a bandwidth-related upcall 2235 */ 2236static void 2237bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2238{ 2239 struct timeval delta; 2240 struct bw_upcall *u; 2241 2242 MFC_LOCK_ASSERT(); 2243 2244 /* 2245 * Compute the measured time interval 2246 */ 2247 delta = *nowp; 2248 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2249 2250 /* 2251 * If there are too many pending upcalls, deliver them now 2252 */ 2253 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2254 bw_upcalls_send(); 2255 2256 /* 2257 * Set the bw_upcall entry 2258 */ 2259 u = &bw_upcalls[bw_upcalls_n++]; 2260 u->bu_src = x->bm_mfc->mfc_origin; 2261 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2262 u->bu_threshold.b_time = x->bm_threshold.b_time; 2263 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2264 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2265 u->bu_measured.b_time = delta; 2266 u->bu_measured.b_packets = x->bm_measured.b_packets; 2267 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2268 u->bu_flags = 0; 2269 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2270 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2271 if (x->bm_flags & BW_METER_UNIT_BYTES) 2272 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2273 if (x->bm_flags & BW_METER_GEQ) 2274 u->bu_flags |= BW_UPCALL_GEQ; 2275 if (x->bm_flags & BW_METER_LEQ) 2276 u->bu_flags |= BW_UPCALL_LEQ; 2277} 2278 2279/* 2280 * Send the pending bandwidth-related upcalls 2281 */ 2282static void 2283bw_upcalls_send(void) 2284{ 2285 struct mbuf *m; 2286 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2287 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2288 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2289 0, /* unused2 */ 2290 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2291 0, /* im_mbz */ 2292 0, /* im_vif */ 2293 0, /* unused3 */ 2294 { 0 }, /* im_src */ 2295 { 0 } }; /* im_dst */ 2296 2297 MFC_LOCK_ASSERT(); 2298 2299 if (bw_upcalls_n == 0) 2300 return; /* No pending upcalls */ 2301 2302 bw_upcalls_n = 0; 2303 2304 /* 2305 * Allocate a new mbuf, initialize it with the header and 2306 * the payload for the pending calls. 2307 */ 2308 MGETHDR(m, M_DONTWAIT, MT_DATA); 2309 if (m == NULL) { 2310 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2311 return; 2312 } 2313 2314 m->m_len = m->m_pkthdr.len = 0; 2315 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2316 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2317 2318 /* 2319 * Send the upcalls 2320 * XXX do we need to set the address in k_igmpsrc ? 2321 */ 2322 mrtstat.mrts_upcalls++; 2323 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2324 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2325 ++mrtstat.mrts_upq_sockfull; 2326 } 2327} 2328 2329/* 2330 * Compute the timeout hash value for the bw_meter entries 2331 */ 2332#define BW_METER_TIMEHASH(bw_meter, hash) \ 2333 do { \ 2334 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2335 \ 2336 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2337 (hash) = next_timeval.tv_sec; \ 2338 if (next_timeval.tv_usec) \ 2339 (hash)++; /* XXX: make sure we don't timeout early */ \ 2340 (hash) %= BW_METER_BUCKETS; \ 2341 } while (0) 2342 2343/* 2344 * Schedule a timer to process periodically bw_meter entry of type "<=" 2345 * by linking the entry in the proper hash bucket. 2346 */ 2347static void 2348schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2349{ 2350 int time_hash; 2351 2352 MFC_LOCK_ASSERT(); 2353 2354 if (!(x->bm_flags & BW_METER_LEQ)) 2355 return; /* XXX: we schedule timers only for "<=" entries */ 2356 2357 /* 2358 * Reset the bw_meter entry 2359 */ 2360 x->bm_start_time = *nowp; 2361 x->bm_measured.b_packets = 0; 2362 x->bm_measured.b_bytes = 0; 2363 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2364 2365 /* 2366 * Compute the timeout hash value and insert the entry 2367 */ 2368 BW_METER_TIMEHASH(x, time_hash); 2369 x->bm_time_next = bw_meter_timers[time_hash]; 2370 bw_meter_timers[time_hash] = x; 2371 x->bm_time_hash = time_hash; 2372} 2373 2374/* 2375 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2376 * by removing the entry from the proper hash bucket. 2377 */ 2378static void 2379unschedule_bw_meter(struct bw_meter *x) 2380{ 2381 int time_hash; 2382 struct bw_meter *prev, *tmp; 2383 2384 MFC_LOCK_ASSERT(); 2385 2386 if (!(x->bm_flags & BW_METER_LEQ)) 2387 return; /* XXX: we schedule timers only for "<=" entries */ 2388 2389 /* 2390 * Compute the timeout hash value and delete the entry 2391 */ 2392 time_hash = x->bm_time_hash; 2393 if (time_hash >= BW_METER_BUCKETS) 2394 return; /* Entry was not scheduled */ 2395 2396 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2397 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2398 if (tmp == x) 2399 break; 2400 2401 if (tmp == NULL) 2402 panic("unschedule_bw_meter: bw_meter entry not found"); 2403 2404 if (prev != NULL) 2405 prev->bm_time_next = x->bm_time_next; 2406 else 2407 bw_meter_timers[time_hash] = x->bm_time_next; 2408 2409 x->bm_time_next = NULL; 2410 x->bm_time_hash = BW_METER_BUCKETS; 2411} 2412 2413 2414/* 2415 * Process all "<=" type of bw_meter that should be processed now, 2416 * and for each entry prepare an upcall if necessary. Each processed 2417 * entry is rescheduled again for the (periodic) processing. 2418 * 2419 * This is run periodically (once per second normally). On each round, 2420 * all the potentially matching entries are in the hash slot that we are 2421 * looking at. 2422 */ 2423static void 2424bw_meter_process() 2425{ 2426 static uint32_t last_tv_sec; /* last time we processed this */ 2427 2428 uint32_t loops; 2429 int i; 2430 struct timeval now, process_endtime; 2431 2432 GET_TIME(now); 2433 if (last_tv_sec == now.tv_sec) 2434 return; /* nothing to do */ 2435 2436 loops = now.tv_sec - last_tv_sec; 2437 last_tv_sec = now.tv_sec; 2438 if (loops > BW_METER_BUCKETS) 2439 loops = BW_METER_BUCKETS; 2440 2441 MFC_LOCK(); 2442 /* 2443 * Process all bins of bw_meter entries from the one after the last 2444 * processed to the current one. On entry, i points to the last bucket 2445 * visited, so we need to increment i at the beginning of the loop. 2446 */ 2447 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2448 struct bw_meter *x, *tmp_list; 2449 2450 if (++i >= BW_METER_BUCKETS) 2451 i = 0; 2452 2453 /* Disconnect the list of bw_meter entries from the bin */ 2454 tmp_list = bw_meter_timers[i]; 2455 bw_meter_timers[i] = NULL; 2456 2457 /* Process the list of bw_meter entries */ 2458 while (tmp_list != NULL) { 2459 x = tmp_list; 2460 tmp_list = tmp_list->bm_time_next; 2461 2462 /* Test if the time interval is over */ 2463 process_endtime = x->bm_start_time; 2464 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2465 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2466 /* Not yet: reschedule, but don't reset */ 2467 int time_hash; 2468 2469 BW_METER_TIMEHASH(x, time_hash); 2470 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2471 /* 2472 * XXX: somehow the bin processing is a bit ahead of time. 2473 * Put the entry in the next bin. 2474 */ 2475 if (++time_hash >= BW_METER_BUCKETS) 2476 time_hash = 0; 2477 } 2478 x->bm_time_next = bw_meter_timers[time_hash]; 2479 bw_meter_timers[time_hash] = x; 2480 x->bm_time_hash = time_hash; 2481 2482 continue; 2483 } 2484 2485 /* 2486 * Test if we should deliver an upcall 2487 */ 2488 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2489 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2490 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2491 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2492 /* Prepare an upcall for delivery */ 2493 bw_meter_prepare_upcall(x, &now); 2494 } 2495 2496 /* 2497 * Reschedule for next processing 2498 */ 2499 schedule_bw_meter(x, &now); 2500 } 2501 } 2502 2503 /* Send all upcalls that are pending delivery */ 2504 bw_upcalls_send(); 2505 2506 MFC_UNLOCK(); 2507} 2508 2509/* 2510 * A periodic function for sending all upcalls that are pending delivery 2511 */ 2512static void 2513expire_bw_upcalls_send(void *unused) 2514{ 2515 MFC_LOCK(); 2516 bw_upcalls_send(); 2517 MFC_UNLOCK(); 2518 2519 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2520 expire_bw_upcalls_send, NULL); 2521} 2522 2523/* 2524 * A periodic function for periodic scanning of the multicast forwarding 2525 * table for processing all "<=" bw_meter entries. 2526 */ 2527static void 2528expire_bw_meter_process(void *unused) 2529{ 2530 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2531 bw_meter_process(); 2532 2533 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2534} 2535 2536/* 2537 * End of bandwidth monitoring code 2538 */ 2539 2540/* 2541 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2542 * 2543 */ 2544static int 2545pim_register_send(struct ip *ip, struct vif *vifp, 2546 struct mbuf *m, struct mfc *rt) 2547{ 2548 struct mbuf *mb_copy, *mm; 2549 2550 if (mrtdebug & DEBUG_PIM) 2551 log(LOG_DEBUG, "pim_register_send: "); 2552 2553 /* 2554 * Do not send IGMP_WHOLEPKT notifications to userland, if the 2555 * rendezvous point was unspecified, and we were told not to. 2556 */ 2557 if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) && 2558 (rt->mfc_rp.s_addr == INADDR_ANY)) 2559 return 0; 2560 2561 mb_copy = pim_register_prepare(ip, m); 2562 if (mb_copy == NULL) 2563 return ENOBUFS; 2564 2565 /* 2566 * Send all the fragments. Note that the mbuf for each fragment 2567 * is freed by the sending machinery. 2568 */ 2569 for (mm = mb_copy; mm; mm = mb_copy) { 2570 mb_copy = mm->m_nextpkt; 2571 mm->m_nextpkt = 0; 2572 mm = m_pullup(mm, sizeof(struct ip)); 2573 if (mm != NULL) { 2574 ip = mtod(mm, struct ip *); 2575 if ((mrt_api_config & MRT_MFC_RP) && 2576 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2577 pim_register_send_rp(ip, vifp, mm, rt); 2578 } else { 2579 pim_register_send_upcall(ip, vifp, mm, rt); 2580 } 2581 } 2582 } 2583 2584 return 0; 2585} 2586 2587/* 2588 * Return a copy of the data packet that is ready for PIM Register 2589 * encapsulation. 2590 * XXX: Note that in the returned copy the IP header is a valid one. 2591 */ 2592static struct mbuf * 2593pim_register_prepare(struct ip *ip, struct mbuf *m) 2594{ 2595 struct mbuf *mb_copy = NULL; 2596 int mtu; 2597 2598 /* Take care of delayed checksums */ 2599 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2600 in_delayed_cksum(m); 2601 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2602 } 2603 2604 /* 2605 * Copy the old packet & pullup its IP header into the 2606 * new mbuf so we can modify it. 2607 */ 2608 mb_copy = m_copypacket(m, M_DONTWAIT); 2609 if (mb_copy == NULL) 2610 return NULL; 2611 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2612 if (mb_copy == NULL) 2613 return NULL; 2614 2615 /* take care of the TTL */ 2616 ip = mtod(mb_copy, struct ip *); 2617 --ip->ip_ttl; 2618 2619 /* Compute the MTU after the PIM Register encapsulation */ 2620 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2621 2622 if (ip->ip_len <= mtu) { 2623 /* Turn the IP header into a valid one */ 2624 ip->ip_len = htons(ip->ip_len); 2625 ip->ip_off = htons(ip->ip_off); 2626 ip->ip_sum = 0; 2627 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2628 } else { 2629 /* Fragment the packet */ 2630 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2631 m_freem(mb_copy); 2632 return NULL; 2633 } 2634 } 2635 return mb_copy; 2636} 2637 2638/* 2639 * Send an upcall with the data packet to the user-level process. 2640 */ 2641static int 2642pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2643 struct mbuf *mb_copy, struct mfc *rt) 2644{ 2645 struct mbuf *mb_first; 2646 int len = ntohs(ip->ip_len); 2647 struct igmpmsg *im; 2648 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2649 2650 VIF_LOCK_ASSERT(); 2651 2652 /* 2653 * Add a new mbuf with an upcall header 2654 */ 2655 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2656 if (mb_first == NULL) { 2657 m_freem(mb_copy); 2658 return ENOBUFS; 2659 } 2660 mb_first->m_data += max_linkhdr; 2661 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2662 mb_first->m_len = sizeof(struct igmpmsg); 2663 mb_first->m_next = mb_copy; 2664 2665 /* Send message to routing daemon */ 2666 im = mtod(mb_first, struct igmpmsg *); 2667 im->im_msgtype = IGMPMSG_WHOLEPKT; 2668 im->im_mbz = 0; 2669 im->im_vif = vifp - viftable; 2670 im->im_src = ip->ip_src; 2671 im->im_dst = ip->ip_dst; 2672 2673 k_igmpsrc.sin_addr = ip->ip_src; 2674 2675 mrtstat.mrts_upcalls++; 2676 2677 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2678 if (mrtdebug & DEBUG_PIM) 2679 log(LOG_WARNING, 2680 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2681 ++mrtstat.mrts_upq_sockfull; 2682 return ENOBUFS; 2683 } 2684 2685 /* Keep statistics */ 2686 pimstat.pims_snd_registers_msgs++; 2687 pimstat.pims_snd_registers_bytes += len; 2688 2689 return 0; 2690} 2691 2692/* 2693 * Encapsulate the data packet in PIM Register message and send it to the RP. 2694 */ 2695static int 2696pim_register_send_rp(struct ip *ip, struct vif *vifp, 2697 struct mbuf *mb_copy, struct mfc *rt) 2698{ 2699 struct mbuf *mb_first; 2700 struct ip *ip_outer; 2701 struct pim_encap_pimhdr *pimhdr; 2702 int len = ntohs(ip->ip_len); 2703 vifi_t vifi = rt->mfc_parent; 2704 2705 VIF_LOCK_ASSERT(); 2706 2707 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2708 m_freem(mb_copy); 2709 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2710 } 2711 2712 /* 2713 * Add a new mbuf with the encapsulating header 2714 */ 2715 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2716 if (mb_first == NULL) { 2717 m_freem(mb_copy); 2718 return ENOBUFS; 2719 } 2720 mb_first->m_data += max_linkhdr; 2721 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2722 mb_first->m_next = mb_copy; 2723 2724 mb_first->m_pkthdr.len = len + mb_first->m_len; 2725 2726 /* 2727 * Fill in the encapsulating IP and PIM header 2728 */ 2729 ip_outer = mtod(mb_first, struct ip *); 2730 *ip_outer = pim_encap_iphdr; 2731 ip_outer->ip_id = ip_newid(); 2732 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2733 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2734 ip_outer->ip_dst = rt->mfc_rp; 2735 /* 2736 * Copy the inner header TOS to the outer header, and take care of the 2737 * IP_DF bit. 2738 */ 2739 ip_outer->ip_tos = ip->ip_tos; 2740 if (ntohs(ip->ip_off) & IP_DF) 2741 ip_outer->ip_off |= IP_DF; 2742 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2743 + sizeof(pim_encap_iphdr)); 2744 *pimhdr = pim_encap_pimhdr; 2745 /* If the iif crosses a border, set the Border-bit */ 2746 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2747 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2748 2749 mb_first->m_data += sizeof(pim_encap_iphdr); 2750 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2751 mb_first->m_data -= sizeof(pim_encap_iphdr); 2752 2753 send_packet(vifp, mb_first); 2754 2755 /* Keep statistics */ 2756 pimstat.pims_snd_registers_msgs++; 2757 pimstat.pims_snd_registers_bytes += len; 2758 2759 return 0; 2760} 2761 2762/* 2763 * pim_encapcheck() is called by the encap[46]_input() path at runtime to 2764 * determine if a packet is for PIM; allowing PIM to be dynamically loaded 2765 * into the kernel. 2766 */ 2767static int 2768pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 2769{ 2770 2771#ifdef DIAGNOSTIC 2772 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); 2773#endif 2774 if (proto != IPPROTO_PIM) 2775 return 0; /* not for us; reject the datagram. */ 2776 2777 return 64; /* claim the datagram. */ 2778} 2779 2780/* 2781 * PIM-SMv2 and PIM-DM messages processing. 2782 * Receives and verifies the PIM control messages, and passes them 2783 * up to the listening socket, using rip_input(). 2784 * The only message with special processing is the PIM_REGISTER message 2785 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2786 * is passed to if_simloop(). 2787 */ 2788void 2789pim_input(struct mbuf *m, int off) 2790{ 2791 struct ip *ip = mtod(m, struct ip *); 2792 struct pim *pim; 2793 int minlen; 2794 int datalen = ip->ip_len; 2795 int ip_tos; 2796 int iphlen = off; 2797 2798 /* Keep statistics */ 2799 pimstat.pims_rcv_total_msgs++; 2800 pimstat.pims_rcv_total_bytes += datalen; 2801 2802 /* 2803 * Validate lengths 2804 */ 2805 if (datalen < PIM_MINLEN) { 2806 pimstat.pims_rcv_tooshort++; 2807 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 2808 datalen, (u_long)ip->ip_src.s_addr); 2809 m_freem(m); 2810 return; 2811 } 2812 2813 /* 2814 * If the packet is at least as big as a REGISTER, go agead 2815 * and grab the PIM REGISTER header size, to avoid another 2816 * possible m_pullup() later. 2817 * 2818 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 2819 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 2820 */ 2821 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 2822 /* 2823 * Get the IP and PIM headers in contiguous memory, and 2824 * possibly the PIM REGISTER header. 2825 */ 2826 if ((m->m_flags & M_EXT || m->m_len < minlen) && 2827 (m = m_pullup(m, minlen)) == 0) { 2828 log(LOG_ERR, "pim_input: m_pullup failure\n"); 2829 return; 2830 } 2831 /* m_pullup() may have given us a new mbuf so reset ip. */ 2832 ip = mtod(m, struct ip *); 2833 ip_tos = ip->ip_tos; 2834 2835 /* adjust mbuf to point to the PIM header */ 2836 m->m_data += iphlen; 2837 m->m_len -= iphlen; 2838 pim = mtod(m, struct pim *); 2839 2840 /* 2841 * Validate checksum. If PIM REGISTER, exclude the data packet. 2842 * 2843 * XXX: some older PIMv2 implementations don't make this distinction, 2844 * so for compatibility reason perform the checksum over part of the 2845 * message, and if error, then over the whole message. 2846 */ 2847 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 2848 /* do nothing, checksum okay */ 2849 } else if (in_cksum(m, datalen)) { 2850 pimstat.pims_rcv_badsum++; 2851 if (mrtdebug & DEBUG_PIM) 2852 log(LOG_DEBUG, "pim_input: invalid checksum"); 2853 m_freem(m); 2854 return; 2855 } 2856 2857 /* PIM version check */ 2858 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 2859 pimstat.pims_rcv_badversion++; 2860 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 2861 PIM_VT_V(pim->pim_vt), PIM_VERSION); 2862 m_freem(m); 2863 return; 2864 } 2865 2866 /* restore mbuf back to the outer IP */ 2867 m->m_data -= iphlen; 2868 m->m_len += iphlen; 2869 2870 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 2871 /* 2872 * Since this is a REGISTER, we'll make a copy of the register 2873 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 2874 * routing daemon. 2875 */ 2876 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 2877 struct mbuf *mcp; 2878 struct ip *encap_ip; 2879 u_int32_t *reghdr; 2880 struct ifnet *vifp; 2881 2882 VIF_LOCK(); 2883 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 2884 VIF_UNLOCK(); 2885 if (mrtdebug & DEBUG_PIM) 2886 log(LOG_DEBUG, 2887 "pim_input: register vif not set: %d\n", reg_vif_num); 2888 m_freem(m); 2889 return; 2890 } 2891 /* XXX need refcnt? */ 2892 vifp = viftable[reg_vif_num].v_ifp; 2893 VIF_UNLOCK(); 2894 2895 /* 2896 * Validate length 2897 */ 2898 if (datalen < PIM_REG_MINLEN) { 2899 pimstat.pims_rcv_tooshort++; 2900 pimstat.pims_rcv_badregisters++; 2901 log(LOG_ERR, 2902 "pim_input: register packet size too small %d from %lx\n", 2903 datalen, (u_long)ip->ip_src.s_addr); 2904 m_freem(m); 2905 return; 2906 } 2907 2908 reghdr = (u_int32_t *)(pim + 1); 2909 encap_ip = (struct ip *)(reghdr + 1); 2910 2911 if (mrtdebug & DEBUG_PIM) { 2912 log(LOG_DEBUG, 2913 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 2914 (u_long)ntohl(encap_ip->ip_src.s_addr), 2915 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2916 ntohs(encap_ip->ip_len)); 2917 } 2918 2919 /* verify the version number of the inner packet */ 2920 if (encap_ip->ip_v != IPVERSION) { 2921 pimstat.pims_rcv_badregisters++; 2922 if (mrtdebug & DEBUG_PIM) { 2923 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 2924 "of the inner packet\n", encap_ip->ip_v); 2925 } 2926 m_freem(m); 2927 return; 2928 } 2929 2930 /* verify the inner packet is destined to a mcast group */ 2931 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 2932 pimstat.pims_rcv_badregisters++; 2933 if (mrtdebug & DEBUG_PIM) 2934 log(LOG_DEBUG, 2935 "pim_input: inner packet of register is not " 2936 "multicast %lx\n", 2937 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 2938 m_freem(m); 2939 return; 2940 } 2941 2942 /* If a NULL_REGISTER, pass it to the daemon */ 2943 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 2944 goto pim_input_to_daemon; 2945 2946 /* 2947 * Copy the TOS from the outer IP header to the inner IP header. 2948 */ 2949 if (encap_ip->ip_tos != ip_tos) { 2950 /* Outer TOS -> inner TOS */ 2951 encap_ip->ip_tos = ip_tos; 2952 /* Recompute the inner header checksum. Sigh... */ 2953 2954 /* adjust mbuf to point to the inner IP header */ 2955 m->m_data += (iphlen + PIM_MINLEN); 2956 m->m_len -= (iphlen + PIM_MINLEN); 2957 2958 encap_ip->ip_sum = 0; 2959 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 2960 2961 /* restore mbuf to point back to the outer IP header */ 2962 m->m_data -= (iphlen + PIM_MINLEN); 2963 m->m_len += (iphlen + PIM_MINLEN); 2964 } 2965 2966 /* 2967 * Decapsulate the inner IP packet and loopback to forward it 2968 * as a normal multicast packet. Also, make a copy of the 2969 * outer_iphdr + pimhdr + reghdr + encap_iphdr 2970 * to pass to the daemon later, so it can take the appropriate 2971 * actions (e.g., send back PIM_REGISTER_STOP). 2972 * XXX: here m->m_data points to the outer IP header. 2973 */ 2974 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 2975 if (mcp == NULL) { 2976 log(LOG_ERR, 2977 "pim_input: pim register: could not copy register head\n"); 2978 m_freem(m); 2979 return; 2980 } 2981 2982 /* Keep statistics */ 2983 /* XXX: registers_bytes include only the encap. mcast pkt */ 2984 pimstat.pims_rcv_registers_msgs++; 2985 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 2986 2987 /* 2988 * forward the inner ip packet; point m_data at the inner ip. 2989 */ 2990 m_adj(m, iphlen + PIM_MINLEN); 2991 2992 if (mrtdebug & DEBUG_PIM) { 2993 log(LOG_DEBUG, 2994 "pim_input: forwarding decapsulated register: " 2995 "src %lx, dst %lx, vif %d\n", 2996 (u_long)ntohl(encap_ip->ip_src.s_addr), 2997 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2998 reg_vif_num); 2999 } 3000 /* NB: vifp was collected above; can it change on us? */ 3001 if_simloop(vifp, m, dst.sin_family, 0); 3002 3003 /* prepare the register head to send to the mrouting daemon */ 3004 m = mcp; 3005 } 3006 3007pim_input_to_daemon: 3008 /* 3009 * Pass the PIM message up to the daemon; if it is a Register message, 3010 * pass the 'head' only up to the daemon. This includes the 3011 * outer IP header, PIM header, PIM-Register header and the 3012 * inner IP header. 3013 * XXX: the outer IP header pkt size of a Register is not adjust to 3014 * reflect the fact that the inner multicast data is truncated. 3015 */ 3016 rip_input(m, iphlen); 3017 3018 return; 3019} 3020 3021/* 3022 * XXX: This is common code for dealing with initialization for both 3023 * the IPv4 and IPv6 multicast forwarding paths. It could do with cleanup. 3024 */ 3025static int 3026ip_mroute_modevent(module_t mod, int type, void *unused) 3027{ 3028 switch (type) { 3029 case MOD_LOAD: 3030 mtx_init(&mrouter_mtx, "mrouter initialization", NULL, MTX_DEF); 3031 MFC_LOCK_INIT(); 3032 VIF_LOCK_INIT(); 3033 ip_mrouter_reset(); 3034 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", 3035 &pim_squelch_wholepkt); 3036 3037 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM, 3038 pim_encapcheck, &in_pim_protosw, NULL); 3039 if (pim_encap_cookie == NULL) { 3040 printf("ip_mroute: unable to attach pim encap\n"); 3041 VIF_LOCK_DESTROY(); 3042 MFC_LOCK_DESTROY(); 3043 mtx_destroy(&mrouter_mtx); 3044 return (EINVAL); 3045 } 3046 3047#ifdef INET6 3048 pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM, 3049 pim_encapcheck, (struct protosw *)&in6_pim_protosw, NULL); 3050 if (pim6_encap_cookie == NULL) { 3051 printf("ip_mroute: unable to attach pim6 encap\n"); 3052 if (pim_encap_cookie) { 3053 encap_detach(pim_encap_cookie); 3054 pim_encap_cookie = NULL; 3055 } 3056 VIF_LOCK_DESTROY(); 3057 MFC_LOCK_DESTROY(); 3058 mtx_destroy(&mrouter_mtx); 3059 return (EINVAL); 3060 } 3061#endif 3062 3063 ip_mcast_src = X_ip_mcast_src; 3064 ip_mforward = X_ip_mforward; 3065 ip_mrouter_done = X_ip_mrouter_done; 3066 ip_mrouter_get = X_ip_mrouter_get; 3067 ip_mrouter_set = X_ip_mrouter_set; 3068 3069#ifdef INET6 3070 ip6_mforward = X_ip6_mforward; 3071 ip6_mrouter_done = X_ip6_mrouter_done; 3072 ip6_mrouter_get = X_ip6_mrouter_get; 3073 ip6_mrouter_set = X_ip6_mrouter_set; 3074 mrt6_ioctl = X_mrt6_ioctl; 3075#endif 3076 3077 ip_rsvp_force_done = X_ip_rsvp_force_done; 3078 ip_rsvp_vif = X_ip_rsvp_vif; 3079 3080 legal_vif_num = X_legal_vif_num; 3081 mrt_ioctl = X_mrt_ioctl; 3082 rsvp_input_p = X_rsvp_input; 3083 break; 3084 3085 case MOD_UNLOAD: 3086 /* 3087 * Typically module unload happens after the user-level 3088 * process has shutdown the kernel services (the check 3089 * below insures someone can't just yank the module out 3090 * from under a running process). But if the module is 3091 * just loaded and then unloaded w/o starting up a user 3092 * process we still need to cleanup. 3093 */ 3094 if (ip_mrouter 3095#ifdef INET6 3096 || ip6_mrouter 3097#endif 3098 ) 3099 return EINVAL; 3100 3101#ifdef INET6 3102 if (pim6_encap_cookie) { 3103 encap_detach(pim6_encap_cookie); 3104 pim6_encap_cookie = NULL; 3105 } 3106 X_ip6_mrouter_done(); 3107 ip6_mforward = NULL; 3108 ip6_mrouter_done = NULL; 3109 ip6_mrouter_get = NULL; 3110 ip6_mrouter_set = NULL; 3111 mrt6_ioctl = NULL; 3112#endif 3113 3114 if (pim_encap_cookie) { 3115 encap_detach(pim_encap_cookie); 3116 pim_encap_cookie = NULL; 3117 } 3118 X_ip_mrouter_done(); 3119 ip_mcast_src = NULL; 3120 ip_mforward = NULL; 3121 ip_mrouter_done = NULL; 3122 ip_mrouter_get = NULL; 3123 ip_mrouter_set = NULL; 3124 3125 ip_rsvp_force_done = NULL; 3126 ip_rsvp_vif = NULL; 3127 3128 legal_vif_num = NULL; 3129 mrt_ioctl = NULL; 3130 rsvp_input_p = NULL; 3131 3132 VIF_LOCK_DESTROY(); 3133 MFC_LOCK_DESTROY(); 3134 mtx_destroy(&mrouter_mtx); 3135 break; 3136 3137 default: 3138 return EOPNOTSUPP; 3139 } 3140 return 0; 3141} 3142 3143static moduledata_t ip_mroutemod = { 3144 "ip_mroute", 3145 ip_mroute_modevent, 3146 0 3147}; 3148DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3149