ip_mroute.c revision 167116
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 * $FreeBSD: head/sys/netinet/ip_mroute.c 167116 2007-02-28 20:02:24Z bms $ 56 */ 57 58#include "opt_inet.h" 59#include "opt_inet6.h" 60#include "opt_mac.h" 61#include "opt_mrouting.h" 62 63#define _PIM_VT 1 64 65#include <sys/param.h> 66#include <sys/kernel.h> 67#include <sys/lock.h> 68#include <sys/malloc.h> 69#include <sys/mbuf.h> 70#include <sys/module.h> 71#include <sys/priv.h> 72#include <sys/protosw.h> 73#include <sys/signalvar.h> 74#include <sys/socket.h> 75#include <sys/socketvar.h> 76#include <sys/sockio.h> 77#include <sys/sx.h> 78#include <sys/sysctl.h> 79#include <sys/syslog.h> 80#include <sys/systm.h> 81#include <sys/time.h> 82#include <net/if.h> 83#include <net/netisr.h> 84#include <net/route.h> 85#include <netinet/in.h> 86#include <netinet/igmp.h> 87#include <netinet/in_systm.h> 88#include <netinet/in_var.h> 89#include <netinet/ip.h> 90#include <netinet/ip_encap.h> 91#include <netinet/ip_mroute.h> 92#include <netinet/ip_var.h> 93#include <netinet/ip_options.h> 94#include <netinet/pim.h> 95#include <netinet/pim_var.h> 96#include <netinet/udp.h> 97#ifdef INET6 98#include <netinet/ip6.h> 99#include <netinet6/in6_var.h> 100#include <netinet6/ip6_mroute.h> 101#include <netinet6/ip6_var.h> 102#endif 103#include <machine/in_cksum.h> 104 105#include <security/mac/mac_framework.h> 106 107/* 108 * Control debugging code for rsvp and multicast routing code. 109 * Can only set them with the debugger. 110 */ 111static u_int rsvpdebug; /* non-zero enables debugging */ 112 113static u_int mrtdebug; /* any set of the flags below */ 114#define DEBUG_MFC 0x02 115#define DEBUG_FORWARD 0x04 116#define DEBUG_EXPIRE 0x08 117#define DEBUG_XMIT 0x10 118#define DEBUG_PIM 0x20 119 120#define VIFI_INVALID ((vifi_t) -1) 121 122#define M_HASCL(m) ((m)->m_flags & M_EXT) 123 124static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 125 126/* 127 * Locking. We use two locks: one for the virtual interface table and 128 * 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 mrouter_mtx; 150#define MROUTER_LOCK() mtx_lock(&mrouter_mtx) 151#define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx) 152#define MROUTER_LOCK_ASSERT() do { \ 153 mtx_assert(&mrouter_mtx, MA_OWNED); \ 154 NET_ASSERT_GIANT(); \ 155} while (0) 156#define MROUTER_LOCK_INIT() \ 157 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF) 158#define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx) 159 160static struct mtx mfc_mtx; 161#define MFC_LOCK() mtx_lock(&mfc_mtx) 162#define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 163#define MFC_LOCK_ASSERT() do { \ 164 mtx_assert(&mfc_mtx, MA_OWNED); \ 165 NET_ASSERT_GIANT(); \ 166} while (0) 167#define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF) 168#define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 169 170static struct vif viftable[MAXVIFS]; 171SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 172 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 173 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 174 175static struct mtx vif_mtx; 176#define VIF_LOCK() mtx_lock(&vif_mtx) 177#define VIF_UNLOCK() mtx_unlock(&vif_mtx) 178#define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 179#define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF) 180#define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 181 182static u_char nexpire[MFCTBLSIZ]; 183 184static eventhandler_tag if_detach_event_tag = NULL; 185 186static struct callout expire_upcalls_ch; 187 188#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 189#define UPCALL_EXPIRE 6 /* number of timeouts */ 190 191#define ENCAP_TTL 64 192 193/* 194 * Bandwidth meter variables and constants 195 */ 196static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 197/* 198 * Pending timeouts are stored in a hash table, the key being the 199 * expiration time. Periodically, the entries are analysed and processed. 200 */ 201#define BW_METER_BUCKETS 1024 202static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 203static struct callout bw_meter_ch; 204#define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 205 206/* 207 * Pending upcalls are stored in a vector which is flushed when 208 * full, or periodically 209 */ 210static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 211static u_int bw_upcalls_n; /* # of pending upcalls */ 212static struct callout bw_upcalls_ch; 213#define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 214 215static struct pimstat pimstat; 216 217SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); 218SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 219 &pimstat, pimstat, 220 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 221 222static u_long pim_squelch_wholepkt = 0; 223SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, 224 &pim_squelch_wholepkt, 0, 225 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); 226 227extern struct domain inetdomain; 228struct protosw in_pim_protosw = { 229 .pr_type = SOCK_RAW, 230 .pr_domain = &inetdomain, 231 .pr_protocol = IPPROTO_PIM, 232 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, 233 .pr_input = pim_input, 234 .pr_output = (pr_output_t*)rip_output, 235 .pr_ctloutput = rip_ctloutput, 236 .pr_usrreqs = &rip_usrreqs 237}; 238static const struct encaptab *pim_encap_cookie; 239 240#ifdef INET6 241/* ip6_mroute.c glue */ 242extern struct in6_protosw in6_pim_protosw; 243static const struct encaptab *pim6_encap_cookie; 244 245extern int X_ip6_mrouter_set(struct socket *, struct sockopt *); 246extern int X_ip6_mrouter_get(struct socket *, struct sockopt *); 247extern int X_ip6_mrouter_done(void); 248extern int X_ip6_mforward(struct ip6_hdr *, struct ifnet *, struct mbuf *); 249extern int X_mrt6_ioctl(int, caddr_t); 250#endif 251 252static int pim_encapcheck(const struct mbuf *, int, int, void *); 253 254/* 255 * Note: the PIM Register encapsulation adds the following in front of a 256 * data packet: 257 * 258 * struct pim_encap_hdr { 259 * struct ip ip; 260 * struct pim_encap_pimhdr pim; 261 * } 262 * 263 */ 264 265struct pim_encap_pimhdr { 266 struct pim pim; 267 uint32_t flags; 268}; 269 270static struct ip pim_encap_iphdr = { 271#if BYTE_ORDER == LITTLE_ENDIAN 272 sizeof(struct ip) >> 2, 273 IPVERSION, 274#else 275 IPVERSION, 276 sizeof(struct ip) >> 2, 277#endif 278 0, /* tos */ 279 sizeof(struct ip), /* total length */ 280 0, /* id */ 281 0, /* frag offset */ 282 ENCAP_TTL, 283 IPPROTO_PIM, 284 0, /* checksum */ 285}; 286 287static struct pim_encap_pimhdr pim_encap_pimhdr = { 288 { 289 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 290 0, /* reserved */ 291 0, /* checksum */ 292 }, 293 0 /* flags */ 294}; 295 296static struct ifnet multicast_register_if; 297static vifi_t reg_vif_num = VIFI_INVALID; 298 299/* 300 * Private variables. 301 */ 302static vifi_t numvifs; 303 304static u_long X_ip_mcast_src(int vifi); 305static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 306 struct mbuf *m, struct ip_moptions *imo); 307static int X_ip_mrouter_done(void); 308static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 309static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 310static int X_legal_vif_num(int vif); 311static int X_mrt_ioctl(int cmd, caddr_t data); 312 313static int get_sg_cnt(struct sioc_sg_req *); 314static int get_vif_cnt(struct sioc_vif_req *); 315static void if_detached_event(void *arg __unused, struct ifnet *); 316static int ip_mrouter_init(struct socket *, int); 317static int add_vif(struct vifctl *); 318static int del_vif_locked(vifi_t); 319static int del_vif(vifi_t); 320static int add_mfc(struct mfcctl2 *); 321static int del_mfc(struct mfcctl2 *); 322static int set_api_config(uint32_t *); /* chose API capabilities */ 323static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 324static int set_assert(int); 325static void expire_upcalls(void *); 326static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 327static void phyint_send(struct ip *, struct vif *, struct mbuf *); 328static void send_packet(struct vif *, struct mbuf *); 329 330/* 331 * Bandwidth monitoring 332 */ 333static void free_bw_list(struct bw_meter *list); 334static int add_bw_upcall(struct bw_upcall *); 335static int del_bw_upcall(struct bw_upcall *); 336static void bw_meter_receive_packet(struct bw_meter *x, int plen, 337 struct timeval *nowp); 338static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 339static void bw_upcalls_send(void); 340static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 341static void unschedule_bw_meter(struct bw_meter *x); 342static void bw_meter_process(void); 343static void expire_bw_upcalls_send(void *); 344static void expire_bw_meter_process(void *); 345 346static int pim_register_send(struct ip *, struct vif *, 347 struct mbuf *, struct mfc *); 348static int pim_register_send_rp(struct ip *, struct vif *, 349 struct mbuf *, struct mfc *); 350static int pim_register_send_upcall(struct ip *, struct vif *, 351 struct mbuf *, struct mfc *); 352static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 353 354/* 355 * whether or not special PIM assert processing is enabled. 356 */ 357static int pim_assert; 358/* 359 * Rate limit for assert notification messages, in usec 360 */ 361#define ASSERT_MSG_TIME 3000000 362 363/* 364 * Kernel multicast routing API capabilities and setup. 365 * If more API capabilities are added to the kernel, they should be 366 * recorded in `mrt_api_support'. 367 */ 368static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 369 MRT_MFC_FLAGS_BORDER_VIF | 370 MRT_MFC_RP | 371 MRT_MFC_BW_UPCALL); 372static uint32_t mrt_api_config = 0; 373 374/* 375 * Hash function for a source, group entry 376 */ 377#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 378 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 379 380/* 381 * Find a route for a given origin IP address and Multicast group address 382 * Type of service parameter to be added in the future!!! 383 * Statistics are updated by the caller if needed 384 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 385 */ 386static struct mfc * 387mfc_find(in_addr_t o, in_addr_t g) 388{ 389 struct mfc *rt; 390 391 MFC_LOCK_ASSERT(); 392 393 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 394 if ((rt->mfc_origin.s_addr == o) && 395 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 396 break; 397 return rt; 398} 399 400/* 401 * Macros to compute elapsed time efficiently 402 * Borrowed from Van Jacobson's scheduling code 403 */ 404#define TV_DELTA(a, b, delta) { \ 405 int xxs; \ 406 delta = (a).tv_usec - (b).tv_usec; \ 407 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 408 switch (xxs) { \ 409 case 2: \ 410 delta += 1000000; \ 411 /* FALLTHROUGH */ \ 412 case 1: \ 413 delta += 1000000; \ 414 break; \ 415 default: \ 416 delta += (1000000 * xxs); \ 417 } \ 418 } \ 419} 420 421#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 422 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 423 424/* 425 * Handle MRT setsockopt commands to modify the multicast routing tables. 426 */ 427static int 428X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 429{ 430 int error, optval; 431 vifi_t vifi; 432 struct vifctl vifc; 433 struct mfcctl2 mfc; 434 struct bw_upcall bw_upcall; 435 uint32_t i; 436 437 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 438 return EPERM; 439 440 error = 0; 441 switch (sopt->sopt_name) { 442 case MRT_INIT: 443 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 444 if (error) 445 break; 446 error = ip_mrouter_init(so, optval); 447 break; 448 449 case MRT_DONE: 450 error = ip_mrouter_done(); 451 break; 452 453 case MRT_ADD_VIF: 454 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 455 if (error) 456 break; 457 error = add_vif(&vifc); 458 break; 459 460 case MRT_DEL_VIF: 461 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 462 if (error) 463 break; 464 error = del_vif(vifi); 465 break; 466 467 case MRT_ADD_MFC: 468 case MRT_DEL_MFC: 469 /* 470 * select data size depending on API version. 471 */ 472 if (sopt->sopt_name == MRT_ADD_MFC && 473 mrt_api_config & MRT_API_FLAGS_ALL) { 474 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 475 sizeof(struct mfcctl2)); 476 } else { 477 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 478 sizeof(struct mfcctl)); 479 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 480 sizeof(mfc) - sizeof(struct mfcctl)); 481 } 482 if (error) 483 break; 484 if (sopt->sopt_name == MRT_ADD_MFC) 485 error = add_mfc(&mfc); 486 else 487 error = del_mfc(&mfc); 488 break; 489 490 case MRT_ASSERT: 491 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 492 if (error) 493 break; 494 set_assert(optval); 495 break; 496 497 case MRT_API_CONFIG: 498 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 499 if (!error) 500 error = set_api_config(&i); 501 if (!error) 502 error = sooptcopyout(sopt, &i, sizeof i); 503 break; 504 505 case MRT_ADD_BW_UPCALL: 506 case MRT_DEL_BW_UPCALL: 507 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 508 sizeof bw_upcall); 509 if (error) 510 break; 511 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 512 error = add_bw_upcall(&bw_upcall); 513 else 514 error = del_bw_upcall(&bw_upcall); 515 break; 516 517 default: 518 error = EOPNOTSUPP; 519 break; 520 } 521 return error; 522} 523 524/* 525 * Handle MRT getsockopt commands 526 */ 527static int 528X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 529{ 530 int error; 531 static int version = 0x0305; /* !!! why is this here? XXX */ 532 533 switch (sopt->sopt_name) { 534 case MRT_VERSION: 535 error = sooptcopyout(sopt, &version, sizeof version); 536 break; 537 538 case MRT_ASSERT: 539 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 540 break; 541 542 case MRT_API_SUPPORT: 543 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 544 break; 545 546 case MRT_API_CONFIG: 547 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 548 break; 549 550 default: 551 error = EOPNOTSUPP; 552 break; 553 } 554 return error; 555} 556 557/* 558 * Handle ioctl commands to obtain information from the cache 559 */ 560static int 561X_mrt_ioctl(int cmd, caddr_t data) 562{ 563 int error = 0; 564 565 /* 566 * Currently the only function calling this ioctl routine is rtioctl(). 567 * Typically, only root can create the raw socket in order to execute 568 * this ioctl method, however the request might be coming from a prison 569 */ 570 error = priv_check(curthread, PRIV_NETINET_MROUTE); 571 if (error) 572 return (error); 573 switch (cmd) { 574 case (SIOCGETVIFCNT): 575 error = get_vif_cnt((struct sioc_vif_req *)data); 576 break; 577 578 case (SIOCGETSGCNT): 579 error = get_sg_cnt((struct sioc_sg_req *)data); 580 break; 581 582 default: 583 error = EINVAL; 584 break; 585 } 586 return error; 587} 588 589/* 590 * returns the packet, byte, rpf-failure count for the source group provided 591 */ 592static int 593get_sg_cnt(struct sioc_sg_req *req) 594{ 595 struct mfc *rt; 596 597 MFC_LOCK(); 598 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 599 if (rt == NULL) { 600 MFC_UNLOCK(); 601 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 602 return EADDRNOTAVAIL; 603 } 604 req->pktcnt = rt->mfc_pkt_cnt; 605 req->bytecnt = rt->mfc_byte_cnt; 606 req->wrong_if = rt->mfc_wrong_if; 607 MFC_UNLOCK(); 608 return 0; 609} 610 611/* 612 * returns the input and output packet and byte counts on the vif provided 613 */ 614static int 615get_vif_cnt(struct sioc_vif_req *req) 616{ 617 vifi_t vifi = req->vifi; 618 619 VIF_LOCK(); 620 if (vifi >= numvifs) { 621 VIF_UNLOCK(); 622 return EINVAL; 623 } 624 625 req->icount = viftable[vifi].v_pkt_in; 626 req->ocount = viftable[vifi].v_pkt_out; 627 req->ibytes = viftable[vifi].v_bytes_in; 628 req->obytes = viftable[vifi].v_bytes_out; 629 VIF_UNLOCK(); 630 631 return 0; 632} 633 634static void 635ip_mrouter_reset(void) 636{ 637 bzero((caddr_t)mfctable, sizeof(mfctable)); 638 bzero((caddr_t)nexpire, sizeof(nexpire)); 639 640 pim_assert = 0; 641 mrt_api_config = 0; 642 643 callout_init(&expire_upcalls_ch, NET_CALLOUT_MPSAFE); 644 645 bw_upcalls_n = 0; 646 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 647 callout_init(&bw_upcalls_ch, NET_CALLOUT_MPSAFE); 648 callout_init(&bw_meter_ch, NET_CALLOUT_MPSAFE); 649} 650 651static void 652if_detached_event(void *arg __unused, struct ifnet *ifp) 653{ 654 vifi_t vifi; 655 int i; 656 struct mfc *mfc; 657 struct mfc *nmfc; 658 struct mfc **ppmfc; /* Pointer to previous node's next-pointer */ 659 struct rtdetq *pq; 660 struct rtdetq *npq; 661 662 MROUTER_LOCK(); 663 if (ip_mrouter == NULL) { 664 MROUTER_UNLOCK(); 665 } 666 667 /* 668 * Tear down multicast forwarder state associated with this ifnet. 669 * 1. Walk the vif list, matching vifs against this ifnet. 670 * 2. Walk the multicast forwarding cache (mfc) looking for 671 * inner matches with this vif's index. 672 * 3. Free any pending mbufs for this mfc. 673 * 4. Free the associated mfc entry and state associated with this vif. 674 * Be very careful about unlinking from a singly-linked list whose 675 * "head node" is a pointer in a simple array. 676 * 5. Free vif state. This should disable ALLMULTI on the interface. 677 */ 678 VIF_LOCK(); 679 MFC_LOCK(); 680 for (vifi = 0; vifi < numvifs; vifi++) { 681 if (viftable[vifi].v_ifp != ifp) 682 continue; 683 for (i = 0; i < MFCTBLSIZ; i++) { 684 ppmfc = &mfctable[i]; 685 for (mfc = mfctable[i]; mfc != NULL; ) { 686 nmfc = mfc->mfc_next; 687 if (mfc->mfc_parent == vifi) { 688 for (pq = mfc->mfc_stall; pq != NULL; ) { 689 npq = pq->next; 690 m_freem(pq->m); 691 free(pq, M_MRTABLE); 692 pq = npq; 693 } 694 free_bw_list(mfc->mfc_bw_meter); 695 free(mfc, M_MRTABLE); 696 *ppmfc = nmfc; 697 } else { 698 ppmfc = &mfc->mfc_next; 699 } 700 mfc = nmfc; 701 } 702 } 703 del_vif_locked(vifi); 704 } 705 MFC_UNLOCK(); 706 VIF_UNLOCK(); 707 708 MROUTER_UNLOCK(); 709} 710 711/* 712 * Enable multicast routing 713 */ 714static int 715ip_mrouter_init(struct socket *so, int version) 716{ 717 if (mrtdebug) 718 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 719 so->so_type, so->so_proto->pr_protocol); 720 721 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 722 return EOPNOTSUPP; 723 724 if (version != 1) 725 return ENOPROTOOPT; 726 727 MROUTER_LOCK(); 728 729 if (ip_mrouter != NULL) { 730 MROUTER_UNLOCK(); 731 return EADDRINUSE; 732 } 733 734 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 735 if_detached_event, NULL, EVENTHANDLER_PRI_ANY); 736 if (if_detach_event_tag == NULL) { 737 MROUTER_UNLOCK(); 738 return (ENOMEM); 739 } 740 741 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 742 743 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 744 expire_bw_upcalls_send, NULL); 745 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 746 747 ip_mrouter = so; 748 749 MROUTER_UNLOCK(); 750 751 if (mrtdebug) 752 log(LOG_DEBUG, "ip_mrouter_init\n"); 753 754 return 0; 755} 756 757/* 758 * Disable multicast routing 759 */ 760static int 761X_ip_mrouter_done(void) 762{ 763 vifi_t vifi; 764 int i; 765 struct ifnet *ifp; 766 struct ifreq ifr; 767 struct mfc *rt; 768 struct rtdetq *rte; 769 770 MROUTER_LOCK(); 771 772 if (ip_mrouter == NULL) { 773 MROUTER_UNLOCK(); 774 return EINVAL; 775 } 776 777 /* 778 * Detach/disable hooks to the reset of the system. 779 */ 780 ip_mrouter = NULL; 781 mrt_api_config = 0; 782 783 VIF_LOCK(); 784 /* 785 * For each phyint in use, disable promiscuous reception of all IP 786 * multicasts. 787 */ 788 for (vifi = 0; vifi < numvifs; vifi++) { 789 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 790 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 791 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 792 793 so->sin_len = sizeof(struct sockaddr_in); 794 so->sin_family = AF_INET; 795 so->sin_addr.s_addr = INADDR_ANY; 796 ifp = viftable[vifi].v_ifp; 797 if_allmulti(ifp, 0); 798 } 799 } 800 bzero((caddr_t)viftable, sizeof(viftable)); 801 numvifs = 0; 802 pim_assert = 0; 803 VIF_UNLOCK(); 804 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 805 806 /* 807 * Free all multicast forwarding cache entries. 808 */ 809 callout_stop(&expire_upcalls_ch); 810 callout_stop(&bw_upcalls_ch); 811 callout_stop(&bw_meter_ch); 812 813 MFC_LOCK(); 814 for (i = 0; i < MFCTBLSIZ; i++) { 815 for (rt = mfctable[i]; rt != NULL; ) { 816 struct mfc *nr = rt->mfc_next; 817 818 for (rte = rt->mfc_stall; rte != NULL; ) { 819 struct rtdetq *n = rte->next; 820 821 m_freem(rte->m); 822 free(rte, M_MRTABLE); 823 rte = n; 824 } 825 free_bw_list(rt->mfc_bw_meter); 826 free(rt, M_MRTABLE); 827 rt = nr; 828 } 829 } 830 bzero((caddr_t)mfctable, sizeof(mfctable)); 831 bzero((caddr_t)nexpire, sizeof(nexpire)); 832 bw_upcalls_n = 0; 833 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 834 MFC_UNLOCK(); 835 836 reg_vif_num = VIFI_INVALID; 837 838 MROUTER_UNLOCK(); 839 840 if (mrtdebug) 841 log(LOG_DEBUG, "ip_mrouter_done\n"); 842 843 return 0; 844} 845 846/* 847 * Set PIM assert processing global 848 */ 849static int 850set_assert(int i) 851{ 852 if ((i != 1) && (i != 0)) 853 return EINVAL; 854 855 pim_assert = i; 856 857 return 0; 858} 859 860/* 861 * Configure API capabilities 862 */ 863int 864set_api_config(uint32_t *apival) 865{ 866 int i; 867 868 /* 869 * We can set the API capabilities only if it is the first operation 870 * after MRT_INIT. I.e.: 871 * - there are no vifs installed 872 * - pim_assert is not enabled 873 * - the MFC table is empty 874 */ 875 if (numvifs > 0) { 876 *apival = 0; 877 return EPERM; 878 } 879 if (pim_assert) { 880 *apival = 0; 881 return EPERM; 882 } 883 for (i = 0; i < MFCTBLSIZ; i++) { 884 if (mfctable[i] != NULL) { 885 *apival = 0; 886 return EPERM; 887 } 888 } 889 890 mrt_api_config = *apival & mrt_api_support; 891 *apival = mrt_api_config; 892 893 return 0; 894} 895 896/* 897 * Add a vif to the vif table 898 */ 899static int 900add_vif(struct vifctl *vifcp) 901{ 902 struct vif *vifp = viftable + vifcp->vifc_vifi; 903 struct sockaddr_in sin = {sizeof sin, AF_INET}; 904 struct ifaddr *ifa; 905 struct ifnet *ifp; 906 int error; 907 908 VIF_LOCK(); 909 if (vifcp->vifc_vifi >= MAXVIFS) { 910 VIF_UNLOCK(); 911 return EINVAL; 912 } 913 /* rate limiting is no longer supported by this code */ 914 if (vifcp->vifc_rate_limit != 0) { 915 log(LOG_ERR, "rate limiting is no longer supported\n"); 916 VIF_UNLOCK(); 917 return EINVAL; 918 } 919 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) { 920 VIF_UNLOCK(); 921 return EADDRINUSE; 922 } 923 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) { 924 VIF_UNLOCK(); 925 return EADDRNOTAVAIL; 926 } 927 928 /* Find the interface with an address in AF_INET family */ 929 if (vifcp->vifc_flags & VIFF_REGISTER) { 930 /* 931 * XXX: Because VIFF_REGISTER does not really need a valid 932 * local interface (e.g. it could be 127.0.0.2), we don't 933 * check its address. 934 */ 935 ifp = NULL; 936 } else { 937 sin.sin_addr = vifcp->vifc_lcl_addr; 938 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 939 if (ifa == NULL) { 940 VIF_UNLOCK(); 941 return EADDRNOTAVAIL; 942 } 943 ifp = ifa->ifa_ifp; 944 } 945 946 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { 947 log(LOG_ERR, "tunnels are no longer supported\n"); 948 VIF_UNLOCK(); 949 return EOPNOTSUPP; 950 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 951 ifp = &multicast_register_if; 952 if (mrtdebug) 953 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 954 (void *)&multicast_register_if); 955 if (reg_vif_num == VIFI_INVALID) { 956 if_initname(&multicast_register_if, "register_vif", 0); 957 multicast_register_if.if_flags = IFF_LOOPBACK; 958 reg_vif_num = vifcp->vifc_vifi; 959 } 960 } else { /* Make sure the interface supports multicast */ 961 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 962 VIF_UNLOCK(); 963 return EOPNOTSUPP; 964 } 965 966 /* Enable promiscuous reception of all IP multicasts from the if */ 967 error = if_allmulti(ifp, 1); 968 if (error) { 969 VIF_UNLOCK(); 970 return error; 971 } 972 } 973 974 vifp->v_flags = vifcp->vifc_flags; 975 vifp->v_threshold = vifcp->vifc_threshold; 976 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 977 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 978 vifp->v_ifp = ifp; 979 vifp->v_rsvp_on = 0; 980 vifp->v_rsvpd = NULL; 981 /* initialize per vif pkt counters */ 982 vifp->v_pkt_in = 0; 983 vifp->v_pkt_out = 0; 984 vifp->v_bytes_in = 0; 985 vifp->v_bytes_out = 0; 986 bzero(&vifp->v_route, sizeof(vifp->v_route)); 987 988 /* Adjust numvifs up if the vifi is higher than numvifs */ 989 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 990 991 VIF_UNLOCK(); 992 993 if (mrtdebug) 994 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x\n", 995 vifcp->vifc_vifi, 996 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 997 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 998 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 999 vifcp->vifc_threshold); 1000 1001 return 0; 1002} 1003 1004/* 1005 * Delete a vif from the vif table 1006 */ 1007static int 1008del_vif_locked(vifi_t vifi) 1009{ 1010 struct vif *vifp; 1011 1012 VIF_LOCK_ASSERT(); 1013 1014 if (vifi >= numvifs) { 1015 return EINVAL; 1016 } 1017 vifp = &viftable[vifi]; 1018 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) { 1019 return EADDRNOTAVAIL; 1020 } 1021 1022 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 1023 if_allmulti(vifp->v_ifp, 0); 1024 1025 if (vifp->v_flags & VIFF_REGISTER) 1026 reg_vif_num = VIFI_INVALID; 1027 1028 bzero((caddr_t)vifp, sizeof (*vifp)); 1029 1030 if (mrtdebug) 1031 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1032 1033 /* Adjust numvifs down */ 1034 for (vifi = numvifs; vifi > 0; vifi--) 1035 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1036 break; 1037 numvifs = vifi; 1038 1039 return 0; 1040} 1041 1042static int 1043del_vif(vifi_t vifi) 1044{ 1045 int cc; 1046 1047 VIF_LOCK(); 1048 cc = del_vif_locked(vifi); 1049 VIF_UNLOCK(); 1050 1051 return cc; 1052} 1053 1054/* 1055 * update an mfc entry without resetting counters and S,G addresses. 1056 */ 1057static void 1058update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1059{ 1060 int i; 1061 1062 rt->mfc_parent = mfccp->mfcc_parent; 1063 for (i = 0; i < numvifs; i++) { 1064 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1065 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1066 MRT_MFC_FLAGS_ALL; 1067 } 1068 /* set the RP address */ 1069 if (mrt_api_config & MRT_MFC_RP) 1070 rt->mfc_rp = mfccp->mfcc_rp; 1071 else 1072 rt->mfc_rp.s_addr = INADDR_ANY; 1073} 1074 1075/* 1076 * fully initialize an mfc entry from the parameter. 1077 */ 1078static void 1079init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1080{ 1081 rt->mfc_origin = mfccp->mfcc_origin; 1082 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1083 1084 update_mfc_params(rt, mfccp); 1085 1086 /* initialize pkt counters per src-grp */ 1087 rt->mfc_pkt_cnt = 0; 1088 rt->mfc_byte_cnt = 0; 1089 rt->mfc_wrong_if = 0; 1090 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1091} 1092 1093 1094/* 1095 * Add an mfc entry 1096 */ 1097static int 1098add_mfc(struct mfcctl2 *mfccp) 1099{ 1100 struct mfc *rt; 1101 u_long hash; 1102 struct rtdetq *rte; 1103 u_short nstl; 1104 1105 VIF_LOCK(); 1106 MFC_LOCK(); 1107 1108 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1109 1110 /* If an entry already exists, just update the fields */ 1111 if (rt) { 1112 if (mrtdebug & DEBUG_MFC) 1113 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1114 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1115 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1116 mfccp->mfcc_parent); 1117 1118 update_mfc_params(rt, mfccp); 1119 MFC_UNLOCK(); 1120 VIF_UNLOCK(); 1121 return 0; 1122 } 1123 1124 /* 1125 * Find the entry for which the upcall was made and update 1126 */ 1127 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1128 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1129 1130 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1131 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1132 (rt->mfc_stall != NULL)) { 1133 1134 if (nstl++) 1135 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1136 "multiple kernel entries", 1137 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1138 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1139 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1140 1141 if (mrtdebug & DEBUG_MFC) 1142 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1143 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1144 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1145 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1146 1147 init_mfc_params(rt, mfccp); 1148 1149 rt->mfc_expire = 0; /* Don't clean this guy up */ 1150 nexpire[hash]--; 1151 1152 /* free packets Qed at the end of this entry */ 1153 for (rte = rt->mfc_stall; rte != NULL; ) { 1154 struct rtdetq *n = rte->next; 1155 1156 ip_mdq(rte->m, rte->ifp, rt, -1); 1157 m_freem(rte->m); 1158 free(rte, M_MRTABLE); 1159 rte = n; 1160 } 1161 rt->mfc_stall = NULL; 1162 } 1163 } 1164 1165 /* 1166 * It is possible that an entry is being inserted without an upcall 1167 */ 1168 if (nstl == 0) { 1169 if (mrtdebug & DEBUG_MFC) 1170 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1171 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1172 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1173 mfccp->mfcc_parent); 1174 1175 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1176 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1177 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1178 init_mfc_params(rt, mfccp); 1179 if (rt->mfc_expire) 1180 nexpire[hash]--; 1181 rt->mfc_expire = 0; 1182 break; /* XXX */ 1183 } 1184 } 1185 if (rt == NULL) { /* no upcall, so make a new entry */ 1186 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1187 if (rt == NULL) { 1188 MFC_UNLOCK(); 1189 VIF_UNLOCK(); 1190 return ENOBUFS; 1191 } 1192 1193 init_mfc_params(rt, mfccp); 1194 rt->mfc_expire = 0; 1195 rt->mfc_stall = NULL; 1196 1197 rt->mfc_bw_meter = NULL; 1198 /* insert new entry at head of hash chain */ 1199 rt->mfc_next = mfctable[hash]; 1200 mfctable[hash] = rt; 1201 } 1202 } 1203 MFC_UNLOCK(); 1204 VIF_UNLOCK(); 1205 return 0; 1206} 1207 1208/* 1209 * Delete an mfc entry 1210 */ 1211static int 1212del_mfc(struct mfcctl2 *mfccp) 1213{ 1214 struct in_addr origin; 1215 struct in_addr mcastgrp; 1216 struct mfc *rt; 1217 struct mfc **nptr; 1218 u_long hash; 1219 struct bw_meter *list; 1220 1221 origin = mfccp->mfcc_origin; 1222 mcastgrp = mfccp->mfcc_mcastgrp; 1223 1224 if (mrtdebug & DEBUG_MFC) 1225 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1226 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1227 1228 MFC_LOCK(); 1229 1230 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1231 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1232 if (origin.s_addr == rt->mfc_origin.s_addr && 1233 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1234 rt->mfc_stall == NULL) 1235 break; 1236 if (rt == NULL) { 1237 MFC_UNLOCK(); 1238 return EADDRNOTAVAIL; 1239 } 1240 1241 *nptr = rt->mfc_next; 1242 1243 /* 1244 * free the bw_meter entries 1245 */ 1246 list = rt->mfc_bw_meter; 1247 rt->mfc_bw_meter = NULL; 1248 1249 free(rt, M_MRTABLE); 1250 1251 free_bw_list(list); 1252 1253 MFC_UNLOCK(); 1254 1255 return 0; 1256} 1257 1258/* 1259 * Send a message to the routing daemon on the multicast routing socket 1260 */ 1261static int 1262socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1263{ 1264 if (s) { 1265 SOCKBUF_LOCK(&s->so_rcv); 1266 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, 1267 NULL) != 0) { 1268 sorwakeup_locked(s); 1269 return 0; 1270 } 1271 SOCKBUF_UNLOCK(&s->so_rcv); 1272 } 1273 m_freem(mm); 1274 return -1; 1275} 1276 1277/* 1278 * IP multicast forwarding function. This function assumes that the packet 1279 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1280 * pointed to by "ifp", and the packet is to be relayed to other networks 1281 * that have members of the packet's destination IP multicast group. 1282 * 1283 * The packet is returned unscathed to the caller, unless it is 1284 * erroneous, in which case a non-zero return value tells the caller to 1285 * discard it. 1286 */ 1287 1288#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1289 1290static int 1291X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1292 struct ip_moptions *imo) 1293{ 1294 struct mfc *rt; 1295 int error; 1296 vifi_t vifi; 1297 1298 if (mrtdebug & DEBUG_FORWARD) 1299 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1300 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1301 (void *)ifp); 1302 1303 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1304 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1305 /* 1306 * Packet arrived via a physical interface or 1307 * an encapsulated tunnel or a register_vif. 1308 */ 1309 } else { 1310 /* 1311 * Packet arrived through a source-route tunnel. 1312 * Source-route tunnels are no longer supported. 1313 */ 1314 static int last_log; 1315 if (last_log != time_uptime) { 1316 last_log = time_uptime; 1317 log(LOG_ERR, 1318 "ip_mforward: received source-routed packet from %lx\n", 1319 (u_long)ntohl(ip->ip_src.s_addr)); 1320 } 1321 return 1; 1322 } 1323 1324 VIF_LOCK(); 1325 MFC_LOCK(); 1326 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1327 if (ip->ip_ttl < MAXTTL) 1328 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1329 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1330 struct vif *vifp = viftable + vifi; 1331 1332 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1333 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1334 vifi, 1335 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1336 vifp->v_ifp->if_xname); 1337 } 1338 error = ip_mdq(m, ifp, NULL, vifi); 1339 MFC_UNLOCK(); 1340 VIF_UNLOCK(); 1341 return error; 1342 } 1343 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1344 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1345 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1346 if (!imo) 1347 printf("In fact, no options were specified at all\n"); 1348 } 1349 1350 /* 1351 * Don't forward a packet with time-to-live of zero or one, 1352 * or a packet destined to a local-only group. 1353 */ 1354 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) { 1355 MFC_UNLOCK(); 1356 VIF_UNLOCK(); 1357 return 0; 1358 } 1359 1360 /* 1361 * Determine forwarding vifs from the forwarding cache table 1362 */ 1363 ++mrtstat.mrts_mfc_lookups; 1364 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1365 1366 /* Entry exists, so forward if necessary */ 1367 if (rt != NULL) { 1368 error = ip_mdq(m, ifp, rt, -1); 1369 MFC_UNLOCK(); 1370 VIF_UNLOCK(); 1371 return error; 1372 } else { 1373 /* 1374 * If we don't have a route for packet's origin, 1375 * Make a copy of the packet & send message to routing daemon 1376 */ 1377 1378 struct mbuf *mb0; 1379 struct rtdetq *rte; 1380 u_long hash; 1381 int hlen = ip->ip_hl << 2; 1382 1383 ++mrtstat.mrts_mfc_misses; 1384 1385 mrtstat.mrts_no_route++; 1386 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1387 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1388 (u_long)ntohl(ip->ip_src.s_addr), 1389 (u_long)ntohl(ip->ip_dst.s_addr)); 1390 1391 /* 1392 * Allocate mbufs early so that we don't do extra work if we are 1393 * just going to fail anyway. Make sure to pullup the header so 1394 * that other people can't step on it. 1395 */ 1396 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1397 if (rte == NULL) { 1398 MFC_UNLOCK(); 1399 VIF_UNLOCK(); 1400 return ENOBUFS; 1401 } 1402 mb0 = m_copypacket(m, M_DONTWAIT); 1403 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1404 mb0 = m_pullup(mb0, hlen); 1405 if (mb0 == NULL) { 1406 free(rte, M_MRTABLE); 1407 MFC_UNLOCK(); 1408 VIF_UNLOCK(); 1409 return ENOBUFS; 1410 } 1411 1412 /* is there an upcall waiting for this flow ? */ 1413 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1414 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1415 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1416 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1417 (rt->mfc_stall != NULL)) 1418 break; 1419 } 1420 1421 if (rt == NULL) { 1422 int i; 1423 struct igmpmsg *im; 1424 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1425 struct mbuf *mm; 1426 1427 /* 1428 * Locate the vifi for the incoming interface for this packet. 1429 * If none found, drop packet. 1430 */ 1431 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1432 ; 1433 if (vifi >= numvifs) /* vif not found, drop packet */ 1434 goto non_fatal; 1435 1436 /* no upcall, so make a new entry */ 1437 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1438 if (rt == NULL) 1439 goto fail; 1440 /* Make a copy of the header to send to the user level process */ 1441 mm = m_copy(mb0, 0, hlen); 1442 if (mm == NULL) 1443 goto fail1; 1444 1445 /* 1446 * Send message to routing daemon to install 1447 * a route into the kernel table 1448 */ 1449 1450 im = mtod(mm, struct igmpmsg *); 1451 im->im_msgtype = IGMPMSG_NOCACHE; 1452 im->im_mbz = 0; 1453 im->im_vif = vifi; 1454 1455 mrtstat.mrts_upcalls++; 1456 1457 k_igmpsrc.sin_addr = ip->ip_src; 1458 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1459 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1460 ++mrtstat.mrts_upq_sockfull; 1461fail1: 1462 free(rt, M_MRTABLE); 1463fail: 1464 free(rte, M_MRTABLE); 1465 m_freem(mb0); 1466 MFC_UNLOCK(); 1467 VIF_UNLOCK(); 1468 return ENOBUFS; 1469 } 1470 1471 /* insert new entry at head of hash chain */ 1472 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1473 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1474 rt->mfc_expire = UPCALL_EXPIRE; 1475 nexpire[hash]++; 1476 for (i = 0; i < numvifs; i++) { 1477 rt->mfc_ttls[i] = 0; 1478 rt->mfc_flags[i] = 0; 1479 } 1480 rt->mfc_parent = -1; 1481 1482 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1483 1484 rt->mfc_bw_meter = NULL; 1485 1486 /* link into table */ 1487 rt->mfc_next = mfctable[hash]; 1488 mfctable[hash] = rt; 1489 rt->mfc_stall = rte; 1490 1491 } else { 1492 /* determine if q has overflowed */ 1493 int npkts = 0; 1494 struct rtdetq **p; 1495 1496 /* 1497 * XXX ouch! we need to append to the list, but we 1498 * only have a pointer to the front, so we have to 1499 * scan the entire list every time. 1500 */ 1501 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1502 npkts++; 1503 1504 if (npkts > MAX_UPQ) { 1505 mrtstat.mrts_upq_ovflw++; 1506non_fatal: 1507 free(rte, M_MRTABLE); 1508 m_freem(mb0); 1509 MFC_UNLOCK(); 1510 VIF_UNLOCK(); 1511 return 0; 1512 } 1513 1514 /* Add this entry to the end of the queue */ 1515 *p = rte; 1516 } 1517 1518 rte->m = mb0; 1519 rte->ifp = ifp; 1520 rte->next = NULL; 1521 1522 MFC_UNLOCK(); 1523 VIF_UNLOCK(); 1524 1525 return 0; 1526 } 1527} 1528 1529/* 1530 * Clean up the cache entry if upcall is not serviced 1531 */ 1532static void 1533expire_upcalls(void *unused) 1534{ 1535 struct rtdetq *rte; 1536 struct mfc *mfc, **nptr; 1537 int i; 1538 1539 MFC_LOCK(); 1540 for (i = 0; i < MFCTBLSIZ; i++) { 1541 if (nexpire[i] == 0) 1542 continue; 1543 nptr = &mfctable[i]; 1544 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1545 /* 1546 * Skip real cache entries 1547 * Make sure it wasn't marked to not expire (shouldn't happen) 1548 * If it expires now 1549 */ 1550 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1551 --mfc->mfc_expire == 0) { 1552 if (mrtdebug & DEBUG_EXPIRE) 1553 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1554 (u_long)ntohl(mfc->mfc_origin.s_addr), 1555 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1556 /* 1557 * drop all the packets 1558 * free the mbuf with the pkt, if, timing info 1559 */ 1560 for (rte = mfc->mfc_stall; rte; ) { 1561 struct rtdetq *n = rte->next; 1562 1563 m_freem(rte->m); 1564 free(rte, M_MRTABLE); 1565 rte = n; 1566 } 1567 ++mrtstat.mrts_cache_cleanups; 1568 nexpire[i]--; 1569 1570 /* 1571 * free the bw_meter entries 1572 */ 1573 while (mfc->mfc_bw_meter != NULL) { 1574 struct bw_meter *x = mfc->mfc_bw_meter; 1575 1576 mfc->mfc_bw_meter = x->bm_mfc_next; 1577 free(x, M_BWMETER); 1578 } 1579 1580 *nptr = mfc->mfc_next; 1581 free(mfc, M_MRTABLE); 1582 } else { 1583 nptr = &mfc->mfc_next; 1584 } 1585 } 1586 } 1587 MFC_UNLOCK(); 1588 1589 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1590} 1591 1592/* 1593 * Packet forwarding routine once entry in the cache is made 1594 */ 1595static int 1596ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1597{ 1598 struct ip *ip = mtod(m, struct ip *); 1599 vifi_t vifi; 1600 int plen = ip->ip_len; 1601 1602 VIF_LOCK_ASSERT(); 1603 1604 /* 1605 * If xmt_vif is not -1, send on only the requested vif. 1606 * 1607 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1608 */ 1609 if (xmt_vif < numvifs) { 1610 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1611 pim_register_send(ip, viftable + xmt_vif, m, rt); 1612 else 1613 phyint_send(ip, viftable + xmt_vif, m); 1614 return 1; 1615 } 1616 1617 /* 1618 * Don't forward if it didn't arrive from the parent vif for its origin. 1619 */ 1620 vifi = rt->mfc_parent; 1621 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1622 /* came in the wrong interface */ 1623 if (mrtdebug & DEBUG_FORWARD) 1624 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1625 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1626 ++mrtstat.mrts_wrong_if; 1627 ++rt->mfc_wrong_if; 1628 /* 1629 * If we are doing PIM assert processing, send a message 1630 * to the routing daemon. 1631 * 1632 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1633 * can complete the SPT switch, regardless of the type 1634 * of the iif (broadcast media, GRE tunnel, etc). 1635 */ 1636 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1637 struct timeval now; 1638 u_long delta; 1639 1640 if (ifp == &multicast_register_if) 1641 pimstat.pims_rcv_registers_wrongiif++; 1642 1643 /* Get vifi for the incoming packet */ 1644 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1645 ; 1646 if (vifi >= numvifs) 1647 return 0; /* The iif is not found: ignore the packet. */ 1648 1649 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1650 return 0; /* WRONGVIF disabled: ignore the packet */ 1651 1652 GET_TIME(now); 1653 1654 TV_DELTA(now, rt->mfc_last_assert, delta); 1655 1656 if (delta > ASSERT_MSG_TIME) { 1657 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1658 struct igmpmsg *im; 1659 int hlen = ip->ip_hl << 2; 1660 struct mbuf *mm = m_copy(m, 0, hlen); 1661 1662 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1663 mm = m_pullup(mm, hlen); 1664 if (mm == NULL) 1665 return ENOBUFS; 1666 1667 rt->mfc_last_assert = now; 1668 1669 im = mtod(mm, struct igmpmsg *); 1670 im->im_msgtype = IGMPMSG_WRONGVIF; 1671 im->im_mbz = 0; 1672 im->im_vif = vifi; 1673 1674 mrtstat.mrts_upcalls++; 1675 1676 k_igmpsrc.sin_addr = im->im_src; 1677 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1678 log(LOG_WARNING, 1679 "ip_mforward: ip_mrouter socket queue full\n"); 1680 ++mrtstat.mrts_upq_sockfull; 1681 return ENOBUFS; 1682 } 1683 } 1684 } 1685 return 0; 1686 } 1687 1688 /* If I sourced this packet, it counts as output, else it was input. */ 1689 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1690 viftable[vifi].v_pkt_out++; 1691 viftable[vifi].v_bytes_out += plen; 1692 } else { 1693 viftable[vifi].v_pkt_in++; 1694 viftable[vifi].v_bytes_in += plen; 1695 } 1696 rt->mfc_pkt_cnt++; 1697 rt->mfc_byte_cnt += plen; 1698 1699 /* 1700 * For each vif, decide if a copy of the packet should be forwarded. 1701 * Forward if: 1702 * - the ttl exceeds the vif's threshold 1703 * - there are group members downstream on interface 1704 */ 1705 for (vifi = 0; vifi < numvifs; vifi++) 1706 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1707 viftable[vifi].v_pkt_out++; 1708 viftable[vifi].v_bytes_out += plen; 1709 if (viftable[vifi].v_flags & VIFF_REGISTER) 1710 pim_register_send(ip, viftable + vifi, m, rt); 1711 else 1712 phyint_send(ip, viftable + vifi, m); 1713 } 1714 1715 /* 1716 * Perform upcall-related bw measuring. 1717 */ 1718 if (rt->mfc_bw_meter != NULL) { 1719 struct bw_meter *x; 1720 struct timeval now; 1721 1722 GET_TIME(now); 1723 MFC_LOCK_ASSERT(); 1724 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1725 bw_meter_receive_packet(x, plen, &now); 1726 } 1727 1728 return 0; 1729} 1730 1731/* 1732 * check if a vif number is legal/ok. This is used by ip_output. 1733 */ 1734static int 1735X_legal_vif_num(int vif) 1736{ 1737 /* XXX unlocked, matter? */ 1738 return (vif >= 0 && vif < numvifs); 1739} 1740 1741/* 1742 * Return the local address used by this vif 1743 */ 1744static u_long 1745X_ip_mcast_src(int vifi) 1746{ 1747 /* XXX unlocked, matter? */ 1748 if (vifi >= 0 && vifi < numvifs) 1749 return viftable[vifi].v_lcl_addr.s_addr; 1750 else 1751 return INADDR_ANY; 1752} 1753 1754static void 1755phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1756{ 1757 struct mbuf *mb_copy; 1758 int hlen = ip->ip_hl << 2; 1759 1760 VIF_LOCK_ASSERT(); 1761 1762 /* 1763 * Make a new reference to the packet; make sure that 1764 * the IP header is actually copied, not just referenced, 1765 * so that ip_output() only scribbles on the copy. 1766 */ 1767 mb_copy = m_copypacket(m, M_DONTWAIT); 1768 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1769 mb_copy = m_pullup(mb_copy, hlen); 1770 if (mb_copy == NULL) 1771 return; 1772 1773 send_packet(vifp, mb_copy); 1774} 1775 1776static void 1777send_packet(struct vif *vifp, struct mbuf *m) 1778{ 1779 struct ip_moptions imo; 1780 struct in_multi *imm[2]; 1781 int error; 1782 1783 VIF_LOCK_ASSERT(); 1784 1785 imo.imo_multicast_ifp = vifp->v_ifp; 1786 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1787 imo.imo_multicast_loop = 1; 1788 imo.imo_multicast_vif = -1; 1789 imo.imo_num_memberships = 0; 1790 imo.imo_max_memberships = 2; 1791 imo.imo_membership = &imm[0]; 1792 1793 /* 1794 * Re-entrancy should not be a problem here, because 1795 * the packets that we send out and are looped back at us 1796 * should get rejected because they appear to come from 1797 * the loopback interface, thus preventing looping. 1798 */ 1799 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL); 1800 if (mrtdebug & DEBUG_XMIT) { 1801 log(LOG_DEBUG, "phyint_send on vif %td err %d\n", 1802 vifp - viftable, error); 1803 } 1804} 1805 1806static int 1807X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 1808{ 1809 int error, vifi; 1810 1811 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1812 return EOPNOTSUPP; 1813 1814 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 1815 if (error) 1816 return error; 1817 1818 VIF_LOCK(); 1819 1820 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 1821 VIF_UNLOCK(); 1822 return EADDRNOTAVAIL; 1823 } 1824 1825 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 1826 /* Check if socket is available. */ 1827 if (viftable[vifi].v_rsvpd != NULL) { 1828 VIF_UNLOCK(); 1829 return EADDRINUSE; 1830 } 1831 1832 viftable[vifi].v_rsvpd = so; 1833 /* This may seem silly, but we need to be sure we don't over-increment 1834 * the RSVP counter, in case something slips up. 1835 */ 1836 if (!viftable[vifi].v_rsvp_on) { 1837 viftable[vifi].v_rsvp_on = 1; 1838 rsvp_on++; 1839 } 1840 } else { /* must be VIF_OFF */ 1841 /* 1842 * XXX as an additional consistency check, one could make sure 1843 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 1844 * first parameter is pretty useless. 1845 */ 1846 viftable[vifi].v_rsvpd = NULL; 1847 /* 1848 * This may seem silly, but we need to be sure we don't over-decrement 1849 * the RSVP counter, in case something slips up. 1850 */ 1851 if (viftable[vifi].v_rsvp_on) { 1852 viftable[vifi].v_rsvp_on = 0; 1853 rsvp_on--; 1854 } 1855 } 1856 VIF_UNLOCK(); 1857 return 0; 1858} 1859 1860static void 1861X_ip_rsvp_force_done(struct socket *so) 1862{ 1863 int vifi; 1864 1865 /* Don't bother if it is not the right type of socket. */ 1866 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1867 return; 1868 1869 VIF_LOCK(); 1870 1871 /* The socket may be attached to more than one vif...this 1872 * is perfectly legal. 1873 */ 1874 for (vifi = 0; vifi < numvifs; vifi++) { 1875 if (viftable[vifi].v_rsvpd == so) { 1876 viftable[vifi].v_rsvpd = NULL; 1877 /* This may seem silly, but we need to be sure we don't 1878 * over-decrement the RSVP counter, in case something slips up. 1879 */ 1880 if (viftable[vifi].v_rsvp_on) { 1881 viftable[vifi].v_rsvp_on = 0; 1882 rsvp_on--; 1883 } 1884 } 1885 } 1886 1887 VIF_UNLOCK(); 1888} 1889 1890static void 1891X_rsvp_input(struct mbuf *m, int off) 1892{ 1893 int vifi; 1894 struct ip *ip = mtod(m, struct ip *); 1895 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 1896 struct ifnet *ifp; 1897 1898 if (rsvpdebug) 1899 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 1900 1901 /* Can still get packets with rsvp_on = 0 if there is a local member 1902 * of the group to which the RSVP packet is addressed. But in this 1903 * case we want to throw the packet away. 1904 */ 1905 if (!rsvp_on) { 1906 m_freem(m); 1907 return; 1908 } 1909 1910 if (rsvpdebug) 1911 printf("rsvp_input: check vifs\n"); 1912 1913#ifdef DIAGNOSTIC 1914 M_ASSERTPKTHDR(m); 1915#endif 1916 1917 ifp = m->m_pkthdr.rcvif; 1918 1919 VIF_LOCK(); 1920 /* Find which vif the packet arrived on. */ 1921 for (vifi = 0; vifi < numvifs; vifi++) 1922 if (viftable[vifi].v_ifp == ifp) 1923 break; 1924 1925 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 1926 /* 1927 * Drop the lock here to avoid holding it across rip_input. 1928 * This could make rsvpdebug printfs wrong. If you care, 1929 * record the state of stuff before dropping the lock. 1930 */ 1931 VIF_UNLOCK(); 1932 /* 1933 * If the old-style non-vif-associated socket is set, 1934 * then use it. Otherwise, drop packet since there 1935 * is no specific socket for this vif. 1936 */ 1937 if (ip_rsvpd != NULL) { 1938 if (rsvpdebug) 1939 printf("rsvp_input: Sending packet up old-style socket\n"); 1940 rip_input(m, off); /* xxx */ 1941 } else { 1942 if (rsvpdebug && vifi == numvifs) 1943 printf("rsvp_input: Can't find vif for packet.\n"); 1944 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 1945 printf("rsvp_input: No socket defined for vif %d\n",vifi); 1946 m_freem(m); 1947 } 1948 return; 1949 } 1950 rsvp_src.sin_addr = ip->ip_src; 1951 1952 if (rsvpdebug && m) 1953 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 1954 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 1955 1956 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 1957 if (rsvpdebug) 1958 printf("rsvp_input: Failed to append to socket\n"); 1959 } else { 1960 if (rsvpdebug) 1961 printf("rsvp_input: send packet up\n"); 1962 } 1963 VIF_UNLOCK(); 1964} 1965 1966/* 1967 * Code for bandwidth monitors 1968 */ 1969 1970/* 1971 * Define common interface for timeval-related methods 1972 */ 1973#define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 1974#define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 1975#define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 1976 1977static uint32_t 1978compute_bw_meter_flags(struct bw_upcall *req) 1979{ 1980 uint32_t flags = 0; 1981 1982 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 1983 flags |= BW_METER_UNIT_PACKETS; 1984 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 1985 flags |= BW_METER_UNIT_BYTES; 1986 if (req->bu_flags & BW_UPCALL_GEQ) 1987 flags |= BW_METER_GEQ; 1988 if (req->bu_flags & BW_UPCALL_LEQ) 1989 flags |= BW_METER_LEQ; 1990 1991 return flags; 1992} 1993 1994/* 1995 * Add a bw_meter entry 1996 */ 1997static int 1998add_bw_upcall(struct bw_upcall *req) 1999{ 2000 struct mfc *mfc; 2001 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2002 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2003 struct timeval now; 2004 struct bw_meter *x; 2005 uint32_t flags; 2006 2007 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2008 return EOPNOTSUPP; 2009 2010 /* Test if the flags are valid */ 2011 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2012 return EINVAL; 2013 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2014 return EINVAL; 2015 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2016 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2017 return EINVAL; 2018 2019 /* Test if the threshold time interval is valid */ 2020 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2021 return EINVAL; 2022 2023 flags = compute_bw_meter_flags(req); 2024 2025 /* 2026 * Find if we have already same bw_meter entry 2027 */ 2028 MFC_LOCK(); 2029 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2030 if (mfc == NULL) { 2031 MFC_UNLOCK(); 2032 return EADDRNOTAVAIL; 2033 } 2034 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2035 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2036 &req->bu_threshold.b_time, ==)) && 2037 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2038 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2039 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2040 MFC_UNLOCK(); 2041 return 0; /* XXX Already installed */ 2042 } 2043 } 2044 2045 /* Allocate the new bw_meter entry */ 2046 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2047 if (x == NULL) { 2048 MFC_UNLOCK(); 2049 return ENOBUFS; 2050 } 2051 2052 /* Set the new bw_meter entry */ 2053 x->bm_threshold.b_time = req->bu_threshold.b_time; 2054 GET_TIME(now); 2055 x->bm_start_time = now; 2056 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2057 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2058 x->bm_measured.b_packets = 0; 2059 x->bm_measured.b_bytes = 0; 2060 x->bm_flags = flags; 2061 x->bm_time_next = NULL; 2062 x->bm_time_hash = BW_METER_BUCKETS; 2063 2064 /* Add the new bw_meter entry to the front of entries for this MFC */ 2065 x->bm_mfc = mfc; 2066 x->bm_mfc_next = mfc->mfc_bw_meter; 2067 mfc->mfc_bw_meter = x; 2068 schedule_bw_meter(x, &now); 2069 MFC_UNLOCK(); 2070 2071 return 0; 2072} 2073 2074static void 2075free_bw_list(struct bw_meter *list) 2076{ 2077 while (list != NULL) { 2078 struct bw_meter *x = list; 2079 2080 list = list->bm_mfc_next; 2081 unschedule_bw_meter(x); 2082 free(x, M_BWMETER); 2083 } 2084} 2085 2086/* 2087 * Delete one or multiple bw_meter entries 2088 */ 2089static int 2090del_bw_upcall(struct bw_upcall *req) 2091{ 2092 struct mfc *mfc; 2093 struct bw_meter *x; 2094 2095 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2096 return EOPNOTSUPP; 2097 2098 MFC_LOCK(); 2099 /* Find the corresponding MFC entry */ 2100 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2101 if (mfc == NULL) { 2102 MFC_UNLOCK(); 2103 return EADDRNOTAVAIL; 2104 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2105 /* 2106 * Delete all bw_meter entries for this mfc 2107 */ 2108 struct bw_meter *list; 2109 2110 list = mfc->mfc_bw_meter; 2111 mfc->mfc_bw_meter = NULL; 2112 free_bw_list(list); 2113 MFC_UNLOCK(); 2114 return 0; 2115 } else { /* Delete a single bw_meter entry */ 2116 struct bw_meter *prev; 2117 uint32_t flags = 0; 2118 2119 flags = compute_bw_meter_flags(req); 2120 2121 /* Find the bw_meter entry to delete */ 2122 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2123 prev = x, x = x->bm_mfc_next) { 2124 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2125 &req->bu_threshold.b_time, ==)) && 2126 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2127 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2128 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2129 break; 2130 } 2131 if (x != NULL) { /* Delete entry from the list for this MFC */ 2132 if (prev != NULL) 2133 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2134 else 2135 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2136 2137 unschedule_bw_meter(x); 2138 MFC_UNLOCK(); 2139 /* Free the bw_meter entry */ 2140 free(x, M_BWMETER); 2141 return 0; 2142 } else { 2143 MFC_UNLOCK(); 2144 return EINVAL; 2145 } 2146 } 2147 /* NOTREACHED */ 2148} 2149 2150/* 2151 * Perform bandwidth measurement processing that may result in an upcall 2152 */ 2153static void 2154bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2155{ 2156 struct timeval delta; 2157 2158 MFC_LOCK_ASSERT(); 2159 2160 delta = *nowp; 2161 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2162 2163 if (x->bm_flags & BW_METER_GEQ) { 2164 /* 2165 * Processing for ">=" type of bw_meter entry 2166 */ 2167 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2168 /* Reset the bw_meter entry */ 2169 x->bm_start_time = *nowp; 2170 x->bm_measured.b_packets = 0; 2171 x->bm_measured.b_bytes = 0; 2172 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2173 } 2174 2175 /* Record that a packet is received */ 2176 x->bm_measured.b_packets++; 2177 x->bm_measured.b_bytes += plen; 2178 2179 /* 2180 * Test if we should deliver an upcall 2181 */ 2182 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2183 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2184 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2185 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2186 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2187 /* Prepare an upcall for delivery */ 2188 bw_meter_prepare_upcall(x, nowp); 2189 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2190 } 2191 } 2192 } else if (x->bm_flags & BW_METER_LEQ) { 2193 /* 2194 * Processing for "<=" type of bw_meter entry 2195 */ 2196 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2197 /* 2198 * We are behind time with the multicast forwarding table 2199 * scanning for "<=" type of bw_meter entries, so test now 2200 * if we should deliver an upcall. 2201 */ 2202 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2203 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2204 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2205 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2206 /* Prepare an upcall for delivery */ 2207 bw_meter_prepare_upcall(x, nowp); 2208 } 2209 /* Reschedule the bw_meter entry */ 2210 unschedule_bw_meter(x); 2211 schedule_bw_meter(x, nowp); 2212 } 2213 2214 /* Record that a packet is received */ 2215 x->bm_measured.b_packets++; 2216 x->bm_measured.b_bytes += plen; 2217 2218 /* 2219 * Test if we should restart the measuring interval 2220 */ 2221 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2222 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2223 (x->bm_flags & BW_METER_UNIT_BYTES && 2224 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2225 /* Don't restart the measuring interval */ 2226 } else { 2227 /* Do restart the measuring interval */ 2228 /* 2229 * XXX: note that we don't unschedule and schedule, because this 2230 * might be too much overhead per packet. Instead, when we process 2231 * all entries for a given timer hash bin, we check whether it is 2232 * really a timeout. If not, we reschedule at that time. 2233 */ 2234 x->bm_start_time = *nowp; 2235 x->bm_measured.b_packets = 0; 2236 x->bm_measured.b_bytes = 0; 2237 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2238 } 2239 } 2240} 2241 2242/* 2243 * Prepare a bandwidth-related upcall 2244 */ 2245static void 2246bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2247{ 2248 struct timeval delta; 2249 struct bw_upcall *u; 2250 2251 MFC_LOCK_ASSERT(); 2252 2253 /* 2254 * Compute the measured time interval 2255 */ 2256 delta = *nowp; 2257 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2258 2259 /* 2260 * If there are too many pending upcalls, deliver them now 2261 */ 2262 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2263 bw_upcalls_send(); 2264 2265 /* 2266 * Set the bw_upcall entry 2267 */ 2268 u = &bw_upcalls[bw_upcalls_n++]; 2269 u->bu_src = x->bm_mfc->mfc_origin; 2270 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2271 u->bu_threshold.b_time = x->bm_threshold.b_time; 2272 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2273 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2274 u->bu_measured.b_time = delta; 2275 u->bu_measured.b_packets = x->bm_measured.b_packets; 2276 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2277 u->bu_flags = 0; 2278 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2279 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2280 if (x->bm_flags & BW_METER_UNIT_BYTES) 2281 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2282 if (x->bm_flags & BW_METER_GEQ) 2283 u->bu_flags |= BW_UPCALL_GEQ; 2284 if (x->bm_flags & BW_METER_LEQ) 2285 u->bu_flags |= BW_UPCALL_LEQ; 2286} 2287 2288/* 2289 * Send the pending bandwidth-related upcalls 2290 */ 2291static void 2292bw_upcalls_send(void) 2293{ 2294 struct mbuf *m; 2295 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2296 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2297 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2298 0, /* unused2 */ 2299 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2300 0, /* im_mbz */ 2301 0, /* im_vif */ 2302 0, /* unused3 */ 2303 { 0 }, /* im_src */ 2304 { 0 } }; /* im_dst */ 2305 2306 MFC_LOCK_ASSERT(); 2307 2308 if (bw_upcalls_n == 0) 2309 return; /* No pending upcalls */ 2310 2311 bw_upcalls_n = 0; 2312 2313 /* 2314 * Allocate a new mbuf, initialize it with the header and 2315 * the payload for the pending calls. 2316 */ 2317 MGETHDR(m, M_DONTWAIT, MT_DATA); 2318 if (m == NULL) { 2319 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2320 return; 2321 } 2322 2323 m->m_len = m->m_pkthdr.len = 0; 2324 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2325 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2326 2327 /* 2328 * Send the upcalls 2329 * XXX do we need to set the address in k_igmpsrc ? 2330 */ 2331 mrtstat.mrts_upcalls++; 2332 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2333 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2334 ++mrtstat.mrts_upq_sockfull; 2335 } 2336} 2337 2338/* 2339 * Compute the timeout hash value for the bw_meter entries 2340 */ 2341#define BW_METER_TIMEHASH(bw_meter, hash) \ 2342 do { \ 2343 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2344 \ 2345 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2346 (hash) = next_timeval.tv_sec; \ 2347 if (next_timeval.tv_usec) \ 2348 (hash)++; /* XXX: make sure we don't timeout early */ \ 2349 (hash) %= BW_METER_BUCKETS; \ 2350 } while (0) 2351 2352/* 2353 * Schedule a timer to process periodically bw_meter entry of type "<=" 2354 * by linking the entry in the proper hash bucket. 2355 */ 2356static void 2357schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2358{ 2359 int time_hash; 2360 2361 MFC_LOCK_ASSERT(); 2362 2363 if (!(x->bm_flags & BW_METER_LEQ)) 2364 return; /* XXX: we schedule timers only for "<=" entries */ 2365 2366 /* 2367 * Reset the bw_meter entry 2368 */ 2369 x->bm_start_time = *nowp; 2370 x->bm_measured.b_packets = 0; 2371 x->bm_measured.b_bytes = 0; 2372 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2373 2374 /* 2375 * Compute the timeout hash value and insert the entry 2376 */ 2377 BW_METER_TIMEHASH(x, time_hash); 2378 x->bm_time_next = bw_meter_timers[time_hash]; 2379 bw_meter_timers[time_hash] = x; 2380 x->bm_time_hash = time_hash; 2381} 2382 2383/* 2384 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2385 * by removing the entry from the proper hash bucket. 2386 */ 2387static void 2388unschedule_bw_meter(struct bw_meter *x) 2389{ 2390 int time_hash; 2391 struct bw_meter *prev, *tmp; 2392 2393 MFC_LOCK_ASSERT(); 2394 2395 if (!(x->bm_flags & BW_METER_LEQ)) 2396 return; /* XXX: we schedule timers only for "<=" entries */ 2397 2398 /* 2399 * Compute the timeout hash value and delete the entry 2400 */ 2401 time_hash = x->bm_time_hash; 2402 if (time_hash >= BW_METER_BUCKETS) 2403 return; /* Entry was not scheduled */ 2404 2405 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2406 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2407 if (tmp == x) 2408 break; 2409 2410 if (tmp == NULL) 2411 panic("unschedule_bw_meter: bw_meter entry not found"); 2412 2413 if (prev != NULL) 2414 prev->bm_time_next = x->bm_time_next; 2415 else 2416 bw_meter_timers[time_hash] = x->bm_time_next; 2417 2418 x->bm_time_next = NULL; 2419 x->bm_time_hash = BW_METER_BUCKETS; 2420} 2421 2422 2423/* 2424 * Process all "<=" type of bw_meter that should be processed now, 2425 * and for each entry prepare an upcall if necessary. Each processed 2426 * entry is rescheduled again for the (periodic) processing. 2427 * 2428 * This is run periodically (once per second normally). On each round, 2429 * all the potentially matching entries are in the hash slot that we are 2430 * looking at. 2431 */ 2432static void 2433bw_meter_process() 2434{ 2435 static uint32_t last_tv_sec; /* last time we processed this */ 2436 2437 uint32_t loops; 2438 int i; 2439 struct timeval now, process_endtime; 2440 2441 GET_TIME(now); 2442 if (last_tv_sec == now.tv_sec) 2443 return; /* nothing to do */ 2444 2445 loops = now.tv_sec - last_tv_sec; 2446 last_tv_sec = now.tv_sec; 2447 if (loops > BW_METER_BUCKETS) 2448 loops = BW_METER_BUCKETS; 2449 2450 MFC_LOCK(); 2451 /* 2452 * Process all bins of bw_meter entries from the one after the last 2453 * processed to the current one. On entry, i points to the last bucket 2454 * visited, so we need to increment i at the beginning of the loop. 2455 */ 2456 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2457 struct bw_meter *x, *tmp_list; 2458 2459 if (++i >= BW_METER_BUCKETS) 2460 i = 0; 2461 2462 /* Disconnect the list of bw_meter entries from the bin */ 2463 tmp_list = bw_meter_timers[i]; 2464 bw_meter_timers[i] = NULL; 2465 2466 /* Process the list of bw_meter entries */ 2467 while (tmp_list != NULL) { 2468 x = tmp_list; 2469 tmp_list = tmp_list->bm_time_next; 2470 2471 /* Test if the time interval is over */ 2472 process_endtime = x->bm_start_time; 2473 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2474 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2475 /* Not yet: reschedule, but don't reset */ 2476 int time_hash; 2477 2478 BW_METER_TIMEHASH(x, time_hash); 2479 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2480 /* 2481 * XXX: somehow the bin processing is a bit ahead of time. 2482 * Put the entry in the next bin. 2483 */ 2484 if (++time_hash >= BW_METER_BUCKETS) 2485 time_hash = 0; 2486 } 2487 x->bm_time_next = bw_meter_timers[time_hash]; 2488 bw_meter_timers[time_hash] = x; 2489 x->bm_time_hash = time_hash; 2490 2491 continue; 2492 } 2493 2494 /* 2495 * Test if we should deliver an upcall 2496 */ 2497 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2498 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2499 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2500 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2501 /* Prepare an upcall for delivery */ 2502 bw_meter_prepare_upcall(x, &now); 2503 } 2504 2505 /* 2506 * Reschedule for next processing 2507 */ 2508 schedule_bw_meter(x, &now); 2509 } 2510 } 2511 2512 /* Send all upcalls that are pending delivery */ 2513 bw_upcalls_send(); 2514 2515 MFC_UNLOCK(); 2516} 2517 2518/* 2519 * A periodic function for sending all upcalls that are pending delivery 2520 */ 2521static void 2522expire_bw_upcalls_send(void *unused) 2523{ 2524 MFC_LOCK(); 2525 bw_upcalls_send(); 2526 MFC_UNLOCK(); 2527 2528 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2529 expire_bw_upcalls_send, NULL); 2530} 2531 2532/* 2533 * A periodic function for periodic scanning of the multicast forwarding 2534 * table for processing all "<=" bw_meter entries. 2535 */ 2536static void 2537expire_bw_meter_process(void *unused) 2538{ 2539 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2540 bw_meter_process(); 2541 2542 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2543} 2544 2545/* 2546 * End of bandwidth monitoring code 2547 */ 2548 2549/* 2550 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2551 * 2552 */ 2553static int 2554pim_register_send(struct ip *ip, struct vif *vifp, 2555 struct mbuf *m, struct mfc *rt) 2556{ 2557 struct mbuf *mb_copy, *mm; 2558 2559 if (mrtdebug & DEBUG_PIM) 2560 log(LOG_DEBUG, "pim_register_send: "); 2561 2562 /* 2563 * Do not send IGMP_WHOLEPKT notifications to userland, if the 2564 * rendezvous point was unspecified, and we were told not to. 2565 */ 2566 if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) && 2567 (rt->mfc_rp.s_addr == INADDR_ANY)) 2568 return 0; 2569 2570 mb_copy = pim_register_prepare(ip, m); 2571 if (mb_copy == NULL) 2572 return ENOBUFS; 2573 2574 /* 2575 * Send all the fragments. Note that the mbuf for each fragment 2576 * is freed by the sending machinery. 2577 */ 2578 for (mm = mb_copy; mm; mm = mb_copy) { 2579 mb_copy = mm->m_nextpkt; 2580 mm->m_nextpkt = 0; 2581 mm = m_pullup(mm, sizeof(struct ip)); 2582 if (mm != NULL) { 2583 ip = mtod(mm, struct ip *); 2584 if ((mrt_api_config & MRT_MFC_RP) && 2585 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2586 pim_register_send_rp(ip, vifp, mm, rt); 2587 } else { 2588 pim_register_send_upcall(ip, vifp, mm, rt); 2589 } 2590 } 2591 } 2592 2593 return 0; 2594} 2595 2596/* 2597 * Return a copy of the data packet that is ready for PIM Register 2598 * encapsulation. 2599 * XXX: Note that in the returned copy the IP header is a valid one. 2600 */ 2601static struct mbuf * 2602pim_register_prepare(struct ip *ip, struct mbuf *m) 2603{ 2604 struct mbuf *mb_copy = NULL; 2605 int mtu; 2606 2607 /* Take care of delayed checksums */ 2608 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2609 in_delayed_cksum(m); 2610 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2611 } 2612 2613 /* 2614 * Copy the old packet & pullup its IP header into the 2615 * new mbuf so we can modify it. 2616 */ 2617 mb_copy = m_copypacket(m, M_DONTWAIT); 2618 if (mb_copy == NULL) 2619 return NULL; 2620 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2621 if (mb_copy == NULL) 2622 return NULL; 2623 2624 /* take care of the TTL */ 2625 ip = mtod(mb_copy, struct ip *); 2626 --ip->ip_ttl; 2627 2628 /* Compute the MTU after the PIM Register encapsulation */ 2629 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2630 2631 if (ip->ip_len <= mtu) { 2632 /* Turn the IP header into a valid one */ 2633 ip->ip_len = htons(ip->ip_len); 2634 ip->ip_off = htons(ip->ip_off); 2635 ip->ip_sum = 0; 2636 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2637 } else { 2638 /* Fragment the packet */ 2639 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2640 m_freem(mb_copy); 2641 return NULL; 2642 } 2643 } 2644 return mb_copy; 2645} 2646 2647/* 2648 * Send an upcall with the data packet to the user-level process. 2649 */ 2650static int 2651pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2652 struct mbuf *mb_copy, struct mfc *rt) 2653{ 2654 struct mbuf *mb_first; 2655 int len = ntohs(ip->ip_len); 2656 struct igmpmsg *im; 2657 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2658 2659 VIF_LOCK_ASSERT(); 2660 2661 /* 2662 * Add a new mbuf with an upcall header 2663 */ 2664 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2665 if (mb_first == NULL) { 2666 m_freem(mb_copy); 2667 return ENOBUFS; 2668 } 2669 mb_first->m_data += max_linkhdr; 2670 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2671 mb_first->m_len = sizeof(struct igmpmsg); 2672 mb_first->m_next = mb_copy; 2673 2674 /* Send message to routing daemon */ 2675 im = mtod(mb_first, struct igmpmsg *); 2676 im->im_msgtype = IGMPMSG_WHOLEPKT; 2677 im->im_mbz = 0; 2678 im->im_vif = vifp - viftable; 2679 im->im_src = ip->ip_src; 2680 im->im_dst = ip->ip_dst; 2681 2682 k_igmpsrc.sin_addr = ip->ip_src; 2683 2684 mrtstat.mrts_upcalls++; 2685 2686 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2687 if (mrtdebug & DEBUG_PIM) 2688 log(LOG_WARNING, 2689 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2690 ++mrtstat.mrts_upq_sockfull; 2691 return ENOBUFS; 2692 } 2693 2694 /* Keep statistics */ 2695 pimstat.pims_snd_registers_msgs++; 2696 pimstat.pims_snd_registers_bytes += len; 2697 2698 return 0; 2699} 2700 2701/* 2702 * Encapsulate the data packet in PIM Register message and send it to the RP. 2703 */ 2704static int 2705pim_register_send_rp(struct ip *ip, struct vif *vifp, 2706 struct mbuf *mb_copy, struct mfc *rt) 2707{ 2708 struct mbuf *mb_first; 2709 struct ip *ip_outer; 2710 struct pim_encap_pimhdr *pimhdr; 2711 int len = ntohs(ip->ip_len); 2712 vifi_t vifi = rt->mfc_parent; 2713 2714 VIF_LOCK_ASSERT(); 2715 2716 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2717 m_freem(mb_copy); 2718 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2719 } 2720 2721 /* 2722 * Add a new mbuf with the encapsulating header 2723 */ 2724 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2725 if (mb_first == NULL) { 2726 m_freem(mb_copy); 2727 return ENOBUFS; 2728 } 2729 mb_first->m_data += max_linkhdr; 2730 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2731 mb_first->m_next = mb_copy; 2732 2733 mb_first->m_pkthdr.len = len + mb_first->m_len; 2734 2735 /* 2736 * Fill in the encapsulating IP and PIM header 2737 */ 2738 ip_outer = mtod(mb_first, struct ip *); 2739 *ip_outer = pim_encap_iphdr; 2740 ip_outer->ip_id = ip_newid(); 2741 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2742 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2743 ip_outer->ip_dst = rt->mfc_rp; 2744 /* 2745 * Copy the inner header TOS to the outer header, and take care of the 2746 * IP_DF bit. 2747 */ 2748 ip_outer->ip_tos = ip->ip_tos; 2749 if (ntohs(ip->ip_off) & IP_DF) 2750 ip_outer->ip_off |= IP_DF; 2751 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2752 + sizeof(pim_encap_iphdr)); 2753 *pimhdr = pim_encap_pimhdr; 2754 /* If the iif crosses a border, set the Border-bit */ 2755 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2756 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2757 2758 mb_first->m_data += sizeof(pim_encap_iphdr); 2759 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2760 mb_first->m_data -= sizeof(pim_encap_iphdr); 2761 2762 send_packet(vifp, mb_first); 2763 2764 /* Keep statistics */ 2765 pimstat.pims_snd_registers_msgs++; 2766 pimstat.pims_snd_registers_bytes += len; 2767 2768 return 0; 2769} 2770 2771/* 2772 * pim_encapcheck() is called by the encap[46]_input() path at runtime to 2773 * determine if a packet is for PIM; allowing PIM to be dynamically loaded 2774 * into the kernel. 2775 */ 2776static int 2777pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 2778{ 2779 2780#ifdef DIAGNOSTIC 2781 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); 2782#endif 2783 if (proto != IPPROTO_PIM) 2784 return 0; /* not for us; reject the datagram. */ 2785 2786 return 64; /* claim the datagram. */ 2787} 2788 2789/* 2790 * PIM-SMv2 and PIM-DM messages processing. 2791 * Receives and verifies the PIM control messages, and passes them 2792 * up to the listening socket, using rip_input(). 2793 * The only message with special processing is the PIM_REGISTER message 2794 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2795 * is passed to if_simloop(). 2796 */ 2797void 2798pim_input(struct mbuf *m, int off) 2799{ 2800 struct ip *ip = mtod(m, struct ip *); 2801 struct pim *pim; 2802 int minlen; 2803 int datalen = ip->ip_len; 2804 int ip_tos; 2805 int iphlen = off; 2806 2807 /* Keep statistics */ 2808 pimstat.pims_rcv_total_msgs++; 2809 pimstat.pims_rcv_total_bytes += datalen; 2810 2811 /* 2812 * Validate lengths 2813 */ 2814 if (datalen < PIM_MINLEN) { 2815 pimstat.pims_rcv_tooshort++; 2816 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 2817 datalen, (u_long)ip->ip_src.s_addr); 2818 m_freem(m); 2819 return; 2820 } 2821 2822 /* 2823 * If the packet is at least as big as a REGISTER, go agead 2824 * and grab the PIM REGISTER header size, to avoid another 2825 * possible m_pullup() later. 2826 * 2827 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 2828 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 2829 */ 2830 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 2831 /* 2832 * Get the IP and PIM headers in contiguous memory, and 2833 * possibly the PIM REGISTER header. 2834 */ 2835 if ((m->m_flags & M_EXT || m->m_len < minlen) && 2836 (m = m_pullup(m, minlen)) == 0) { 2837 log(LOG_ERR, "pim_input: m_pullup failure\n"); 2838 return; 2839 } 2840 /* m_pullup() may have given us a new mbuf so reset ip. */ 2841 ip = mtod(m, struct ip *); 2842 ip_tos = ip->ip_tos; 2843 2844 /* adjust mbuf to point to the PIM header */ 2845 m->m_data += iphlen; 2846 m->m_len -= iphlen; 2847 pim = mtod(m, struct pim *); 2848 2849 /* 2850 * Validate checksum. If PIM REGISTER, exclude the data packet. 2851 * 2852 * XXX: some older PIMv2 implementations don't make this distinction, 2853 * so for compatibility reason perform the checksum over part of the 2854 * message, and if error, then over the whole message. 2855 */ 2856 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 2857 /* do nothing, checksum okay */ 2858 } else if (in_cksum(m, datalen)) { 2859 pimstat.pims_rcv_badsum++; 2860 if (mrtdebug & DEBUG_PIM) 2861 log(LOG_DEBUG, "pim_input: invalid checksum"); 2862 m_freem(m); 2863 return; 2864 } 2865 2866 /* PIM version check */ 2867 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 2868 pimstat.pims_rcv_badversion++; 2869 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 2870 PIM_VT_V(pim->pim_vt), PIM_VERSION); 2871 m_freem(m); 2872 return; 2873 } 2874 2875 /* restore mbuf back to the outer IP */ 2876 m->m_data -= iphlen; 2877 m->m_len += iphlen; 2878 2879 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 2880 /* 2881 * Since this is a REGISTER, we'll make a copy of the register 2882 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 2883 * routing daemon. 2884 */ 2885 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 2886 struct mbuf *mcp; 2887 struct ip *encap_ip; 2888 u_int32_t *reghdr; 2889 struct ifnet *vifp; 2890 2891 VIF_LOCK(); 2892 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 2893 VIF_UNLOCK(); 2894 if (mrtdebug & DEBUG_PIM) 2895 log(LOG_DEBUG, 2896 "pim_input: register vif not set: %d\n", reg_vif_num); 2897 m_freem(m); 2898 return; 2899 } 2900 /* XXX need refcnt? */ 2901 vifp = viftable[reg_vif_num].v_ifp; 2902 VIF_UNLOCK(); 2903 2904 /* 2905 * Validate length 2906 */ 2907 if (datalen < PIM_REG_MINLEN) { 2908 pimstat.pims_rcv_tooshort++; 2909 pimstat.pims_rcv_badregisters++; 2910 log(LOG_ERR, 2911 "pim_input: register packet size too small %d from %lx\n", 2912 datalen, (u_long)ip->ip_src.s_addr); 2913 m_freem(m); 2914 return; 2915 } 2916 2917 reghdr = (u_int32_t *)(pim + 1); 2918 encap_ip = (struct ip *)(reghdr + 1); 2919 2920 if (mrtdebug & DEBUG_PIM) { 2921 log(LOG_DEBUG, 2922 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 2923 (u_long)ntohl(encap_ip->ip_src.s_addr), 2924 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2925 ntohs(encap_ip->ip_len)); 2926 } 2927 2928 /* verify the version number of the inner packet */ 2929 if (encap_ip->ip_v != IPVERSION) { 2930 pimstat.pims_rcv_badregisters++; 2931 if (mrtdebug & DEBUG_PIM) { 2932 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 2933 "of the inner packet\n", encap_ip->ip_v); 2934 } 2935 m_freem(m); 2936 return; 2937 } 2938 2939 /* verify the inner packet is destined to a mcast group */ 2940 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 2941 pimstat.pims_rcv_badregisters++; 2942 if (mrtdebug & DEBUG_PIM) 2943 log(LOG_DEBUG, 2944 "pim_input: inner packet of register is not " 2945 "multicast %lx\n", 2946 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 2947 m_freem(m); 2948 return; 2949 } 2950 2951 /* If a NULL_REGISTER, pass it to the daemon */ 2952 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 2953 goto pim_input_to_daemon; 2954 2955 /* 2956 * Copy the TOS from the outer IP header to the inner IP header. 2957 */ 2958 if (encap_ip->ip_tos != ip_tos) { 2959 /* Outer TOS -> inner TOS */ 2960 encap_ip->ip_tos = ip_tos; 2961 /* Recompute the inner header checksum. Sigh... */ 2962 2963 /* adjust mbuf to point to the inner IP header */ 2964 m->m_data += (iphlen + PIM_MINLEN); 2965 m->m_len -= (iphlen + PIM_MINLEN); 2966 2967 encap_ip->ip_sum = 0; 2968 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 2969 2970 /* restore mbuf to point back to the outer IP header */ 2971 m->m_data -= (iphlen + PIM_MINLEN); 2972 m->m_len += (iphlen + PIM_MINLEN); 2973 } 2974 2975 /* 2976 * Decapsulate the inner IP packet and loopback to forward it 2977 * as a normal multicast packet. Also, make a copy of the 2978 * outer_iphdr + pimhdr + reghdr + encap_iphdr 2979 * to pass to the daemon later, so it can take the appropriate 2980 * actions (e.g., send back PIM_REGISTER_STOP). 2981 * XXX: here m->m_data points to the outer IP header. 2982 */ 2983 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 2984 if (mcp == NULL) { 2985 log(LOG_ERR, 2986 "pim_input: pim register: could not copy register head\n"); 2987 m_freem(m); 2988 return; 2989 } 2990 2991 /* Keep statistics */ 2992 /* XXX: registers_bytes include only the encap. mcast pkt */ 2993 pimstat.pims_rcv_registers_msgs++; 2994 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 2995 2996 /* 2997 * forward the inner ip packet; point m_data at the inner ip. 2998 */ 2999 m_adj(m, iphlen + PIM_MINLEN); 3000 3001 if (mrtdebug & DEBUG_PIM) { 3002 log(LOG_DEBUG, 3003 "pim_input: forwarding decapsulated register: " 3004 "src %lx, dst %lx, vif %d\n", 3005 (u_long)ntohl(encap_ip->ip_src.s_addr), 3006 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3007 reg_vif_num); 3008 } 3009 /* NB: vifp was collected above; can it change on us? */ 3010 if_simloop(vifp, m, dst.sin_family, 0); 3011 3012 /* prepare the register head to send to the mrouting daemon */ 3013 m = mcp; 3014 } 3015 3016pim_input_to_daemon: 3017 /* 3018 * Pass the PIM message up to the daemon; if it is a Register message, 3019 * pass the 'head' only up to the daemon. This includes the 3020 * outer IP header, PIM header, PIM-Register header and the 3021 * inner IP header. 3022 * XXX: the outer IP header pkt size of a Register is not adjust to 3023 * reflect the fact that the inner multicast data is truncated. 3024 */ 3025 rip_input(m, iphlen); 3026 3027 return; 3028} 3029 3030/* 3031 * XXX: This is common code for dealing with initialization for both 3032 * the IPv4 and IPv6 multicast forwarding paths. It could do with cleanup. 3033 */ 3034static int 3035ip_mroute_modevent(module_t mod, int type, void *unused) 3036{ 3037 switch (type) { 3038 case MOD_LOAD: 3039 MROUTER_LOCK_INIT(); 3040 MFC_LOCK_INIT(); 3041 VIF_LOCK_INIT(); 3042 ip_mrouter_reset(); 3043 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", 3044 &pim_squelch_wholepkt); 3045 3046 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM, 3047 pim_encapcheck, &in_pim_protosw, NULL); 3048 if (pim_encap_cookie == NULL) { 3049 printf("ip_mroute: unable to attach pim encap\n"); 3050 VIF_LOCK_DESTROY(); 3051 MFC_LOCK_DESTROY(); 3052 MROUTER_LOCK_DESTROY(); 3053 return (EINVAL); 3054 } 3055 3056#ifdef INET6 3057 pim6_encap_cookie = encap_attach_func(AF_INET6, IPPROTO_PIM, 3058 pim_encapcheck, (struct protosw *)&in6_pim_protosw, NULL); 3059 if (pim6_encap_cookie == NULL) { 3060 printf("ip_mroute: unable to attach pim6 encap\n"); 3061 if (pim_encap_cookie) { 3062 encap_detach(pim_encap_cookie); 3063 pim_encap_cookie = NULL; 3064 } 3065 VIF_LOCK_DESTROY(); 3066 MFC_LOCK_DESTROY(); 3067 MROUTER_LOCK_DESTROY(); 3068 return (EINVAL); 3069 } 3070#endif 3071 3072 ip_mcast_src = X_ip_mcast_src; 3073 ip_mforward = X_ip_mforward; 3074 ip_mrouter_done = X_ip_mrouter_done; 3075 ip_mrouter_get = X_ip_mrouter_get; 3076 ip_mrouter_set = X_ip_mrouter_set; 3077 3078#ifdef INET6 3079 ip6_mforward = X_ip6_mforward; 3080 ip6_mrouter_done = X_ip6_mrouter_done; 3081 ip6_mrouter_get = X_ip6_mrouter_get; 3082 ip6_mrouter_set = X_ip6_mrouter_set; 3083 mrt6_ioctl = X_mrt6_ioctl; 3084#endif 3085 3086 ip_rsvp_force_done = X_ip_rsvp_force_done; 3087 ip_rsvp_vif = X_ip_rsvp_vif; 3088 3089 legal_vif_num = X_legal_vif_num; 3090 mrt_ioctl = X_mrt_ioctl; 3091 rsvp_input_p = X_rsvp_input; 3092 break; 3093 3094 case MOD_UNLOAD: 3095 /* 3096 * Typically module unload happens after the user-level 3097 * process has shutdown the kernel services (the check 3098 * below insures someone can't just yank the module out 3099 * from under a running process). But if the module is 3100 * just loaded and then unloaded w/o starting up a user 3101 * process we still need to cleanup. 3102 */ 3103 if (ip_mrouter 3104#ifdef INET6 3105 || ip6_mrouter 3106#endif 3107 ) 3108 return EINVAL; 3109 3110#ifdef INET6 3111 if (pim6_encap_cookie) { 3112 encap_detach(pim6_encap_cookie); 3113 pim6_encap_cookie = NULL; 3114 } 3115 X_ip6_mrouter_done(); 3116 ip6_mforward = NULL; 3117 ip6_mrouter_done = NULL; 3118 ip6_mrouter_get = NULL; 3119 ip6_mrouter_set = NULL; 3120 mrt6_ioctl = NULL; 3121#endif 3122 3123 if (pim_encap_cookie) { 3124 encap_detach(pim_encap_cookie); 3125 pim_encap_cookie = NULL; 3126 } 3127 X_ip_mrouter_done(); 3128 ip_mcast_src = NULL; 3129 ip_mforward = NULL; 3130 ip_mrouter_done = NULL; 3131 ip_mrouter_get = NULL; 3132 ip_mrouter_set = NULL; 3133 3134 ip_rsvp_force_done = NULL; 3135 ip_rsvp_vif = NULL; 3136 3137 legal_vif_num = NULL; 3138 mrt_ioctl = NULL; 3139 rsvp_input_p = NULL; 3140 3141 VIF_LOCK_DESTROY(); 3142 MFC_LOCK_DESTROY(); 3143 MROUTER_LOCK_DESTROY(); 3144 break; 3145 3146 default: 3147 return EOPNOTSUPP; 3148 } 3149 return 0; 3150} 3151 3152static moduledata_t ip_mroutemod = { 3153 "ip_mroute", 3154 ip_mroute_modevent, 3155 0 3156}; 3157DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3158