ip_mroute.c revision 166629
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 166629 2007-02-10 23:15:28Z bms $ 56 */ 57 58#include "opt_mac.h" 59#include "opt_mrouting.h" 60 61#define _PIM_VT 1 62 63#include <sys/param.h> 64#include <sys/kernel.h> 65#include <sys/lock.h> 66#include <sys/malloc.h> 67#include <sys/mbuf.h> 68#include <sys/module.h> 69#include <sys/priv.h> 70#include <sys/protosw.h> 71#include <sys/signalvar.h> 72#include <sys/socket.h> 73#include <sys/socketvar.h> 74#include <sys/sockio.h> 75#include <sys/sx.h> 76#include <sys/sysctl.h> 77#include <sys/syslog.h> 78#include <sys/systm.h> 79#include <sys/time.h> 80#include <net/if.h> 81#include <net/netisr.h> 82#include <net/route.h> 83#include <netinet/in.h> 84#include <netinet/igmp.h> 85#include <netinet/in_systm.h> 86#include <netinet/in_var.h> 87#include <netinet/ip.h> 88#include <netinet/ip_encap.h> 89#include <netinet/ip_mroute.h> 90#include <netinet/ip_var.h> 91#include <netinet/ip_options.h> 92#include <netinet/pim.h> 93#include <netinet/pim_var.h> 94#include <netinet/udp.h> 95#include <machine/in_cksum.h> 96 97#include <security/mac/mac_framework.h> 98 99/* 100 * Control debugging code for rsvp and multicast routing code. 101 * Can only set them with the debugger. 102 */ 103static u_int rsvpdebug; /* non-zero enables debugging */ 104 105static u_int mrtdebug; /* any set of the flags below */ 106#define DEBUG_MFC 0x02 107#define DEBUG_FORWARD 0x04 108#define DEBUG_EXPIRE 0x08 109#define DEBUG_XMIT 0x10 110#define DEBUG_PIM 0x20 111 112#define VIFI_INVALID ((vifi_t) -1) 113 114#define M_HASCL(m) ((m)->m_flags & M_EXT) 115 116static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast routing tables"); 117 118/* 119 * Locking. We use two locks: one for the virtual interface table and 120 * one for the forwarding table. These locks may be nested in which case 121 * the VIF lock must always be taken first. Note that each lock is used 122 * to cover not only the specific data structure but also related data 123 * structures. It may be better to add more fine-grained locking later; 124 * it's not clear how performance-critical this code is. 125 * 126 * XXX: This module could particularly benefit from being cleaned 127 * up to use the <sys/queue.h> macros. 128 * 129 */ 130 131static struct mrtstat mrtstat; 132SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 133 &mrtstat, mrtstat, 134 "Multicast Routing Statistics (struct mrtstat, netinet/ip_mroute.h)"); 135 136static struct mfc *mfctable[MFCTBLSIZ]; 137SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 138 &mfctable, sizeof(mfctable), "S,*mfc[MFCTBLSIZ]", 139 "Multicast Forwarding Table (struct *mfc[MFCTBLSIZ], netinet/ip_mroute.h)"); 140 141static struct mtx mfc_mtx; 142#define MFC_LOCK() mtx_lock(&mfc_mtx) 143#define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 144#define MFC_LOCK_ASSERT() do { \ 145 mtx_assert(&mfc_mtx, MA_OWNED); \ 146 NET_ASSERT_GIANT(); \ 147} while (0) 148#define MFC_LOCK_INIT() mtx_init(&mfc_mtx, "mroute mfc table", NULL, MTX_DEF) 149#define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 150 151static struct vif viftable[MAXVIFS]; 152SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 153 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 154 "Multicast Virtual Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 155 156static struct mtx vif_mtx; 157#define VIF_LOCK() mtx_lock(&vif_mtx) 158#define VIF_UNLOCK() mtx_unlock(&vif_mtx) 159#define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 160#define VIF_LOCK_INIT() mtx_init(&vif_mtx, "mroute vif table", NULL, MTX_DEF) 161#define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 162 163static u_char nexpire[MFCTBLSIZ]; 164 165static eventhandler_tag if_detach_event_tag = NULL; 166 167static struct callout expire_upcalls_ch; 168 169#define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 170#define UPCALL_EXPIRE 6 /* number of timeouts */ 171 172#define ENCAP_TTL 64 173 174/* 175 * Bandwidth meter variables and constants 176 */ 177static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 178/* 179 * Pending timeouts are stored in a hash table, the key being the 180 * expiration time. Periodically, the entries are analysed and processed. 181 */ 182#define BW_METER_BUCKETS 1024 183static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 184static struct callout bw_meter_ch; 185#define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 186 187/* 188 * Pending upcalls are stored in a vector which is flushed when 189 * full, or periodically 190 */ 191static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 192static u_int bw_upcalls_n; /* # of pending upcalls */ 193static struct callout bw_upcalls_ch; 194#define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 195 196static struct pimstat pimstat; 197 198SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); 199SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 200 &pimstat, pimstat, 201 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 202 203static u_long pim_squelch_wholepkt = 0; 204SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, 205 &pim_squelch_wholepkt, 0, 206 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); 207 208extern struct domain inetdomain; 209struct protosw in_pim_protosw = { 210 .pr_type = SOCK_RAW, 211 .pr_domain = &inetdomain, 212 .pr_protocol = IPPROTO_PIM, 213 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, 214 .pr_input = pim_input, 215 .pr_output = (pr_output_t*)rip_output, 216 .pr_ctloutput = rip_ctloutput, 217 .pr_usrreqs = &rip_usrreqs 218}; 219static const struct encaptab *pim_encap_cookie; 220static int pim_encapcheck(const struct mbuf *, int, int, void *); 221 222/* 223 * Note: the PIM Register encapsulation adds the following in front of a 224 * data packet: 225 * 226 * struct pim_encap_hdr { 227 * struct ip ip; 228 * struct pim_encap_pimhdr pim; 229 * } 230 * 231 */ 232 233struct pim_encap_pimhdr { 234 struct pim pim; 235 uint32_t flags; 236}; 237 238static struct ip pim_encap_iphdr = { 239#if BYTE_ORDER == LITTLE_ENDIAN 240 sizeof(struct ip) >> 2, 241 IPVERSION, 242#else 243 IPVERSION, 244 sizeof(struct ip) >> 2, 245#endif 246 0, /* tos */ 247 sizeof(struct ip), /* total length */ 248 0, /* id */ 249 0, /* frag offset */ 250 ENCAP_TTL, 251 IPPROTO_PIM, 252 0, /* checksum */ 253}; 254 255static struct pim_encap_pimhdr pim_encap_pimhdr = { 256 { 257 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 258 0, /* reserved */ 259 0, /* checksum */ 260 }, 261 0 /* flags */ 262}; 263 264static struct ifnet multicast_register_if; 265static vifi_t reg_vif_num = VIFI_INVALID; 266 267/* 268 * Private variables. 269 */ 270static vifi_t numvifs; 271 272static u_long X_ip_mcast_src(int vifi); 273static int X_ip_mforward(struct ip *ip, struct ifnet *ifp, 274 struct mbuf *m, struct ip_moptions *imo); 275static int X_ip_mrouter_done(void); 276static int X_ip_mrouter_get(struct socket *so, struct sockopt *m); 277static int X_ip_mrouter_set(struct socket *so, struct sockopt *m); 278static int X_legal_vif_num(int vif); 279static int X_mrt_ioctl(int cmd, caddr_t data); 280 281static int get_sg_cnt(struct sioc_sg_req *); 282static int get_vif_cnt(struct sioc_vif_req *); 283static void if_detached_event(void *arg __unused, struct ifnet *); 284static int ip_mrouter_init(struct socket *, int); 285static int add_vif(struct vifctl *); 286static int del_vif_locked(vifi_t); 287static int del_vif(vifi_t); 288static int add_mfc(struct mfcctl2 *); 289static int del_mfc(struct mfcctl2 *); 290static int set_api_config(uint32_t *); /* chose API capabilities */ 291static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 292static int set_assert(int); 293static void expire_upcalls(void *); 294static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 295static void phyint_send(struct ip *, struct vif *, struct mbuf *); 296static void send_packet(struct vif *, struct mbuf *); 297 298/* 299 * Bandwidth monitoring 300 */ 301static void free_bw_list(struct bw_meter *list); 302static int add_bw_upcall(struct bw_upcall *); 303static int del_bw_upcall(struct bw_upcall *); 304static void bw_meter_receive_packet(struct bw_meter *x, int plen, 305 struct timeval *nowp); 306static void bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp); 307static void bw_upcalls_send(void); 308static void schedule_bw_meter(struct bw_meter *x, struct timeval *nowp); 309static void unschedule_bw_meter(struct bw_meter *x); 310static void bw_meter_process(void); 311static void expire_bw_upcalls_send(void *); 312static void expire_bw_meter_process(void *); 313 314static int pim_register_send(struct ip *, struct vif *, 315 struct mbuf *, struct mfc *); 316static int pim_register_send_rp(struct ip *, struct vif *, 317 struct mbuf *, struct mfc *); 318static int pim_register_send_upcall(struct ip *, struct vif *, 319 struct mbuf *, struct mfc *); 320static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 321 322/* 323 * whether or not special PIM assert processing is enabled. 324 */ 325static int pim_assert; 326/* 327 * Rate limit for assert notification messages, in usec 328 */ 329#define ASSERT_MSG_TIME 3000000 330 331/* 332 * Kernel multicast routing API capabilities and setup. 333 * If more API capabilities are added to the kernel, they should be 334 * recorded in `mrt_api_support'. 335 */ 336static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 337 MRT_MFC_FLAGS_BORDER_VIF | 338 MRT_MFC_RP | 339 MRT_MFC_BW_UPCALL); 340static uint32_t mrt_api_config = 0; 341 342/* 343 * Hash function for a source, group entry 344 */ 345#define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 346 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 347 348/* 349 * Find a route for a given origin IP address and Multicast group address 350 * Type of service parameter to be added in the future!!! 351 * Statistics are updated by the caller if needed 352 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 353 */ 354static struct mfc * 355mfc_find(in_addr_t o, in_addr_t g) 356{ 357 struct mfc *rt; 358 359 MFC_LOCK_ASSERT(); 360 361 for (rt = mfctable[MFCHASH(o,g)]; rt; rt = rt->mfc_next) 362 if ((rt->mfc_origin.s_addr == o) && 363 (rt->mfc_mcastgrp.s_addr == g) && (rt->mfc_stall == NULL)) 364 break; 365 return rt; 366} 367 368/* 369 * Macros to compute elapsed time efficiently 370 * Borrowed from Van Jacobson's scheduling code 371 */ 372#define TV_DELTA(a, b, delta) { \ 373 int xxs; \ 374 delta = (a).tv_usec - (b).tv_usec; \ 375 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 376 switch (xxs) { \ 377 case 2: \ 378 delta += 1000000; \ 379 /* FALLTHROUGH */ \ 380 case 1: \ 381 delta += 1000000; \ 382 break; \ 383 default: \ 384 delta += (1000000 * xxs); \ 385 } \ 386 } \ 387} 388 389#define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 390 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 391 392/* 393 * Handle MRT setsockopt commands to modify the multicast routing tables. 394 */ 395static int 396X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 397{ 398 int error, optval; 399 vifi_t vifi; 400 struct vifctl vifc; 401 struct mfcctl2 mfc; 402 struct bw_upcall bw_upcall; 403 uint32_t i; 404 405 if (so != ip_mrouter && sopt->sopt_name != MRT_INIT) 406 return EPERM; 407 408 error = 0; 409 switch (sopt->sopt_name) { 410 case MRT_INIT: 411 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 412 if (error) 413 break; 414 error = ip_mrouter_init(so, optval); 415 break; 416 417 case MRT_DONE: 418 error = ip_mrouter_done(); 419 break; 420 421 case MRT_ADD_VIF: 422 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 423 if (error) 424 break; 425 error = add_vif(&vifc); 426 break; 427 428 case MRT_DEL_VIF: 429 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 430 if (error) 431 break; 432 error = del_vif(vifi); 433 break; 434 435 case MRT_ADD_MFC: 436 case MRT_DEL_MFC: 437 /* 438 * select data size depending on API version. 439 */ 440 if (sopt->sopt_name == MRT_ADD_MFC && 441 mrt_api_config & MRT_API_FLAGS_ALL) { 442 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 443 sizeof(struct mfcctl2)); 444 } else { 445 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 446 sizeof(struct mfcctl)); 447 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 448 sizeof(mfc) - sizeof(struct mfcctl)); 449 } 450 if (error) 451 break; 452 if (sopt->sopt_name == MRT_ADD_MFC) 453 error = add_mfc(&mfc); 454 else 455 error = del_mfc(&mfc); 456 break; 457 458 case MRT_ASSERT: 459 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 460 if (error) 461 break; 462 set_assert(optval); 463 break; 464 465 case MRT_API_CONFIG: 466 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 467 if (!error) 468 error = set_api_config(&i); 469 if (!error) 470 error = sooptcopyout(sopt, &i, sizeof i); 471 break; 472 473 case MRT_ADD_BW_UPCALL: 474 case MRT_DEL_BW_UPCALL: 475 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 476 sizeof bw_upcall); 477 if (error) 478 break; 479 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 480 error = add_bw_upcall(&bw_upcall); 481 else 482 error = del_bw_upcall(&bw_upcall); 483 break; 484 485 default: 486 error = EOPNOTSUPP; 487 break; 488 } 489 return error; 490} 491 492/* 493 * Handle MRT getsockopt commands 494 */ 495static int 496X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 497{ 498 int error; 499 static int version = 0x0305; /* !!! why is this here? XXX */ 500 501 switch (sopt->sopt_name) { 502 case MRT_VERSION: 503 error = sooptcopyout(sopt, &version, sizeof version); 504 break; 505 506 case MRT_ASSERT: 507 error = sooptcopyout(sopt, &pim_assert, sizeof pim_assert); 508 break; 509 510 case MRT_API_SUPPORT: 511 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 512 break; 513 514 case MRT_API_CONFIG: 515 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 516 break; 517 518 default: 519 error = EOPNOTSUPP; 520 break; 521 } 522 return error; 523} 524 525/* 526 * Handle ioctl commands to obtain information from the cache 527 */ 528static int 529X_mrt_ioctl(int cmd, caddr_t data) 530{ 531 int error = 0; 532 533 /* 534 * Currently the only function calling this ioctl routine is rtioctl(). 535 * Typically, only root can create the raw socket in order to execute 536 * this ioctl method, however the request might be coming from a prison 537 */ 538 error = priv_check(curthread, PRIV_NETINET_MROUTE); 539 if (error) 540 return (error); 541 switch (cmd) { 542 case (SIOCGETVIFCNT): 543 error = get_vif_cnt((struct sioc_vif_req *)data); 544 break; 545 546 case (SIOCGETSGCNT): 547 error = get_sg_cnt((struct sioc_sg_req *)data); 548 break; 549 550 default: 551 error = EINVAL; 552 break; 553 } 554 return error; 555} 556 557/* 558 * returns the packet, byte, rpf-failure count for the source group provided 559 */ 560static int 561get_sg_cnt(struct sioc_sg_req *req) 562{ 563 struct mfc *rt; 564 565 MFC_LOCK(); 566 rt = mfc_find(req->src.s_addr, req->grp.s_addr); 567 if (rt == NULL) { 568 MFC_UNLOCK(); 569 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 570 return EADDRNOTAVAIL; 571 } 572 req->pktcnt = rt->mfc_pkt_cnt; 573 req->bytecnt = rt->mfc_byte_cnt; 574 req->wrong_if = rt->mfc_wrong_if; 575 MFC_UNLOCK(); 576 return 0; 577} 578 579/* 580 * returns the input and output packet and byte counts on the vif provided 581 */ 582static int 583get_vif_cnt(struct sioc_vif_req *req) 584{ 585 vifi_t vifi = req->vifi; 586 587 VIF_LOCK(); 588 if (vifi >= numvifs) { 589 VIF_UNLOCK(); 590 return EINVAL; 591 } 592 593 req->icount = viftable[vifi].v_pkt_in; 594 req->ocount = viftable[vifi].v_pkt_out; 595 req->ibytes = viftable[vifi].v_bytes_in; 596 req->obytes = viftable[vifi].v_bytes_out; 597 VIF_UNLOCK(); 598 599 return 0; 600} 601 602static void 603ip_mrouter_reset(void) 604{ 605 bzero((caddr_t)mfctable, sizeof(mfctable)); 606 bzero((caddr_t)nexpire, sizeof(nexpire)); 607 608 pim_assert = 0; 609 mrt_api_config = 0; 610 611 callout_init(&expire_upcalls_ch, NET_CALLOUT_MPSAFE); 612 613 bw_upcalls_n = 0; 614 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 615 callout_init(&bw_upcalls_ch, NET_CALLOUT_MPSAFE); 616 callout_init(&bw_meter_ch, NET_CALLOUT_MPSAFE); 617} 618 619static struct mtx mrouter_mtx; 620 621static void 622if_detached_event(void *arg __unused, struct ifnet *ifp) 623{ 624 vifi_t vifi; 625 int i; 626 struct mfc *mfc; 627 struct mfc *nmfc; 628 struct mfc **ppmfc; /* Pointer to previous node's next-pointer */ 629 struct rtdetq *pq; 630 struct rtdetq *npq; 631 632 mtx_lock(&mrouter_mtx); 633 if (ip_mrouter == NULL) { 634 mtx_unlock(&mrouter_mtx); 635 } 636 637 /* 638 * Tear down multicast forwarder state associated with this ifnet. 639 * 1. Walk the vif list, matching vifs against this ifnet. 640 * 2. Walk the multicast forwarding cache (mfc) looking for 641 * inner matches with this vif's index. 642 * 3. Free any pending mbufs for this mfc. 643 * 4. Free the associated mfc entry and state associated with this vif. 644 * Be very careful about unlinking from a singly-linked list whose 645 * "head node" is a pointer in a simple array. 646 * 5. Free vif state. This should disable ALLMULTI on the interface. 647 */ 648 VIF_LOCK(); 649 MFC_LOCK(); 650 for (vifi = 0; vifi < numvifs; vifi++) { 651 if (viftable[vifi].v_ifp != ifp) 652 continue; 653 for (i = 0; i < MFCTBLSIZ; i++) { 654 ppmfc = &mfctable[i]; 655 for (mfc = mfctable[i]; mfc != NULL; ) { 656 nmfc = mfc->mfc_next; 657 if (mfc->mfc_parent == vifi) { 658 for (pq = mfc->mfc_stall; pq != NULL; ) { 659 npq = pq->next; 660 m_freem(pq->m); 661 free(pq, M_MRTABLE); 662 pq = npq; 663 } 664 free_bw_list(mfc->mfc_bw_meter); 665 free(mfc, M_MRTABLE); 666 *ppmfc = nmfc; 667 } else { 668 ppmfc = &mfc->mfc_next; 669 } 670 mfc = nmfc; 671 } 672 } 673 del_vif_locked(vifi); 674 } 675 MFC_UNLOCK(); 676 VIF_UNLOCK(); 677 678 mtx_unlock(&mrouter_mtx); 679} 680 681/* 682 * Enable multicast routing 683 */ 684static int 685ip_mrouter_init(struct socket *so, int version) 686{ 687 if (mrtdebug) 688 log(LOG_DEBUG, "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 689 so->so_type, so->so_proto->pr_protocol); 690 691 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 692 return EOPNOTSUPP; 693 694 if (version != 1) 695 return ENOPROTOOPT; 696 697 mtx_lock(&mrouter_mtx); 698 699 if (ip_mrouter != NULL) { 700 mtx_unlock(&mrouter_mtx); 701 return EADDRINUSE; 702 } 703 704 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 705 if_detached_event, NULL, EVENTHANDLER_PRI_ANY); 706 if (if_detach_event_tag == NULL) 707 return (ENOMEM); 708 709 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 710 711 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 712 expire_bw_upcalls_send, NULL); 713 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 714 715 ip_mrouter = so; 716 717 mtx_unlock(&mrouter_mtx); 718 719 if (mrtdebug) 720 log(LOG_DEBUG, "ip_mrouter_init\n"); 721 722 return 0; 723} 724 725/* 726 * Disable multicast routing 727 */ 728static int 729X_ip_mrouter_done(void) 730{ 731 vifi_t vifi; 732 int i; 733 struct ifnet *ifp; 734 struct ifreq ifr; 735 struct mfc *rt; 736 struct rtdetq *rte; 737 738 mtx_lock(&mrouter_mtx); 739 740 if (ip_mrouter == NULL) { 741 mtx_unlock(&mrouter_mtx); 742 return EINVAL; 743 } 744 745 /* 746 * Detach/disable hooks to the reset of the system. 747 */ 748 ip_mrouter = NULL; 749 mrt_api_config = 0; 750 751 VIF_LOCK(); 752 /* 753 * For each phyint in use, disable promiscuous reception of all IP 754 * multicasts. 755 */ 756 for (vifi = 0; vifi < numvifs; vifi++) { 757 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 758 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 759 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 760 761 so->sin_len = sizeof(struct sockaddr_in); 762 so->sin_family = AF_INET; 763 so->sin_addr.s_addr = INADDR_ANY; 764 ifp = viftable[vifi].v_ifp; 765 if_allmulti(ifp, 0); 766 } 767 } 768 bzero((caddr_t)viftable, sizeof(viftable)); 769 numvifs = 0; 770 pim_assert = 0; 771 VIF_UNLOCK(); 772 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 773 774 /* 775 * Free all multicast forwarding cache entries. 776 */ 777 callout_stop(&expire_upcalls_ch); 778 callout_stop(&bw_upcalls_ch); 779 callout_stop(&bw_meter_ch); 780 781 MFC_LOCK(); 782 for (i = 0; i < MFCTBLSIZ; i++) { 783 for (rt = mfctable[i]; rt != NULL; ) { 784 struct mfc *nr = rt->mfc_next; 785 786 for (rte = rt->mfc_stall; rte != NULL; ) { 787 struct rtdetq *n = rte->next; 788 789 m_freem(rte->m); 790 free(rte, M_MRTABLE); 791 rte = n; 792 } 793 free_bw_list(rt->mfc_bw_meter); 794 free(rt, M_MRTABLE); 795 rt = nr; 796 } 797 } 798 bzero((caddr_t)mfctable, sizeof(mfctable)); 799 bzero((caddr_t)nexpire, sizeof(nexpire)); 800 bw_upcalls_n = 0; 801 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 802 MFC_UNLOCK(); 803 804 reg_vif_num = VIFI_INVALID; 805 806 mtx_unlock(&mrouter_mtx); 807 808 if (mrtdebug) 809 log(LOG_DEBUG, "ip_mrouter_done\n"); 810 811 return 0; 812} 813 814/* 815 * Set PIM assert processing global 816 */ 817static int 818set_assert(int i) 819{ 820 if ((i != 1) && (i != 0)) 821 return EINVAL; 822 823 pim_assert = i; 824 825 return 0; 826} 827 828/* 829 * Configure API capabilities 830 */ 831int 832set_api_config(uint32_t *apival) 833{ 834 int i; 835 836 /* 837 * We can set the API capabilities only if it is the first operation 838 * after MRT_INIT. I.e.: 839 * - there are no vifs installed 840 * - pim_assert is not enabled 841 * - the MFC table is empty 842 */ 843 if (numvifs > 0) { 844 *apival = 0; 845 return EPERM; 846 } 847 if (pim_assert) { 848 *apival = 0; 849 return EPERM; 850 } 851 for (i = 0; i < MFCTBLSIZ; i++) { 852 if (mfctable[i] != NULL) { 853 *apival = 0; 854 return EPERM; 855 } 856 } 857 858 mrt_api_config = *apival & mrt_api_support; 859 *apival = mrt_api_config; 860 861 return 0; 862} 863 864/* 865 * Add a vif to the vif table 866 */ 867static int 868add_vif(struct vifctl *vifcp) 869{ 870 struct vif *vifp = viftable + vifcp->vifc_vifi; 871 struct sockaddr_in sin = {sizeof sin, AF_INET}; 872 struct ifaddr *ifa; 873 struct ifnet *ifp; 874 int error; 875 876 VIF_LOCK(); 877 if (vifcp->vifc_vifi >= MAXVIFS) { 878 VIF_UNLOCK(); 879 return EINVAL; 880 } 881 /* rate limiting is no longer supported by this code */ 882 if (vifcp->vifc_rate_limit != 0) { 883 log(LOG_ERR, "rate limiting is no longer supported\n"); 884 VIF_UNLOCK(); 885 return EINVAL; 886 } 887 if (vifp->v_lcl_addr.s_addr != INADDR_ANY) { 888 VIF_UNLOCK(); 889 return EADDRINUSE; 890 } 891 if (vifcp->vifc_lcl_addr.s_addr == INADDR_ANY) { 892 VIF_UNLOCK(); 893 return EADDRNOTAVAIL; 894 } 895 896 /* Find the interface with an address in AF_INET family */ 897 if (vifcp->vifc_flags & VIFF_REGISTER) { 898 /* 899 * XXX: Because VIFF_REGISTER does not really need a valid 900 * local interface (e.g. it could be 127.0.0.2), we don't 901 * check its address. 902 */ 903 ifp = NULL; 904 } else { 905 sin.sin_addr = vifcp->vifc_lcl_addr; 906 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 907 if (ifa == NULL) { 908 VIF_UNLOCK(); 909 return EADDRNOTAVAIL; 910 } 911 ifp = ifa->ifa_ifp; 912 } 913 914 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { 915 log(LOG_ERR, "tunnels are no longer supported\n"); 916 VIF_UNLOCK(); 917 return EOPNOTSUPP; 918 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 919 ifp = &multicast_register_if; 920 if (mrtdebug) 921 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 922 (void *)&multicast_register_if); 923 if (reg_vif_num == VIFI_INVALID) { 924 if_initname(&multicast_register_if, "register_vif", 0); 925 multicast_register_if.if_flags = IFF_LOOPBACK; 926 reg_vif_num = vifcp->vifc_vifi; 927 } 928 } else { /* Make sure the interface supports multicast */ 929 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 930 VIF_UNLOCK(); 931 return EOPNOTSUPP; 932 } 933 934 /* Enable promiscuous reception of all IP multicasts from the if */ 935 error = if_allmulti(ifp, 1); 936 if (error) { 937 VIF_UNLOCK(); 938 return error; 939 } 940 } 941 942 vifp->v_flags = vifcp->vifc_flags; 943 vifp->v_threshold = vifcp->vifc_threshold; 944 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 945 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 946 vifp->v_ifp = ifp; 947 vifp->v_rsvp_on = 0; 948 vifp->v_rsvpd = NULL; 949 /* initialize per vif pkt counters */ 950 vifp->v_pkt_in = 0; 951 vifp->v_pkt_out = 0; 952 vifp->v_bytes_in = 0; 953 vifp->v_bytes_out = 0; 954 bzero(&vifp->v_route, sizeof(vifp->v_route)); 955 956 /* Adjust numvifs up if the vifi is higher than numvifs */ 957 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 958 959 VIF_UNLOCK(); 960 961 if (mrtdebug) 962 log(LOG_DEBUG, "add_vif #%d, lcladdr %lx, %s %lx, thresh %x\n", 963 vifcp->vifc_vifi, 964 (u_long)ntohl(vifcp->vifc_lcl_addr.s_addr), 965 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 966 (u_long)ntohl(vifcp->vifc_rmt_addr.s_addr), 967 vifcp->vifc_threshold); 968 969 return 0; 970} 971 972/* 973 * Delete a vif from the vif table 974 */ 975static int 976del_vif_locked(vifi_t vifi) 977{ 978 struct vif *vifp; 979 980 VIF_LOCK_ASSERT(); 981 982 if (vifi >= numvifs) { 983 return EINVAL; 984 } 985 vifp = &viftable[vifi]; 986 if (vifp->v_lcl_addr.s_addr == INADDR_ANY) { 987 return EADDRNOTAVAIL; 988 } 989 990 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 991 if_allmulti(vifp->v_ifp, 0); 992 993 if (vifp->v_flags & VIFF_REGISTER) 994 reg_vif_num = VIFI_INVALID; 995 996 bzero((caddr_t)vifp, sizeof (*vifp)); 997 998 if (mrtdebug) 999 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", vifi, numvifs); 1000 1001 /* Adjust numvifs down */ 1002 for (vifi = numvifs; vifi > 0; vifi--) 1003 if (viftable[vifi-1].v_lcl_addr.s_addr != INADDR_ANY) 1004 break; 1005 numvifs = vifi; 1006 1007 return 0; 1008} 1009 1010static int 1011del_vif(vifi_t vifi) 1012{ 1013 int cc; 1014 1015 VIF_LOCK(); 1016 cc = del_vif_locked(vifi); 1017 VIF_UNLOCK(); 1018 1019 return cc; 1020} 1021 1022/* 1023 * update an mfc entry without resetting counters and S,G addresses. 1024 */ 1025static void 1026update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1027{ 1028 int i; 1029 1030 rt->mfc_parent = mfccp->mfcc_parent; 1031 for (i = 0; i < numvifs; i++) { 1032 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1033 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1034 MRT_MFC_FLAGS_ALL; 1035 } 1036 /* set the RP address */ 1037 if (mrt_api_config & MRT_MFC_RP) 1038 rt->mfc_rp = mfccp->mfcc_rp; 1039 else 1040 rt->mfc_rp.s_addr = INADDR_ANY; 1041} 1042 1043/* 1044 * fully initialize an mfc entry from the parameter. 1045 */ 1046static void 1047init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1048{ 1049 rt->mfc_origin = mfccp->mfcc_origin; 1050 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1051 1052 update_mfc_params(rt, mfccp); 1053 1054 /* initialize pkt counters per src-grp */ 1055 rt->mfc_pkt_cnt = 0; 1056 rt->mfc_byte_cnt = 0; 1057 rt->mfc_wrong_if = 0; 1058 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 1059} 1060 1061 1062/* 1063 * Add an mfc entry 1064 */ 1065static int 1066add_mfc(struct mfcctl2 *mfccp) 1067{ 1068 struct mfc *rt; 1069 u_long hash; 1070 struct rtdetq *rte; 1071 u_short nstl; 1072 1073 VIF_LOCK(); 1074 MFC_LOCK(); 1075 1076 rt = mfc_find(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1077 1078 /* If an entry already exists, just update the fields */ 1079 if (rt) { 1080 if (mrtdebug & DEBUG_MFC) 1081 log(LOG_DEBUG,"add_mfc update o %lx g %lx p %x\n", 1082 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1083 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1084 mfccp->mfcc_parent); 1085 1086 update_mfc_params(rt, mfccp); 1087 MFC_UNLOCK(); 1088 VIF_UNLOCK(); 1089 return 0; 1090 } 1091 1092 /* 1093 * Find the entry for which the upcall was made and update 1094 */ 1095 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 1096 for (rt = mfctable[hash], nstl = 0; rt; rt = rt->mfc_next) { 1097 1098 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1099 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 1100 (rt->mfc_stall != NULL)) { 1101 1102 if (nstl++) 1103 log(LOG_ERR, "add_mfc %s o %lx g %lx p %x dbx %p\n", 1104 "multiple kernel entries", 1105 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1106 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1107 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1108 1109 if (mrtdebug & DEBUG_MFC) 1110 log(LOG_DEBUG,"add_mfc o %lx g %lx p %x dbg %p\n", 1111 (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1112 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1113 mfccp->mfcc_parent, (void *)rt->mfc_stall); 1114 1115 init_mfc_params(rt, mfccp); 1116 1117 rt->mfc_expire = 0; /* Don't clean this guy up */ 1118 nexpire[hash]--; 1119 1120 /* free packets Qed at the end of this entry */ 1121 for (rte = rt->mfc_stall; rte != NULL; ) { 1122 struct rtdetq *n = rte->next; 1123 1124 ip_mdq(rte->m, rte->ifp, rt, -1); 1125 m_freem(rte->m); 1126 free(rte, M_MRTABLE); 1127 rte = n; 1128 } 1129 rt->mfc_stall = NULL; 1130 } 1131 } 1132 1133 /* 1134 * It is possible that an entry is being inserted without an upcall 1135 */ 1136 if (nstl == 0) { 1137 if (mrtdebug & DEBUG_MFC) 1138 log(LOG_DEBUG,"add_mfc no upcall h %lu o %lx g %lx p %x\n", 1139 hash, (u_long)ntohl(mfccp->mfcc_origin.s_addr), 1140 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1141 mfccp->mfcc_parent); 1142 1143 for (rt = mfctable[hash]; rt != NULL; rt = rt->mfc_next) { 1144 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 1145 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 1146 init_mfc_params(rt, mfccp); 1147 if (rt->mfc_expire) 1148 nexpire[hash]--; 1149 rt->mfc_expire = 0; 1150 break; /* XXX */ 1151 } 1152 } 1153 if (rt == NULL) { /* no upcall, so make a new entry */ 1154 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1155 if (rt == NULL) { 1156 MFC_UNLOCK(); 1157 VIF_UNLOCK(); 1158 return ENOBUFS; 1159 } 1160 1161 init_mfc_params(rt, mfccp); 1162 rt->mfc_expire = 0; 1163 rt->mfc_stall = NULL; 1164 1165 rt->mfc_bw_meter = NULL; 1166 /* insert new entry at head of hash chain */ 1167 rt->mfc_next = mfctable[hash]; 1168 mfctable[hash] = rt; 1169 } 1170 } 1171 MFC_UNLOCK(); 1172 VIF_UNLOCK(); 1173 return 0; 1174} 1175 1176/* 1177 * Delete an mfc entry 1178 */ 1179static int 1180del_mfc(struct mfcctl2 *mfccp) 1181{ 1182 struct in_addr origin; 1183 struct in_addr mcastgrp; 1184 struct mfc *rt; 1185 struct mfc **nptr; 1186 u_long hash; 1187 struct bw_meter *list; 1188 1189 origin = mfccp->mfcc_origin; 1190 mcastgrp = mfccp->mfcc_mcastgrp; 1191 1192 if (mrtdebug & DEBUG_MFC) 1193 log(LOG_DEBUG,"del_mfc orig %lx mcastgrp %lx\n", 1194 (u_long)ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1195 1196 MFC_LOCK(); 1197 1198 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1199 for (nptr = &mfctable[hash]; (rt = *nptr) != NULL; nptr = &rt->mfc_next) 1200 if (origin.s_addr == rt->mfc_origin.s_addr && 1201 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1202 rt->mfc_stall == NULL) 1203 break; 1204 if (rt == NULL) { 1205 MFC_UNLOCK(); 1206 return EADDRNOTAVAIL; 1207 } 1208 1209 *nptr = rt->mfc_next; 1210 1211 /* 1212 * free the bw_meter entries 1213 */ 1214 list = rt->mfc_bw_meter; 1215 rt->mfc_bw_meter = NULL; 1216 1217 free(rt, M_MRTABLE); 1218 1219 free_bw_list(list); 1220 1221 MFC_UNLOCK(); 1222 1223 return 0; 1224} 1225 1226/* 1227 * Send a message to the routing daemon on the multicast routing socket 1228 */ 1229static int 1230socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1231{ 1232 if (s) { 1233 SOCKBUF_LOCK(&s->so_rcv); 1234 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, 1235 NULL) != 0) { 1236 sorwakeup_locked(s); 1237 return 0; 1238 } 1239 SOCKBUF_UNLOCK(&s->so_rcv); 1240 } 1241 m_freem(mm); 1242 return -1; 1243} 1244 1245/* 1246 * IP multicast forwarding function. This function assumes that the packet 1247 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1248 * pointed to by "ifp", and the packet is to be relayed to other networks 1249 * that have members of the packet's destination IP multicast group. 1250 * 1251 * The packet is returned unscathed to the caller, unless it is 1252 * erroneous, in which case a non-zero return value tells the caller to 1253 * discard it. 1254 */ 1255 1256#define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1257 1258static int 1259X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1260 struct ip_moptions *imo) 1261{ 1262 struct mfc *rt; 1263 int error; 1264 vifi_t vifi; 1265 1266 if (mrtdebug & DEBUG_FORWARD) 1267 log(LOG_DEBUG, "ip_mforward: src %lx, dst %lx, ifp %p\n", 1268 (u_long)ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), 1269 (void *)ifp); 1270 1271 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1272 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1273 /* 1274 * Packet arrived via a physical interface or 1275 * an encapsulated tunnel or a register_vif. 1276 */ 1277 } else { 1278 /* 1279 * Packet arrived through a source-route tunnel. 1280 * Source-route tunnels are no longer supported. 1281 */ 1282 static int last_log; 1283 if (last_log != time_uptime) { 1284 last_log = time_uptime; 1285 log(LOG_ERR, 1286 "ip_mforward: received source-routed packet from %lx\n", 1287 (u_long)ntohl(ip->ip_src.s_addr)); 1288 } 1289 return 1; 1290 } 1291 1292 VIF_LOCK(); 1293 MFC_LOCK(); 1294 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1295 if (ip->ip_ttl < MAXTTL) 1296 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1297 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1298 struct vif *vifp = viftable + vifi; 1299 1300 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s)\n", 1301 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr), 1302 vifi, 1303 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1304 vifp->v_ifp->if_xname); 1305 } 1306 error = ip_mdq(m, ifp, NULL, vifi); 1307 MFC_UNLOCK(); 1308 VIF_UNLOCK(); 1309 return error; 1310 } 1311 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1312 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1313 (long)ntohl(ip->ip_src.s_addr), (long)ntohl(ip->ip_dst.s_addr)); 1314 if (!imo) 1315 printf("In fact, no options were specified at all\n"); 1316 } 1317 1318 /* 1319 * Don't forward a packet with time-to-live of zero or one, 1320 * or a packet destined to a local-only group. 1321 */ 1322 if (ip->ip_ttl <= 1 || ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) { 1323 MFC_UNLOCK(); 1324 VIF_UNLOCK(); 1325 return 0; 1326 } 1327 1328 /* 1329 * Determine forwarding vifs from the forwarding cache table 1330 */ 1331 ++mrtstat.mrts_mfc_lookups; 1332 rt = mfc_find(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1333 1334 /* Entry exists, so forward if necessary */ 1335 if (rt != NULL) { 1336 error = ip_mdq(m, ifp, rt, -1); 1337 MFC_UNLOCK(); 1338 VIF_UNLOCK(); 1339 return error; 1340 } else { 1341 /* 1342 * If we don't have a route for packet's origin, 1343 * Make a copy of the packet & send message to routing daemon 1344 */ 1345 1346 struct mbuf *mb0; 1347 struct rtdetq *rte; 1348 u_long hash; 1349 int hlen = ip->ip_hl << 2; 1350 1351 ++mrtstat.mrts_mfc_misses; 1352 1353 mrtstat.mrts_no_route++; 1354 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1355 log(LOG_DEBUG, "ip_mforward: no rte s %lx g %lx\n", 1356 (u_long)ntohl(ip->ip_src.s_addr), 1357 (u_long)ntohl(ip->ip_dst.s_addr)); 1358 1359 /* 1360 * Allocate mbufs early so that we don't do extra work if we are 1361 * just going to fail anyway. Make sure to pullup the header so 1362 * that other people can't step on it. 1363 */ 1364 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, M_NOWAIT); 1365 if (rte == NULL) { 1366 MFC_UNLOCK(); 1367 VIF_UNLOCK(); 1368 return ENOBUFS; 1369 } 1370 mb0 = m_copypacket(m, M_DONTWAIT); 1371 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1372 mb0 = m_pullup(mb0, hlen); 1373 if (mb0 == NULL) { 1374 free(rte, M_MRTABLE); 1375 MFC_UNLOCK(); 1376 VIF_UNLOCK(); 1377 return ENOBUFS; 1378 } 1379 1380 /* is there an upcall waiting for this flow ? */ 1381 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1382 for (rt = mfctable[hash]; rt; rt = rt->mfc_next) { 1383 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1384 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1385 (rt->mfc_stall != NULL)) 1386 break; 1387 } 1388 1389 if (rt == NULL) { 1390 int i; 1391 struct igmpmsg *im; 1392 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1393 struct mbuf *mm; 1394 1395 /* 1396 * Locate the vifi for the incoming interface for this packet. 1397 * If none found, drop packet. 1398 */ 1399 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1400 ; 1401 if (vifi >= numvifs) /* vif not found, drop packet */ 1402 goto non_fatal; 1403 1404 /* no upcall, so make a new entry */ 1405 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1406 if (rt == NULL) 1407 goto fail; 1408 /* Make a copy of the header to send to the user level process */ 1409 mm = m_copy(mb0, 0, hlen); 1410 if (mm == NULL) 1411 goto fail1; 1412 1413 /* 1414 * Send message to routing daemon to install 1415 * a route into the kernel table 1416 */ 1417 1418 im = mtod(mm, struct igmpmsg *); 1419 im->im_msgtype = IGMPMSG_NOCACHE; 1420 im->im_mbz = 0; 1421 im->im_vif = vifi; 1422 1423 mrtstat.mrts_upcalls++; 1424 1425 k_igmpsrc.sin_addr = ip->ip_src; 1426 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1427 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1428 ++mrtstat.mrts_upq_sockfull; 1429fail1: 1430 free(rt, M_MRTABLE); 1431fail: 1432 free(rte, M_MRTABLE); 1433 m_freem(mb0); 1434 MFC_UNLOCK(); 1435 VIF_UNLOCK(); 1436 return ENOBUFS; 1437 } 1438 1439 /* insert new entry at head of hash chain */ 1440 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1441 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1442 rt->mfc_expire = UPCALL_EXPIRE; 1443 nexpire[hash]++; 1444 for (i = 0; i < numvifs; i++) { 1445 rt->mfc_ttls[i] = 0; 1446 rt->mfc_flags[i] = 0; 1447 } 1448 rt->mfc_parent = -1; 1449 1450 rt->mfc_rp.s_addr = INADDR_ANY; /* clear the RP address */ 1451 1452 rt->mfc_bw_meter = NULL; 1453 1454 /* link into table */ 1455 rt->mfc_next = mfctable[hash]; 1456 mfctable[hash] = rt; 1457 rt->mfc_stall = rte; 1458 1459 } else { 1460 /* determine if q has overflowed */ 1461 int npkts = 0; 1462 struct rtdetq **p; 1463 1464 /* 1465 * XXX ouch! we need to append to the list, but we 1466 * only have a pointer to the front, so we have to 1467 * scan the entire list every time. 1468 */ 1469 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1470 npkts++; 1471 1472 if (npkts > MAX_UPQ) { 1473 mrtstat.mrts_upq_ovflw++; 1474non_fatal: 1475 free(rte, M_MRTABLE); 1476 m_freem(mb0); 1477 MFC_UNLOCK(); 1478 VIF_UNLOCK(); 1479 return 0; 1480 } 1481 1482 /* Add this entry to the end of the queue */ 1483 *p = rte; 1484 } 1485 1486 rte->m = mb0; 1487 rte->ifp = ifp; 1488 rte->next = NULL; 1489 1490 MFC_UNLOCK(); 1491 VIF_UNLOCK(); 1492 1493 return 0; 1494 } 1495} 1496 1497/* 1498 * Clean up the cache entry if upcall is not serviced 1499 */ 1500static void 1501expire_upcalls(void *unused) 1502{ 1503 struct rtdetq *rte; 1504 struct mfc *mfc, **nptr; 1505 int i; 1506 1507 MFC_LOCK(); 1508 for (i = 0; i < MFCTBLSIZ; i++) { 1509 if (nexpire[i] == 0) 1510 continue; 1511 nptr = &mfctable[i]; 1512 for (mfc = *nptr; mfc != NULL; mfc = *nptr) { 1513 /* 1514 * Skip real cache entries 1515 * Make sure it wasn't marked to not expire (shouldn't happen) 1516 * If it expires now 1517 */ 1518 if (mfc->mfc_stall != NULL && mfc->mfc_expire != 0 && 1519 --mfc->mfc_expire == 0) { 1520 if (mrtdebug & DEBUG_EXPIRE) 1521 log(LOG_DEBUG, "expire_upcalls: expiring (%lx %lx)\n", 1522 (u_long)ntohl(mfc->mfc_origin.s_addr), 1523 (u_long)ntohl(mfc->mfc_mcastgrp.s_addr)); 1524 /* 1525 * drop all the packets 1526 * free the mbuf with the pkt, if, timing info 1527 */ 1528 for (rte = mfc->mfc_stall; rte; ) { 1529 struct rtdetq *n = rte->next; 1530 1531 m_freem(rte->m); 1532 free(rte, M_MRTABLE); 1533 rte = n; 1534 } 1535 ++mrtstat.mrts_cache_cleanups; 1536 nexpire[i]--; 1537 1538 /* 1539 * free the bw_meter entries 1540 */ 1541 while (mfc->mfc_bw_meter != NULL) { 1542 struct bw_meter *x = mfc->mfc_bw_meter; 1543 1544 mfc->mfc_bw_meter = x->bm_mfc_next; 1545 free(x, M_BWMETER); 1546 } 1547 1548 *nptr = mfc->mfc_next; 1549 free(mfc, M_MRTABLE); 1550 } else { 1551 nptr = &mfc->mfc_next; 1552 } 1553 } 1554 } 1555 MFC_UNLOCK(); 1556 1557 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1558} 1559 1560/* 1561 * Packet forwarding routine once entry in the cache is made 1562 */ 1563static int 1564ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1565{ 1566 struct ip *ip = mtod(m, struct ip *); 1567 vifi_t vifi; 1568 int plen = ip->ip_len; 1569 1570 VIF_LOCK_ASSERT(); 1571 1572 /* 1573 * If xmt_vif is not -1, send on only the requested vif. 1574 * 1575 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1576 */ 1577 if (xmt_vif < numvifs) { 1578 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1579 pim_register_send(ip, viftable + xmt_vif, m, rt); 1580 else 1581 phyint_send(ip, viftable + xmt_vif, m); 1582 return 1; 1583 } 1584 1585 /* 1586 * Don't forward if it didn't arrive from the parent vif for its origin. 1587 */ 1588 vifi = rt->mfc_parent; 1589 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1590 /* came in the wrong interface */ 1591 if (mrtdebug & DEBUG_FORWARD) 1592 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1593 (void *)ifp, vifi, (void *)viftable[vifi].v_ifp); 1594 ++mrtstat.mrts_wrong_if; 1595 ++rt->mfc_wrong_if; 1596 /* 1597 * If we are doing PIM assert processing, send a message 1598 * to the routing daemon. 1599 * 1600 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1601 * can complete the SPT switch, regardless of the type 1602 * of the iif (broadcast media, GRE tunnel, etc). 1603 */ 1604 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1605 struct timeval now; 1606 u_long delta; 1607 1608 if (ifp == &multicast_register_if) 1609 pimstat.pims_rcv_registers_wrongiif++; 1610 1611 /* Get vifi for the incoming packet */ 1612 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1613 ; 1614 if (vifi >= numvifs) 1615 return 0; /* The iif is not found: ignore the packet. */ 1616 1617 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1618 return 0; /* WRONGVIF disabled: ignore the packet */ 1619 1620 GET_TIME(now); 1621 1622 TV_DELTA(now, rt->mfc_last_assert, delta); 1623 1624 if (delta > ASSERT_MSG_TIME) { 1625 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1626 struct igmpmsg *im; 1627 int hlen = ip->ip_hl << 2; 1628 struct mbuf *mm = m_copy(m, 0, hlen); 1629 1630 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1631 mm = m_pullup(mm, hlen); 1632 if (mm == NULL) 1633 return ENOBUFS; 1634 1635 rt->mfc_last_assert = now; 1636 1637 im = mtod(mm, struct igmpmsg *); 1638 im->im_msgtype = IGMPMSG_WRONGVIF; 1639 im->im_mbz = 0; 1640 im->im_vif = vifi; 1641 1642 mrtstat.mrts_upcalls++; 1643 1644 k_igmpsrc.sin_addr = im->im_src; 1645 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1646 log(LOG_WARNING, 1647 "ip_mforward: ip_mrouter socket queue full\n"); 1648 ++mrtstat.mrts_upq_sockfull; 1649 return ENOBUFS; 1650 } 1651 } 1652 } 1653 return 0; 1654 } 1655 1656 /* If I sourced this packet, it counts as output, else it was input. */ 1657 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1658 viftable[vifi].v_pkt_out++; 1659 viftable[vifi].v_bytes_out += plen; 1660 } else { 1661 viftable[vifi].v_pkt_in++; 1662 viftable[vifi].v_bytes_in += plen; 1663 } 1664 rt->mfc_pkt_cnt++; 1665 rt->mfc_byte_cnt += plen; 1666 1667 /* 1668 * For each vif, decide if a copy of the packet should be forwarded. 1669 * Forward if: 1670 * - the ttl exceeds the vif's threshold 1671 * - there are group members downstream on interface 1672 */ 1673 for (vifi = 0; vifi < numvifs; vifi++) 1674 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1675 viftable[vifi].v_pkt_out++; 1676 viftable[vifi].v_bytes_out += plen; 1677 if (viftable[vifi].v_flags & VIFF_REGISTER) 1678 pim_register_send(ip, viftable + vifi, m, rt); 1679 else 1680 phyint_send(ip, viftable + vifi, m); 1681 } 1682 1683 /* 1684 * Perform upcall-related bw measuring. 1685 */ 1686 if (rt->mfc_bw_meter != NULL) { 1687 struct bw_meter *x; 1688 struct timeval now; 1689 1690 GET_TIME(now); 1691 MFC_LOCK_ASSERT(); 1692 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1693 bw_meter_receive_packet(x, plen, &now); 1694 } 1695 1696 return 0; 1697} 1698 1699/* 1700 * check if a vif number is legal/ok. This is used by ip_output. 1701 */ 1702static int 1703X_legal_vif_num(int vif) 1704{ 1705 /* XXX unlocked, matter? */ 1706 return (vif >= 0 && vif < numvifs); 1707} 1708 1709/* 1710 * Return the local address used by this vif 1711 */ 1712static u_long 1713X_ip_mcast_src(int vifi) 1714{ 1715 /* XXX unlocked, matter? */ 1716 if (vifi >= 0 && vifi < numvifs) 1717 return viftable[vifi].v_lcl_addr.s_addr; 1718 else 1719 return INADDR_ANY; 1720} 1721 1722static void 1723phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1724{ 1725 struct mbuf *mb_copy; 1726 int hlen = ip->ip_hl << 2; 1727 1728 VIF_LOCK_ASSERT(); 1729 1730 /* 1731 * Make a new reference to the packet; make sure that 1732 * the IP header is actually copied, not just referenced, 1733 * so that ip_output() only scribbles on the copy. 1734 */ 1735 mb_copy = m_copypacket(m, M_DONTWAIT); 1736 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1737 mb_copy = m_pullup(mb_copy, hlen); 1738 if (mb_copy == NULL) 1739 return; 1740 1741 send_packet(vifp, mb_copy); 1742} 1743 1744static void 1745send_packet(struct vif *vifp, struct mbuf *m) 1746{ 1747 struct ip_moptions imo; 1748 struct in_multi *imm[2]; 1749 int error; 1750 1751 VIF_LOCK_ASSERT(); 1752 1753 imo.imo_multicast_ifp = vifp->v_ifp; 1754 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1755 imo.imo_multicast_loop = 1; 1756 imo.imo_multicast_vif = -1; 1757 imo.imo_num_memberships = 0; 1758 imo.imo_max_memberships = 2; 1759 imo.imo_membership = &imm[0]; 1760 1761 /* 1762 * Re-entrancy should not be a problem here, because 1763 * the packets that we send out and are looped back at us 1764 * should get rejected because they appear to come from 1765 * the loopback interface, thus preventing looping. 1766 */ 1767 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL); 1768 if (mrtdebug & DEBUG_XMIT) { 1769 log(LOG_DEBUG, "phyint_send on vif %td err %d\n", 1770 vifp - viftable, error); 1771 } 1772} 1773 1774static int 1775X_ip_rsvp_vif(struct socket *so, struct sockopt *sopt) 1776{ 1777 int error, vifi; 1778 1779 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1780 return EOPNOTSUPP; 1781 1782 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 1783 if (error) 1784 return error; 1785 1786 VIF_LOCK(); 1787 1788 if (vifi < 0 || vifi >= numvifs) { /* Error if vif is invalid */ 1789 VIF_UNLOCK(); 1790 return EADDRNOTAVAIL; 1791 } 1792 1793 if (sopt->sopt_name == IP_RSVP_VIF_ON) { 1794 /* Check if socket is available. */ 1795 if (viftable[vifi].v_rsvpd != NULL) { 1796 VIF_UNLOCK(); 1797 return EADDRINUSE; 1798 } 1799 1800 viftable[vifi].v_rsvpd = so; 1801 /* This may seem silly, but we need to be sure we don't over-increment 1802 * the RSVP counter, in case something slips up. 1803 */ 1804 if (!viftable[vifi].v_rsvp_on) { 1805 viftable[vifi].v_rsvp_on = 1; 1806 rsvp_on++; 1807 } 1808 } else { /* must be VIF_OFF */ 1809 /* 1810 * XXX as an additional consistency check, one could make sure 1811 * that viftable[vifi].v_rsvpd == so, otherwise passing so as 1812 * first parameter is pretty useless. 1813 */ 1814 viftable[vifi].v_rsvpd = NULL; 1815 /* 1816 * This may seem silly, but we need to be sure we don't over-decrement 1817 * the RSVP counter, in case something slips up. 1818 */ 1819 if (viftable[vifi].v_rsvp_on) { 1820 viftable[vifi].v_rsvp_on = 0; 1821 rsvp_on--; 1822 } 1823 } 1824 VIF_UNLOCK(); 1825 return 0; 1826} 1827 1828static void 1829X_ip_rsvp_force_done(struct socket *so) 1830{ 1831 int vifi; 1832 1833 /* Don't bother if it is not the right type of socket. */ 1834 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 1835 return; 1836 1837 VIF_LOCK(); 1838 1839 /* The socket may be attached to more than one vif...this 1840 * is perfectly legal. 1841 */ 1842 for (vifi = 0; vifi < numvifs; vifi++) { 1843 if (viftable[vifi].v_rsvpd == so) { 1844 viftable[vifi].v_rsvpd = NULL; 1845 /* This may seem silly, but we need to be sure we don't 1846 * over-decrement the RSVP counter, in case something slips up. 1847 */ 1848 if (viftable[vifi].v_rsvp_on) { 1849 viftable[vifi].v_rsvp_on = 0; 1850 rsvp_on--; 1851 } 1852 } 1853 } 1854 1855 VIF_UNLOCK(); 1856} 1857 1858static void 1859X_rsvp_input(struct mbuf *m, int off) 1860{ 1861 int vifi; 1862 struct ip *ip = mtod(m, struct ip *); 1863 struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 1864 struct ifnet *ifp; 1865 1866 if (rsvpdebug) 1867 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 1868 1869 /* Can still get packets with rsvp_on = 0 if there is a local member 1870 * of the group to which the RSVP packet is addressed. But in this 1871 * case we want to throw the packet away. 1872 */ 1873 if (!rsvp_on) { 1874 m_freem(m); 1875 return; 1876 } 1877 1878 if (rsvpdebug) 1879 printf("rsvp_input: check vifs\n"); 1880 1881#ifdef DIAGNOSTIC 1882 M_ASSERTPKTHDR(m); 1883#endif 1884 1885 ifp = m->m_pkthdr.rcvif; 1886 1887 VIF_LOCK(); 1888 /* Find which vif the packet arrived on. */ 1889 for (vifi = 0; vifi < numvifs; vifi++) 1890 if (viftable[vifi].v_ifp == ifp) 1891 break; 1892 1893 if (vifi == numvifs || viftable[vifi].v_rsvpd == NULL) { 1894 /* 1895 * Drop the lock here to avoid holding it across rip_input. 1896 * This could make rsvpdebug printfs wrong. If you care, 1897 * record the state of stuff before dropping the lock. 1898 */ 1899 VIF_UNLOCK(); 1900 /* 1901 * If the old-style non-vif-associated socket is set, 1902 * then use it. Otherwise, drop packet since there 1903 * is no specific socket for this vif. 1904 */ 1905 if (ip_rsvpd != NULL) { 1906 if (rsvpdebug) 1907 printf("rsvp_input: Sending packet up old-style socket\n"); 1908 rip_input(m, off); /* xxx */ 1909 } else { 1910 if (rsvpdebug && vifi == numvifs) 1911 printf("rsvp_input: Can't find vif for packet.\n"); 1912 else if (rsvpdebug && viftable[vifi].v_rsvpd == NULL) 1913 printf("rsvp_input: No socket defined for vif %d\n",vifi); 1914 m_freem(m); 1915 } 1916 return; 1917 } 1918 rsvp_src.sin_addr = ip->ip_src; 1919 1920 if (rsvpdebug && m) 1921 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 1922 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 1923 1924 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) { 1925 if (rsvpdebug) 1926 printf("rsvp_input: Failed to append to socket\n"); 1927 } else { 1928 if (rsvpdebug) 1929 printf("rsvp_input: send packet up\n"); 1930 } 1931 VIF_UNLOCK(); 1932} 1933 1934/* 1935 * Code for bandwidth monitors 1936 */ 1937 1938/* 1939 * Define common interface for timeval-related methods 1940 */ 1941#define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 1942#define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 1943#define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 1944 1945static uint32_t 1946compute_bw_meter_flags(struct bw_upcall *req) 1947{ 1948 uint32_t flags = 0; 1949 1950 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 1951 flags |= BW_METER_UNIT_PACKETS; 1952 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 1953 flags |= BW_METER_UNIT_BYTES; 1954 if (req->bu_flags & BW_UPCALL_GEQ) 1955 flags |= BW_METER_GEQ; 1956 if (req->bu_flags & BW_UPCALL_LEQ) 1957 flags |= BW_METER_LEQ; 1958 1959 return flags; 1960} 1961 1962/* 1963 * Add a bw_meter entry 1964 */ 1965static int 1966add_bw_upcall(struct bw_upcall *req) 1967{ 1968 struct mfc *mfc; 1969 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 1970 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 1971 struct timeval now; 1972 struct bw_meter *x; 1973 uint32_t flags; 1974 1975 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 1976 return EOPNOTSUPP; 1977 1978 /* Test if the flags are valid */ 1979 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 1980 return EINVAL; 1981 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 1982 return EINVAL; 1983 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 1984 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 1985 return EINVAL; 1986 1987 /* Test if the threshold time interval is valid */ 1988 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 1989 return EINVAL; 1990 1991 flags = compute_bw_meter_flags(req); 1992 1993 /* 1994 * Find if we have already same bw_meter entry 1995 */ 1996 MFC_LOCK(); 1997 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 1998 if (mfc == NULL) { 1999 MFC_UNLOCK(); 2000 return EADDRNOTAVAIL; 2001 } 2002 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2003 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2004 &req->bu_threshold.b_time, ==)) && 2005 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2006 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2007 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2008 MFC_UNLOCK(); 2009 return 0; /* XXX Already installed */ 2010 } 2011 } 2012 2013 /* Allocate the new bw_meter entry */ 2014 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2015 if (x == NULL) { 2016 MFC_UNLOCK(); 2017 return ENOBUFS; 2018 } 2019 2020 /* Set the new bw_meter entry */ 2021 x->bm_threshold.b_time = req->bu_threshold.b_time; 2022 GET_TIME(now); 2023 x->bm_start_time = now; 2024 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2025 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2026 x->bm_measured.b_packets = 0; 2027 x->bm_measured.b_bytes = 0; 2028 x->bm_flags = flags; 2029 x->bm_time_next = NULL; 2030 x->bm_time_hash = BW_METER_BUCKETS; 2031 2032 /* Add the new bw_meter entry to the front of entries for this MFC */ 2033 x->bm_mfc = mfc; 2034 x->bm_mfc_next = mfc->mfc_bw_meter; 2035 mfc->mfc_bw_meter = x; 2036 schedule_bw_meter(x, &now); 2037 MFC_UNLOCK(); 2038 2039 return 0; 2040} 2041 2042static void 2043free_bw_list(struct bw_meter *list) 2044{ 2045 while (list != NULL) { 2046 struct bw_meter *x = list; 2047 2048 list = list->bm_mfc_next; 2049 unschedule_bw_meter(x); 2050 free(x, M_BWMETER); 2051 } 2052} 2053 2054/* 2055 * Delete one or multiple bw_meter entries 2056 */ 2057static int 2058del_bw_upcall(struct bw_upcall *req) 2059{ 2060 struct mfc *mfc; 2061 struct bw_meter *x; 2062 2063 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2064 return EOPNOTSUPP; 2065 2066 MFC_LOCK(); 2067 /* Find the corresponding MFC entry */ 2068 mfc = mfc_find(req->bu_src.s_addr, req->bu_dst.s_addr); 2069 if (mfc == NULL) { 2070 MFC_UNLOCK(); 2071 return EADDRNOTAVAIL; 2072 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2073 /* 2074 * Delete all bw_meter entries for this mfc 2075 */ 2076 struct bw_meter *list; 2077 2078 list = mfc->mfc_bw_meter; 2079 mfc->mfc_bw_meter = NULL; 2080 free_bw_list(list); 2081 MFC_UNLOCK(); 2082 return 0; 2083 } else { /* Delete a single bw_meter entry */ 2084 struct bw_meter *prev; 2085 uint32_t flags = 0; 2086 2087 flags = compute_bw_meter_flags(req); 2088 2089 /* Find the bw_meter entry to delete */ 2090 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2091 prev = x, x = x->bm_mfc_next) { 2092 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2093 &req->bu_threshold.b_time, ==)) && 2094 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2095 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2096 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2097 break; 2098 } 2099 if (x != NULL) { /* Delete entry from the list for this MFC */ 2100 if (prev != NULL) 2101 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2102 else 2103 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2104 2105 unschedule_bw_meter(x); 2106 MFC_UNLOCK(); 2107 /* Free the bw_meter entry */ 2108 free(x, M_BWMETER); 2109 return 0; 2110 } else { 2111 MFC_UNLOCK(); 2112 return EINVAL; 2113 } 2114 } 2115 /* NOTREACHED */ 2116} 2117 2118/* 2119 * Perform bandwidth measurement processing that may result in an upcall 2120 */ 2121static void 2122bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2123{ 2124 struct timeval delta; 2125 2126 MFC_LOCK_ASSERT(); 2127 2128 delta = *nowp; 2129 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2130 2131 if (x->bm_flags & BW_METER_GEQ) { 2132 /* 2133 * Processing for ">=" type of bw_meter entry 2134 */ 2135 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2136 /* Reset the bw_meter entry */ 2137 x->bm_start_time = *nowp; 2138 x->bm_measured.b_packets = 0; 2139 x->bm_measured.b_bytes = 0; 2140 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2141 } 2142 2143 /* Record that a packet is received */ 2144 x->bm_measured.b_packets++; 2145 x->bm_measured.b_bytes += plen; 2146 2147 /* 2148 * Test if we should deliver an upcall 2149 */ 2150 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2151 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2152 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2153 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2154 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2155 /* Prepare an upcall for delivery */ 2156 bw_meter_prepare_upcall(x, nowp); 2157 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2158 } 2159 } 2160 } else if (x->bm_flags & BW_METER_LEQ) { 2161 /* 2162 * Processing for "<=" type of bw_meter entry 2163 */ 2164 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2165 /* 2166 * We are behind time with the multicast forwarding table 2167 * scanning for "<=" type of bw_meter entries, so test now 2168 * if we should deliver an upcall. 2169 */ 2170 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2171 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2172 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2173 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2174 /* Prepare an upcall for delivery */ 2175 bw_meter_prepare_upcall(x, nowp); 2176 } 2177 /* Reschedule the bw_meter entry */ 2178 unschedule_bw_meter(x); 2179 schedule_bw_meter(x, nowp); 2180 } 2181 2182 /* Record that a packet is received */ 2183 x->bm_measured.b_packets++; 2184 x->bm_measured.b_bytes += plen; 2185 2186 /* 2187 * Test if we should restart the measuring interval 2188 */ 2189 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2190 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2191 (x->bm_flags & BW_METER_UNIT_BYTES && 2192 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2193 /* Don't restart the measuring interval */ 2194 } else { 2195 /* Do restart the measuring interval */ 2196 /* 2197 * XXX: note that we don't unschedule and schedule, because this 2198 * might be too much overhead per packet. Instead, when we process 2199 * all entries for a given timer hash bin, we check whether it is 2200 * really a timeout. If not, we reschedule at that time. 2201 */ 2202 x->bm_start_time = *nowp; 2203 x->bm_measured.b_packets = 0; 2204 x->bm_measured.b_bytes = 0; 2205 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2206 } 2207 } 2208} 2209 2210/* 2211 * Prepare a bandwidth-related upcall 2212 */ 2213static void 2214bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2215{ 2216 struct timeval delta; 2217 struct bw_upcall *u; 2218 2219 MFC_LOCK_ASSERT(); 2220 2221 /* 2222 * Compute the measured time interval 2223 */ 2224 delta = *nowp; 2225 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2226 2227 /* 2228 * If there are too many pending upcalls, deliver them now 2229 */ 2230 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2231 bw_upcalls_send(); 2232 2233 /* 2234 * Set the bw_upcall entry 2235 */ 2236 u = &bw_upcalls[bw_upcalls_n++]; 2237 u->bu_src = x->bm_mfc->mfc_origin; 2238 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2239 u->bu_threshold.b_time = x->bm_threshold.b_time; 2240 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2241 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2242 u->bu_measured.b_time = delta; 2243 u->bu_measured.b_packets = x->bm_measured.b_packets; 2244 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2245 u->bu_flags = 0; 2246 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2247 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2248 if (x->bm_flags & BW_METER_UNIT_BYTES) 2249 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2250 if (x->bm_flags & BW_METER_GEQ) 2251 u->bu_flags |= BW_UPCALL_GEQ; 2252 if (x->bm_flags & BW_METER_LEQ) 2253 u->bu_flags |= BW_UPCALL_LEQ; 2254} 2255 2256/* 2257 * Send the pending bandwidth-related upcalls 2258 */ 2259static void 2260bw_upcalls_send(void) 2261{ 2262 struct mbuf *m; 2263 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2264 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2265 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2266 0, /* unused2 */ 2267 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2268 0, /* im_mbz */ 2269 0, /* im_vif */ 2270 0, /* unused3 */ 2271 { 0 }, /* im_src */ 2272 { 0 } }; /* im_dst */ 2273 2274 MFC_LOCK_ASSERT(); 2275 2276 if (bw_upcalls_n == 0) 2277 return; /* No pending upcalls */ 2278 2279 bw_upcalls_n = 0; 2280 2281 /* 2282 * Allocate a new mbuf, initialize it with the header and 2283 * the payload for the pending calls. 2284 */ 2285 MGETHDR(m, M_DONTWAIT, MT_DATA); 2286 if (m == NULL) { 2287 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2288 return; 2289 } 2290 2291 m->m_len = m->m_pkthdr.len = 0; 2292 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2293 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2294 2295 /* 2296 * Send the upcalls 2297 * XXX do we need to set the address in k_igmpsrc ? 2298 */ 2299 mrtstat.mrts_upcalls++; 2300 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2301 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2302 ++mrtstat.mrts_upq_sockfull; 2303 } 2304} 2305 2306/* 2307 * Compute the timeout hash value for the bw_meter entries 2308 */ 2309#define BW_METER_TIMEHASH(bw_meter, hash) \ 2310 do { \ 2311 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2312 \ 2313 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2314 (hash) = next_timeval.tv_sec; \ 2315 if (next_timeval.tv_usec) \ 2316 (hash)++; /* XXX: make sure we don't timeout early */ \ 2317 (hash) %= BW_METER_BUCKETS; \ 2318 } while (0) 2319 2320/* 2321 * Schedule a timer to process periodically bw_meter entry of type "<=" 2322 * by linking the entry in the proper hash bucket. 2323 */ 2324static void 2325schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2326{ 2327 int time_hash; 2328 2329 MFC_LOCK_ASSERT(); 2330 2331 if (!(x->bm_flags & BW_METER_LEQ)) 2332 return; /* XXX: we schedule timers only for "<=" entries */ 2333 2334 /* 2335 * Reset the bw_meter entry 2336 */ 2337 x->bm_start_time = *nowp; 2338 x->bm_measured.b_packets = 0; 2339 x->bm_measured.b_bytes = 0; 2340 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2341 2342 /* 2343 * Compute the timeout hash value and insert the entry 2344 */ 2345 BW_METER_TIMEHASH(x, time_hash); 2346 x->bm_time_next = bw_meter_timers[time_hash]; 2347 bw_meter_timers[time_hash] = x; 2348 x->bm_time_hash = time_hash; 2349} 2350 2351/* 2352 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2353 * by removing the entry from the proper hash bucket. 2354 */ 2355static void 2356unschedule_bw_meter(struct bw_meter *x) 2357{ 2358 int time_hash; 2359 struct bw_meter *prev, *tmp; 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 * Compute the timeout hash value and delete the entry 2368 */ 2369 time_hash = x->bm_time_hash; 2370 if (time_hash >= BW_METER_BUCKETS) 2371 return; /* Entry was not scheduled */ 2372 2373 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2374 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2375 if (tmp == x) 2376 break; 2377 2378 if (tmp == NULL) 2379 panic("unschedule_bw_meter: bw_meter entry not found"); 2380 2381 if (prev != NULL) 2382 prev->bm_time_next = x->bm_time_next; 2383 else 2384 bw_meter_timers[time_hash] = x->bm_time_next; 2385 2386 x->bm_time_next = NULL; 2387 x->bm_time_hash = BW_METER_BUCKETS; 2388} 2389 2390 2391/* 2392 * Process all "<=" type of bw_meter that should be processed now, 2393 * and for each entry prepare an upcall if necessary. Each processed 2394 * entry is rescheduled again for the (periodic) processing. 2395 * 2396 * This is run periodically (once per second normally). On each round, 2397 * all the potentially matching entries are in the hash slot that we are 2398 * looking at. 2399 */ 2400static void 2401bw_meter_process() 2402{ 2403 static uint32_t last_tv_sec; /* last time we processed this */ 2404 2405 uint32_t loops; 2406 int i; 2407 struct timeval now, process_endtime; 2408 2409 GET_TIME(now); 2410 if (last_tv_sec == now.tv_sec) 2411 return; /* nothing to do */ 2412 2413 loops = now.tv_sec - last_tv_sec; 2414 last_tv_sec = now.tv_sec; 2415 if (loops > BW_METER_BUCKETS) 2416 loops = BW_METER_BUCKETS; 2417 2418 MFC_LOCK(); 2419 /* 2420 * Process all bins of bw_meter entries from the one after the last 2421 * processed to the current one. On entry, i points to the last bucket 2422 * visited, so we need to increment i at the beginning of the loop. 2423 */ 2424 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2425 struct bw_meter *x, *tmp_list; 2426 2427 if (++i >= BW_METER_BUCKETS) 2428 i = 0; 2429 2430 /* Disconnect the list of bw_meter entries from the bin */ 2431 tmp_list = bw_meter_timers[i]; 2432 bw_meter_timers[i] = NULL; 2433 2434 /* Process the list of bw_meter entries */ 2435 while (tmp_list != NULL) { 2436 x = tmp_list; 2437 tmp_list = tmp_list->bm_time_next; 2438 2439 /* Test if the time interval is over */ 2440 process_endtime = x->bm_start_time; 2441 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2442 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2443 /* Not yet: reschedule, but don't reset */ 2444 int time_hash; 2445 2446 BW_METER_TIMEHASH(x, time_hash); 2447 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2448 /* 2449 * XXX: somehow the bin processing is a bit ahead of time. 2450 * Put the entry in the next bin. 2451 */ 2452 if (++time_hash >= BW_METER_BUCKETS) 2453 time_hash = 0; 2454 } 2455 x->bm_time_next = bw_meter_timers[time_hash]; 2456 bw_meter_timers[time_hash] = x; 2457 x->bm_time_hash = time_hash; 2458 2459 continue; 2460 } 2461 2462 /* 2463 * Test if we should deliver an upcall 2464 */ 2465 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2466 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2467 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2468 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2469 /* Prepare an upcall for delivery */ 2470 bw_meter_prepare_upcall(x, &now); 2471 } 2472 2473 /* 2474 * Reschedule for next processing 2475 */ 2476 schedule_bw_meter(x, &now); 2477 } 2478 } 2479 2480 /* Send all upcalls that are pending delivery */ 2481 bw_upcalls_send(); 2482 2483 MFC_UNLOCK(); 2484} 2485 2486/* 2487 * A periodic function for sending all upcalls that are pending delivery 2488 */ 2489static void 2490expire_bw_upcalls_send(void *unused) 2491{ 2492 MFC_LOCK(); 2493 bw_upcalls_send(); 2494 MFC_UNLOCK(); 2495 2496 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2497 expire_bw_upcalls_send, NULL); 2498} 2499 2500/* 2501 * A periodic function for periodic scanning of the multicast forwarding 2502 * table for processing all "<=" bw_meter entries. 2503 */ 2504static void 2505expire_bw_meter_process(void *unused) 2506{ 2507 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2508 bw_meter_process(); 2509 2510 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2511} 2512 2513/* 2514 * End of bandwidth monitoring code 2515 */ 2516 2517/* 2518 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2519 * 2520 */ 2521static int 2522pim_register_send(struct ip *ip, struct vif *vifp, 2523 struct mbuf *m, struct mfc *rt) 2524{ 2525 struct mbuf *mb_copy, *mm; 2526 2527 if (mrtdebug & DEBUG_PIM) 2528 log(LOG_DEBUG, "pim_register_send: "); 2529 2530 /* 2531 * Do not send IGMP_WHOLEPKT notifications to userland, if the 2532 * rendezvous point was unspecified, and we were told not to. 2533 */ 2534 if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) && 2535 (rt->mfc_rp.s_addr == INADDR_ANY)) 2536 return 0; 2537 2538 mb_copy = pim_register_prepare(ip, m); 2539 if (mb_copy == NULL) 2540 return ENOBUFS; 2541 2542 /* 2543 * Send all the fragments. Note that the mbuf for each fragment 2544 * is freed by the sending machinery. 2545 */ 2546 for (mm = mb_copy; mm; mm = mb_copy) { 2547 mb_copy = mm->m_nextpkt; 2548 mm->m_nextpkt = 0; 2549 mm = m_pullup(mm, sizeof(struct ip)); 2550 if (mm != NULL) { 2551 ip = mtod(mm, struct ip *); 2552 if ((mrt_api_config & MRT_MFC_RP) && 2553 (rt->mfc_rp.s_addr != INADDR_ANY)) { 2554 pim_register_send_rp(ip, vifp, mm, rt); 2555 } else { 2556 pim_register_send_upcall(ip, vifp, mm, rt); 2557 } 2558 } 2559 } 2560 2561 return 0; 2562} 2563 2564/* 2565 * Return a copy of the data packet that is ready for PIM Register 2566 * encapsulation. 2567 * XXX: Note that in the returned copy the IP header is a valid one. 2568 */ 2569static struct mbuf * 2570pim_register_prepare(struct ip *ip, struct mbuf *m) 2571{ 2572 struct mbuf *mb_copy = NULL; 2573 int mtu; 2574 2575 /* Take care of delayed checksums */ 2576 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2577 in_delayed_cksum(m); 2578 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2579 } 2580 2581 /* 2582 * Copy the old packet & pullup its IP header into the 2583 * new mbuf so we can modify it. 2584 */ 2585 mb_copy = m_copypacket(m, M_DONTWAIT); 2586 if (mb_copy == NULL) 2587 return NULL; 2588 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2589 if (mb_copy == NULL) 2590 return NULL; 2591 2592 /* take care of the TTL */ 2593 ip = mtod(mb_copy, struct ip *); 2594 --ip->ip_ttl; 2595 2596 /* Compute the MTU after the PIM Register encapsulation */ 2597 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2598 2599 if (ip->ip_len <= mtu) { 2600 /* Turn the IP header into a valid one */ 2601 ip->ip_len = htons(ip->ip_len); 2602 ip->ip_off = htons(ip->ip_off); 2603 ip->ip_sum = 0; 2604 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2605 } else { 2606 /* Fragment the packet */ 2607 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2608 m_freem(mb_copy); 2609 return NULL; 2610 } 2611 } 2612 return mb_copy; 2613} 2614 2615/* 2616 * Send an upcall with the data packet to the user-level process. 2617 */ 2618static int 2619pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2620 struct mbuf *mb_copy, struct mfc *rt) 2621{ 2622 struct mbuf *mb_first; 2623 int len = ntohs(ip->ip_len); 2624 struct igmpmsg *im; 2625 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2626 2627 VIF_LOCK_ASSERT(); 2628 2629 /* 2630 * Add a new mbuf with an upcall header 2631 */ 2632 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2633 if (mb_first == NULL) { 2634 m_freem(mb_copy); 2635 return ENOBUFS; 2636 } 2637 mb_first->m_data += max_linkhdr; 2638 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2639 mb_first->m_len = sizeof(struct igmpmsg); 2640 mb_first->m_next = mb_copy; 2641 2642 /* Send message to routing daemon */ 2643 im = mtod(mb_first, struct igmpmsg *); 2644 im->im_msgtype = IGMPMSG_WHOLEPKT; 2645 im->im_mbz = 0; 2646 im->im_vif = vifp - viftable; 2647 im->im_src = ip->ip_src; 2648 im->im_dst = ip->ip_dst; 2649 2650 k_igmpsrc.sin_addr = ip->ip_src; 2651 2652 mrtstat.mrts_upcalls++; 2653 2654 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2655 if (mrtdebug & DEBUG_PIM) 2656 log(LOG_WARNING, 2657 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 2658 ++mrtstat.mrts_upq_sockfull; 2659 return ENOBUFS; 2660 } 2661 2662 /* Keep statistics */ 2663 pimstat.pims_snd_registers_msgs++; 2664 pimstat.pims_snd_registers_bytes += len; 2665 2666 return 0; 2667} 2668 2669/* 2670 * Encapsulate the data packet in PIM Register message and send it to the RP. 2671 */ 2672static int 2673pim_register_send_rp(struct ip *ip, struct vif *vifp, 2674 struct mbuf *mb_copy, struct mfc *rt) 2675{ 2676 struct mbuf *mb_first; 2677 struct ip *ip_outer; 2678 struct pim_encap_pimhdr *pimhdr; 2679 int len = ntohs(ip->ip_len); 2680 vifi_t vifi = rt->mfc_parent; 2681 2682 VIF_LOCK_ASSERT(); 2683 2684 if ((vifi >= numvifs) || (viftable[vifi].v_lcl_addr.s_addr == 0)) { 2685 m_freem(mb_copy); 2686 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2687 } 2688 2689 /* 2690 * Add a new mbuf with the encapsulating header 2691 */ 2692 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2693 if (mb_first == NULL) { 2694 m_freem(mb_copy); 2695 return ENOBUFS; 2696 } 2697 mb_first->m_data += max_linkhdr; 2698 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2699 mb_first->m_next = mb_copy; 2700 2701 mb_first->m_pkthdr.len = len + mb_first->m_len; 2702 2703 /* 2704 * Fill in the encapsulating IP and PIM header 2705 */ 2706 ip_outer = mtod(mb_first, struct ip *); 2707 *ip_outer = pim_encap_iphdr; 2708 ip_outer->ip_id = ip_newid(); 2709 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2710 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2711 ip_outer->ip_dst = rt->mfc_rp; 2712 /* 2713 * Copy the inner header TOS to the outer header, and take care of the 2714 * IP_DF bit. 2715 */ 2716 ip_outer->ip_tos = ip->ip_tos; 2717 if (ntohs(ip->ip_off) & IP_DF) 2718 ip_outer->ip_off |= IP_DF; 2719 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2720 + sizeof(pim_encap_iphdr)); 2721 *pimhdr = pim_encap_pimhdr; 2722 /* If the iif crosses a border, set the Border-bit */ 2723 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2724 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2725 2726 mb_first->m_data += sizeof(pim_encap_iphdr); 2727 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2728 mb_first->m_data -= sizeof(pim_encap_iphdr); 2729 2730 send_packet(vifp, mb_first); 2731 2732 /* Keep statistics */ 2733 pimstat.pims_snd_registers_msgs++; 2734 pimstat.pims_snd_registers_bytes += len; 2735 2736 return 0; 2737} 2738 2739/* 2740 * pim_encapcheck() is called by the encap4_input() path at runtime to 2741 * determine if a packet is for PIM; allowing PIM to be dynamically loaded 2742 * into the kernel. 2743 */ 2744static int 2745pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 2746{ 2747 2748#ifdef DIAGNOSTIC 2749 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); 2750#endif 2751 if (proto != IPPROTO_PIM) 2752 return 0; /* not for us; reject the datagram. */ 2753 2754 return 64; /* claim the datagram. */ 2755} 2756 2757/* 2758 * PIM-SMv2 and PIM-DM messages processing. 2759 * Receives and verifies the PIM control messages, and passes them 2760 * up to the listening socket, using rip_input(). 2761 * The only message with special processing is the PIM_REGISTER message 2762 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2763 * is passed to if_simloop(). 2764 */ 2765void 2766pim_input(struct mbuf *m, int off) 2767{ 2768 struct ip *ip = mtod(m, struct ip *); 2769 struct pim *pim; 2770 int minlen; 2771 int datalen = ip->ip_len; 2772 int ip_tos; 2773 int iphlen = off; 2774 2775 /* Keep statistics */ 2776 pimstat.pims_rcv_total_msgs++; 2777 pimstat.pims_rcv_total_bytes += datalen; 2778 2779 /* 2780 * Validate lengths 2781 */ 2782 if (datalen < PIM_MINLEN) { 2783 pimstat.pims_rcv_tooshort++; 2784 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 2785 datalen, (u_long)ip->ip_src.s_addr); 2786 m_freem(m); 2787 return; 2788 } 2789 2790 /* 2791 * If the packet is at least as big as a REGISTER, go agead 2792 * and grab the PIM REGISTER header size, to avoid another 2793 * possible m_pullup() later. 2794 * 2795 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 2796 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 2797 */ 2798 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 2799 /* 2800 * Get the IP and PIM headers in contiguous memory, and 2801 * possibly the PIM REGISTER header. 2802 */ 2803 if ((m->m_flags & M_EXT || m->m_len < minlen) && 2804 (m = m_pullup(m, minlen)) == 0) { 2805 log(LOG_ERR, "pim_input: m_pullup failure\n"); 2806 return; 2807 } 2808 /* m_pullup() may have given us a new mbuf so reset ip. */ 2809 ip = mtod(m, struct ip *); 2810 ip_tos = ip->ip_tos; 2811 2812 /* adjust mbuf to point to the PIM header */ 2813 m->m_data += iphlen; 2814 m->m_len -= iphlen; 2815 pim = mtod(m, struct pim *); 2816 2817 /* 2818 * Validate checksum. If PIM REGISTER, exclude the data packet. 2819 * 2820 * XXX: some older PIMv2 implementations don't make this distinction, 2821 * so for compatibility reason perform the checksum over part of the 2822 * message, and if error, then over the whole message. 2823 */ 2824 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 2825 /* do nothing, checksum okay */ 2826 } else if (in_cksum(m, datalen)) { 2827 pimstat.pims_rcv_badsum++; 2828 if (mrtdebug & DEBUG_PIM) 2829 log(LOG_DEBUG, "pim_input: invalid checksum"); 2830 m_freem(m); 2831 return; 2832 } 2833 2834 /* PIM version check */ 2835 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 2836 pimstat.pims_rcv_badversion++; 2837 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 2838 PIM_VT_V(pim->pim_vt), PIM_VERSION); 2839 m_freem(m); 2840 return; 2841 } 2842 2843 /* restore mbuf back to the outer IP */ 2844 m->m_data -= iphlen; 2845 m->m_len += iphlen; 2846 2847 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 2848 /* 2849 * Since this is a REGISTER, we'll make a copy of the register 2850 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 2851 * routing daemon. 2852 */ 2853 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 2854 struct mbuf *mcp; 2855 struct ip *encap_ip; 2856 u_int32_t *reghdr; 2857 struct ifnet *vifp; 2858 2859 VIF_LOCK(); 2860 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 2861 VIF_UNLOCK(); 2862 if (mrtdebug & DEBUG_PIM) 2863 log(LOG_DEBUG, 2864 "pim_input: register vif not set: %d\n", reg_vif_num); 2865 m_freem(m); 2866 return; 2867 } 2868 /* XXX need refcnt? */ 2869 vifp = viftable[reg_vif_num].v_ifp; 2870 VIF_UNLOCK(); 2871 2872 /* 2873 * Validate length 2874 */ 2875 if (datalen < PIM_REG_MINLEN) { 2876 pimstat.pims_rcv_tooshort++; 2877 pimstat.pims_rcv_badregisters++; 2878 log(LOG_ERR, 2879 "pim_input: register packet size too small %d from %lx\n", 2880 datalen, (u_long)ip->ip_src.s_addr); 2881 m_freem(m); 2882 return; 2883 } 2884 2885 reghdr = (u_int32_t *)(pim + 1); 2886 encap_ip = (struct ip *)(reghdr + 1); 2887 2888 if (mrtdebug & DEBUG_PIM) { 2889 log(LOG_DEBUG, 2890 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 2891 (u_long)ntohl(encap_ip->ip_src.s_addr), 2892 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2893 ntohs(encap_ip->ip_len)); 2894 } 2895 2896 /* verify the version number of the inner packet */ 2897 if (encap_ip->ip_v != IPVERSION) { 2898 pimstat.pims_rcv_badregisters++; 2899 if (mrtdebug & DEBUG_PIM) { 2900 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 2901 "of the inner packet\n", encap_ip->ip_v); 2902 } 2903 m_freem(m); 2904 return; 2905 } 2906 2907 /* verify the inner packet is destined to a mcast group */ 2908 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 2909 pimstat.pims_rcv_badregisters++; 2910 if (mrtdebug & DEBUG_PIM) 2911 log(LOG_DEBUG, 2912 "pim_input: inner packet of register is not " 2913 "multicast %lx\n", 2914 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 2915 m_freem(m); 2916 return; 2917 } 2918 2919 /* If a NULL_REGISTER, pass it to the daemon */ 2920 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 2921 goto pim_input_to_daemon; 2922 2923 /* 2924 * Copy the TOS from the outer IP header to the inner IP header. 2925 */ 2926 if (encap_ip->ip_tos != ip_tos) { 2927 /* Outer TOS -> inner TOS */ 2928 encap_ip->ip_tos = ip_tos; 2929 /* Recompute the inner header checksum. Sigh... */ 2930 2931 /* adjust mbuf to point to the inner IP header */ 2932 m->m_data += (iphlen + PIM_MINLEN); 2933 m->m_len -= (iphlen + PIM_MINLEN); 2934 2935 encap_ip->ip_sum = 0; 2936 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 2937 2938 /* restore mbuf to point back to the outer IP header */ 2939 m->m_data -= (iphlen + PIM_MINLEN); 2940 m->m_len += (iphlen + PIM_MINLEN); 2941 } 2942 2943 /* 2944 * Decapsulate the inner IP packet and loopback to forward it 2945 * as a normal multicast packet. Also, make a copy of the 2946 * outer_iphdr + pimhdr + reghdr + encap_iphdr 2947 * to pass to the daemon later, so it can take the appropriate 2948 * actions (e.g., send back PIM_REGISTER_STOP). 2949 * XXX: here m->m_data points to the outer IP header. 2950 */ 2951 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 2952 if (mcp == NULL) { 2953 log(LOG_ERR, 2954 "pim_input: pim register: could not copy register head\n"); 2955 m_freem(m); 2956 return; 2957 } 2958 2959 /* Keep statistics */ 2960 /* XXX: registers_bytes include only the encap. mcast pkt */ 2961 pimstat.pims_rcv_registers_msgs++; 2962 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 2963 2964 /* 2965 * forward the inner ip packet; point m_data at the inner ip. 2966 */ 2967 m_adj(m, iphlen + PIM_MINLEN); 2968 2969 if (mrtdebug & DEBUG_PIM) { 2970 log(LOG_DEBUG, 2971 "pim_input: forwarding decapsulated register: " 2972 "src %lx, dst %lx, vif %d\n", 2973 (u_long)ntohl(encap_ip->ip_src.s_addr), 2974 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2975 reg_vif_num); 2976 } 2977 /* NB: vifp was collected above; can it change on us? */ 2978 if_simloop(vifp, m, dst.sin_family, 0); 2979 2980 /* prepare the register head to send to the mrouting daemon */ 2981 m = mcp; 2982 } 2983 2984pim_input_to_daemon: 2985 /* 2986 * Pass the PIM message up to the daemon; if it is a Register message, 2987 * pass the 'head' only up to the daemon. This includes the 2988 * outer IP header, PIM header, PIM-Register header and the 2989 * inner IP header. 2990 * XXX: the outer IP header pkt size of a Register is not adjust to 2991 * reflect the fact that the inner multicast data is truncated. 2992 */ 2993 rip_input(m, iphlen); 2994 2995 return; 2996} 2997 2998static int 2999ip_mroute_modevent(module_t mod, int type, void *unused) 3000{ 3001 switch (type) { 3002 case MOD_LOAD: 3003 mtx_init(&mrouter_mtx, "mrouter initialization", NULL, MTX_DEF); 3004 MFC_LOCK_INIT(); 3005 VIF_LOCK_INIT(); 3006 ip_mrouter_reset(); 3007 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", 3008 &pim_squelch_wholepkt); 3009 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM, 3010 pim_encapcheck, &in_pim_protosw, NULL); 3011 if (pim_encap_cookie == NULL) { 3012 printf("ip_mroute: unable to attach pim encap\n"); 3013 VIF_LOCK_DESTROY(); 3014 MFC_LOCK_DESTROY(); 3015 mtx_destroy(&mrouter_mtx); 3016 return (EINVAL); 3017 } 3018 ip_mcast_src = X_ip_mcast_src; 3019 ip_mforward = X_ip_mforward; 3020 ip_mrouter_done = X_ip_mrouter_done; 3021 ip_mrouter_get = X_ip_mrouter_get; 3022 ip_mrouter_set = X_ip_mrouter_set; 3023 ip_rsvp_force_done = X_ip_rsvp_force_done; 3024 ip_rsvp_vif = X_ip_rsvp_vif; 3025 legal_vif_num = X_legal_vif_num; 3026 mrt_ioctl = X_mrt_ioctl; 3027 rsvp_input_p = X_rsvp_input; 3028 break; 3029 3030 case MOD_UNLOAD: 3031 /* 3032 * Typically module unload happens after the user-level 3033 * process has shutdown the kernel services (the check 3034 * below insures someone can't just yank the module out 3035 * from under a running process). But if the module is 3036 * just loaded and then unloaded w/o starting up a user 3037 * process we still need to cleanup. 3038 */ 3039 if (ip_mrouter) 3040 return EINVAL; 3041 3042 if (pim_encap_cookie) { 3043 encap_detach(pim_encap_cookie); 3044 pim_encap_cookie = NULL; 3045 } 3046 3047 X_ip_mrouter_done(); 3048 ip_mcast_src = NULL; 3049 ip_mforward = NULL; 3050 ip_mrouter_done = NULL; 3051 ip_mrouter_get = NULL; 3052 ip_mrouter_set = NULL; 3053 ip_rsvp_force_done = NULL; 3054 ip_rsvp_vif = NULL; 3055 legal_vif_num = NULL; 3056 mrt_ioctl = NULL; 3057 rsvp_input_p = NULL; 3058 VIF_LOCK_DESTROY(); 3059 MFC_LOCK_DESTROY(); 3060 mtx_destroy(&mrouter_mtx); 3061 break; 3062 default: 3063 return EOPNOTSUPP; 3064 } 3065 return 0; 3066} 3067 3068static moduledata_t ip_mroutemod = { 3069 "ip_mroute", 3070 ip_mroute_modevent, 3071 0 3072}; 3073DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 3074