ip_mroute.c revision 1.88
1 /* $NetBSD: ip_mroute.c,v 1.88 2005/02/02 21:41:55 perry Exp $ */ 2 3 /* 4 * Copyright (c) 1992, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * This code is derived from software contributed to Berkeley by 8 * Stephen Deering of Stanford University. 9 * 10 * Redistribution and use in source and binary forms, with or without 11 * modification, are permitted provided that the following conditions 12 * are met: 13 * 1. Redistributions of source code must retain the above copyright 14 * notice, this list of conditions and the following disclaimer. 15 * 2. Redistributions in binary form must reproduce the above copyright 16 * notice, this list of conditions and the following disclaimer in the 17 * documentation and/or other materials provided with the distribution. 18 * 3. Neither the name of the University nor the names of its contributors 19 * may be used to endorse or promote products derived from this software 20 * without specific prior written permission. 21 * 22 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 23 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 24 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 25 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 26 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 27 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 28 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 29 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 30 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 31 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 32 * SUCH DAMAGE. 33 * 34 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 35 */ 36 37 /* 38 * Copyright (c) 1989 Stephen Deering 39 * 40 * This code is derived from software contributed to Berkeley by 41 * Stephen Deering of Stanford University. 42 * 43 * Redistribution and use in source and binary forms, with or without 44 * modification, are permitted provided that the following conditions 45 * are met: 46 * 1. Redistributions of source code must retain the above copyright 47 * notice, this list of conditions and the following disclaimer. 48 * 2. Redistributions in binary form must reproduce the above copyright 49 * notice, this list of conditions and the following disclaimer in the 50 * documentation and/or other materials provided with the distribution. 51 * 3. All advertising materials mentioning features or use of this software 52 * must display the following acknowledgement: 53 * This product includes software developed by the University of 54 * California, Berkeley and its contributors. 55 * 4. Neither the name of the University nor the names of its contributors 56 * may be used to endorse or promote products derived from this software 57 * without specific prior written permission. 58 * 59 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 60 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 61 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 62 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 63 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 64 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 65 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 66 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 67 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 68 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 69 * SUCH DAMAGE. 70 * 71 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 72 */ 73 74 /* 75 * IP multicast forwarding procedures 76 * 77 * Written by David Waitzman, BBN Labs, August 1988. 78 * Modified by Steve Deering, Stanford, February 1989. 79 * Modified by Mark J. Steiglitz, Stanford, May, 1991 80 * Modified by Van Jacobson, LBL, January 1993 81 * Modified by Ajit Thyagarajan, PARC, August 1993 82 * Modified by Bill Fenner, PARC, April 1994 83 * Modified by Charles M. Hannum, NetBSD, May 1995. 84 * Modified by Ahmed Helmy, SGI, June 1996 85 * Modified by George Edmond Eddy (Rusty), ISI, February 1998 86 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000 87 * Modified by Hitoshi Asaeda, WIDE, August 2000 88 * Modified by Pavlin Radoslavov, ICSI, October 2002 89 * 90 * MROUTING Revision: 1.2 91 * and PIM-SMv2 and PIM-DM support, advanced API support, 92 * bandwidth metering and signaling 93 */ 94 95 #include <sys/cdefs.h> 96 __KERNEL_RCSID(0, "$NetBSD: ip_mroute.c,v 1.88 2005/02/02 21:41:55 perry Exp $"); 97 98 #include "opt_inet.h" 99 #include "opt_ipsec.h" 100 #include "opt_pim.h" 101 102 #ifdef PIM 103 #define _PIM_VT 1 104 #endif 105 106 #include <sys/param.h> 107 #include <sys/systm.h> 108 #include <sys/callout.h> 109 #include <sys/mbuf.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/protosw.h> 113 #include <sys/errno.h> 114 #include <sys/time.h> 115 #include <sys/kernel.h> 116 #include <sys/ioctl.h> 117 #include <sys/syslog.h> 118 119 #include <net/if.h> 120 #include <net/route.h> 121 #include <net/raw_cb.h> 122 123 #include <netinet/in.h> 124 #include <netinet/in_var.h> 125 #include <netinet/in_systm.h> 126 #include <netinet/ip.h> 127 #include <netinet/ip_var.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/udp.h> 130 #include <netinet/igmp.h> 131 #include <netinet/igmp_var.h> 132 #include <netinet/ip_mroute.h> 133 #ifdef PIM 134 #include <netinet/pim.h> 135 #include <netinet/pim_var.h> 136 #endif 137 #include <netinet/ip_encap.h> 138 139 #ifdef IPSEC 140 #include <netinet6/ipsec.h> 141 #include <netkey/key.h> 142 #endif 143 144 #ifdef FAST_IPSEC 145 #include <netipsec/ipsec.h> 146 #include <netipsec/key.h> 147 #endif 148 149 #include <machine/stdarg.h> 150 151 #define IP_MULTICASTOPTS 0 152 #define M_PULLUP(m, len) \ 153 do { \ 154 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \ 155 (m) = m_pullup((m), (len)); \ 156 } while (/*CONSTCOND*/ 0) 157 158 /* 159 * Globals. All but ip_mrouter and ip_mrtproto could be static, 160 * except for netstat or debugging purposes. 161 */ 162 struct socket *ip_mrouter = NULL; 163 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 164 165 #define NO_RTE_FOUND 0x1 166 #define RTE_FOUND 0x2 167 168 #define MFCHASH(a, g) \ 169 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ 170 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash) 171 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl; 172 u_long mfchash; 173 174 u_char nexpire[MFCTBLSIZ]; 175 struct vif viftable[MAXVIFS]; 176 struct mrtstat mrtstat; 177 u_int mrtdebug = 0; /* debug level */ 178 #define DEBUG_MFC 0x02 179 #define DEBUG_FORWARD 0x04 180 #define DEBUG_EXPIRE 0x08 181 #define DEBUG_XMIT 0x10 182 #define DEBUG_PIM 0x20 183 184 #define VIFI_INVALID ((vifi_t) -1) 185 186 u_int tbfdebug = 0; /* tbf debug level */ 187 #ifdef RSVP_ISI 188 u_int rsvpdebug = 0; /* rsvp debug level */ 189 extern struct socket *ip_rsvpd; 190 extern int rsvp_on; 191 #endif /* RSVP_ISI */ 192 193 /* vif attachment using sys/netinet/ip_encap.c */ 194 static void vif_input(struct mbuf *, ...); 195 static int vif_encapcheck(const struct mbuf *, int, int, void *); 196 197 static const struct protosw vif_protosw = 198 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 199 vif_input, rip_output, 0, rip_ctloutput, 200 rip_usrreq, 201 0, 0, 0, 0, 202 }; 203 204 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 205 #define UPCALL_EXPIRE 6 /* number of timeouts */ 206 207 /* 208 * Define the token bucket filter structures 209 */ 210 211 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 212 213 static int get_sg_cnt(struct sioc_sg_req *); 214 static int get_vif_cnt(struct sioc_vif_req *); 215 static int ip_mrouter_init(struct socket *, struct mbuf *); 216 static int get_version(struct mbuf *); 217 static int set_assert(struct mbuf *); 218 static int get_assert(struct mbuf *); 219 static int add_vif(struct mbuf *); 220 static int del_vif(struct mbuf *); 221 static void update_mfc_params(struct mfc *, struct mfcctl2 *); 222 static void init_mfc_params(struct mfc *, struct mfcctl2 *); 223 static void expire_mfc(struct mfc *); 224 static int add_mfc(struct mbuf *); 225 #ifdef UPCALL_TIMING 226 static void collate(struct timeval *); 227 #endif 228 static int del_mfc(struct mbuf *); 229 static int set_api_config(struct mbuf *); /* chose API capabilities */ 230 static int get_api_support(struct mbuf *); 231 static int get_api_config(struct mbuf *); 232 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 233 static void expire_upcalls(void *); 234 #ifdef RSVP_ISI 235 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 236 #else 237 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *); 238 #endif 239 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 240 static void encap_send(struct ip *, struct vif *, struct mbuf *); 241 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t); 242 static void tbf_queue(struct vif *, struct mbuf *); 243 static void tbf_process_q(struct vif *); 244 static void tbf_reprocess_q(void *); 245 static int tbf_dq_sel(struct vif *, struct ip *); 246 static void tbf_send_packet(struct vif *, struct mbuf *); 247 static void tbf_update_tokens(struct vif *); 248 static int priority(struct vif *, struct ip *); 249 250 /* 251 * Bandwidth monitoring 252 */ 253 static void free_bw_list(struct bw_meter *); 254 static int add_bw_upcall(struct mbuf *); 255 static int del_bw_upcall(struct mbuf *); 256 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *); 257 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *); 258 static void bw_upcalls_send(void); 259 static void schedule_bw_meter(struct bw_meter *, struct timeval *); 260 static void unschedule_bw_meter(struct bw_meter *); 261 static void bw_meter_process(void); 262 static void expire_bw_upcalls_send(void *); 263 static void expire_bw_meter_process(void *); 264 265 #ifdef PIM 266 static int pim_register_send(struct ip *, struct vif *, 267 struct mbuf *, struct mfc *); 268 static int pim_register_send_rp(struct ip *, struct vif *, 269 struct mbuf *, struct mfc *); 270 static int pim_register_send_upcall(struct ip *, struct vif *, 271 struct mbuf *, struct mfc *); 272 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 273 #endif 274 275 /* 276 * 'Interfaces' associated with decapsulator (so we can tell 277 * packets that went through it from ones that get reflected 278 * by a broken gateway). These interfaces are never linked into 279 * the system ifnet list & no routes point to them. I.e., packets 280 * can't be sent this way. They only exist as a placeholder for 281 * multicast source verification. 282 */ 283 #if 0 284 struct ifnet multicast_decap_if[MAXVIFS]; 285 #endif 286 287 #define ENCAP_TTL 64 288 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 289 290 /* prototype IP hdr for encapsulated packets */ 291 struct ip multicast_encap_iphdr = { 292 #if BYTE_ORDER == LITTLE_ENDIAN 293 sizeof(struct ip) >> 2, IPVERSION, 294 #else 295 IPVERSION, sizeof(struct ip) >> 2, 296 #endif 297 0, /* tos */ 298 sizeof(struct ip), /* total length */ 299 0, /* id */ 300 0, /* frag offset */ 301 ENCAP_TTL, ENCAP_PROTO, 302 0, /* checksum */ 303 }; 304 305 /* 306 * Bandwidth meter variables and constants 307 */ 308 309 /* 310 * Pending timeouts are stored in a hash table, the key being the 311 * expiration time. Periodically, the entries are analysed and processed. 312 */ 313 #define BW_METER_BUCKETS 1024 314 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 315 struct callout bw_meter_ch; 316 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 317 318 /* 319 * Pending upcalls are stored in a vector which is flushed when 320 * full, or periodically 321 */ 322 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 323 static u_int bw_upcalls_n; /* # of pending upcalls */ 324 struct callout bw_upcalls_ch; 325 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 326 327 #ifdef PIM 328 struct pimstat pimstat; 329 330 /* 331 * Note: the PIM Register encapsulation adds the following in front of a 332 * data packet: 333 * 334 * struct pim_encap_hdr { 335 * struct ip ip; 336 * struct pim_encap_pimhdr pim; 337 * } 338 * 339 */ 340 341 struct pim_encap_pimhdr { 342 struct pim pim; 343 uint32_t flags; 344 }; 345 346 static struct ip pim_encap_iphdr = { 347 #if BYTE_ORDER == LITTLE_ENDIAN 348 sizeof(struct ip) >> 2, 349 IPVERSION, 350 #else 351 IPVERSION, 352 sizeof(struct ip) >> 2, 353 #endif 354 0, /* tos */ 355 sizeof(struct ip), /* total length */ 356 0, /* id */ 357 0, /* frag offset */ 358 ENCAP_TTL, 359 IPPROTO_PIM, 360 0, /* checksum */ 361 }; 362 363 static struct pim_encap_pimhdr pim_encap_pimhdr = { 364 { 365 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 366 0, /* reserved */ 367 0, /* checksum */ 368 }, 369 0 /* flags */ 370 }; 371 372 static struct ifnet multicast_register_if; 373 static vifi_t reg_vif_num = VIFI_INVALID; 374 #endif /* PIM */ 375 376 377 /* 378 * Private variables. 379 */ 380 static vifi_t numvifs = 0; 381 382 static struct callout expire_upcalls_ch; 383 384 /* 385 * one-back cache used by vif_encapcheck to locate a tunnel's vif 386 * given a datagram's src ip address. 387 */ 388 static struct in_addr last_encap_src; 389 static struct vif *last_encap_vif; 390 391 /* 392 * whether or not special PIM assert processing is enabled. 393 */ 394 static int pim_assert; 395 /* 396 * Rate limit for assert notification messages, in usec 397 */ 398 #define ASSERT_MSG_TIME 3000000 399 400 /* 401 * Kernel multicast routing API capabilities and setup. 402 * If more API capabilities are added to the kernel, they should be 403 * recorded in `mrt_api_support'. 404 */ 405 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 406 MRT_MFC_FLAGS_BORDER_VIF | 407 MRT_MFC_RP | 408 MRT_MFC_BW_UPCALL); 409 static u_int32_t mrt_api_config = 0; 410 411 /* 412 * Find a route for a given origin IP address and Multicast group address 413 * Type of service parameter to be added in the future!!! 414 * Statistics are updated by the caller if needed 415 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 416 */ 417 static struct mfc * 418 mfc_find(struct in_addr *o, struct in_addr *g) 419 { 420 struct mfc *rt; 421 422 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { 423 if (in_hosteq(rt->mfc_origin, *o) && 424 in_hosteq(rt->mfc_mcastgrp, *g) && 425 (rt->mfc_stall == NULL)) 426 break; 427 } 428 429 return (rt); 430 } 431 432 /* 433 * Macros to compute elapsed time efficiently 434 * Borrowed from Van Jacobson's scheduling code 435 */ 436 #define TV_DELTA(a, b, delta) do { \ 437 int xxs; \ 438 delta = (a).tv_usec - (b).tv_usec; \ 439 xxs = (a).tv_sec - (b).tv_sec; \ 440 switch (xxs) { \ 441 case 2: \ 442 delta += 1000000; \ 443 /* fall through */ \ 444 case 1: \ 445 delta += 1000000; \ 446 /* fall through */ \ 447 case 0: \ 448 break; \ 449 default: \ 450 delta += (1000000 * xxs); \ 451 break; \ 452 } \ 453 } while (/*CONSTCOND*/ 0) 454 455 #ifdef UPCALL_TIMING 456 u_int32_t upcall_data[51]; 457 #endif /* UPCALL_TIMING */ 458 459 /* 460 * Handle MRT setsockopt commands to modify the multicast routing tables. 461 */ 462 int 463 ip_mrouter_set(so, optname, m) 464 struct socket *so; 465 int optname; 466 struct mbuf **m; 467 { 468 int error; 469 470 if (optname != MRT_INIT && so != ip_mrouter) 471 error = ENOPROTOOPT; 472 else 473 switch (optname) { 474 case MRT_INIT: 475 error = ip_mrouter_init(so, *m); 476 break; 477 case MRT_DONE: 478 error = ip_mrouter_done(); 479 break; 480 case MRT_ADD_VIF: 481 error = add_vif(*m); 482 break; 483 case MRT_DEL_VIF: 484 error = del_vif(*m); 485 break; 486 case MRT_ADD_MFC: 487 error = add_mfc(*m); 488 break; 489 case MRT_DEL_MFC: 490 error = del_mfc(*m); 491 break; 492 case MRT_ASSERT: 493 error = set_assert(*m); 494 break; 495 case MRT_API_CONFIG: 496 error = set_api_config(*m); 497 break; 498 case MRT_ADD_BW_UPCALL: 499 error = add_bw_upcall(*m); 500 break; 501 case MRT_DEL_BW_UPCALL: 502 error = del_bw_upcall(*m); 503 break; 504 default: 505 error = ENOPROTOOPT; 506 break; 507 } 508 509 if (*m) 510 m_free(*m); 511 return (error); 512 } 513 514 /* 515 * Handle MRT getsockopt commands 516 */ 517 int 518 ip_mrouter_get(so, optname, m) 519 struct socket *so; 520 int optname; 521 struct mbuf **m; 522 { 523 int error; 524 525 if (so != ip_mrouter) 526 error = ENOPROTOOPT; 527 else { 528 *m = m_get(M_WAIT, MT_SOOPTS); 529 MCLAIM(*m, so->so_mowner); 530 531 switch (optname) { 532 case MRT_VERSION: 533 error = get_version(*m); 534 break; 535 case MRT_ASSERT: 536 error = get_assert(*m); 537 break; 538 case MRT_API_SUPPORT: 539 error = get_api_support(*m); 540 break; 541 case MRT_API_CONFIG: 542 error = get_api_config(*m); 543 break; 544 default: 545 error = ENOPROTOOPT; 546 break; 547 } 548 549 if (error) 550 m_free(*m); 551 } 552 553 return (error); 554 } 555 556 /* 557 * Handle ioctl commands to obtain information from the cache 558 */ 559 int 560 mrt_ioctl(so, cmd, data) 561 struct socket *so; 562 u_long cmd; 563 caddr_t data; 564 { 565 int error; 566 567 if (so != ip_mrouter) 568 error = EINVAL; 569 else 570 switch (cmd) { 571 case SIOCGETVIFCNT: 572 error = get_vif_cnt((struct sioc_vif_req *)data); 573 break; 574 case SIOCGETSGCNT: 575 error = get_sg_cnt((struct sioc_sg_req *)data); 576 break; 577 default: 578 error = EINVAL; 579 break; 580 } 581 582 return (error); 583 } 584 585 /* 586 * returns the packet, byte, rpf-failure count for the source group provided 587 */ 588 static int 589 get_sg_cnt(req) 590 struct sioc_sg_req *req; 591 { 592 int s; 593 struct mfc *rt; 594 595 s = splsoftnet(); 596 rt = mfc_find(&req->src, &req->grp); 597 if (rt == NULL) { 598 splx(s); 599 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 600 return (EADDRNOTAVAIL); 601 } 602 req->pktcnt = rt->mfc_pkt_cnt; 603 req->bytecnt = rt->mfc_byte_cnt; 604 req->wrong_if = rt->mfc_wrong_if; 605 splx(s); 606 607 return (0); 608 } 609 610 /* 611 * returns the input and output packet and byte counts on the vif provided 612 */ 613 static int 614 get_vif_cnt(req) 615 struct sioc_vif_req *req; 616 { 617 vifi_t vifi = req->vifi; 618 619 if (vifi >= numvifs) 620 return (EINVAL); 621 622 req->icount = viftable[vifi].v_pkt_in; 623 req->ocount = viftable[vifi].v_pkt_out; 624 req->ibytes = viftable[vifi].v_bytes_in; 625 req->obytes = viftable[vifi].v_bytes_out; 626 627 return (0); 628 } 629 630 /* 631 * Enable multicast routing 632 */ 633 static int 634 ip_mrouter_init(so, m) 635 struct socket *so; 636 struct mbuf *m; 637 { 638 int *v; 639 640 if (mrtdebug) 641 log(LOG_DEBUG, 642 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 643 so->so_type, so->so_proto->pr_protocol); 644 645 if (so->so_type != SOCK_RAW || 646 so->so_proto->pr_protocol != IPPROTO_IGMP) 647 return (EOPNOTSUPP); 648 649 if (m == NULL || m->m_len < sizeof(int)) 650 return (EINVAL); 651 652 v = mtod(m, int *); 653 if (*v != 1) 654 return (EINVAL); 655 656 if (ip_mrouter != NULL) 657 return (EADDRINUSE); 658 659 ip_mrouter = so; 660 661 mfchashtbl = 662 hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash); 663 bzero((caddr_t)nexpire, sizeof(nexpire)); 664 665 pim_assert = 0; 666 667 callout_init(&expire_upcalls_ch); 668 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 669 expire_upcalls, NULL); 670 671 callout_init(&bw_upcalls_ch); 672 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 673 expire_bw_upcalls_send, NULL); 674 675 callout_init(&bw_meter_ch); 676 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 677 expire_bw_meter_process, NULL); 678 679 if (mrtdebug) 680 log(LOG_DEBUG, "ip_mrouter_init\n"); 681 682 return (0); 683 } 684 685 /* 686 * Disable multicast routing 687 */ 688 int 689 ip_mrouter_done() 690 { 691 vifi_t vifi; 692 struct vif *vifp; 693 int i; 694 int s; 695 696 s = splsoftnet(); 697 698 /* Clear out all the vifs currently in use. */ 699 for (vifi = 0; vifi < numvifs; vifi++) { 700 vifp = &viftable[vifi]; 701 if (!in_nullhost(vifp->v_lcl_addr)) 702 reset_vif(vifp); 703 } 704 705 numvifs = 0; 706 pim_assert = 0; 707 mrt_api_config = 0; 708 709 callout_stop(&expire_upcalls_ch); 710 callout_stop(&bw_upcalls_ch); 711 callout_stop(&bw_meter_ch); 712 713 /* 714 * Free all multicast forwarding cache entries. 715 */ 716 for (i = 0; i < MFCTBLSIZ; i++) { 717 struct mfc *rt, *nrt; 718 719 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 720 nrt = LIST_NEXT(rt, mfc_hash); 721 722 expire_mfc(rt); 723 } 724 } 725 726 bzero((caddr_t)nexpire, sizeof(nexpire)); 727 free(mfchashtbl, M_MRTABLE); 728 mfchashtbl = NULL; 729 730 bw_upcalls_n = 0; 731 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 732 733 /* Reset de-encapsulation cache. */ 734 735 ip_mrouter = NULL; 736 737 splx(s); 738 739 if (mrtdebug) 740 log(LOG_DEBUG, "ip_mrouter_done\n"); 741 742 return (0); 743 } 744 745 void 746 ip_mrouter_detach(ifp) 747 struct ifnet *ifp; 748 { 749 int vifi, i; 750 struct vif *vifp; 751 struct mfc *rt; 752 struct rtdetq *rte; 753 754 /* XXX not sure about side effect to userland routing daemon */ 755 for (vifi = 0; vifi < numvifs; vifi++) { 756 vifp = &viftable[vifi]; 757 if (vifp->v_ifp == ifp) 758 reset_vif(vifp); 759 } 760 for (i = 0; i < MFCTBLSIZ; i++) { 761 if (nexpire[i] == 0) 762 continue; 763 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) { 764 for (rte = rt->mfc_stall; rte; rte = rte->next) { 765 if (rte->ifp == ifp) 766 rte->ifp = NULL; 767 } 768 } 769 } 770 } 771 772 static int 773 get_version(m) 774 struct mbuf *m; 775 { 776 int *v = mtod(m, int *); 777 778 *v = 0x0305; /* XXX !!!! */ 779 m->m_len = sizeof(int); 780 return (0); 781 } 782 783 /* 784 * Set PIM assert processing global 785 */ 786 static int 787 set_assert(m) 788 struct mbuf *m; 789 { 790 int *i; 791 792 if (m == NULL || m->m_len < sizeof(int)) 793 return (EINVAL); 794 795 i = mtod(m, int *); 796 pim_assert = !!*i; 797 return (0); 798 } 799 800 /* 801 * Get PIM assert processing global 802 */ 803 static int 804 get_assert(m) 805 struct mbuf *m; 806 { 807 int *i = mtod(m, int *); 808 809 *i = pim_assert; 810 m->m_len = sizeof(int); 811 return (0); 812 } 813 814 /* 815 * Configure API capabilities 816 */ 817 static int 818 set_api_config(struct mbuf *m) 819 { 820 int i; 821 u_int32_t *apival; 822 823 if (m == NULL || m->m_len < sizeof(u_int32_t)) 824 return (EINVAL); 825 826 apival = mtod(m, u_int32_t *); 827 828 /* 829 * We can set the API capabilities only if it is the first operation 830 * after MRT_INIT. I.e.: 831 * - there are no vifs installed 832 * - pim_assert is not enabled 833 * - the MFC table is empty 834 */ 835 if (numvifs > 0) { 836 *apival = 0; 837 return (EPERM); 838 } 839 if (pim_assert) { 840 *apival = 0; 841 return (EPERM); 842 } 843 for (i = 0; i < MFCTBLSIZ; i++) { 844 if (LIST_FIRST(&mfchashtbl[i]) != NULL) { 845 *apival = 0; 846 return (EPERM); 847 } 848 } 849 850 mrt_api_config = *apival & mrt_api_support; 851 *apival = mrt_api_config; 852 853 return (0); 854 } 855 856 /* 857 * Get API capabilities 858 */ 859 static int 860 get_api_support(struct mbuf *m) 861 { 862 u_int32_t *apival; 863 864 if (m == NULL || m->m_len < sizeof(u_int32_t)) 865 return (EINVAL); 866 867 apival = mtod(m, u_int32_t *); 868 869 *apival = mrt_api_support; 870 871 return (0); 872 } 873 874 /* 875 * Get API configured capabilities 876 */ 877 static int 878 get_api_config(struct mbuf *m) 879 { 880 u_int32_t *apival; 881 882 if (m == NULL || m->m_len < sizeof(u_int32_t)) 883 return (EINVAL); 884 885 apival = mtod(m, u_int32_t *); 886 887 *apival = mrt_api_config; 888 889 return (0); 890 } 891 892 static struct sockaddr_in sin = { sizeof(sin), AF_INET }; 893 894 /* 895 * Add a vif to the vif table 896 */ 897 static int 898 add_vif(m) 899 struct mbuf *m; 900 { 901 struct vifctl *vifcp; 902 struct vif *vifp; 903 struct ifaddr *ifa; 904 struct ifnet *ifp; 905 struct ifreq ifr; 906 int error, s; 907 908 if (m == NULL || m->m_len < sizeof(struct vifctl)) 909 return (EINVAL); 910 911 vifcp = mtod(m, struct vifctl *); 912 if (vifcp->vifc_vifi >= MAXVIFS) 913 return (EINVAL); 914 if (in_nullhost(vifcp->vifc_lcl_addr)) 915 return (EADDRNOTAVAIL); 916 917 vifp = &viftable[vifcp->vifc_vifi]; 918 if (!in_nullhost(vifp->v_lcl_addr)) 919 return (EADDRINUSE); 920 921 /* Find the interface with an address in AF_INET family. */ 922 #ifdef PIM 923 if (vifcp->vifc_flags & VIFF_REGISTER) { 924 /* 925 * XXX: Because VIFF_REGISTER does not really need a valid 926 * local interface (e.g. it could be 127.0.0.2), we don't 927 * check its address. 928 */ 929 ifp = NULL; 930 } else 931 #endif 932 { 933 sin.sin_addr = vifcp->vifc_lcl_addr; 934 ifa = ifa_ifwithaddr(sintosa(&sin)); 935 if (ifa == NULL) 936 return (EADDRNOTAVAIL); 937 ifp = ifa->ifa_ifp; 938 } 939 940 if (vifcp->vifc_flags & VIFF_TUNNEL) { 941 if (vifcp->vifc_flags & VIFF_SRCRT) { 942 log(LOG_ERR, "source routed tunnels not supported\n"); 943 return (EOPNOTSUPP); 944 } 945 946 /* attach this vif to decapsulator dispatch table */ 947 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 948 vif_encapcheck, &vif_protosw, vifp); 949 if (!vifp->v_encap_cookie) 950 return (EINVAL); 951 952 /* Create a fake encapsulation interface. */ 953 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 954 bzero(ifp, sizeof(*ifp)); 955 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 956 "mdecap%d", vifcp->vifc_vifi); 957 958 /* Prepare cached route entry. */ 959 bzero(&vifp->v_route, sizeof(vifp->v_route)); 960 #ifdef PIM 961 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 962 ifp = &multicast_register_if; 963 if (mrtdebug) 964 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 965 (void *)ifp); 966 if (reg_vif_num == VIFI_INVALID) { 967 bzero(ifp, sizeof(*ifp)); 968 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 969 "register_vif"); 970 ifp->if_flags = IFF_LOOPBACK; 971 bzero(&vifp->v_route, sizeof(vifp->v_route)); 972 reg_vif_num = vifcp->vifc_vifi; 973 } 974 #endif 975 } else { 976 /* Make sure the interface supports multicast. */ 977 if ((ifp->if_flags & IFF_MULTICAST) == 0) 978 return (EOPNOTSUPP); 979 980 /* Enable promiscuous reception of all IP multicasts. */ 981 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 982 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 983 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 984 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 985 if (error) 986 return (error); 987 } 988 989 s = splsoftnet(); 990 991 /* Define parameters for the tbf structure. */ 992 vifp->tbf_q = NULL; 993 vifp->tbf_t = &vifp->tbf_q; 994 microtime(&vifp->tbf_last_pkt_t); 995 vifp->tbf_n_tok = 0; 996 vifp->tbf_q_len = 0; 997 vifp->tbf_max_q_len = MAXQSIZE; 998 999 vifp->v_flags = vifcp->vifc_flags; 1000 vifp->v_threshold = vifcp->vifc_threshold; 1001 /* scaling up here allows division by 1024 in critical code */ 1002 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 1003 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 1004 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 1005 vifp->v_ifp = ifp; 1006 /* Initialize per vif pkt counters. */ 1007 vifp->v_pkt_in = 0; 1008 vifp->v_pkt_out = 0; 1009 vifp->v_bytes_in = 0; 1010 vifp->v_bytes_out = 0; 1011 1012 callout_init(&vifp->v_repq_ch); 1013 1014 #ifdef RSVP_ISI 1015 vifp->v_rsvp_on = 0; 1016 vifp->v_rsvpd = NULL; 1017 #endif /* RSVP_ISI */ 1018 1019 splx(s); 1020 1021 /* Adjust numvifs up if the vifi is higher than numvifs. */ 1022 if (numvifs <= vifcp->vifc_vifi) 1023 numvifs = vifcp->vifc_vifi + 1; 1024 1025 if (mrtdebug) 1026 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 1027 vifcp->vifc_vifi, 1028 ntohl(vifcp->vifc_lcl_addr.s_addr), 1029 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 1030 ntohl(vifcp->vifc_rmt_addr.s_addr), 1031 vifcp->vifc_threshold, 1032 vifcp->vifc_rate_limit); 1033 1034 return (0); 1035 } 1036 1037 void 1038 reset_vif(vifp) 1039 struct vif *vifp; 1040 { 1041 struct mbuf *m, *n; 1042 struct ifnet *ifp; 1043 struct ifreq ifr; 1044 1045 callout_stop(&vifp->v_repq_ch); 1046 1047 /* detach this vif from decapsulator dispatch table */ 1048 encap_detach(vifp->v_encap_cookie); 1049 vifp->v_encap_cookie = NULL; 1050 1051 /* 1052 * Free packets queued at the interface 1053 */ 1054 for (m = vifp->tbf_q; m != NULL; m = n) { 1055 n = m->m_nextpkt; 1056 m_freem(m); 1057 } 1058 1059 if (vifp->v_flags & VIFF_TUNNEL) { 1060 free(vifp->v_ifp, M_MRTABLE); 1061 if (vifp == last_encap_vif) { 1062 last_encap_vif = NULL; 1063 last_encap_src = zeroin_addr; 1064 } 1065 } else if (vifp->v_flags & VIFF_REGISTER) { 1066 #ifdef PIM 1067 reg_vif_num = VIFI_INVALID; 1068 #endif 1069 } else { 1070 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 1071 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 1072 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 1073 ifp = vifp->v_ifp; 1074 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 1075 } 1076 bzero((caddr_t)vifp, sizeof(*vifp)); 1077 } 1078 1079 /* 1080 * Delete a vif from the vif table 1081 */ 1082 static int 1083 del_vif(m) 1084 struct mbuf *m; 1085 { 1086 vifi_t *vifip; 1087 struct vif *vifp; 1088 vifi_t vifi; 1089 int s; 1090 1091 if (m == NULL || m->m_len < sizeof(vifi_t)) 1092 return (EINVAL); 1093 1094 vifip = mtod(m, vifi_t *); 1095 if (*vifip >= numvifs) 1096 return (EINVAL); 1097 1098 vifp = &viftable[*vifip]; 1099 if (in_nullhost(vifp->v_lcl_addr)) 1100 return (EADDRNOTAVAIL); 1101 1102 s = splsoftnet(); 1103 1104 reset_vif(vifp); 1105 1106 /* Adjust numvifs down */ 1107 for (vifi = numvifs; vifi > 0; vifi--) 1108 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr)) 1109 break; 1110 numvifs = vifi; 1111 1112 splx(s); 1113 1114 if (mrtdebug) 1115 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 1116 1117 return (0); 1118 } 1119 1120 /* 1121 * update an mfc entry without resetting counters and S,G addresses. 1122 */ 1123 static void 1124 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1125 { 1126 int i; 1127 1128 rt->mfc_parent = mfccp->mfcc_parent; 1129 for (i = 0; i < numvifs; i++) { 1130 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1131 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1132 MRT_MFC_FLAGS_ALL; 1133 } 1134 /* set the RP address */ 1135 if (mrt_api_config & MRT_MFC_RP) 1136 rt->mfc_rp = mfccp->mfcc_rp; 1137 else 1138 rt->mfc_rp = zeroin_addr; 1139 } 1140 1141 /* 1142 * fully initialize an mfc entry from the parameter. 1143 */ 1144 static void 1145 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1146 { 1147 rt->mfc_origin = mfccp->mfcc_origin; 1148 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1149 1150 update_mfc_params(rt, mfccp); 1151 1152 /* initialize pkt counters per src-grp */ 1153 rt->mfc_pkt_cnt = 0; 1154 rt->mfc_byte_cnt = 0; 1155 rt->mfc_wrong_if = 0; 1156 timerclear(&rt->mfc_last_assert); 1157 } 1158 1159 static void 1160 expire_mfc(rt) 1161 struct mfc *rt; 1162 { 1163 struct rtdetq *rte, *nrte; 1164 1165 free_bw_list(rt->mfc_bw_meter); 1166 1167 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 1168 nrte = rte->next; 1169 m_freem(rte->m); 1170 free(rte, M_MRTABLE); 1171 } 1172 1173 LIST_REMOVE(rt, mfc_hash); 1174 free(rt, M_MRTABLE); 1175 } 1176 1177 /* 1178 * Add an mfc entry 1179 */ 1180 static int 1181 add_mfc(m) 1182 struct mbuf *m; 1183 { 1184 struct mfcctl2 mfcctl2; 1185 struct mfcctl2 *mfccp; 1186 struct mfc *rt; 1187 u_int32_t hash = 0; 1188 struct rtdetq *rte, *nrte; 1189 u_short nstl; 1190 int s; 1191 int mfcctl_size = sizeof(struct mfcctl); 1192 1193 if (mrt_api_config & MRT_API_FLAGS_ALL) 1194 mfcctl_size = sizeof(struct mfcctl2); 1195 1196 if (m == NULL || m->m_len < mfcctl_size) 1197 return (EINVAL); 1198 1199 /* 1200 * select data size depending on API version. 1201 */ 1202 if (mrt_api_config & MRT_API_FLAGS_ALL) { 1203 struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *); 1204 bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2)); 1205 } else { 1206 struct mfcctl *mp = mtod(m, struct mfcctl *); 1207 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp)); 1208 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl), 1209 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1210 } 1211 mfccp = &mfcctl2; 1212 1213 s = splsoftnet(); 1214 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1215 1216 /* If an entry already exists, just update the fields */ 1217 if (rt) { 1218 if (mrtdebug & DEBUG_MFC) 1219 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n", 1220 ntohl(mfccp->mfcc_origin.s_addr), 1221 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1222 mfccp->mfcc_parent); 1223 1224 update_mfc_params(rt, mfccp); 1225 1226 splx(s); 1227 return (0); 1228 } 1229 1230 /* 1231 * Find the entry for which the upcall was made and update 1232 */ 1233 nstl = 0; 1234 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1235 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1236 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1237 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1238 rt->mfc_stall != NULL) { 1239 if (nstl++) 1240 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 1241 "multiple kernel entries", 1242 ntohl(mfccp->mfcc_origin.s_addr), 1243 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1244 mfccp->mfcc_parent, rt->mfc_stall); 1245 1246 if (mrtdebug & DEBUG_MFC) 1247 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n", 1248 ntohl(mfccp->mfcc_origin.s_addr), 1249 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1250 mfccp->mfcc_parent, rt->mfc_stall); 1251 1252 rte = rt->mfc_stall; 1253 init_mfc_params(rt, mfccp); 1254 rt->mfc_stall = NULL; 1255 1256 rt->mfc_expire = 0; /* Don't clean this guy up */ 1257 nexpire[hash]--; 1258 1259 /* free packets Qed at the end of this entry */ 1260 for (; rte != NULL; rte = nrte) { 1261 nrte = rte->next; 1262 if (rte->ifp) { 1263 #ifdef RSVP_ISI 1264 ip_mdq(rte->m, rte->ifp, rt, -1); 1265 #else 1266 ip_mdq(rte->m, rte->ifp, rt); 1267 #endif /* RSVP_ISI */ 1268 } 1269 m_freem(rte->m); 1270 #ifdef UPCALL_TIMING 1271 collate(&rte->t); 1272 #endif /* UPCALL_TIMING */ 1273 free(rte, M_MRTABLE); 1274 } 1275 } 1276 } 1277 1278 /* 1279 * It is possible that an entry is being inserted without an upcall 1280 */ 1281 if (nstl == 0) { 1282 /* 1283 * No mfc; make a new one 1284 */ 1285 if (mrtdebug & DEBUG_MFC) 1286 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n", 1287 ntohl(mfccp->mfcc_origin.s_addr), 1288 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1289 mfccp->mfcc_parent); 1290 1291 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1292 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1293 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1294 init_mfc_params(rt, mfccp); 1295 if (rt->mfc_expire) 1296 nexpire[hash]--; 1297 rt->mfc_expire = 0; 1298 break; /* XXX */ 1299 } 1300 } 1301 if (rt == NULL) { /* no upcall, so make a new entry */ 1302 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1303 M_NOWAIT); 1304 if (rt == NULL) { 1305 splx(s); 1306 return (ENOBUFS); 1307 } 1308 1309 init_mfc_params(rt, mfccp); 1310 rt->mfc_expire = 0; 1311 rt->mfc_stall = NULL; 1312 rt->mfc_bw_meter = NULL; 1313 1314 /* insert new entry at head of hash chain */ 1315 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1316 } 1317 } 1318 1319 splx(s); 1320 return (0); 1321 } 1322 1323 #ifdef UPCALL_TIMING 1324 /* 1325 * collect delay statistics on the upcalls 1326 */ 1327 static void 1328 collate(t) 1329 struct timeval *t; 1330 { 1331 u_int32_t d; 1332 struct timeval tp; 1333 u_int32_t delta; 1334 1335 microtime(&tp); 1336 1337 if (timercmp(t, &tp, <)) { 1338 TV_DELTA(tp, *t, delta); 1339 1340 d = delta >> 10; 1341 if (d > 50) 1342 d = 50; 1343 1344 ++upcall_data[d]; 1345 } 1346 } 1347 #endif /* UPCALL_TIMING */ 1348 1349 /* 1350 * Delete an mfc entry 1351 */ 1352 static int 1353 del_mfc(m) 1354 struct mbuf *m; 1355 { 1356 struct mfcctl2 mfcctl2; 1357 struct mfcctl2 *mfccp; 1358 struct mfc *rt; 1359 int s; 1360 int mfcctl_size = sizeof(struct mfcctl); 1361 struct mfcctl *mp = mtod(m, struct mfcctl *); 1362 1363 /* 1364 * XXX: for deleting MFC entries the information in entries 1365 * of size "struct mfcctl" is sufficient. 1366 */ 1367 1368 if (m == NULL || m->m_len < mfcctl_size) 1369 return (EINVAL); 1370 1371 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp)); 1372 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl), 1373 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1374 1375 mfccp = &mfcctl2; 1376 1377 if (mrtdebug & DEBUG_MFC) 1378 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 1379 ntohl(mfccp->mfcc_origin.s_addr), 1380 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 1381 1382 s = splsoftnet(); 1383 1384 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1385 if (rt == NULL) { 1386 splx(s); 1387 return (EADDRNOTAVAIL); 1388 } 1389 1390 /* 1391 * free the bw_meter entries 1392 */ 1393 free_bw_list(rt->mfc_bw_meter); 1394 rt->mfc_bw_meter = NULL; 1395 1396 LIST_REMOVE(rt, mfc_hash); 1397 free(rt, M_MRTABLE); 1398 1399 splx(s); 1400 return (0); 1401 } 1402 1403 static int 1404 socket_send(s, mm, src) 1405 struct socket *s; 1406 struct mbuf *mm; 1407 struct sockaddr_in *src; 1408 { 1409 if (s) { 1410 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, 1411 (struct mbuf *)NULL) != 0) { 1412 sorwakeup(s); 1413 return (0); 1414 } 1415 } 1416 m_freem(mm); 1417 return (-1); 1418 } 1419 1420 /* 1421 * IP multicast forwarding function. This function assumes that the packet 1422 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1423 * pointed to by "ifp", and the packet is to be relayed to other networks 1424 * that have members of the packet's destination IP multicast group. 1425 * 1426 * The packet is returned unscathed to the caller, unless it is 1427 * erroneous, in which case a non-zero return value tells the caller to 1428 * discard it. 1429 */ 1430 1431 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1432 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1433 1434 int 1435 #ifdef RSVP_ISI 1436 ip_mforward(m, ifp, imo) 1437 #else 1438 ip_mforward(m, ifp) 1439 #endif /* RSVP_ISI */ 1440 struct mbuf *m; 1441 struct ifnet *ifp; 1442 #ifdef RSVP_ISI 1443 struct ip_moptions *imo; 1444 #endif /* RSVP_ISI */ 1445 { 1446 struct ip *ip = mtod(m, struct ip *); 1447 struct mfc *rt; 1448 static int srctun = 0; 1449 struct mbuf *mm; 1450 int s; 1451 vifi_t vifi; 1452 1453 if (mrtdebug & DEBUG_FORWARD) 1454 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1455 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1456 1457 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1458 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1459 /* 1460 * Packet arrived via a physical interface or 1461 * an encapsulated tunnel or a register_vif. 1462 */ 1463 } else { 1464 /* 1465 * Packet arrived through a source-route tunnel. 1466 * Source-route tunnels are no longer supported. 1467 */ 1468 if ((srctun++ % 1000) == 0) 1469 log(LOG_ERR, 1470 "ip_mforward: received source-routed packet from %x\n", 1471 ntohl(ip->ip_src.s_addr)); 1472 1473 return (1); 1474 } 1475 1476 #ifdef RSVP_ISI 1477 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1478 if (ip->ip_ttl < 255) 1479 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1480 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1481 struct vif *vifp = viftable + vifi; 1482 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1483 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1484 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1485 vifp->v_ifp->if_xname); 1486 } 1487 return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi)); 1488 } 1489 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1490 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1491 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1492 } 1493 #endif /* RSVP_ISI */ 1494 1495 /* 1496 * Don't forward a packet with time-to-live of zero or one, 1497 * or a packet destined to a local-only group. 1498 */ 1499 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1500 return (0); 1501 1502 /* 1503 * Determine forwarding vifs from the forwarding cache table 1504 */ 1505 s = splsoftnet(); 1506 ++mrtstat.mrts_mfc_lookups; 1507 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1508 1509 /* Entry exists, so forward if necessary */ 1510 if (rt != NULL) { 1511 splx(s); 1512 #ifdef RSVP_ISI 1513 return (ip_mdq(m, ifp, rt, -1)); 1514 #else 1515 return (ip_mdq(m, ifp, rt)); 1516 #endif /* RSVP_ISI */ 1517 } else { 1518 /* 1519 * If we don't have a route for packet's origin, 1520 * Make a copy of the packet & send message to routing daemon 1521 */ 1522 1523 struct mbuf *mb0; 1524 struct rtdetq *rte; 1525 u_int32_t hash; 1526 int hlen = ip->ip_hl << 2; 1527 #ifdef UPCALL_TIMING 1528 struct timeval tp; 1529 1530 microtime(&tp); 1531 #endif /* UPCALL_TIMING */ 1532 1533 ++mrtstat.mrts_mfc_misses; 1534 1535 mrtstat.mrts_no_route++; 1536 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1537 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1538 ntohl(ip->ip_src.s_addr), 1539 ntohl(ip->ip_dst.s_addr)); 1540 1541 /* 1542 * Allocate mbufs early so that we don't do extra work if we are 1543 * just going to fail anyway. Make sure to pullup the header so 1544 * that other people can't step on it. 1545 */ 1546 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1547 M_NOWAIT); 1548 if (rte == NULL) { 1549 splx(s); 1550 return (ENOBUFS); 1551 } 1552 mb0 = m_copy(m, 0, M_COPYALL); 1553 M_PULLUP(mb0, hlen); 1554 if (mb0 == NULL) { 1555 free(rte, M_MRTABLE); 1556 splx(s); 1557 return (ENOBUFS); 1558 } 1559 1560 /* is there an upcall waiting for this flow? */ 1561 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1562 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1563 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1564 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1565 rt->mfc_stall != NULL) 1566 break; 1567 } 1568 1569 if (rt == NULL) { 1570 int i; 1571 struct igmpmsg *im; 1572 1573 /* 1574 * Locate the vifi for the incoming interface for 1575 * this packet. 1576 * If none found, drop packet. 1577 */ 1578 for (vifi = 0; vifi < numvifs && 1579 viftable[vifi].v_ifp != ifp; vifi++) 1580 ; 1581 if (vifi >= numvifs) /* vif not found, drop packet */ 1582 goto non_fatal; 1583 1584 /* no upcall, so make a new entry */ 1585 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1586 M_NOWAIT); 1587 if (rt == NULL) 1588 goto fail; 1589 1590 /* 1591 * Make a copy of the header to send to the user level 1592 * process 1593 */ 1594 mm = m_copy(m, 0, hlen); 1595 M_PULLUP(mm, hlen); 1596 if (mm == NULL) 1597 goto fail1; 1598 1599 /* 1600 * Send message to routing daemon to install 1601 * a route into the kernel table 1602 */ 1603 1604 im = mtod(mm, struct igmpmsg *); 1605 im->im_msgtype = IGMPMSG_NOCACHE; 1606 im->im_mbz = 0; 1607 im->im_vif = vifi; 1608 1609 mrtstat.mrts_upcalls++; 1610 1611 sin.sin_addr = ip->ip_src; 1612 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1613 log(LOG_WARNING, 1614 "ip_mforward: ip_mrouter socket queue full\n"); 1615 ++mrtstat.mrts_upq_sockfull; 1616 fail1: 1617 free(rt, M_MRTABLE); 1618 fail: 1619 free(rte, M_MRTABLE); 1620 m_freem(mb0); 1621 splx(s); 1622 return (ENOBUFS); 1623 } 1624 1625 /* insert new entry at head of hash chain */ 1626 rt->mfc_origin = ip->ip_src; 1627 rt->mfc_mcastgrp = ip->ip_dst; 1628 rt->mfc_pkt_cnt = 0; 1629 rt->mfc_byte_cnt = 0; 1630 rt->mfc_wrong_if = 0; 1631 rt->mfc_expire = UPCALL_EXPIRE; 1632 nexpire[hash]++; 1633 for (i = 0; i < numvifs; i++) { 1634 rt->mfc_ttls[i] = 0; 1635 rt->mfc_flags[i] = 0; 1636 } 1637 rt->mfc_parent = -1; 1638 1639 /* clear the RP address */ 1640 rt->mfc_rp = zeroin_addr; 1641 1642 rt->mfc_bw_meter = NULL; 1643 1644 /* link into table */ 1645 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1646 /* Add this entry to the end of the queue */ 1647 rt->mfc_stall = rte; 1648 } else { 1649 /* determine if q has overflowed */ 1650 struct rtdetq **p; 1651 int npkts = 0; 1652 1653 /* 1654 * XXX ouch! we need to append to the list, but we 1655 * only have a pointer to the front, so we have to 1656 * scan the entire list every time. 1657 */ 1658 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1659 if (++npkts > MAX_UPQ) { 1660 mrtstat.mrts_upq_ovflw++; 1661 non_fatal: 1662 free(rte, M_MRTABLE); 1663 m_freem(mb0); 1664 splx(s); 1665 return (0); 1666 } 1667 1668 /* Add this entry to the end of the queue */ 1669 *p = rte; 1670 } 1671 1672 rte->next = NULL; 1673 rte->m = mb0; 1674 rte->ifp = ifp; 1675 #ifdef UPCALL_TIMING 1676 rte->t = tp; 1677 #endif /* UPCALL_TIMING */ 1678 1679 splx(s); 1680 1681 return (0); 1682 } 1683 } 1684 1685 1686 /*ARGSUSED*/ 1687 static void 1688 expire_upcalls(v) 1689 void *v; 1690 { 1691 int i; 1692 int s; 1693 1694 s = splsoftnet(); 1695 1696 for (i = 0; i < MFCTBLSIZ; i++) { 1697 struct mfc *rt, *nrt; 1698 1699 if (nexpire[i] == 0) 1700 continue; 1701 1702 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1703 nrt = LIST_NEXT(rt, mfc_hash); 1704 1705 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1706 continue; 1707 nexpire[i]--; 1708 1709 /* 1710 * free the bw_meter entries 1711 */ 1712 while (rt->mfc_bw_meter != NULL) { 1713 struct bw_meter *x = rt->mfc_bw_meter; 1714 1715 rt->mfc_bw_meter = x->bm_mfc_next; 1716 free(x, M_BWMETER); 1717 } 1718 1719 ++mrtstat.mrts_cache_cleanups; 1720 if (mrtdebug & DEBUG_EXPIRE) 1721 log(LOG_DEBUG, 1722 "expire_upcalls: expiring (%x %x)\n", 1723 ntohl(rt->mfc_origin.s_addr), 1724 ntohl(rt->mfc_mcastgrp.s_addr)); 1725 1726 expire_mfc(rt); 1727 } 1728 } 1729 1730 splx(s); 1731 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1732 expire_upcalls, NULL); 1733 } 1734 1735 /* 1736 * Packet forwarding routine once entry in the cache is made 1737 */ 1738 static int 1739 #ifdef RSVP_ISI 1740 ip_mdq(m, ifp, rt, xmt_vif) 1741 #else 1742 ip_mdq(m, ifp, rt) 1743 #endif /* RSVP_ISI */ 1744 struct mbuf *m; 1745 struct ifnet *ifp; 1746 struct mfc *rt; 1747 #ifdef RSVP_ISI 1748 vifi_t xmt_vif; 1749 #endif /* RSVP_ISI */ 1750 { 1751 struct ip *ip = mtod(m, struct ip *); 1752 vifi_t vifi; 1753 struct vif *vifp; 1754 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2); 1755 1756 /* 1757 * Macro to send packet on vif. Since RSVP packets don't get counted on 1758 * input, they shouldn't get counted on output, so statistics keeping is 1759 * separate. 1760 */ 1761 #define MC_SEND(ip, vifp, m) do { \ 1762 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1763 encap_send((ip), (vifp), (m)); \ 1764 else \ 1765 phyint_send((ip), (vifp), (m)); \ 1766 } while (/*CONSTCOND*/ 0) 1767 1768 #ifdef RSVP_ISI 1769 /* 1770 * If xmt_vif is not -1, send on only the requested vif. 1771 * 1772 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1773 */ 1774 if (xmt_vif < numvifs) { 1775 #ifdef PIM 1776 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1777 pim_register_send(ip, viftable + xmt_vif, m, rt); 1778 else 1779 #endif 1780 MC_SEND(ip, viftable + xmt_vif, m); 1781 return (1); 1782 } 1783 #endif /* RSVP_ISI */ 1784 1785 /* 1786 * Don't forward if it didn't arrive from the parent vif for its origin. 1787 */ 1788 vifi = rt->mfc_parent; 1789 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1790 /* came in the wrong interface */ 1791 if (mrtdebug & DEBUG_FORWARD) 1792 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1793 ifp, vifi, 1794 vifi >= numvifs ? 0 : viftable[vifi].v_ifp); 1795 ++mrtstat.mrts_wrong_if; 1796 ++rt->mfc_wrong_if; 1797 /* 1798 * If we are doing PIM assert processing, send a message 1799 * to the routing daemon. 1800 * 1801 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1802 * can complete the SPT switch, regardless of the type 1803 * of the iif (broadcast media, GRE tunnel, etc). 1804 */ 1805 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1806 struct timeval now; 1807 u_int32_t delta; 1808 1809 #ifdef PIM 1810 if (ifp == &multicast_register_if) 1811 pimstat.pims_rcv_registers_wrongiif++; 1812 #endif 1813 1814 /* Get vifi for the incoming packet */ 1815 for (vifi = 0; 1816 vifi < numvifs && viftable[vifi].v_ifp != ifp; 1817 vifi++) 1818 ; 1819 if (vifi >= numvifs) { 1820 /* The iif is not found: ignore the packet. */ 1821 return (0); 1822 } 1823 1824 if (rt->mfc_flags[vifi] & 1825 MRT_MFC_FLAGS_DISABLE_WRONGVIF) { 1826 /* WRONGVIF disabled: ignore the packet */ 1827 return (0); 1828 } 1829 1830 microtime(&now); 1831 1832 TV_DELTA(rt->mfc_last_assert, now, delta); 1833 1834 if (delta > ASSERT_MSG_TIME) { 1835 struct igmpmsg *im; 1836 int hlen = ip->ip_hl << 2; 1837 struct mbuf *mm = m_copy(m, 0, hlen); 1838 1839 M_PULLUP(mm, hlen); 1840 if (mm == NULL) 1841 return (ENOBUFS); 1842 1843 rt->mfc_last_assert = now; 1844 1845 im = mtod(mm, struct igmpmsg *); 1846 im->im_msgtype = IGMPMSG_WRONGVIF; 1847 im->im_mbz = 0; 1848 im->im_vif = vifi; 1849 1850 mrtstat.mrts_upcalls++; 1851 1852 sin.sin_addr = im->im_src; 1853 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1854 log(LOG_WARNING, 1855 "ip_mforward: ip_mrouter socket queue full\n"); 1856 ++mrtstat.mrts_upq_sockfull; 1857 return (ENOBUFS); 1858 } 1859 } 1860 } 1861 return (0); 1862 } 1863 1864 /* If I sourced this packet, it counts as output, else it was input. */ 1865 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1866 viftable[vifi].v_pkt_out++; 1867 viftable[vifi].v_bytes_out += plen; 1868 } else { 1869 viftable[vifi].v_pkt_in++; 1870 viftable[vifi].v_bytes_in += plen; 1871 } 1872 rt->mfc_pkt_cnt++; 1873 rt->mfc_byte_cnt += plen; 1874 1875 /* 1876 * For each vif, decide if a copy of the packet should be forwarded. 1877 * Forward if: 1878 * - the ttl exceeds the vif's threshold 1879 * - there are group members downstream on interface 1880 */ 1881 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1882 if ((rt->mfc_ttls[vifi] > 0) && 1883 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1884 vifp->v_pkt_out++; 1885 vifp->v_bytes_out += plen; 1886 #ifdef PIM 1887 if (vifp->v_flags & VIFF_REGISTER) 1888 pim_register_send(ip, vifp, m, rt); 1889 else 1890 #endif 1891 MC_SEND(ip, vifp, m); 1892 } 1893 1894 /* 1895 * Perform upcall-related bw measuring. 1896 */ 1897 if (rt->mfc_bw_meter != NULL) { 1898 struct bw_meter *x; 1899 struct timeval now; 1900 1901 microtime(&now); 1902 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1903 bw_meter_receive_packet(x, plen, &now); 1904 } 1905 1906 return (0); 1907 } 1908 1909 #ifdef RSVP_ISI 1910 /* 1911 * check if a vif number is legal/ok. This is used by ip_output. 1912 */ 1913 int 1914 legal_vif_num(vif) 1915 int vif; 1916 { 1917 if (vif >= 0 && vif < numvifs) 1918 return (1); 1919 else 1920 return (0); 1921 } 1922 #endif /* RSVP_ISI */ 1923 1924 static void 1925 phyint_send(ip, vifp, m) 1926 struct ip *ip; 1927 struct vif *vifp; 1928 struct mbuf *m; 1929 { 1930 struct mbuf *mb_copy; 1931 int hlen = ip->ip_hl << 2; 1932 1933 /* 1934 * Make a new reference to the packet; make sure that 1935 * the IP header is actually copied, not just referenced, 1936 * so that ip_output() only scribbles on the copy. 1937 */ 1938 mb_copy = m_copy(m, 0, M_COPYALL); 1939 M_PULLUP(mb_copy, hlen); 1940 if (mb_copy == NULL) 1941 return; 1942 1943 if (vifp->v_rate_limit <= 0) 1944 tbf_send_packet(vifp, mb_copy); 1945 else 1946 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1947 ntohs(ip->ip_len)); 1948 } 1949 1950 static void 1951 encap_send(ip, vifp, m) 1952 struct ip *ip; 1953 struct vif *vifp; 1954 struct mbuf *m; 1955 { 1956 struct mbuf *mb_copy; 1957 struct ip *ip_copy; 1958 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1959 1960 /* Take care of delayed checksums */ 1961 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1962 in_delayed_cksum(m); 1963 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1964 } 1965 1966 /* 1967 * copy the old packet & pullup it's IP header into the 1968 * new mbuf so we can modify it. Try to fill the new 1969 * mbuf since if we don't the ethernet driver will. 1970 */ 1971 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1972 if (mb_copy == NULL) 1973 return; 1974 mb_copy->m_data += max_linkhdr; 1975 mb_copy->m_pkthdr.len = len; 1976 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1977 1978 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1979 m_freem(mb_copy); 1980 return; 1981 } 1982 i = MHLEN - max_linkhdr; 1983 if (i > len) 1984 i = len; 1985 mb_copy = m_pullup(mb_copy, i); 1986 if (mb_copy == NULL) 1987 return; 1988 1989 /* 1990 * fill in the encapsulating IP header. 1991 */ 1992 ip_copy = mtod(mb_copy, struct ip *); 1993 *ip_copy = multicast_encap_iphdr; 1994 ip_copy->ip_id = ip_newid(); 1995 ip_copy->ip_len = htons(len); 1996 ip_copy->ip_src = vifp->v_lcl_addr; 1997 ip_copy->ip_dst = vifp->v_rmt_addr; 1998 1999 /* 2000 * turn the encapsulated IP header back into a valid one. 2001 */ 2002 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 2003 --ip->ip_ttl; 2004 ip->ip_sum = 0; 2005 mb_copy->m_data += sizeof(multicast_encap_iphdr); 2006 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2007 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 2008 2009 if (vifp->v_rate_limit <= 0) 2010 tbf_send_packet(vifp, mb_copy); 2011 else 2012 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len)); 2013 } 2014 2015 /* 2016 * De-encapsulate a packet and feed it back through ip input. 2017 */ 2018 static void 2019 vif_input(struct mbuf *m, ...) 2020 { 2021 int off, proto; 2022 va_list ap; 2023 struct vif *vifp; 2024 int s; 2025 struct ifqueue *ifq; 2026 2027 va_start(ap, m); 2028 off = va_arg(ap, int); 2029 proto = va_arg(ap, int); 2030 va_end(ap); 2031 2032 vifp = (struct vif *)encap_getarg(m); 2033 if (!vifp || proto != AF_INET) { 2034 m_freem(m); 2035 mrtstat.mrts_bad_tunnel++; 2036 return; 2037 } 2038 2039 m_adj(m, off); 2040 m->m_pkthdr.rcvif = vifp->v_ifp; 2041 ifq = &ipintrq; 2042 s = splnet(); 2043 if (IF_QFULL(ifq)) { 2044 IF_DROP(ifq); 2045 m_freem(m); 2046 } else { 2047 IF_ENQUEUE(ifq, m); 2048 /* 2049 * normally we would need a "schednetisr(NETISR_IP)" 2050 * here but we were called by ip_input and it is going 2051 * to loop back & try to dequeue the packet we just 2052 * queued as soon as we return so we avoid the 2053 * unnecessary software interrrupt. 2054 */ 2055 } 2056 splx(s); 2057 } 2058 2059 /* 2060 * Check if the packet should be grabbed by us. 2061 */ 2062 static int 2063 vif_encapcheck(m, off, proto, arg) 2064 const struct mbuf *m; 2065 int off; 2066 int proto; 2067 void *arg; 2068 { 2069 struct vif *vifp; 2070 struct ip ip; 2071 2072 #ifdef DIAGNOSTIC 2073 if (!arg || proto != IPPROTO_IPV4) 2074 panic("unexpected arg in vif_encapcheck"); 2075 #endif 2076 2077 /* 2078 * do not grab the packet if it's not to a multicast destination or if 2079 * we don't have an encapsulating tunnel with the source. 2080 * Note: This code assumes that the remote site IP address 2081 * uniquely identifies the tunnel (i.e., that this site has 2082 * at most one tunnel with the remote site). 2083 */ 2084 2085 /* LINTED const cast */ 2086 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip); 2087 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 2088 return 0; 2089 2090 /* LINTED const cast */ 2091 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip); 2092 if (!in_hosteq(ip.ip_src, last_encap_src)) { 2093 vifp = (struct vif *)arg; 2094 if (vifp->v_flags & VIFF_TUNNEL && 2095 in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 2096 ; 2097 else 2098 return 0; 2099 last_encap_vif = vifp; 2100 last_encap_src = ip.ip_src; 2101 } else 2102 vifp = last_encap_vif; 2103 2104 /* 32bit match, since we have checked ip_src only */ 2105 return 32; 2106 } 2107 2108 /* 2109 * Token bucket filter module 2110 */ 2111 static void 2112 tbf_control(vifp, m, ip, len) 2113 struct vif *vifp; 2114 struct mbuf *m; 2115 struct ip *ip; 2116 u_int32_t len; 2117 { 2118 2119 if (len > MAX_BKT_SIZE) { 2120 /* drop if packet is too large */ 2121 mrtstat.mrts_pkt2large++; 2122 m_freem(m); 2123 return; 2124 } 2125 2126 tbf_update_tokens(vifp); 2127 2128 /* 2129 * If there are enough tokens, and the queue is empty, send this packet 2130 * out immediately. Otherwise, try to insert it on this vif's queue. 2131 */ 2132 if (vifp->tbf_q_len == 0) { 2133 if (len <= vifp->tbf_n_tok) { 2134 vifp->tbf_n_tok -= len; 2135 tbf_send_packet(vifp, m); 2136 } else { 2137 /* queue packet and timeout till later */ 2138 tbf_queue(vifp, m); 2139 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2140 tbf_reprocess_q, vifp); 2141 } 2142 } else { 2143 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 2144 !tbf_dq_sel(vifp, ip)) { 2145 /* queue full, and couldn't make room */ 2146 mrtstat.mrts_q_overflow++; 2147 m_freem(m); 2148 } else { 2149 /* queue length low enough, or made room */ 2150 tbf_queue(vifp, m); 2151 tbf_process_q(vifp); 2152 } 2153 } 2154 } 2155 2156 /* 2157 * adds a packet to the queue at the interface 2158 */ 2159 static void 2160 tbf_queue(vifp, m) 2161 struct vif *vifp; 2162 struct mbuf *m; 2163 { 2164 int s = splsoftnet(); 2165 2166 /* insert at tail */ 2167 *vifp->tbf_t = m; 2168 vifp->tbf_t = &m->m_nextpkt; 2169 vifp->tbf_q_len++; 2170 2171 splx(s); 2172 } 2173 2174 2175 /* 2176 * processes the queue at the interface 2177 */ 2178 static void 2179 tbf_process_q(vifp) 2180 struct vif *vifp; 2181 { 2182 struct mbuf *m; 2183 int len; 2184 int s = splsoftnet(); 2185 2186 /* 2187 * Loop through the queue at the interface and send as many packets 2188 * as possible. 2189 */ 2190 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) { 2191 len = ntohs(mtod(m, struct ip *)->ip_len); 2192 2193 /* determine if the packet can be sent */ 2194 if (len <= vifp->tbf_n_tok) { 2195 /* if so, 2196 * reduce no of tokens, dequeue the packet, 2197 * send the packet. 2198 */ 2199 if ((vifp->tbf_q = m->m_nextpkt) == NULL) 2200 vifp->tbf_t = &vifp->tbf_q; 2201 --vifp->tbf_q_len; 2202 2203 m->m_nextpkt = NULL; 2204 vifp->tbf_n_tok -= len; 2205 tbf_send_packet(vifp, m); 2206 } else 2207 break; 2208 } 2209 splx(s); 2210 } 2211 2212 static void 2213 tbf_reprocess_q(arg) 2214 void *arg; 2215 { 2216 struct vif *vifp = arg; 2217 2218 if (ip_mrouter == NULL) 2219 return; 2220 2221 tbf_update_tokens(vifp); 2222 tbf_process_q(vifp); 2223 2224 if (vifp->tbf_q_len != 0) 2225 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2226 tbf_reprocess_q, vifp); 2227 } 2228 2229 /* function that will selectively discard a member of the queue 2230 * based on the precedence value and the priority 2231 */ 2232 static int 2233 tbf_dq_sel(vifp, ip) 2234 struct vif *vifp; 2235 struct ip *ip; 2236 { 2237 u_int p; 2238 struct mbuf **mp, *m; 2239 int s = splsoftnet(); 2240 2241 p = priority(vifp, ip); 2242 2243 for (mp = &vifp->tbf_q, m = *mp; 2244 m != NULL; 2245 mp = &m->m_nextpkt, m = *mp) { 2246 if (p > priority(vifp, mtod(m, struct ip *))) { 2247 if ((*mp = m->m_nextpkt) == NULL) 2248 vifp->tbf_t = mp; 2249 --vifp->tbf_q_len; 2250 2251 m_freem(m); 2252 mrtstat.mrts_drop_sel++; 2253 splx(s); 2254 return (1); 2255 } 2256 } 2257 splx(s); 2258 return (0); 2259 } 2260 2261 static void 2262 tbf_send_packet(vifp, m) 2263 struct vif *vifp; 2264 struct mbuf *m; 2265 { 2266 int error; 2267 int s = splsoftnet(); 2268 2269 if (vifp->v_flags & VIFF_TUNNEL) { 2270 /* If tunnel options */ 2271 ip_output(m, (struct mbuf *)NULL, &vifp->v_route, 2272 IP_FORWARDING, (struct ip_moptions *)NULL, 2273 (struct socket *)NULL); 2274 } else { 2275 /* if physical interface option, extract the options and then send */ 2276 struct ip_moptions imo; 2277 2278 imo.imo_multicast_ifp = vifp->v_ifp; 2279 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 2280 imo.imo_multicast_loop = 1; 2281 #ifdef RSVP_ISI 2282 imo.imo_multicast_vif = -1; 2283 #endif 2284 2285 error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL, 2286 IP_FORWARDING|IP_MULTICASTOPTS, &imo, 2287 (struct socket *)NULL); 2288 2289 if (mrtdebug & DEBUG_XMIT) 2290 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 2291 (long)(vifp - viftable), error); 2292 } 2293 splx(s); 2294 } 2295 2296 /* determine the current time and then 2297 * the elapsed time (between the last time and time now) 2298 * in milliseconds & update the no. of tokens in the bucket 2299 */ 2300 static void 2301 tbf_update_tokens(vifp) 2302 struct vif *vifp; 2303 { 2304 struct timeval tp; 2305 u_int32_t tm; 2306 int s = splsoftnet(); 2307 2308 microtime(&tp); 2309 2310 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 2311 2312 /* 2313 * This formula is actually 2314 * "time in seconds" * "bytes/second". 2315 * 2316 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2317 * 2318 * The (1000/1024) was introduced in add_vif to optimize 2319 * this divide into a shift. 2320 */ 2321 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 2322 vifp->tbf_last_pkt_t = tp; 2323 2324 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 2325 vifp->tbf_n_tok = MAX_BKT_SIZE; 2326 2327 splx(s); 2328 } 2329 2330 static int 2331 priority(vifp, ip) 2332 struct vif *vifp; 2333 struct ip *ip; 2334 { 2335 int prio = 50; /* the lowest priority -- default case */ 2336 2337 /* temporary hack; may add general packet classifier some day */ 2338 2339 /* 2340 * The UDP port space is divided up into four priority ranges: 2341 * [0, 16384) : unclassified - lowest priority 2342 * [16384, 32768) : audio - highest priority 2343 * [32768, 49152) : whiteboard - medium priority 2344 * [49152, 65536) : video - low priority 2345 */ 2346 if (ip->ip_p == IPPROTO_UDP) { 2347 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2348 2349 switch (ntohs(udp->uh_dport) & 0xc000) { 2350 case 0x4000: 2351 prio = 70; 2352 break; 2353 case 0x8000: 2354 prio = 60; 2355 break; 2356 case 0xc000: 2357 prio = 55; 2358 break; 2359 } 2360 2361 if (tbfdebug > 1) 2362 log(LOG_DEBUG, "port %x prio %d\n", 2363 ntohs(udp->uh_dport), prio); 2364 } 2365 2366 return (prio); 2367 } 2368 2369 /* 2370 * End of token bucket filter modifications 2371 */ 2372 #ifdef RSVP_ISI 2373 int 2374 ip_rsvp_vif_init(so, m) 2375 struct socket *so; 2376 struct mbuf *m; 2377 { 2378 int vifi, s; 2379 2380 if (rsvpdebug) 2381 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2382 so->so_type, so->so_proto->pr_protocol); 2383 2384 if (so->so_type != SOCK_RAW || 2385 so->so_proto->pr_protocol != IPPROTO_RSVP) 2386 return (EOPNOTSUPP); 2387 2388 /* Check mbuf. */ 2389 if (m == NULL || m->m_len != sizeof(int)) { 2390 return (EINVAL); 2391 } 2392 vifi = *(mtod(m, int *)); 2393 2394 if (rsvpdebug) 2395 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", 2396 vifi, rsvp_on); 2397 2398 s = splsoftnet(); 2399 2400 /* Check vif. */ 2401 if (!legal_vif_num(vifi)) { 2402 splx(s); 2403 return (EADDRNOTAVAIL); 2404 } 2405 2406 /* Check if socket is available. */ 2407 if (viftable[vifi].v_rsvpd != NULL) { 2408 splx(s); 2409 return (EADDRINUSE); 2410 } 2411 2412 viftable[vifi].v_rsvpd = so; 2413 /* 2414 * This may seem silly, but we need to be sure we don't over-increment 2415 * the RSVP counter, in case something slips up. 2416 */ 2417 if (!viftable[vifi].v_rsvp_on) { 2418 viftable[vifi].v_rsvp_on = 1; 2419 rsvp_on++; 2420 } 2421 2422 splx(s); 2423 return (0); 2424 } 2425 2426 int 2427 ip_rsvp_vif_done(so, m) 2428 struct socket *so; 2429 struct mbuf *m; 2430 { 2431 int vifi, s; 2432 2433 if (rsvpdebug) 2434 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2435 so->so_type, so->so_proto->pr_protocol); 2436 2437 if (so->so_type != SOCK_RAW || 2438 so->so_proto->pr_protocol != IPPROTO_RSVP) 2439 return (EOPNOTSUPP); 2440 2441 /* Check mbuf. */ 2442 if (m == NULL || m->m_len != sizeof(int)) { 2443 return (EINVAL); 2444 } 2445 vifi = *(mtod(m, int *)); 2446 2447 s = splsoftnet(); 2448 2449 /* Check vif. */ 2450 if (!legal_vif_num(vifi)) { 2451 splx(s); 2452 return (EADDRNOTAVAIL); 2453 } 2454 2455 if (rsvpdebug) 2456 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 2457 viftable[vifi].v_rsvpd, so); 2458 2459 viftable[vifi].v_rsvpd = NULL; 2460 /* 2461 * This may seem silly, but we need to be sure we don't over-decrement 2462 * the RSVP counter, in case something slips up. 2463 */ 2464 if (viftable[vifi].v_rsvp_on) { 2465 viftable[vifi].v_rsvp_on = 0; 2466 rsvp_on--; 2467 } 2468 2469 splx(s); 2470 return (0); 2471 } 2472 2473 void 2474 ip_rsvp_force_done(so) 2475 struct socket *so; 2476 { 2477 int vifi, s; 2478 2479 /* Don't bother if it is not the right type of socket. */ 2480 if (so->so_type != SOCK_RAW || 2481 so->so_proto->pr_protocol != IPPROTO_RSVP) 2482 return; 2483 2484 s = splsoftnet(); 2485 2486 /* 2487 * The socket may be attached to more than one vif...this 2488 * is perfectly legal. 2489 */ 2490 for (vifi = 0; vifi < numvifs; vifi++) { 2491 if (viftable[vifi].v_rsvpd == so) { 2492 viftable[vifi].v_rsvpd = NULL; 2493 /* 2494 * This may seem silly, but we need to be sure we don't 2495 * over-decrement the RSVP counter, in case something 2496 * slips up. 2497 */ 2498 if (viftable[vifi].v_rsvp_on) { 2499 viftable[vifi].v_rsvp_on = 0; 2500 rsvp_on--; 2501 } 2502 } 2503 } 2504 2505 splx(s); 2506 return; 2507 } 2508 2509 void 2510 rsvp_input(m, ifp) 2511 struct mbuf *m; 2512 struct ifnet *ifp; 2513 { 2514 int vifi, s; 2515 struct ip *ip = mtod(m, struct ip *); 2516 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 2517 2518 if (rsvpdebug) 2519 printf("rsvp_input: rsvp_on %d\n", rsvp_on); 2520 2521 /* 2522 * Can still get packets with rsvp_on = 0 if there is a local member 2523 * of the group to which the RSVP packet is addressed. But in this 2524 * case we want to throw the packet away. 2525 */ 2526 if (!rsvp_on) { 2527 m_freem(m); 2528 return; 2529 } 2530 2531 /* 2532 * If the old-style non-vif-associated socket is set, then use 2533 * it and ignore the new ones. 2534 */ 2535 if (ip_rsvpd != NULL) { 2536 if (rsvpdebug) 2537 printf("rsvp_input: " 2538 "Sending packet up old-style socket\n"); 2539 rip_input(m); /*XXX*/ 2540 return; 2541 } 2542 2543 s = splsoftnet(); 2544 2545 if (rsvpdebug) 2546 printf("rsvp_input: check vifs\n"); 2547 2548 /* Find which vif the packet arrived on. */ 2549 for (vifi = 0; vifi < numvifs; vifi++) { 2550 if (viftable[vifi].v_ifp == ifp) 2551 break; 2552 } 2553 2554 if (vifi == numvifs) { 2555 /* Can't find vif packet arrived on. Drop packet. */ 2556 if (rsvpdebug) 2557 printf("rsvp_input: " 2558 "Can't find vif for packet...dropping it.\n"); 2559 m_freem(m); 2560 splx(s); 2561 return; 2562 } 2563 2564 if (rsvpdebug) 2565 printf("rsvp_input: check socket\n"); 2566 2567 if (viftable[vifi].v_rsvpd == NULL) { 2568 /* 2569 * drop packet, since there is no specific socket for this 2570 * interface 2571 */ 2572 if (rsvpdebug) 2573 printf("rsvp_input: No socket defined for vif %d\n", 2574 vifi); 2575 m_freem(m); 2576 splx(s); 2577 return; 2578 } 2579 2580 rsvp_src.sin_addr = ip->ip_src; 2581 2582 if (rsvpdebug && m) 2583 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2584 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2585 2586 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2587 if (rsvpdebug) 2588 printf("rsvp_input: Failed to append to socket\n"); 2589 else 2590 if (rsvpdebug) 2591 printf("rsvp_input: send packet up\n"); 2592 2593 splx(s); 2594 } 2595 #endif /* RSVP_ISI */ 2596 2597 /* 2598 * Code for bandwidth monitors 2599 */ 2600 2601 /* 2602 * Define common interface for timeval-related methods 2603 */ 2604 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp) 2605 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp)) 2606 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp)) 2607 2608 static uint32_t 2609 compute_bw_meter_flags(struct bw_upcall *req) 2610 { 2611 uint32_t flags = 0; 2612 2613 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2614 flags |= BW_METER_UNIT_PACKETS; 2615 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2616 flags |= BW_METER_UNIT_BYTES; 2617 if (req->bu_flags & BW_UPCALL_GEQ) 2618 flags |= BW_METER_GEQ; 2619 if (req->bu_flags & BW_UPCALL_LEQ) 2620 flags |= BW_METER_LEQ; 2621 2622 return flags; 2623 } 2624 2625 /* 2626 * Add a bw_meter entry 2627 */ 2628 static int 2629 add_bw_upcall(struct mbuf *m) 2630 { 2631 int s; 2632 struct mfc *mfc; 2633 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2634 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2635 struct timeval now; 2636 struct bw_meter *x; 2637 uint32_t flags; 2638 struct bw_upcall *req; 2639 2640 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2641 return EINVAL; 2642 2643 req = mtod(m, struct bw_upcall *); 2644 2645 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2646 return EOPNOTSUPP; 2647 2648 /* Test if the flags are valid */ 2649 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2650 return EINVAL; 2651 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2652 return EINVAL; 2653 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2654 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2655 return EINVAL; 2656 2657 /* Test if the threshold time interval is valid */ 2658 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2659 return EINVAL; 2660 2661 flags = compute_bw_meter_flags(req); 2662 2663 /* 2664 * Find if we have already same bw_meter entry 2665 */ 2666 s = splsoftnet(); 2667 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2668 if (mfc == NULL) { 2669 splx(s); 2670 return EADDRNOTAVAIL; 2671 } 2672 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2673 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2674 &req->bu_threshold.b_time, ==)) && 2675 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2676 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2677 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2678 splx(s); 2679 return 0; /* XXX Already installed */ 2680 } 2681 } 2682 2683 /* Allocate the new bw_meter entry */ 2684 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2685 if (x == NULL) { 2686 splx(s); 2687 return ENOBUFS; 2688 } 2689 2690 /* Set the new bw_meter entry */ 2691 x->bm_threshold.b_time = req->bu_threshold.b_time; 2692 microtime(&now); 2693 x->bm_start_time = now; 2694 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2695 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2696 x->bm_measured.b_packets = 0; 2697 x->bm_measured.b_bytes = 0; 2698 x->bm_flags = flags; 2699 x->bm_time_next = NULL; 2700 x->bm_time_hash = BW_METER_BUCKETS; 2701 2702 /* Add the new bw_meter entry to the front of entries for this MFC */ 2703 x->bm_mfc = mfc; 2704 x->bm_mfc_next = mfc->mfc_bw_meter; 2705 mfc->mfc_bw_meter = x; 2706 schedule_bw_meter(x, &now); 2707 splx(s); 2708 2709 return 0; 2710 } 2711 2712 static void 2713 free_bw_list(struct bw_meter *list) 2714 { 2715 while (list != NULL) { 2716 struct bw_meter *x = list; 2717 2718 list = list->bm_mfc_next; 2719 unschedule_bw_meter(x); 2720 free(x, M_BWMETER); 2721 } 2722 } 2723 2724 /* 2725 * Delete one or multiple bw_meter entries 2726 */ 2727 static int 2728 del_bw_upcall(struct mbuf *m) 2729 { 2730 int s; 2731 struct mfc *mfc; 2732 struct bw_meter *x; 2733 struct bw_upcall *req; 2734 2735 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2736 return EINVAL; 2737 2738 req = mtod(m, struct bw_upcall *); 2739 2740 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2741 return EOPNOTSUPP; 2742 2743 s = splsoftnet(); 2744 /* Find the corresponding MFC entry */ 2745 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2746 if (mfc == NULL) { 2747 splx(s); 2748 return EADDRNOTAVAIL; 2749 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2750 /* 2751 * Delete all bw_meter entries for this mfc 2752 */ 2753 struct bw_meter *list; 2754 2755 list = mfc->mfc_bw_meter; 2756 mfc->mfc_bw_meter = NULL; 2757 free_bw_list(list); 2758 splx(s); 2759 return 0; 2760 } else { /* Delete a single bw_meter entry */ 2761 struct bw_meter *prev; 2762 uint32_t flags = 0; 2763 2764 flags = compute_bw_meter_flags(req); 2765 2766 /* Find the bw_meter entry to delete */ 2767 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2768 prev = x, x = x->bm_mfc_next) { 2769 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2770 &req->bu_threshold.b_time, ==)) && 2771 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2772 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2773 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2774 break; 2775 } 2776 if (x != NULL) { /* Delete entry from the list for this MFC */ 2777 if (prev != NULL) 2778 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2779 else 2780 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2781 2782 unschedule_bw_meter(x); 2783 splx(s); 2784 /* Free the bw_meter entry */ 2785 free(x, M_BWMETER); 2786 return 0; 2787 } else { 2788 splx(s); 2789 return EINVAL; 2790 } 2791 } 2792 /* NOTREACHED */ 2793 } 2794 2795 /* 2796 * Perform bandwidth measurement processing that may result in an upcall 2797 */ 2798 static void 2799 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2800 { 2801 struct timeval delta; 2802 2803 delta = *nowp; 2804 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2805 2806 if (x->bm_flags & BW_METER_GEQ) { 2807 /* 2808 * Processing for ">=" type of bw_meter entry 2809 */ 2810 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2811 /* Reset the bw_meter entry */ 2812 x->bm_start_time = *nowp; 2813 x->bm_measured.b_packets = 0; 2814 x->bm_measured.b_bytes = 0; 2815 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2816 } 2817 2818 /* Record that a packet is received */ 2819 x->bm_measured.b_packets++; 2820 x->bm_measured.b_bytes += plen; 2821 2822 /* 2823 * Test if we should deliver an upcall 2824 */ 2825 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2826 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2827 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2828 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2829 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2830 /* Prepare an upcall for delivery */ 2831 bw_meter_prepare_upcall(x, nowp); 2832 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2833 } 2834 } 2835 } else if (x->bm_flags & BW_METER_LEQ) { 2836 /* 2837 * Processing for "<=" type of bw_meter entry 2838 */ 2839 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2840 /* 2841 * We are behind time with the multicast forwarding table 2842 * scanning for "<=" type of bw_meter entries, so test now 2843 * if we should deliver an upcall. 2844 */ 2845 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2846 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2847 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2848 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2849 /* Prepare an upcall for delivery */ 2850 bw_meter_prepare_upcall(x, nowp); 2851 } 2852 /* Reschedule the bw_meter entry */ 2853 unschedule_bw_meter(x); 2854 schedule_bw_meter(x, nowp); 2855 } 2856 2857 /* Record that a packet is received */ 2858 x->bm_measured.b_packets++; 2859 x->bm_measured.b_bytes += plen; 2860 2861 /* 2862 * Test if we should restart the measuring interval 2863 */ 2864 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2865 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2866 (x->bm_flags & BW_METER_UNIT_BYTES && 2867 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2868 /* Don't restart the measuring interval */ 2869 } else { 2870 /* Do restart the measuring interval */ 2871 /* 2872 * XXX: note that we don't unschedule and schedule, because this 2873 * might be too much overhead per packet. Instead, when we process 2874 * all entries for a given timer hash bin, we check whether it is 2875 * really a timeout. If not, we reschedule at that time. 2876 */ 2877 x->bm_start_time = *nowp; 2878 x->bm_measured.b_packets = 0; 2879 x->bm_measured.b_bytes = 0; 2880 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2881 } 2882 } 2883 } 2884 2885 /* 2886 * Prepare a bandwidth-related upcall 2887 */ 2888 static void 2889 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2890 { 2891 struct timeval delta; 2892 struct bw_upcall *u; 2893 2894 /* 2895 * Compute the measured time interval 2896 */ 2897 delta = *nowp; 2898 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2899 2900 /* 2901 * If there are too many pending upcalls, deliver them now 2902 */ 2903 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2904 bw_upcalls_send(); 2905 2906 /* 2907 * Set the bw_upcall entry 2908 */ 2909 u = &bw_upcalls[bw_upcalls_n++]; 2910 u->bu_src = x->bm_mfc->mfc_origin; 2911 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2912 u->bu_threshold.b_time = x->bm_threshold.b_time; 2913 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2914 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2915 u->bu_measured.b_time = delta; 2916 u->bu_measured.b_packets = x->bm_measured.b_packets; 2917 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2918 u->bu_flags = 0; 2919 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2920 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2921 if (x->bm_flags & BW_METER_UNIT_BYTES) 2922 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2923 if (x->bm_flags & BW_METER_GEQ) 2924 u->bu_flags |= BW_UPCALL_GEQ; 2925 if (x->bm_flags & BW_METER_LEQ) 2926 u->bu_flags |= BW_UPCALL_LEQ; 2927 } 2928 2929 /* 2930 * Send the pending bandwidth-related upcalls 2931 */ 2932 static void 2933 bw_upcalls_send(void) 2934 { 2935 struct mbuf *m; 2936 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2937 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2938 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2939 0, /* unused2 */ 2940 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2941 0, /* im_mbz */ 2942 0, /* im_vif */ 2943 0, /* unused3 */ 2944 { 0 }, /* im_src */ 2945 { 0 } }; /* im_dst */ 2946 2947 if (bw_upcalls_n == 0) 2948 return; /* No pending upcalls */ 2949 2950 bw_upcalls_n = 0; 2951 2952 /* 2953 * Allocate a new mbuf, initialize it with the header and 2954 * the payload for the pending calls. 2955 */ 2956 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2957 if (m == NULL) { 2958 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2959 return; 2960 } 2961 2962 m->m_len = m->m_pkthdr.len = 0; 2963 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2964 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2965 2966 /* 2967 * Send the upcalls 2968 * XXX do we need to set the address in k_igmpsrc ? 2969 */ 2970 mrtstat.mrts_upcalls++; 2971 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2972 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2973 ++mrtstat.mrts_upq_sockfull; 2974 } 2975 } 2976 2977 /* 2978 * Compute the timeout hash value for the bw_meter entries 2979 */ 2980 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2981 do { \ 2982 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2983 \ 2984 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2985 (hash) = next_timeval.tv_sec; \ 2986 if (next_timeval.tv_usec) \ 2987 (hash)++; /* XXX: make sure we don't timeout early */ \ 2988 (hash) %= BW_METER_BUCKETS; \ 2989 } while (/*CONSTCOND*/ 0) 2990 2991 /* 2992 * Schedule a timer to process periodically bw_meter entry of type "<=" 2993 * by linking the entry in the proper hash bucket. 2994 */ 2995 static void 2996 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2997 { 2998 int time_hash; 2999 3000 if (!(x->bm_flags & BW_METER_LEQ)) 3001 return; /* XXX: we schedule timers only for "<=" entries */ 3002 3003 /* 3004 * Reset the bw_meter entry 3005 */ 3006 x->bm_start_time = *nowp; 3007 x->bm_measured.b_packets = 0; 3008 x->bm_measured.b_bytes = 0; 3009 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 3010 3011 /* 3012 * Compute the timeout hash value and insert the entry 3013 */ 3014 BW_METER_TIMEHASH(x, time_hash); 3015 x->bm_time_next = bw_meter_timers[time_hash]; 3016 bw_meter_timers[time_hash] = x; 3017 x->bm_time_hash = time_hash; 3018 } 3019 3020 /* 3021 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 3022 * by removing the entry from the proper hash bucket. 3023 */ 3024 static void 3025 unschedule_bw_meter(struct bw_meter *x) 3026 { 3027 int time_hash; 3028 struct bw_meter *prev, *tmp; 3029 3030 if (!(x->bm_flags & BW_METER_LEQ)) 3031 return; /* XXX: we schedule timers only for "<=" entries */ 3032 3033 /* 3034 * Compute the timeout hash value and delete the entry 3035 */ 3036 time_hash = x->bm_time_hash; 3037 if (time_hash >= BW_METER_BUCKETS) 3038 return; /* Entry was not scheduled */ 3039 3040 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 3041 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 3042 if (tmp == x) 3043 break; 3044 3045 if (tmp == NULL) 3046 panic("unschedule_bw_meter: bw_meter entry not found"); 3047 3048 if (prev != NULL) 3049 prev->bm_time_next = x->bm_time_next; 3050 else 3051 bw_meter_timers[time_hash] = x->bm_time_next; 3052 3053 x->bm_time_next = NULL; 3054 x->bm_time_hash = BW_METER_BUCKETS; 3055 } 3056 3057 /* 3058 * Process all "<=" type of bw_meter that should be processed now, 3059 * and for each entry prepare an upcall if necessary. Each processed 3060 * entry is rescheduled again for the (periodic) processing. 3061 * 3062 * This is run periodically (once per second normally). On each round, 3063 * all the potentially matching entries are in the hash slot that we are 3064 * looking at. 3065 */ 3066 static void 3067 bw_meter_process() 3068 { 3069 int s; 3070 static uint32_t last_tv_sec; /* last time we processed this */ 3071 3072 uint32_t loops; 3073 int i; 3074 struct timeval now, process_endtime; 3075 3076 microtime(&now); 3077 if (last_tv_sec == now.tv_sec) 3078 return; /* nothing to do */ 3079 3080 loops = now.tv_sec - last_tv_sec; 3081 last_tv_sec = now.tv_sec; 3082 if (loops > BW_METER_BUCKETS) 3083 loops = BW_METER_BUCKETS; 3084 3085 s = splsoftnet(); 3086 /* 3087 * Process all bins of bw_meter entries from the one after the last 3088 * processed to the current one. On entry, i points to the last bucket 3089 * visited, so we need to increment i at the beginning of the loop. 3090 */ 3091 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 3092 struct bw_meter *x, *tmp_list; 3093 3094 if (++i >= BW_METER_BUCKETS) 3095 i = 0; 3096 3097 /* Disconnect the list of bw_meter entries from the bin */ 3098 tmp_list = bw_meter_timers[i]; 3099 bw_meter_timers[i] = NULL; 3100 3101 /* Process the list of bw_meter entries */ 3102 while (tmp_list != NULL) { 3103 x = tmp_list; 3104 tmp_list = tmp_list->bm_time_next; 3105 3106 /* Test if the time interval is over */ 3107 process_endtime = x->bm_start_time; 3108 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 3109 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 3110 /* Not yet: reschedule, but don't reset */ 3111 int time_hash; 3112 3113 BW_METER_TIMEHASH(x, time_hash); 3114 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 3115 /* 3116 * XXX: somehow the bin processing is a bit ahead of time. 3117 * Put the entry in the next bin. 3118 */ 3119 if (++time_hash >= BW_METER_BUCKETS) 3120 time_hash = 0; 3121 } 3122 x->bm_time_next = bw_meter_timers[time_hash]; 3123 bw_meter_timers[time_hash] = x; 3124 x->bm_time_hash = time_hash; 3125 3126 continue; 3127 } 3128 3129 /* 3130 * Test if we should deliver an upcall 3131 */ 3132 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 3133 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 3134 ((x->bm_flags & BW_METER_UNIT_BYTES) && 3135 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 3136 /* Prepare an upcall for delivery */ 3137 bw_meter_prepare_upcall(x, &now); 3138 } 3139 3140 /* 3141 * Reschedule for next processing 3142 */ 3143 schedule_bw_meter(x, &now); 3144 } 3145 } 3146 3147 /* Send all upcalls that are pending delivery */ 3148 bw_upcalls_send(); 3149 3150 splx(s); 3151 } 3152 3153 /* 3154 * A periodic function for sending all upcalls that are pending delivery 3155 */ 3156 static void 3157 expire_bw_upcalls_send(void *unused) 3158 { 3159 int s; 3160 3161 s = splsoftnet(); 3162 bw_upcalls_send(); 3163 splx(s); 3164 3165 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 3166 expire_bw_upcalls_send, NULL); 3167 } 3168 3169 /* 3170 * A periodic function for periodic scanning of the multicast forwarding 3171 * table for processing all "<=" bw_meter entries. 3172 */ 3173 static void 3174 expire_bw_meter_process(void *unused) 3175 { 3176 if (mrt_api_config & MRT_MFC_BW_UPCALL) 3177 bw_meter_process(); 3178 3179 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 3180 expire_bw_meter_process, NULL); 3181 } 3182 3183 /* 3184 * End of bandwidth monitoring code 3185 */ 3186 3187 #ifdef PIM 3188 /* 3189 * Send the packet up to the user daemon, or eventually do kernel encapsulation 3190 */ 3191 static int 3192 pim_register_send(struct ip *ip, struct vif *vifp, 3193 struct mbuf *m, struct mfc *rt) 3194 { 3195 struct mbuf *mb_copy, *mm; 3196 3197 if (mrtdebug & DEBUG_PIM) 3198 log(LOG_DEBUG, "pim_register_send: "); 3199 3200 mb_copy = pim_register_prepare(ip, m); 3201 if (mb_copy == NULL) 3202 return ENOBUFS; 3203 3204 /* 3205 * Send all the fragments. Note that the mbuf for each fragment 3206 * is freed by the sending machinery. 3207 */ 3208 for (mm = mb_copy; mm; mm = mb_copy) { 3209 mb_copy = mm->m_nextpkt; 3210 mm->m_nextpkt = NULL; 3211 mm = m_pullup(mm, sizeof(struct ip)); 3212 if (mm != NULL) { 3213 ip = mtod(mm, struct ip *); 3214 if ((mrt_api_config & MRT_MFC_RP) && 3215 !in_nullhost(rt->mfc_rp)) { 3216 pim_register_send_rp(ip, vifp, mm, rt); 3217 } else { 3218 pim_register_send_upcall(ip, vifp, mm, rt); 3219 } 3220 } 3221 } 3222 3223 return 0; 3224 } 3225 3226 /* 3227 * Return a copy of the data packet that is ready for PIM Register 3228 * encapsulation. 3229 * XXX: Note that in the returned copy the IP header is a valid one. 3230 */ 3231 static struct mbuf * 3232 pim_register_prepare(struct ip *ip, struct mbuf *m) 3233 { 3234 struct mbuf *mb_copy = NULL; 3235 int mtu; 3236 3237 /* Take care of delayed checksums */ 3238 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 3239 in_delayed_cksum(m); 3240 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 3241 } 3242 3243 /* 3244 * Copy the old packet & pullup its IP header into the 3245 * new mbuf so we can modify it. 3246 */ 3247 mb_copy = m_copy(m, 0, M_COPYALL); 3248 if (mb_copy == NULL) 3249 return NULL; 3250 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 3251 if (mb_copy == NULL) 3252 return NULL; 3253 3254 /* take care of the TTL */ 3255 ip = mtod(mb_copy, struct ip *); 3256 --ip->ip_ttl; 3257 3258 /* Compute the MTU after the PIM Register encapsulation */ 3259 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 3260 3261 if (ntohs(ip->ip_len) <= mtu) { 3262 /* Turn the IP header into a valid one */ 3263 ip->ip_sum = 0; 3264 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 3265 } else { 3266 /* Fragment the packet */ 3267 if (ip_fragment(mb_copy, NULL, mtu) != 0) { 3268 /* XXX: mb_copy was freed by ip_fragment() */ 3269 return NULL; 3270 } 3271 } 3272 return mb_copy; 3273 } 3274 3275 /* 3276 * Send an upcall with the data packet to the user-level process. 3277 */ 3278 static int 3279 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 3280 struct mbuf *mb_copy, struct mfc *rt) 3281 { 3282 struct mbuf *mb_first; 3283 int len = ntohs(ip->ip_len); 3284 struct igmpmsg *im; 3285 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 3286 3287 /* 3288 * Add a new mbuf with an upcall header 3289 */ 3290 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3291 if (mb_first == NULL) { 3292 m_freem(mb_copy); 3293 return ENOBUFS; 3294 } 3295 mb_first->m_data += max_linkhdr; 3296 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3297 mb_first->m_len = sizeof(struct igmpmsg); 3298 mb_first->m_next = mb_copy; 3299 3300 /* Send message to routing daemon */ 3301 im = mtod(mb_first, struct igmpmsg *); 3302 im->im_msgtype = IGMPMSG_WHOLEPKT; 3303 im->im_mbz = 0; 3304 im->im_vif = vifp - viftable; 3305 im->im_src = ip->ip_src; 3306 im->im_dst = ip->ip_dst; 3307 3308 k_igmpsrc.sin_addr = ip->ip_src; 3309 3310 mrtstat.mrts_upcalls++; 3311 3312 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3313 if (mrtdebug & DEBUG_PIM) 3314 log(LOG_WARNING, 3315 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3316 ++mrtstat.mrts_upq_sockfull; 3317 return ENOBUFS; 3318 } 3319 3320 /* Keep statistics */ 3321 pimstat.pims_snd_registers_msgs++; 3322 pimstat.pims_snd_registers_bytes += len; 3323 3324 return 0; 3325 } 3326 3327 /* 3328 * Encapsulate the data packet in PIM Register message and send it to the RP. 3329 */ 3330 static int 3331 pim_register_send_rp(struct ip *ip, struct vif *vifp, 3332 struct mbuf *mb_copy, struct mfc *rt) 3333 { 3334 struct mbuf *mb_first; 3335 struct ip *ip_outer; 3336 struct pim_encap_pimhdr *pimhdr; 3337 int len = ntohs(ip->ip_len); 3338 vifi_t vifi = rt->mfc_parent; 3339 3340 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) { 3341 m_freem(mb_copy); 3342 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3343 } 3344 3345 /* 3346 * Add a new mbuf with the encapsulating header 3347 */ 3348 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3349 if (mb_first == NULL) { 3350 m_freem(mb_copy); 3351 return ENOBUFS; 3352 } 3353 mb_first->m_data += max_linkhdr; 3354 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3355 mb_first->m_next = mb_copy; 3356 3357 mb_first->m_pkthdr.len = len + mb_first->m_len; 3358 3359 /* 3360 * Fill in the encapsulating IP and PIM header 3361 */ 3362 ip_outer = mtod(mb_first, struct ip *); 3363 *ip_outer = pim_encap_iphdr; 3364 ip_outer->ip_id = ip_newid(); 3365 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + 3366 sizeof(pim_encap_pimhdr)); 3367 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3368 ip_outer->ip_dst = rt->mfc_rp; 3369 /* 3370 * Copy the inner header TOS to the outer header, and take care of the 3371 * IP_DF bit. 3372 */ 3373 ip_outer->ip_tos = ip->ip_tos; 3374 if (ntohs(ip->ip_off) & IP_DF) 3375 ip_outer->ip_off |= IP_DF; 3376 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 3377 + sizeof(pim_encap_iphdr)); 3378 *pimhdr = pim_encap_pimhdr; 3379 /* If the iif crosses a border, set the Border-bit */ 3380 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3381 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3382 3383 mb_first->m_data += sizeof(pim_encap_iphdr); 3384 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3385 mb_first->m_data -= sizeof(pim_encap_iphdr); 3386 3387 if (vifp->v_rate_limit == 0) 3388 tbf_send_packet(vifp, mb_first); 3389 else 3390 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len)); 3391 3392 /* Keep statistics */ 3393 pimstat.pims_snd_registers_msgs++; 3394 pimstat.pims_snd_registers_bytes += len; 3395 3396 return 0; 3397 } 3398 3399 /* 3400 * PIM-SMv2 and PIM-DM messages processing. 3401 * Receives and verifies the PIM control messages, and passes them 3402 * up to the listening socket, using rip_input(). 3403 * The only message with special processing is the PIM_REGISTER message 3404 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3405 * is passed to if_simloop(). 3406 */ 3407 void 3408 pim_input(struct mbuf *m, ...) 3409 { 3410 struct ip *ip = mtod(m, struct ip *); 3411 struct pim *pim; 3412 int minlen; 3413 int datalen; 3414 int ip_tos; 3415 int proto; 3416 int iphlen; 3417 va_list ap; 3418 3419 va_start(ap, m); 3420 iphlen = va_arg(ap, int); 3421 proto = va_arg(ap, int); 3422 va_end(ap); 3423 3424 datalen = ntohs(ip->ip_len) - iphlen; 3425 3426 /* Keep statistics */ 3427 pimstat.pims_rcv_total_msgs++; 3428 pimstat.pims_rcv_total_bytes += datalen; 3429 3430 /* 3431 * Validate lengths 3432 */ 3433 if (datalen < PIM_MINLEN) { 3434 pimstat.pims_rcv_tooshort++; 3435 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3436 datalen, (u_long)ip->ip_src.s_addr); 3437 m_freem(m); 3438 return; 3439 } 3440 3441 /* 3442 * If the packet is at least as big as a REGISTER, go agead 3443 * and grab the PIM REGISTER header size, to avoid another 3444 * possible m_pullup() later. 3445 * 3446 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3447 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3448 */ 3449 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3450 /* 3451 * Get the IP and PIM headers in contiguous memory, and 3452 * possibly the PIM REGISTER header. 3453 */ 3454 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3455 (m = m_pullup(m, minlen)) == NULL) { 3456 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3457 return; 3458 } 3459 /* m_pullup() may have given us a new mbuf so reset ip. */ 3460 ip = mtod(m, struct ip *); 3461 ip_tos = ip->ip_tos; 3462 3463 /* adjust mbuf to point to the PIM header */ 3464 m->m_data += iphlen; 3465 m->m_len -= iphlen; 3466 pim = mtod(m, struct pim *); 3467 3468 /* 3469 * Validate checksum. If PIM REGISTER, exclude the data packet. 3470 * 3471 * XXX: some older PIMv2 implementations don't make this distinction, 3472 * so for compatibility reason perform the checksum over part of the 3473 * message, and if error, then over the whole message. 3474 */ 3475 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3476 /* do nothing, checksum okay */ 3477 } else if (in_cksum(m, datalen)) { 3478 pimstat.pims_rcv_badsum++; 3479 if (mrtdebug & DEBUG_PIM) 3480 log(LOG_DEBUG, "pim_input: invalid checksum"); 3481 m_freem(m); 3482 return; 3483 } 3484 3485 /* PIM version check */ 3486 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3487 pimstat.pims_rcv_badversion++; 3488 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3489 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3490 m_freem(m); 3491 return; 3492 } 3493 3494 /* restore mbuf back to the outer IP */ 3495 m->m_data -= iphlen; 3496 m->m_len += iphlen; 3497 3498 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3499 /* 3500 * Since this is a REGISTER, we'll make a copy of the register 3501 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3502 * routing daemon. 3503 */ 3504 int s; 3505 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3506 struct mbuf *mcp; 3507 struct ip *encap_ip; 3508 u_int32_t *reghdr; 3509 struct ifnet *vifp; 3510 3511 s = splsoftnet(); 3512 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3513 splx(s); 3514 if (mrtdebug & DEBUG_PIM) 3515 log(LOG_DEBUG, 3516 "pim_input: register vif not set: %d\n", reg_vif_num); 3517 m_freem(m); 3518 return; 3519 } 3520 /* XXX need refcnt? */ 3521 vifp = viftable[reg_vif_num].v_ifp; 3522 splx(s); 3523 3524 /* 3525 * Validate length 3526 */ 3527 if (datalen < PIM_REG_MINLEN) { 3528 pimstat.pims_rcv_tooshort++; 3529 pimstat.pims_rcv_badregisters++; 3530 log(LOG_ERR, 3531 "pim_input: register packet size too small %d from %lx\n", 3532 datalen, (u_long)ip->ip_src.s_addr); 3533 m_freem(m); 3534 return; 3535 } 3536 3537 reghdr = (u_int32_t *)(pim + 1); 3538 encap_ip = (struct ip *)(reghdr + 1); 3539 3540 if (mrtdebug & DEBUG_PIM) { 3541 log(LOG_DEBUG, 3542 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3543 (u_long)ntohl(encap_ip->ip_src.s_addr), 3544 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3545 ntohs(encap_ip->ip_len)); 3546 } 3547 3548 /* verify the version number of the inner packet */ 3549 if (encap_ip->ip_v != IPVERSION) { 3550 pimstat.pims_rcv_badregisters++; 3551 if (mrtdebug & DEBUG_PIM) { 3552 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3553 "of the inner packet\n", encap_ip->ip_v); 3554 } 3555 m_freem(m); 3556 return; 3557 } 3558 3559 /* verify the inner packet is destined to a mcast group */ 3560 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) { 3561 pimstat.pims_rcv_badregisters++; 3562 if (mrtdebug & DEBUG_PIM) 3563 log(LOG_DEBUG, 3564 "pim_input: inner packet of register is not " 3565 "multicast %lx\n", 3566 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3567 m_freem(m); 3568 return; 3569 } 3570 3571 /* If a NULL_REGISTER, pass it to the daemon */ 3572 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3573 goto pim_input_to_daemon; 3574 3575 /* 3576 * Copy the TOS from the outer IP header to the inner IP header. 3577 */ 3578 if (encap_ip->ip_tos != ip_tos) { 3579 /* Outer TOS -> inner TOS */ 3580 encap_ip->ip_tos = ip_tos; 3581 /* Recompute the inner header checksum. Sigh... */ 3582 3583 /* adjust mbuf to point to the inner IP header */ 3584 m->m_data += (iphlen + PIM_MINLEN); 3585 m->m_len -= (iphlen + PIM_MINLEN); 3586 3587 encap_ip->ip_sum = 0; 3588 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3589 3590 /* restore mbuf to point back to the outer IP header */ 3591 m->m_data -= (iphlen + PIM_MINLEN); 3592 m->m_len += (iphlen + PIM_MINLEN); 3593 } 3594 3595 /* 3596 * Decapsulate the inner IP packet and loopback to forward it 3597 * as a normal multicast packet. Also, make a copy of the 3598 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3599 * to pass to the daemon later, so it can take the appropriate 3600 * actions (e.g., send back PIM_REGISTER_STOP). 3601 * XXX: here m->m_data points to the outer IP header. 3602 */ 3603 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3604 if (mcp == NULL) { 3605 log(LOG_ERR, 3606 "pim_input: pim register: could not copy register head\n"); 3607 m_freem(m); 3608 return; 3609 } 3610 3611 /* Keep statistics */ 3612 /* XXX: registers_bytes include only the encap. mcast pkt */ 3613 pimstat.pims_rcv_registers_msgs++; 3614 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3615 3616 /* 3617 * forward the inner ip packet; point m_data at the inner ip. 3618 */ 3619 m_adj(m, iphlen + PIM_MINLEN); 3620 3621 if (mrtdebug & DEBUG_PIM) { 3622 log(LOG_DEBUG, 3623 "pim_input: forwarding decapsulated register: " 3624 "src %lx, dst %lx, vif %d\n", 3625 (u_long)ntohl(encap_ip->ip_src.s_addr), 3626 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3627 reg_vif_num); 3628 } 3629 /* NB: vifp was collected above; can it change on us? */ 3630 looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL); 3631 3632 /* prepare the register head to send to the mrouting daemon */ 3633 m = mcp; 3634 } 3635 3636 pim_input_to_daemon: 3637 /* 3638 * Pass the PIM message up to the daemon; if it is a Register message, 3639 * pass the 'head' only up to the daemon. This includes the 3640 * outer IP header, PIM header, PIM-Register header and the 3641 * inner IP header. 3642 * XXX: the outer IP header pkt size of a Register is not adjust to 3643 * reflect the fact that the inner multicast data is truncated. 3644 */ 3645 rip_input(m, iphlen, proto); 3646 3647 return; 3648 } 3649 #endif /* PIM */ 3650
Indexes created Fri Sep 26 08:10:20 GMT 2025