ip_mroute.c revision 1.87
1 /* $NetBSD: ip_mroute.c,v 1.87 2005/01/15 06:50:47 manu 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 1.87 2005/01/15 06:50:47 manu 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 __P((struct mbuf *, ...)); 195 static int vif_encapcheck __P((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 __P((struct sioc_sg_req *)); 214 static int get_vif_cnt __P((struct sioc_vif_req *)); 215 static int ip_mrouter_init __P((struct socket *, struct mbuf *)); 216 static int get_version __P((struct mbuf *)); 217 static int set_assert __P((struct mbuf *)); 218 static int get_assert __P((struct mbuf *)); 219 static int add_vif __P((struct mbuf *)); 220 static int del_vif __P((struct mbuf *)); 221 static void update_mfc_params __P((struct mfc *, struct mfcctl2 *)); 222 static void init_mfc_params __P((struct mfc *, struct mfcctl2 *)); 223 static void expire_mfc __P((struct mfc *)); 224 static int add_mfc __P((struct mbuf *)); 225 #ifdef UPCALL_TIMING 226 static void collate __P((struct timeval *)); 227 #endif 228 static int del_mfc __P((struct mbuf *)); 229 static int set_api_config __P((struct mbuf *)); /* chose API capabilities */ 230 static int get_api_support __P((struct mbuf *)); 231 static int get_api_config __P((struct mbuf *)); 232 static int socket_send __P((struct socket *, struct mbuf *, 233 struct sockaddr_in *)); 234 static void expire_upcalls __P((void *)); 235 #ifdef RSVP_ISI 236 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *, vifi_t)); 237 #else 238 static int ip_mdq __P((struct mbuf *, struct ifnet *, struct mfc *)); 239 #endif 240 static void phyint_send __P((struct ip *, struct vif *, struct mbuf *)); 241 static void encap_send __P((struct ip *, struct vif *, struct mbuf *)); 242 static void tbf_control __P((struct vif *, struct mbuf *, struct ip *, 243 u_int32_t)); 244 static void tbf_queue __P((struct vif *, struct mbuf *)); 245 static void tbf_process_q __P((struct vif *)); 246 static void tbf_reprocess_q __P((void *)); 247 static int tbf_dq_sel __P((struct vif *, struct ip *)); 248 static void tbf_send_packet __P((struct vif *, struct mbuf *)); 249 static void tbf_update_tokens __P((struct vif *)); 250 static int priority __P((struct vif *, struct ip *)); 251 252 /* 253 * Bandwidth monitoring 254 */ 255 static void free_bw_list __P((struct bw_meter *)); 256 static int add_bw_upcall __P((struct mbuf *)); 257 static int del_bw_upcall __P((struct mbuf *)); 258 static void bw_meter_receive_packet __P((struct bw_meter *, int , struct timeval *)); 259 static void bw_meter_prepare_upcall __P((struct bw_meter *, struct timeval *)); 260 static void bw_upcalls_send __P((void)); 261 static void schedule_bw_meter __P((struct bw_meter *, struct timeval *)); 262 static void unschedule_bw_meter __P((struct bw_meter *)); 263 static void bw_meter_process __P((void)); 264 static void expire_bw_upcalls_send __P((void *)); 265 static void expire_bw_meter_process __P((void *)); 266 267 #ifdef PIM 268 static int pim_register_send __P((struct ip *, struct vif *, 269 struct mbuf *, struct mfc *)); 270 static int pim_register_send_rp __P((struct ip *, struct vif *, 271 struct mbuf *, struct mfc *)); 272 static int pim_register_send_upcall __P((struct ip *, struct vif *, 273 struct mbuf *, struct mfc *)); 274 static struct mbuf *pim_register_prepare __P((struct ip *, struct mbuf *)); 275 #endif 276 277 /* 278 * 'Interfaces' associated with decapsulator (so we can tell 279 * packets that went through it from ones that get reflected 280 * by a broken gateway). These interfaces are never linked into 281 * the system ifnet list & no routes point to them. I.e., packets 282 * can't be sent this way. They only exist as a placeholder for 283 * multicast source verification. 284 */ 285 #if 0 286 struct ifnet multicast_decap_if[MAXVIFS]; 287 #endif 288 289 #define ENCAP_TTL 64 290 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 291 292 /* prototype IP hdr for encapsulated packets */ 293 struct ip multicast_encap_iphdr = { 294 #if BYTE_ORDER == LITTLE_ENDIAN 295 sizeof(struct ip) >> 2, IPVERSION, 296 #else 297 IPVERSION, sizeof(struct ip) >> 2, 298 #endif 299 0, /* tos */ 300 sizeof(struct ip), /* total length */ 301 0, /* id */ 302 0, /* frag offset */ 303 ENCAP_TTL, ENCAP_PROTO, 304 0, /* checksum */ 305 }; 306 307 /* 308 * Bandwidth meter variables and constants 309 */ 310 311 /* 312 * Pending timeouts are stored in a hash table, the key being the 313 * expiration time. Periodically, the entries are analysed and processed. 314 */ 315 #define BW_METER_BUCKETS 1024 316 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 317 struct callout bw_meter_ch; 318 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 319 320 /* 321 * Pending upcalls are stored in a vector which is flushed when 322 * full, or periodically 323 */ 324 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 325 static u_int bw_upcalls_n; /* # of pending upcalls */ 326 struct callout bw_upcalls_ch; 327 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 328 329 #ifdef PIM 330 struct pimstat pimstat; 331 332 /* 333 * Note: the PIM Register encapsulation adds the following in front of a 334 * data packet: 335 * 336 * struct pim_encap_hdr { 337 * struct ip ip; 338 * struct pim_encap_pimhdr pim; 339 * } 340 * 341 */ 342 343 struct pim_encap_pimhdr { 344 struct pim pim; 345 uint32_t flags; 346 }; 347 348 static struct ip pim_encap_iphdr = { 349 #if BYTE_ORDER == LITTLE_ENDIAN 350 sizeof(struct ip) >> 2, 351 IPVERSION, 352 #else 353 IPVERSION, 354 sizeof(struct ip) >> 2, 355 #endif 356 0, /* tos */ 357 sizeof(struct ip), /* total length */ 358 0, /* id */ 359 0, /* frag offset */ 360 ENCAP_TTL, 361 IPPROTO_PIM, 362 0, /* checksum */ 363 }; 364 365 static struct pim_encap_pimhdr pim_encap_pimhdr = { 366 { 367 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 368 0, /* reserved */ 369 0, /* checksum */ 370 }, 371 0 /* flags */ 372 }; 373 374 static struct ifnet multicast_register_if; 375 static vifi_t reg_vif_num = VIFI_INVALID; 376 #endif /* PIM */ 377 378 379 /* 380 * Private variables. 381 */ 382 static vifi_t numvifs = 0; 383 384 static struct callout expire_upcalls_ch; 385 386 /* 387 * one-back cache used by vif_encapcheck to locate a tunnel's vif 388 * given a datagram's src ip address. 389 */ 390 static struct in_addr last_encap_src; 391 static struct vif *last_encap_vif; 392 393 /* 394 * whether or not special PIM assert processing is enabled. 395 */ 396 static int pim_assert; 397 /* 398 * Rate limit for assert notification messages, in usec 399 */ 400 #define ASSERT_MSG_TIME 3000000 401 402 /* 403 * Kernel multicast routing API capabilities and setup. 404 * If more API capabilities are added to the kernel, they should be 405 * recorded in `mrt_api_support'. 406 */ 407 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 408 MRT_MFC_FLAGS_BORDER_VIF | 409 MRT_MFC_RP | 410 MRT_MFC_BW_UPCALL); 411 static u_int32_t mrt_api_config = 0; 412 413 /* 414 * Find a route for a given origin IP address and Multicast group address 415 * Type of service parameter to be added in the future!!! 416 * Statistics are updated by the caller if needed 417 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 418 */ 419 static struct mfc * 420 mfc_find(struct in_addr *o, struct in_addr *g) 421 { 422 struct mfc *rt; 423 424 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { 425 if (in_hosteq(rt->mfc_origin, *o) && 426 in_hosteq(rt->mfc_mcastgrp, *g) && 427 (rt->mfc_stall == NULL)) 428 break; 429 } 430 431 return (rt); 432 } 433 434 /* 435 * Macros to compute elapsed time efficiently 436 * Borrowed from Van Jacobson's scheduling code 437 */ 438 #define TV_DELTA(a, b, delta) do { \ 439 int xxs; \ 440 delta = (a).tv_usec - (b).tv_usec; \ 441 xxs = (a).tv_sec - (b).tv_sec; \ 442 switch (xxs) { \ 443 case 2: \ 444 delta += 1000000; \ 445 /* fall through */ \ 446 case 1: \ 447 delta += 1000000; \ 448 /* fall through */ \ 449 case 0: \ 450 break; \ 451 default: \ 452 delta += (1000000 * xxs); \ 453 break; \ 454 } \ 455 } while (/*CONSTCOND*/ 0) 456 457 #ifdef UPCALL_TIMING 458 u_int32_t upcall_data[51]; 459 #endif /* UPCALL_TIMING */ 460 461 /* 462 * Handle MRT setsockopt commands to modify the multicast routing tables. 463 */ 464 int 465 ip_mrouter_set(so, optname, m) 466 struct socket *so; 467 int optname; 468 struct mbuf **m; 469 { 470 int error; 471 472 if (optname != MRT_INIT && so != ip_mrouter) 473 error = ENOPROTOOPT; 474 else 475 switch (optname) { 476 case MRT_INIT: 477 error = ip_mrouter_init(so, *m); 478 break; 479 case MRT_DONE: 480 error = ip_mrouter_done(); 481 break; 482 case MRT_ADD_VIF: 483 error = add_vif(*m); 484 break; 485 case MRT_DEL_VIF: 486 error = del_vif(*m); 487 break; 488 case MRT_ADD_MFC: 489 error = add_mfc(*m); 490 break; 491 case MRT_DEL_MFC: 492 error = del_mfc(*m); 493 break; 494 case MRT_ASSERT: 495 error = set_assert(*m); 496 break; 497 case MRT_API_CONFIG: 498 error = set_api_config(*m); 499 break; 500 case MRT_ADD_BW_UPCALL: 501 error = add_bw_upcall(*m); 502 break; 503 case MRT_DEL_BW_UPCALL: 504 error = del_bw_upcall(*m); 505 break; 506 default: 507 error = ENOPROTOOPT; 508 break; 509 } 510 511 if (*m) 512 m_free(*m); 513 return (error); 514 } 515 516 /* 517 * Handle MRT getsockopt commands 518 */ 519 int 520 ip_mrouter_get(so, optname, m) 521 struct socket *so; 522 int optname; 523 struct mbuf **m; 524 { 525 int error; 526 527 if (so != ip_mrouter) 528 error = ENOPROTOOPT; 529 else { 530 *m = m_get(M_WAIT, MT_SOOPTS); 531 MCLAIM(*m, so->so_mowner); 532 533 switch (optname) { 534 case MRT_VERSION: 535 error = get_version(*m); 536 break; 537 case MRT_ASSERT: 538 error = get_assert(*m); 539 break; 540 case MRT_API_SUPPORT: 541 error = get_api_support(*m); 542 break; 543 case MRT_API_CONFIG: 544 error = get_api_config(*m); 545 break; 546 default: 547 error = ENOPROTOOPT; 548 break; 549 } 550 551 if (error) 552 m_free(*m); 553 } 554 555 return (error); 556 } 557 558 /* 559 * Handle ioctl commands to obtain information from the cache 560 */ 561 int 562 mrt_ioctl(so, cmd, data) 563 struct socket *so; 564 u_long cmd; 565 caddr_t data; 566 { 567 int error; 568 569 if (so != ip_mrouter) 570 error = EINVAL; 571 else 572 switch (cmd) { 573 case SIOCGETVIFCNT: 574 error = get_vif_cnt((struct sioc_vif_req *)data); 575 break; 576 case SIOCGETSGCNT: 577 error = get_sg_cnt((struct sioc_sg_req *)data); 578 break; 579 default: 580 error = EINVAL; 581 break; 582 } 583 584 return (error); 585 } 586 587 /* 588 * returns the packet, byte, rpf-failure count for the source group provided 589 */ 590 static int 591 get_sg_cnt(req) 592 struct sioc_sg_req *req; 593 { 594 int s; 595 struct mfc *rt; 596 597 s = splsoftnet(); 598 rt = mfc_find(&req->src, &req->grp); 599 if (rt == NULL) { 600 splx(s); 601 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 602 return (EADDRNOTAVAIL); 603 } 604 req->pktcnt = rt->mfc_pkt_cnt; 605 req->bytecnt = rt->mfc_byte_cnt; 606 req->wrong_if = rt->mfc_wrong_if; 607 splx(s); 608 609 return (0); 610 } 611 612 /* 613 * returns the input and output packet and byte counts on the vif provided 614 */ 615 static int 616 get_vif_cnt(req) 617 struct sioc_vif_req *req; 618 { 619 vifi_t vifi = req->vifi; 620 621 if (vifi >= numvifs) 622 return (EINVAL); 623 624 req->icount = viftable[vifi].v_pkt_in; 625 req->ocount = viftable[vifi].v_pkt_out; 626 req->ibytes = viftable[vifi].v_bytes_in; 627 req->obytes = viftable[vifi].v_bytes_out; 628 629 return (0); 630 } 631 632 /* 633 * Enable multicast routing 634 */ 635 static int 636 ip_mrouter_init(so, m) 637 struct socket *so; 638 struct mbuf *m; 639 { 640 int *v; 641 642 if (mrtdebug) 643 log(LOG_DEBUG, 644 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 645 so->so_type, so->so_proto->pr_protocol); 646 647 if (so->so_type != SOCK_RAW || 648 so->so_proto->pr_protocol != IPPROTO_IGMP) 649 return (EOPNOTSUPP); 650 651 if (m == NULL || m->m_len < sizeof(int)) 652 return (EINVAL); 653 654 v = mtod(m, int *); 655 if (*v != 1) 656 return (EINVAL); 657 658 if (ip_mrouter != NULL) 659 return (EADDRINUSE); 660 661 ip_mrouter = so; 662 663 mfchashtbl = 664 hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash); 665 bzero((caddr_t)nexpire, sizeof(nexpire)); 666 667 pim_assert = 0; 668 669 callout_init(&expire_upcalls_ch); 670 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 671 expire_upcalls, NULL); 672 673 callout_init(&bw_upcalls_ch); 674 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 675 expire_bw_upcalls_send, NULL); 676 677 callout_init(&bw_meter_ch); 678 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 679 expire_bw_meter_process, NULL); 680 681 if (mrtdebug) 682 log(LOG_DEBUG, "ip_mrouter_init\n"); 683 684 return (0); 685 } 686 687 /* 688 * Disable multicast routing 689 */ 690 int 691 ip_mrouter_done() 692 { 693 vifi_t vifi; 694 struct vif *vifp; 695 int i; 696 int s; 697 698 s = splsoftnet(); 699 700 /* Clear out all the vifs currently in use. */ 701 for (vifi = 0; vifi < numvifs; vifi++) { 702 vifp = &viftable[vifi]; 703 if (!in_nullhost(vifp->v_lcl_addr)) 704 reset_vif(vifp); 705 } 706 707 numvifs = 0; 708 pim_assert = 0; 709 mrt_api_config = 0; 710 711 callout_stop(&expire_upcalls_ch); 712 callout_stop(&bw_upcalls_ch); 713 callout_stop(&bw_meter_ch); 714 715 /* 716 * Free all multicast forwarding cache entries. 717 */ 718 for (i = 0; i < MFCTBLSIZ; i++) { 719 struct mfc *rt, *nrt; 720 721 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 722 nrt = LIST_NEXT(rt, mfc_hash); 723 724 expire_mfc(rt); 725 } 726 } 727 728 bzero((caddr_t)nexpire, sizeof(nexpire)); 729 free(mfchashtbl, M_MRTABLE); 730 mfchashtbl = NULL; 731 732 bw_upcalls_n = 0; 733 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 734 735 /* Reset de-encapsulation cache. */ 736 737 ip_mrouter = NULL; 738 739 splx(s); 740 741 if (mrtdebug) 742 log(LOG_DEBUG, "ip_mrouter_done\n"); 743 744 return (0); 745 } 746 747 void 748 ip_mrouter_detach(ifp) 749 struct ifnet *ifp; 750 { 751 int vifi, i; 752 struct vif *vifp; 753 struct mfc *rt; 754 struct rtdetq *rte; 755 756 /* XXX not sure about side effect to userland routing daemon */ 757 for (vifi = 0; vifi < numvifs; vifi++) { 758 vifp = &viftable[vifi]; 759 if (vifp->v_ifp == ifp) 760 reset_vif(vifp); 761 } 762 for (i = 0; i < MFCTBLSIZ; i++) { 763 if (nexpire[i] == 0) 764 continue; 765 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) { 766 for (rte = rt->mfc_stall; rte; rte = rte->next) { 767 if (rte->ifp == ifp) 768 rte->ifp = NULL; 769 } 770 } 771 } 772 } 773 774 static int 775 get_version(m) 776 struct mbuf *m; 777 { 778 int *v = mtod(m, int *); 779 780 *v = 0x0305; /* XXX !!!! */ 781 m->m_len = sizeof(int); 782 return (0); 783 } 784 785 /* 786 * Set PIM assert processing global 787 */ 788 static int 789 set_assert(m) 790 struct mbuf *m; 791 { 792 int *i; 793 794 if (m == NULL || m->m_len < sizeof(int)) 795 return (EINVAL); 796 797 i = mtod(m, int *); 798 pim_assert = !!*i; 799 return (0); 800 } 801 802 /* 803 * Get PIM assert processing global 804 */ 805 static int 806 get_assert(m) 807 struct mbuf *m; 808 { 809 int *i = mtod(m, int *); 810 811 *i = pim_assert; 812 m->m_len = sizeof(int); 813 return (0); 814 } 815 816 /* 817 * Configure API capabilities 818 */ 819 static int 820 set_api_config(struct mbuf *m) 821 { 822 int i; 823 u_int32_t *apival; 824 825 if (m == NULL || m->m_len < sizeof(u_int32_t)) 826 return (EINVAL); 827 828 apival = mtod(m, u_int32_t *); 829 830 /* 831 * We can set the API capabilities only if it is the first operation 832 * after MRT_INIT. I.e.: 833 * - there are no vifs installed 834 * - pim_assert is not enabled 835 * - the MFC table is empty 836 */ 837 if (numvifs > 0) { 838 *apival = 0; 839 return (EPERM); 840 } 841 if (pim_assert) { 842 *apival = 0; 843 return (EPERM); 844 } 845 for (i = 0; i < MFCTBLSIZ; i++) { 846 if (LIST_FIRST(&mfchashtbl[i]) != NULL) { 847 *apival = 0; 848 return (EPERM); 849 } 850 } 851 852 mrt_api_config = *apival & mrt_api_support; 853 *apival = mrt_api_config; 854 855 return (0); 856 } 857 858 /* 859 * Get API capabilities 860 */ 861 static int 862 get_api_support(struct mbuf *m) 863 { 864 u_int32_t *apival; 865 866 if (m == NULL || m->m_len < sizeof(u_int32_t)) 867 return (EINVAL); 868 869 apival = mtod(m, u_int32_t *); 870 871 *apival = mrt_api_support; 872 873 return (0); 874 } 875 876 /* 877 * Get API configured capabilities 878 */ 879 static int 880 get_api_config(struct mbuf *m) 881 { 882 u_int32_t *apival; 883 884 if (m == NULL || m->m_len < sizeof(u_int32_t)) 885 return (EINVAL); 886 887 apival = mtod(m, u_int32_t *); 888 889 *apival = mrt_api_config; 890 891 return (0); 892 } 893 894 static struct sockaddr_in sin = { sizeof(sin), AF_INET }; 895 896 /* 897 * Add a vif to the vif table 898 */ 899 static int 900 add_vif(m) 901 struct mbuf *m; 902 { 903 struct vifctl *vifcp; 904 struct vif *vifp; 905 struct ifaddr *ifa; 906 struct ifnet *ifp; 907 struct ifreq ifr; 908 int error, s; 909 910 if (m == NULL || m->m_len < sizeof(struct vifctl)) 911 return (EINVAL); 912 913 vifcp = mtod(m, struct vifctl *); 914 if (vifcp->vifc_vifi >= MAXVIFS) 915 return (EINVAL); 916 if (in_nullhost(vifcp->vifc_lcl_addr)) 917 return (EADDRNOTAVAIL); 918 919 vifp = &viftable[vifcp->vifc_vifi]; 920 if (!in_nullhost(vifp->v_lcl_addr)) 921 return (EADDRINUSE); 922 923 /* Find the interface with an address in AF_INET family. */ 924 #ifdef PIM 925 if (vifcp->vifc_flags & VIFF_REGISTER) { 926 /* 927 * XXX: Because VIFF_REGISTER does not really need a valid 928 * local interface (e.g. it could be 127.0.0.2), we don't 929 * check its address. 930 */ 931 ifp = NULL; 932 } else 933 #endif 934 { 935 sin.sin_addr = vifcp->vifc_lcl_addr; 936 ifa = ifa_ifwithaddr(sintosa(&sin)); 937 if (ifa == NULL) 938 return (EADDRNOTAVAIL); 939 ifp = ifa->ifa_ifp; 940 } 941 942 if (vifcp->vifc_flags & VIFF_TUNNEL) { 943 if (vifcp->vifc_flags & VIFF_SRCRT) { 944 log(LOG_ERR, "source routed tunnels not supported\n"); 945 return (EOPNOTSUPP); 946 } 947 948 /* attach this vif to decapsulator dispatch table */ 949 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 950 vif_encapcheck, &vif_protosw, vifp); 951 if (!vifp->v_encap_cookie) 952 return (EINVAL); 953 954 /* Create a fake encapsulation interface. */ 955 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 956 bzero(ifp, sizeof(*ifp)); 957 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 958 "mdecap%d", vifcp->vifc_vifi); 959 960 /* Prepare cached route entry. */ 961 bzero(&vifp->v_route, sizeof(vifp->v_route)); 962 #ifdef PIM 963 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 964 ifp = &multicast_register_if; 965 if (mrtdebug) 966 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 967 (void *)ifp); 968 if (reg_vif_num == VIFI_INVALID) { 969 bzero(ifp, sizeof(*ifp)); 970 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 971 "register_vif"); 972 ifp->if_flags = IFF_LOOPBACK; 973 bzero(&vifp->v_route, sizeof(vifp->v_route)); 974 reg_vif_num = vifcp->vifc_vifi; 975 } 976 #endif 977 } else { 978 /* Make sure the interface supports multicast. */ 979 if ((ifp->if_flags & IFF_MULTICAST) == 0) 980 return (EOPNOTSUPP); 981 982 /* Enable promiscuous reception of all IP multicasts. */ 983 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 984 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 985 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 986 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 987 if (error) 988 return (error); 989 } 990 991 s = splsoftnet(); 992 993 /* Define parameters for the tbf structure. */ 994 vifp->tbf_q = NULL; 995 vifp->tbf_t = &vifp->tbf_q; 996 microtime(&vifp->tbf_last_pkt_t); 997 vifp->tbf_n_tok = 0; 998 vifp->tbf_q_len = 0; 999 vifp->tbf_max_q_len = MAXQSIZE; 1000 1001 vifp->v_flags = vifcp->vifc_flags; 1002 vifp->v_threshold = vifcp->vifc_threshold; 1003 /* scaling up here allows division by 1024 in critical code */ 1004 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 1005 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 1006 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 1007 vifp->v_ifp = ifp; 1008 /* Initialize per vif pkt counters. */ 1009 vifp->v_pkt_in = 0; 1010 vifp->v_pkt_out = 0; 1011 vifp->v_bytes_in = 0; 1012 vifp->v_bytes_out = 0; 1013 1014 callout_init(&vifp->v_repq_ch); 1015 1016 #ifdef RSVP_ISI 1017 vifp->v_rsvp_on = 0; 1018 vifp->v_rsvpd = NULL; 1019 #endif /* RSVP_ISI */ 1020 1021 splx(s); 1022 1023 /* Adjust numvifs up if the vifi is higher than numvifs. */ 1024 if (numvifs <= vifcp->vifc_vifi) 1025 numvifs = vifcp->vifc_vifi + 1; 1026 1027 if (mrtdebug) 1028 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 1029 vifcp->vifc_vifi, 1030 ntohl(vifcp->vifc_lcl_addr.s_addr), 1031 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 1032 ntohl(vifcp->vifc_rmt_addr.s_addr), 1033 vifcp->vifc_threshold, 1034 vifcp->vifc_rate_limit); 1035 1036 return (0); 1037 } 1038 1039 void 1040 reset_vif(vifp) 1041 struct vif *vifp; 1042 { 1043 struct mbuf *m, *n; 1044 struct ifnet *ifp; 1045 struct ifreq ifr; 1046 1047 callout_stop(&vifp->v_repq_ch); 1048 1049 /* detach this vif from decapsulator dispatch table */ 1050 encap_detach(vifp->v_encap_cookie); 1051 vifp->v_encap_cookie = NULL; 1052 1053 /* 1054 * Free packets queued at the interface 1055 */ 1056 for (m = vifp->tbf_q; m != NULL; m = n) { 1057 n = m->m_nextpkt; 1058 m_freem(m); 1059 } 1060 1061 if (vifp->v_flags & VIFF_TUNNEL) { 1062 free(vifp->v_ifp, M_MRTABLE); 1063 if (vifp == last_encap_vif) { 1064 last_encap_vif = NULL; 1065 last_encap_src = zeroin_addr; 1066 } 1067 } else if (vifp->v_flags & VIFF_REGISTER) { 1068 #ifdef PIM 1069 reg_vif_num = VIFI_INVALID; 1070 #endif 1071 } else { 1072 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 1073 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 1074 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 1075 ifp = vifp->v_ifp; 1076 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 1077 } 1078 bzero((caddr_t)vifp, sizeof(*vifp)); 1079 } 1080 1081 /* 1082 * Delete a vif from the vif table 1083 */ 1084 static int 1085 del_vif(m) 1086 struct mbuf *m; 1087 { 1088 vifi_t *vifip; 1089 struct vif *vifp; 1090 vifi_t vifi; 1091 int s; 1092 1093 if (m == NULL || m->m_len < sizeof(vifi_t)) 1094 return (EINVAL); 1095 1096 vifip = mtod(m, vifi_t *); 1097 if (*vifip >= numvifs) 1098 return (EINVAL); 1099 1100 vifp = &viftable[*vifip]; 1101 if (in_nullhost(vifp->v_lcl_addr)) 1102 return (EADDRNOTAVAIL); 1103 1104 s = splsoftnet(); 1105 1106 reset_vif(vifp); 1107 1108 /* Adjust numvifs down */ 1109 for (vifi = numvifs; vifi > 0; vifi--) 1110 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr)) 1111 break; 1112 numvifs = vifi; 1113 1114 splx(s); 1115 1116 if (mrtdebug) 1117 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 1118 1119 return (0); 1120 } 1121 1122 /* 1123 * update an mfc entry without resetting counters and S,G addresses. 1124 */ 1125 static void 1126 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1127 { 1128 int i; 1129 1130 rt->mfc_parent = mfccp->mfcc_parent; 1131 for (i = 0; i < numvifs; i++) { 1132 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1133 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1134 MRT_MFC_FLAGS_ALL; 1135 } 1136 /* set the RP address */ 1137 if (mrt_api_config & MRT_MFC_RP) 1138 rt->mfc_rp = mfccp->mfcc_rp; 1139 else 1140 rt->mfc_rp = zeroin_addr; 1141 } 1142 1143 /* 1144 * fully initialize an mfc entry from the parameter. 1145 */ 1146 static void 1147 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1148 { 1149 rt->mfc_origin = mfccp->mfcc_origin; 1150 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1151 1152 update_mfc_params(rt, mfccp); 1153 1154 /* initialize pkt counters per src-grp */ 1155 rt->mfc_pkt_cnt = 0; 1156 rt->mfc_byte_cnt = 0; 1157 rt->mfc_wrong_if = 0; 1158 timerclear(&rt->mfc_last_assert); 1159 } 1160 1161 static void 1162 expire_mfc(rt) 1163 struct mfc *rt; 1164 { 1165 struct rtdetq *rte, *nrte; 1166 1167 free_bw_list(rt->mfc_bw_meter); 1168 1169 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 1170 nrte = rte->next; 1171 m_freem(rte->m); 1172 free(rte, M_MRTABLE); 1173 } 1174 1175 LIST_REMOVE(rt, mfc_hash); 1176 free(rt, M_MRTABLE); 1177 } 1178 1179 /* 1180 * Add an mfc entry 1181 */ 1182 static int 1183 add_mfc(m) 1184 struct mbuf *m; 1185 { 1186 struct mfcctl2 mfcctl2; 1187 struct mfcctl2 *mfccp; 1188 struct mfc *rt; 1189 u_int32_t hash = 0; 1190 struct rtdetq *rte, *nrte; 1191 u_short nstl; 1192 int s; 1193 int mfcctl_size = sizeof(struct mfcctl); 1194 1195 if (mrt_api_config & MRT_API_FLAGS_ALL) 1196 mfcctl_size = sizeof(struct mfcctl2); 1197 1198 if (m == NULL || m->m_len < mfcctl_size) 1199 return (EINVAL); 1200 1201 /* 1202 * select data size depending on API version. 1203 */ 1204 if (mrt_api_config & MRT_API_FLAGS_ALL) { 1205 struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *); 1206 bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2)); 1207 } else { 1208 struct mfcctl *mp = mtod(m, struct mfcctl *); 1209 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp)); 1210 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl), 1211 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1212 } 1213 mfccp = &mfcctl2; 1214 1215 s = splsoftnet(); 1216 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1217 1218 /* If an entry already exists, just update the fields */ 1219 if (rt) { 1220 if (mrtdebug & DEBUG_MFC) 1221 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n", 1222 ntohl(mfccp->mfcc_origin.s_addr), 1223 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1224 mfccp->mfcc_parent); 1225 1226 update_mfc_params(rt, mfccp); 1227 1228 splx(s); 1229 return (0); 1230 } 1231 1232 /* 1233 * Find the entry for which the upcall was made and update 1234 */ 1235 nstl = 0; 1236 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1237 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1238 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1239 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1240 rt->mfc_stall != NULL) { 1241 if (nstl++) 1242 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 1243 "multiple kernel entries", 1244 ntohl(mfccp->mfcc_origin.s_addr), 1245 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1246 mfccp->mfcc_parent, rt->mfc_stall); 1247 1248 if (mrtdebug & DEBUG_MFC) 1249 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n", 1250 ntohl(mfccp->mfcc_origin.s_addr), 1251 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1252 mfccp->mfcc_parent, rt->mfc_stall); 1253 1254 rte = rt->mfc_stall; 1255 init_mfc_params(rt, mfccp); 1256 rt->mfc_stall = NULL; 1257 1258 rt->mfc_expire = 0; /* Don't clean this guy up */ 1259 nexpire[hash]--; 1260 1261 /* free packets Qed at the end of this entry */ 1262 for (; rte != NULL; rte = nrte) { 1263 nrte = rte->next; 1264 if (rte->ifp) { 1265 #ifdef RSVP_ISI 1266 ip_mdq(rte->m, rte->ifp, rt, -1); 1267 #else 1268 ip_mdq(rte->m, rte->ifp, rt); 1269 #endif /* RSVP_ISI */ 1270 } 1271 m_freem(rte->m); 1272 #ifdef UPCALL_TIMING 1273 collate(&rte->t); 1274 #endif /* UPCALL_TIMING */ 1275 free(rte, M_MRTABLE); 1276 } 1277 } 1278 } 1279 1280 /* 1281 * It is possible that an entry is being inserted without an upcall 1282 */ 1283 if (nstl == 0) { 1284 /* 1285 * No mfc; make a new one 1286 */ 1287 if (mrtdebug & DEBUG_MFC) 1288 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n", 1289 ntohl(mfccp->mfcc_origin.s_addr), 1290 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1291 mfccp->mfcc_parent); 1292 1293 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1294 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1295 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1296 init_mfc_params(rt, mfccp); 1297 if (rt->mfc_expire) 1298 nexpire[hash]--; 1299 rt->mfc_expire = 0; 1300 break; /* XXX */ 1301 } 1302 } 1303 if (rt == NULL) { /* no upcall, so make a new entry */ 1304 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1305 M_NOWAIT); 1306 if (rt == NULL) { 1307 splx(s); 1308 return (ENOBUFS); 1309 } 1310 1311 init_mfc_params(rt, mfccp); 1312 rt->mfc_expire = 0; 1313 rt->mfc_stall = NULL; 1314 rt->mfc_bw_meter = NULL; 1315 1316 /* insert new entry at head of hash chain */ 1317 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1318 } 1319 } 1320 1321 splx(s); 1322 return (0); 1323 } 1324 1325 #ifdef UPCALL_TIMING 1326 /* 1327 * collect delay statistics on the upcalls 1328 */ 1329 static void 1330 collate(t) 1331 struct timeval *t; 1332 { 1333 u_int32_t d; 1334 struct timeval tp; 1335 u_int32_t delta; 1336 1337 microtime(&tp); 1338 1339 if (timercmp(t, &tp, <)) { 1340 TV_DELTA(tp, *t, delta); 1341 1342 d = delta >> 10; 1343 if (d > 50) 1344 d = 50; 1345 1346 ++upcall_data[d]; 1347 } 1348 } 1349 #endif /* UPCALL_TIMING */ 1350 1351 /* 1352 * Delete an mfc entry 1353 */ 1354 static int 1355 del_mfc(m) 1356 struct mbuf *m; 1357 { 1358 struct mfcctl2 mfcctl2; 1359 struct mfcctl2 *mfccp; 1360 struct mfc *rt; 1361 int s; 1362 int mfcctl_size = sizeof(struct mfcctl); 1363 struct mfcctl *mp = mtod(m, struct mfcctl *); 1364 1365 /* 1366 * XXX: for deleting MFC entries the information in entries 1367 * of size "struct mfcctl" is sufficient. 1368 */ 1369 1370 if (m == NULL || m->m_len < mfcctl_size) 1371 return (EINVAL); 1372 1373 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp)); 1374 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl), 1375 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1376 1377 mfccp = &mfcctl2; 1378 1379 if (mrtdebug & DEBUG_MFC) 1380 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 1381 ntohl(mfccp->mfcc_origin.s_addr), 1382 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 1383 1384 s = splsoftnet(); 1385 1386 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1387 if (rt == NULL) { 1388 splx(s); 1389 return (EADDRNOTAVAIL); 1390 } 1391 1392 /* 1393 * free the bw_meter entries 1394 */ 1395 free_bw_list(rt->mfc_bw_meter); 1396 rt->mfc_bw_meter = NULL; 1397 1398 LIST_REMOVE(rt, mfc_hash); 1399 free(rt, M_MRTABLE); 1400 1401 splx(s); 1402 return (0); 1403 } 1404 1405 static int 1406 socket_send(s, mm, src) 1407 struct socket *s; 1408 struct mbuf *mm; 1409 struct sockaddr_in *src; 1410 { 1411 if (s) { 1412 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, 1413 (struct mbuf *)NULL) != 0) { 1414 sorwakeup(s); 1415 return (0); 1416 } 1417 } 1418 m_freem(mm); 1419 return (-1); 1420 } 1421 1422 /* 1423 * IP multicast forwarding function. This function assumes that the packet 1424 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1425 * pointed to by "ifp", and the packet is to be relayed to other networks 1426 * that have members of the packet's destination IP multicast group. 1427 * 1428 * The packet is returned unscathed to the caller, unless it is 1429 * erroneous, in which case a non-zero return value tells the caller to 1430 * discard it. 1431 */ 1432 1433 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1434 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1435 1436 int 1437 #ifdef RSVP_ISI 1438 ip_mforward(m, ifp, imo) 1439 #else 1440 ip_mforward(m, ifp) 1441 #endif /* RSVP_ISI */ 1442 struct mbuf *m; 1443 struct ifnet *ifp; 1444 #ifdef RSVP_ISI 1445 struct ip_moptions *imo; 1446 #endif /* RSVP_ISI */ 1447 { 1448 struct ip *ip = mtod(m, struct ip *); 1449 struct mfc *rt; 1450 static int srctun = 0; 1451 struct mbuf *mm; 1452 int s; 1453 vifi_t vifi; 1454 1455 if (mrtdebug & DEBUG_FORWARD) 1456 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1457 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1458 1459 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1460 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1461 /* 1462 * Packet arrived via a physical interface or 1463 * an encapsulated tunnel or a register_vif. 1464 */ 1465 } else { 1466 /* 1467 * Packet arrived through a source-route tunnel. 1468 * Source-route tunnels are no longer supported. 1469 */ 1470 if ((srctun++ % 1000) == 0) 1471 log(LOG_ERR, 1472 "ip_mforward: received source-routed packet from %x\n", 1473 ntohl(ip->ip_src.s_addr)); 1474 1475 return (1); 1476 } 1477 1478 #ifdef RSVP_ISI 1479 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1480 if (ip->ip_ttl < 255) 1481 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1482 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1483 struct vif *vifp = viftable + vifi; 1484 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1485 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1486 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1487 vifp->v_ifp->if_xname); 1488 } 1489 return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi)); 1490 } 1491 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1492 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1493 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1494 } 1495 #endif /* RSVP_ISI */ 1496 1497 /* 1498 * Don't forward a packet with time-to-live of zero or one, 1499 * or a packet destined to a local-only group. 1500 */ 1501 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1502 return (0); 1503 1504 /* 1505 * Determine forwarding vifs from the forwarding cache table 1506 */ 1507 s = splsoftnet(); 1508 ++mrtstat.mrts_mfc_lookups; 1509 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1510 1511 /* Entry exists, so forward if necessary */ 1512 if (rt != NULL) { 1513 splx(s); 1514 #ifdef RSVP_ISI 1515 return (ip_mdq(m, ifp, rt, -1)); 1516 #else 1517 return (ip_mdq(m, ifp, rt)); 1518 #endif /* RSVP_ISI */ 1519 } else { 1520 /* 1521 * If we don't have a route for packet's origin, 1522 * Make a copy of the packet & send message to routing daemon 1523 */ 1524 1525 struct mbuf *mb0; 1526 struct rtdetq *rte; 1527 u_int32_t hash; 1528 int hlen = ip->ip_hl << 2; 1529 #ifdef UPCALL_TIMING 1530 struct timeval tp; 1531 1532 microtime(&tp); 1533 #endif /* UPCALL_TIMING */ 1534 1535 ++mrtstat.mrts_mfc_misses; 1536 1537 mrtstat.mrts_no_route++; 1538 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1539 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1540 ntohl(ip->ip_src.s_addr), 1541 ntohl(ip->ip_dst.s_addr)); 1542 1543 /* 1544 * Allocate mbufs early so that we don't do extra work if we are 1545 * just going to fail anyway. Make sure to pullup the header so 1546 * that other people can't step on it. 1547 */ 1548 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1549 M_NOWAIT); 1550 if (rte == NULL) { 1551 splx(s); 1552 return (ENOBUFS); 1553 } 1554 mb0 = m_copy(m, 0, M_COPYALL); 1555 M_PULLUP(mb0, hlen); 1556 if (mb0 == NULL) { 1557 free(rte, M_MRTABLE); 1558 splx(s); 1559 return (ENOBUFS); 1560 } 1561 1562 /* is there an upcall waiting for this flow? */ 1563 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1564 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1565 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1566 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1567 rt->mfc_stall != NULL) 1568 break; 1569 } 1570 1571 if (rt == NULL) { 1572 int i; 1573 struct igmpmsg *im; 1574 1575 /* 1576 * Locate the vifi for the incoming interface for 1577 * this packet. 1578 * If none found, drop packet. 1579 */ 1580 for (vifi = 0; vifi < numvifs && 1581 viftable[vifi].v_ifp != ifp; vifi++) 1582 ; 1583 if (vifi >= numvifs) /* vif not found, drop packet */ 1584 goto non_fatal; 1585 1586 /* no upcall, so make a new entry */ 1587 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1588 M_NOWAIT); 1589 if (rt == NULL) 1590 goto fail; 1591 1592 /* 1593 * Make a copy of the header to send to the user level 1594 * process 1595 */ 1596 mm = m_copy(m, 0, hlen); 1597 M_PULLUP(mm, hlen); 1598 if (mm == NULL) 1599 goto fail1; 1600 1601 /* 1602 * Send message to routing daemon to install 1603 * a route into the kernel table 1604 */ 1605 1606 im = mtod(mm, struct igmpmsg *); 1607 im->im_msgtype = IGMPMSG_NOCACHE; 1608 im->im_mbz = 0; 1609 im->im_vif = vifi; 1610 1611 mrtstat.mrts_upcalls++; 1612 1613 sin.sin_addr = ip->ip_src; 1614 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1615 log(LOG_WARNING, 1616 "ip_mforward: ip_mrouter socket queue full\n"); 1617 ++mrtstat.mrts_upq_sockfull; 1618 fail1: 1619 free(rt, M_MRTABLE); 1620 fail: 1621 free(rte, M_MRTABLE); 1622 m_freem(mb0); 1623 splx(s); 1624 return (ENOBUFS); 1625 } 1626 1627 /* insert new entry at head of hash chain */ 1628 rt->mfc_origin = ip->ip_src; 1629 rt->mfc_mcastgrp = ip->ip_dst; 1630 rt->mfc_pkt_cnt = 0; 1631 rt->mfc_byte_cnt = 0; 1632 rt->mfc_wrong_if = 0; 1633 rt->mfc_expire = UPCALL_EXPIRE; 1634 nexpire[hash]++; 1635 for (i = 0; i < numvifs; i++) { 1636 rt->mfc_ttls[i] = 0; 1637 rt->mfc_flags[i] = 0; 1638 } 1639 rt->mfc_parent = -1; 1640 1641 /* clear the RP address */ 1642 rt->mfc_rp = zeroin_addr; 1643 1644 rt->mfc_bw_meter = NULL; 1645 1646 /* link into table */ 1647 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1648 /* Add this entry to the end of the queue */ 1649 rt->mfc_stall = rte; 1650 } else { 1651 /* determine if q has overflowed */ 1652 struct rtdetq **p; 1653 int npkts = 0; 1654 1655 /* 1656 * XXX ouch! we need to append to the list, but we 1657 * only have a pointer to the front, so we have to 1658 * scan the entire list every time. 1659 */ 1660 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1661 if (++npkts > MAX_UPQ) { 1662 mrtstat.mrts_upq_ovflw++; 1663 non_fatal: 1664 free(rte, M_MRTABLE); 1665 m_freem(mb0); 1666 splx(s); 1667 return (0); 1668 } 1669 1670 /* Add this entry to the end of the queue */ 1671 *p = rte; 1672 } 1673 1674 rte->next = NULL; 1675 rte->m = mb0; 1676 rte->ifp = ifp; 1677 #ifdef UPCALL_TIMING 1678 rte->t = tp; 1679 #endif /* UPCALL_TIMING */ 1680 1681 splx(s); 1682 1683 return (0); 1684 } 1685 } 1686 1687 1688 /*ARGSUSED*/ 1689 static void 1690 expire_upcalls(v) 1691 void *v; 1692 { 1693 int i; 1694 int s; 1695 1696 s = splsoftnet(); 1697 1698 for (i = 0; i < MFCTBLSIZ; i++) { 1699 struct mfc *rt, *nrt; 1700 1701 if (nexpire[i] == 0) 1702 continue; 1703 1704 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1705 nrt = LIST_NEXT(rt, mfc_hash); 1706 1707 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1708 continue; 1709 nexpire[i]--; 1710 1711 /* 1712 * free the bw_meter entries 1713 */ 1714 while (rt->mfc_bw_meter != NULL) { 1715 struct bw_meter *x = rt->mfc_bw_meter; 1716 1717 rt->mfc_bw_meter = x->bm_mfc_next; 1718 free(x, M_BWMETER); 1719 } 1720 1721 ++mrtstat.mrts_cache_cleanups; 1722 if (mrtdebug & DEBUG_EXPIRE) 1723 log(LOG_DEBUG, 1724 "expire_upcalls: expiring (%x %x)\n", 1725 ntohl(rt->mfc_origin.s_addr), 1726 ntohl(rt->mfc_mcastgrp.s_addr)); 1727 1728 expire_mfc(rt); 1729 } 1730 } 1731 1732 splx(s); 1733 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1734 expire_upcalls, NULL); 1735 } 1736 1737 /* 1738 * Packet forwarding routine once entry in the cache is made 1739 */ 1740 static int 1741 #ifdef RSVP_ISI 1742 ip_mdq(m, ifp, rt, xmt_vif) 1743 #else 1744 ip_mdq(m, ifp, rt) 1745 #endif /* RSVP_ISI */ 1746 struct mbuf *m; 1747 struct ifnet *ifp; 1748 struct mfc *rt; 1749 #ifdef RSVP_ISI 1750 vifi_t xmt_vif; 1751 #endif /* RSVP_ISI */ 1752 { 1753 struct ip *ip = mtod(m, struct ip *); 1754 vifi_t vifi; 1755 struct vif *vifp; 1756 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2); 1757 1758 /* 1759 * Macro to send packet on vif. Since RSVP packets don't get counted on 1760 * input, they shouldn't get counted on output, so statistics keeping is 1761 * separate. 1762 */ 1763 #define MC_SEND(ip, vifp, m) do { \ 1764 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1765 encap_send((ip), (vifp), (m)); \ 1766 else \ 1767 phyint_send((ip), (vifp), (m)); \ 1768 } while (/*CONSTCOND*/ 0) 1769 1770 #ifdef RSVP_ISI 1771 /* 1772 * If xmt_vif is not -1, send on only the requested vif. 1773 * 1774 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1775 */ 1776 if (xmt_vif < numvifs) { 1777 #ifdef PIM 1778 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1779 pim_register_send(ip, viftable + xmt_vif, m, rt); 1780 else 1781 #endif 1782 MC_SEND(ip, viftable + xmt_vif, m); 1783 return (1); 1784 } 1785 #endif /* RSVP_ISI */ 1786 1787 /* 1788 * Don't forward if it didn't arrive from the parent vif for its origin. 1789 */ 1790 vifi = rt->mfc_parent; 1791 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1792 /* came in the wrong interface */ 1793 if (mrtdebug & DEBUG_FORWARD) 1794 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1795 ifp, vifi, 1796 vifi >= numvifs ? 0 : viftable[vifi].v_ifp); 1797 ++mrtstat.mrts_wrong_if; 1798 ++rt->mfc_wrong_if; 1799 /* 1800 * If we are doing PIM assert processing, send a message 1801 * to the routing daemon. 1802 * 1803 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1804 * can complete the SPT switch, regardless of the type 1805 * of the iif (broadcast media, GRE tunnel, etc). 1806 */ 1807 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1808 struct timeval now; 1809 u_int32_t delta; 1810 1811 #ifdef PIM 1812 if (ifp == &multicast_register_if) 1813 pimstat.pims_rcv_registers_wrongiif++; 1814 #endif 1815 1816 /* Get vifi for the incoming packet */ 1817 for (vifi = 0; 1818 vifi < numvifs && viftable[vifi].v_ifp != ifp; 1819 vifi++) 1820 ; 1821 if (vifi >= numvifs) { 1822 /* The iif is not found: ignore the packet. */ 1823 return (0); 1824 } 1825 1826 if (rt->mfc_flags[vifi] & 1827 MRT_MFC_FLAGS_DISABLE_WRONGVIF) { 1828 /* WRONGVIF disabled: ignore the packet */ 1829 return (0); 1830 } 1831 1832 microtime(&now); 1833 1834 TV_DELTA(rt->mfc_last_assert, now, delta); 1835 1836 if (delta > ASSERT_MSG_TIME) { 1837 struct igmpmsg *im; 1838 int hlen = ip->ip_hl << 2; 1839 struct mbuf *mm = m_copy(m, 0, hlen); 1840 1841 M_PULLUP(mm, hlen); 1842 if (mm == NULL) 1843 return (ENOBUFS); 1844 1845 rt->mfc_last_assert = now; 1846 1847 im = mtod(mm, struct igmpmsg *); 1848 im->im_msgtype = IGMPMSG_WRONGVIF; 1849 im->im_mbz = 0; 1850 im->im_vif = vifi; 1851 1852 mrtstat.mrts_upcalls++; 1853 1854 sin.sin_addr = im->im_src; 1855 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1856 log(LOG_WARNING, 1857 "ip_mforward: ip_mrouter socket queue full\n"); 1858 ++mrtstat.mrts_upq_sockfull; 1859 return (ENOBUFS); 1860 } 1861 } 1862 } 1863 return (0); 1864 } 1865 1866 /* If I sourced this packet, it counts as output, else it was input. */ 1867 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1868 viftable[vifi].v_pkt_out++; 1869 viftable[vifi].v_bytes_out += plen; 1870 } else { 1871 viftable[vifi].v_pkt_in++; 1872 viftable[vifi].v_bytes_in += plen; 1873 } 1874 rt->mfc_pkt_cnt++; 1875 rt->mfc_byte_cnt += plen; 1876 1877 /* 1878 * For each vif, decide if a copy of the packet should be forwarded. 1879 * Forward if: 1880 * - the ttl exceeds the vif's threshold 1881 * - there are group members downstream on interface 1882 */ 1883 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1884 if ((rt->mfc_ttls[vifi] > 0) && 1885 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1886 vifp->v_pkt_out++; 1887 vifp->v_bytes_out += plen; 1888 #ifdef PIM 1889 if (vifp->v_flags & VIFF_REGISTER) 1890 pim_register_send(ip, vifp, m, rt); 1891 else 1892 #endif 1893 MC_SEND(ip, vifp, m); 1894 } 1895 1896 /* 1897 * Perform upcall-related bw measuring. 1898 */ 1899 if (rt->mfc_bw_meter != NULL) { 1900 struct bw_meter *x; 1901 struct timeval now; 1902 1903 microtime(&now); 1904 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1905 bw_meter_receive_packet(x, plen, &now); 1906 } 1907 1908 return (0); 1909 } 1910 1911 #ifdef RSVP_ISI 1912 /* 1913 * check if a vif number is legal/ok. This is used by ip_output. 1914 */ 1915 int 1916 legal_vif_num(vif) 1917 int vif; 1918 { 1919 if (vif >= 0 && vif < numvifs) 1920 return (1); 1921 else 1922 return (0); 1923 } 1924 #endif /* RSVP_ISI */ 1925 1926 static void 1927 phyint_send(ip, vifp, m) 1928 struct ip *ip; 1929 struct vif *vifp; 1930 struct mbuf *m; 1931 { 1932 struct mbuf *mb_copy; 1933 int hlen = ip->ip_hl << 2; 1934 1935 /* 1936 * Make a new reference to the packet; make sure that 1937 * the IP header is actually copied, not just referenced, 1938 * so that ip_output() only scribbles on the copy. 1939 */ 1940 mb_copy = m_copy(m, 0, M_COPYALL); 1941 M_PULLUP(mb_copy, hlen); 1942 if (mb_copy == NULL) 1943 return; 1944 1945 if (vifp->v_rate_limit <= 0) 1946 tbf_send_packet(vifp, mb_copy); 1947 else 1948 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1949 ntohs(ip->ip_len)); 1950 } 1951 1952 static void 1953 encap_send(ip, vifp, m) 1954 struct ip *ip; 1955 struct vif *vifp; 1956 struct mbuf *m; 1957 { 1958 struct mbuf *mb_copy; 1959 struct ip *ip_copy; 1960 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1961 1962 /* Take care of delayed checksums */ 1963 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1964 in_delayed_cksum(m); 1965 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1966 } 1967 1968 /* 1969 * copy the old packet & pullup it's IP header into the 1970 * new mbuf so we can modify it. Try to fill the new 1971 * mbuf since if we don't the ethernet driver will. 1972 */ 1973 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1974 if (mb_copy == NULL) 1975 return; 1976 mb_copy->m_data += max_linkhdr; 1977 mb_copy->m_pkthdr.len = len; 1978 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1979 1980 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1981 m_freem(mb_copy); 1982 return; 1983 } 1984 i = MHLEN - max_linkhdr; 1985 if (i > len) 1986 i = len; 1987 mb_copy = m_pullup(mb_copy, i); 1988 if (mb_copy == NULL) 1989 return; 1990 1991 /* 1992 * fill in the encapsulating IP header. 1993 */ 1994 ip_copy = mtod(mb_copy, struct ip *); 1995 *ip_copy = multicast_encap_iphdr; 1996 ip_copy->ip_id = ip_newid(); 1997 ip_copy->ip_len = htons(len); 1998 ip_copy->ip_src = vifp->v_lcl_addr; 1999 ip_copy->ip_dst = vifp->v_rmt_addr; 2000 2001 /* 2002 * turn the encapsulated IP header back into a valid one. 2003 */ 2004 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 2005 --ip->ip_ttl; 2006 ip->ip_sum = 0; 2007 mb_copy->m_data += sizeof(multicast_encap_iphdr); 2008 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2009 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 2010 2011 if (vifp->v_rate_limit <= 0) 2012 tbf_send_packet(vifp, mb_copy); 2013 else 2014 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len)); 2015 } 2016 2017 /* 2018 * De-encapsulate a packet and feed it back through ip input. 2019 */ 2020 static void 2021 vif_input(struct mbuf *m, ...) 2022 { 2023 int off, proto; 2024 va_list ap; 2025 struct vif *vifp; 2026 int s; 2027 struct ifqueue *ifq; 2028 2029 va_start(ap, m); 2030 off = va_arg(ap, int); 2031 proto = va_arg(ap, int); 2032 va_end(ap); 2033 2034 vifp = (struct vif *)encap_getarg(m); 2035 if (!vifp || proto != AF_INET) { 2036 m_freem(m); 2037 mrtstat.mrts_bad_tunnel++; 2038 return; 2039 } 2040 2041 m_adj(m, off); 2042 m->m_pkthdr.rcvif = vifp->v_ifp; 2043 ifq = &ipintrq; 2044 s = splnet(); 2045 if (IF_QFULL(ifq)) { 2046 IF_DROP(ifq); 2047 m_freem(m); 2048 } else { 2049 IF_ENQUEUE(ifq, m); 2050 /* 2051 * normally we would need a "schednetisr(NETISR_IP)" 2052 * here but we were called by ip_input and it is going 2053 * to loop back & try to dequeue the packet we just 2054 * queued as soon as we return so we avoid the 2055 * unnecessary software interrrupt. 2056 */ 2057 } 2058 splx(s); 2059 } 2060 2061 /* 2062 * Check if the packet should be grabbed by us. 2063 */ 2064 static int 2065 vif_encapcheck(m, off, proto, arg) 2066 const struct mbuf *m; 2067 int off; 2068 int proto; 2069 void *arg; 2070 { 2071 struct vif *vifp; 2072 struct ip ip; 2073 2074 #ifdef DIAGNOSTIC 2075 if (!arg || proto != IPPROTO_IPV4) 2076 panic("unexpected arg in vif_encapcheck"); 2077 #endif 2078 2079 /* 2080 * do not grab the packet if it's not to a multicast destination or if 2081 * we don't have an encapsulating tunnel with the source. 2082 * Note: This code assumes that the remote site IP address 2083 * uniquely identifies the tunnel (i.e., that this site has 2084 * at most one tunnel with the remote site). 2085 */ 2086 2087 /* LINTED const cast */ 2088 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip); 2089 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 2090 return 0; 2091 2092 /* LINTED const cast */ 2093 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip); 2094 if (!in_hosteq(ip.ip_src, last_encap_src)) { 2095 vifp = (struct vif *)arg; 2096 if (vifp->v_flags & VIFF_TUNNEL && 2097 in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 2098 ; 2099 else 2100 return 0; 2101 last_encap_vif = vifp; 2102 last_encap_src = ip.ip_src; 2103 } else 2104 vifp = last_encap_vif; 2105 2106 /* 32bit match, since we have checked ip_src only */ 2107 return 32; 2108 } 2109 2110 /* 2111 * Token bucket filter module 2112 */ 2113 static void 2114 tbf_control(vifp, m, ip, len) 2115 struct vif *vifp; 2116 struct mbuf *m; 2117 struct ip *ip; 2118 u_int32_t len; 2119 { 2120 2121 if (len > MAX_BKT_SIZE) { 2122 /* drop if packet is too large */ 2123 mrtstat.mrts_pkt2large++; 2124 m_freem(m); 2125 return; 2126 } 2127 2128 tbf_update_tokens(vifp); 2129 2130 /* 2131 * If there are enough tokens, and the queue is empty, send this packet 2132 * out immediately. Otherwise, try to insert it on this vif's queue. 2133 */ 2134 if (vifp->tbf_q_len == 0) { 2135 if (len <= vifp->tbf_n_tok) { 2136 vifp->tbf_n_tok -= len; 2137 tbf_send_packet(vifp, m); 2138 } else { 2139 /* queue packet and timeout till later */ 2140 tbf_queue(vifp, m); 2141 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2142 tbf_reprocess_q, vifp); 2143 } 2144 } else { 2145 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 2146 !tbf_dq_sel(vifp, ip)) { 2147 /* queue full, and couldn't make room */ 2148 mrtstat.mrts_q_overflow++; 2149 m_freem(m); 2150 } else { 2151 /* queue length low enough, or made room */ 2152 tbf_queue(vifp, m); 2153 tbf_process_q(vifp); 2154 } 2155 } 2156 } 2157 2158 /* 2159 * adds a packet to the queue at the interface 2160 */ 2161 static void 2162 tbf_queue(vifp, m) 2163 struct vif *vifp; 2164 struct mbuf *m; 2165 { 2166 int s = splsoftnet(); 2167 2168 /* insert at tail */ 2169 *vifp->tbf_t = m; 2170 vifp->tbf_t = &m->m_nextpkt; 2171 vifp->tbf_q_len++; 2172 2173 splx(s); 2174 } 2175 2176 2177 /* 2178 * processes the queue at the interface 2179 */ 2180 static void 2181 tbf_process_q(vifp) 2182 struct vif *vifp; 2183 { 2184 struct mbuf *m; 2185 int len; 2186 int s = splsoftnet(); 2187 2188 /* 2189 * Loop through the queue at the interface and send as many packets 2190 * as possible. 2191 */ 2192 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) { 2193 len = ntohs(mtod(m, struct ip *)->ip_len); 2194 2195 /* determine if the packet can be sent */ 2196 if (len <= vifp->tbf_n_tok) { 2197 /* if so, 2198 * reduce no of tokens, dequeue the packet, 2199 * send the packet. 2200 */ 2201 if ((vifp->tbf_q = m->m_nextpkt) == NULL) 2202 vifp->tbf_t = &vifp->tbf_q; 2203 --vifp->tbf_q_len; 2204 2205 m->m_nextpkt = NULL; 2206 vifp->tbf_n_tok -= len; 2207 tbf_send_packet(vifp, m); 2208 } else 2209 break; 2210 } 2211 splx(s); 2212 } 2213 2214 static void 2215 tbf_reprocess_q(arg) 2216 void *arg; 2217 { 2218 struct vif *vifp = arg; 2219 2220 if (ip_mrouter == NULL) 2221 return; 2222 2223 tbf_update_tokens(vifp); 2224 tbf_process_q(vifp); 2225 2226 if (vifp->tbf_q_len != 0) 2227 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2228 tbf_reprocess_q, vifp); 2229 } 2230 2231 /* function that will selectively discard a member of the queue 2232 * based on the precedence value and the priority 2233 */ 2234 static int 2235 tbf_dq_sel(vifp, ip) 2236 struct vif *vifp; 2237 struct ip *ip; 2238 { 2239 u_int p; 2240 struct mbuf **mp, *m; 2241 int s = splsoftnet(); 2242 2243 p = priority(vifp, ip); 2244 2245 for (mp = &vifp->tbf_q, m = *mp; 2246 m != NULL; 2247 mp = &m->m_nextpkt, m = *mp) { 2248 if (p > priority(vifp, mtod(m, struct ip *))) { 2249 if ((*mp = m->m_nextpkt) == NULL) 2250 vifp->tbf_t = mp; 2251 --vifp->tbf_q_len; 2252 2253 m_freem(m); 2254 mrtstat.mrts_drop_sel++; 2255 splx(s); 2256 return (1); 2257 } 2258 } 2259 splx(s); 2260 return (0); 2261 } 2262 2263 static void 2264 tbf_send_packet(vifp, m) 2265 struct vif *vifp; 2266 struct mbuf *m; 2267 { 2268 int error; 2269 int s = splsoftnet(); 2270 2271 if (vifp->v_flags & VIFF_TUNNEL) { 2272 /* If tunnel options */ 2273 ip_output(m, (struct mbuf *)NULL, &vifp->v_route, 2274 IP_FORWARDING, (struct ip_moptions *)NULL, 2275 (struct socket *)NULL); 2276 } else { 2277 /* if physical interface option, extract the options and then send */ 2278 struct ip_moptions imo; 2279 2280 imo.imo_multicast_ifp = vifp->v_ifp; 2281 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 2282 imo.imo_multicast_loop = 1; 2283 #ifdef RSVP_ISI 2284 imo.imo_multicast_vif = -1; 2285 #endif 2286 2287 error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL, 2288 IP_FORWARDING|IP_MULTICASTOPTS, &imo, 2289 (struct socket *)NULL); 2290 2291 if (mrtdebug & DEBUG_XMIT) 2292 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 2293 (long)(vifp - viftable), error); 2294 } 2295 splx(s); 2296 } 2297 2298 /* determine the current time and then 2299 * the elapsed time (between the last time and time now) 2300 * in milliseconds & update the no. of tokens in the bucket 2301 */ 2302 static void 2303 tbf_update_tokens(vifp) 2304 struct vif *vifp; 2305 { 2306 struct timeval tp; 2307 u_int32_t tm; 2308 int s = splsoftnet(); 2309 2310 microtime(&tp); 2311 2312 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 2313 2314 /* 2315 * This formula is actually 2316 * "time in seconds" * "bytes/second". 2317 * 2318 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2319 * 2320 * The (1000/1024) was introduced in add_vif to optimize 2321 * this divide into a shift. 2322 */ 2323 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 2324 vifp->tbf_last_pkt_t = tp; 2325 2326 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 2327 vifp->tbf_n_tok = MAX_BKT_SIZE; 2328 2329 splx(s); 2330 } 2331 2332 static int 2333 priority(vifp, ip) 2334 struct vif *vifp; 2335 struct ip *ip; 2336 { 2337 int prio = 50; /* the lowest priority -- default case */ 2338 2339 /* temporary hack; may add general packet classifier some day */ 2340 2341 /* 2342 * The UDP port space is divided up into four priority ranges: 2343 * [0, 16384) : unclassified - lowest priority 2344 * [16384, 32768) : audio - highest priority 2345 * [32768, 49152) : whiteboard - medium priority 2346 * [49152, 65536) : video - low priority 2347 */ 2348 if (ip->ip_p == IPPROTO_UDP) { 2349 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2350 2351 switch (ntohs(udp->uh_dport) & 0xc000) { 2352 case 0x4000: 2353 prio = 70; 2354 break; 2355 case 0x8000: 2356 prio = 60; 2357 break; 2358 case 0xc000: 2359 prio = 55; 2360 break; 2361 } 2362 2363 if (tbfdebug > 1) 2364 log(LOG_DEBUG, "port %x prio %d\n", 2365 ntohs(udp->uh_dport), prio); 2366 } 2367 2368 return (prio); 2369 } 2370 2371 /* 2372 * End of token bucket filter modifications 2373 */ 2374 #ifdef RSVP_ISI 2375 int 2376 ip_rsvp_vif_init(so, m) 2377 struct socket *so; 2378 struct mbuf *m; 2379 { 2380 int vifi, s; 2381 2382 if (rsvpdebug) 2383 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2384 so->so_type, so->so_proto->pr_protocol); 2385 2386 if (so->so_type != SOCK_RAW || 2387 so->so_proto->pr_protocol != IPPROTO_RSVP) 2388 return (EOPNOTSUPP); 2389 2390 /* Check mbuf. */ 2391 if (m == NULL || m->m_len != sizeof(int)) { 2392 return (EINVAL); 2393 } 2394 vifi = *(mtod(m, int *)); 2395 2396 if (rsvpdebug) 2397 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", 2398 vifi, rsvp_on); 2399 2400 s = splsoftnet(); 2401 2402 /* Check vif. */ 2403 if (!legal_vif_num(vifi)) { 2404 splx(s); 2405 return (EADDRNOTAVAIL); 2406 } 2407 2408 /* Check if socket is available. */ 2409 if (viftable[vifi].v_rsvpd != NULL) { 2410 splx(s); 2411 return (EADDRINUSE); 2412 } 2413 2414 viftable[vifi].v_rsvpd = so; 2415 /* 2416 * This may seem silly, but we need to be sure we don't over-increment 2417 * the RSVP counter, in case something slips up. 2418 */ 2419 if (!viftable[vifi].v_rsvp_on) { 2420 viftable[vifi].v_rsvp_on = 1; 2421 rsvp_on++; 2422 } 2423 2424 splx(s); 2425 return (0); 2426 } 2427 2428 int 2429 ip_rsvp_vif_done(so, m) 2430 struct socket *so; 2431 struct mbuf *m; 2432 { 2433 int vifi, s; 2434 2435 if (rsvpdebug) 2436 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2437 so->so_type, so->so_proto->pr_protocol); 2438 2439 if (so->so_type != SOCK_RAW || 2440 so->so_proto->pr_protocol != IPPROTO_RSVP) 2441 return (EOPNOTSUPP); 2442 2443 /* Check mbuf. */ 2444 if (m == NULL || m->m_len != sizeof(int)) { 2445 return (EINVAL); 2446 } 2447 vifi = *(mtod(m, int *)); 2448 2449 s = splsoftnet(); 2450 2451 /* Check vif. */ 2452 if (!legal_vif_num(vifi)) { 2453 splx(s); 2454 return (EADDRNOTAVAIL); 2455 } 2456 2457 if (rsvpdebug) 2458 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 2459 viftable[vifi].v_rsvpd, so); 2460 2461 viftable[vifi].v_rsvpd = NULL; 2462 /* 2463 * This may seem silly, but we need to be sure we don't over-decrement 2464 * the RSVP counter, in case something slips up. 2465 */ 2466 if (viftable[vifi].v_rsvp_on) { 2467 viftable[vifi].v_rsvp_on = 0; 2468 rsvp_on--; 2469 } 2470 2471 splx(s); 2472 return (0); 2473 } 2474 2475 void 2476 ip_rsvp_force_done(so) 2477 struct socket *so; 2478 { 2479 int vifi, s; 2480 2481 /* Don't bother if it is not the right type of socket. */ 2482 if (so->so_type != SOCK_RAW || 2483 so->so_proto->pr_protocol != IPPROTO_RSVP) 2484 return; 2485 2486 s = splsoftnet(); 2487 2488 /* 2489 * The socket may be attached to more than one vif...this 2490 * is perfectly legal. 2491 */ 2492 for (vifi = 0; vifi < numvifs; vifi++) { 2493 if (viftable[vifi].v_rsvpd == so) { 2494 viftable[vifi].v_rsvpd = NULL; 2495 /* 2496 * This may seem silly, but we need to be sure we don't 2497 * over-decrement the RSVP counter, in case something 2498 * slips up. 2499 */ 2500 if (viftable[vifi].v_rsvp_on) { 2501 viftable[vifi].v_rsvp_on = 0; 2502 rsvp_on--; 2503 } 2504 } 2505 } 2506 2507 splx(s); 2508 return; 2509 } 2510 2511 void 2512 rsvp_input(m, ifp) 2513 struct mbuf *m; 2514 struct ifnet *ifp; 2515 { 2516 int vifi, s; 2517 struct ip *ip = mtod(m, struct ip *); 2518 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 2519 2520 if (rsvpdebug) 2521 printf("rsvp_input: rsvp_on %d\n", rsvp_on); 2522 2523 /* 2524 * Can still get packets with rsvp_on = 0 if there is a local member 2525 * of the group to which the RSVP packet is addressed. But in this 2526 * case we want to throw the packet away. 2527 */ 2528 if (!rsvp_on) { 2529 m_freem(m); 2530 return; 2531 } 2532 2533 /* 2534 * If the old-style non-vif-associated socket is set, then use 2535 * it and ignore the new ones. 2536 */ 2537 if (ip_rsvpd != NULL) { 2538 if (rsvpdebug) 2539 printf("rsvp_input: " 2540 "Sending packet up old-style socket\n"); 2541 rip_input(m); /*XXX*/ 2542 return; 2543 } 2544 2545 s = splsoftnet(); 2546 2547 if (rsvpdebug) 2548 printf("rsvp_input: check vifs\n"); 2549 2550 /* Find which vif the packet arrived on. */ 2551 for (vifi = 0; vifi < numvifs; vifi++) { 2552 if (viftable[vifi].v_ifp == ifp) 2553 break; 2554 } 2555 2556 if (vifi == numvifs) { 2557 /* Can't find vif packet arrived on. Drop packet. */ 2558 if (rsvpdebug) 2559 printf("rsvp_input: " 2560 "Can't find vif for packet...dropping it.\n"); 2561 m_freem(m); 2562 splx(s); 2563 return; 2564 } 2565 2566 if (rsvpdebug) 2567 printf("rsvp_input: check socket\n"); 2568 2569 if (viftable[vifi].v_rsvpd == NULL) { 2570 /* 2571 * drop packet, since there is no specific socket for this 2572 * interface 2573 */ 2574 if (rsvpdebug) 2575 printf("rsvp_input: No socket defined for vif %d\n", 2576 vifi); 2577 m_freem(m); 2578 splx(s); 2579 return; 2580 } 2581 2582 rsvp_src.sin_addr = ip->ip_src; 2583 2584 if (rsvpdebug && m) 2585 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2586 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2587 2588 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2589 if (rsvpdebug) 2590 printf("rsvp_input: Failed to append to socket\n"); 2591 else 2592 if (rsvpdebug) 2593 printf("rsvp_input: send packet up\n"); 2594 2595 splx(s); 2596 } 2597 #endif /* RSVP_ISI */ 2598 2599 /* 2600 * Code for bandwidth monitors 2601 */ 2602 2603 /* 2604 * Define common interface for timeval-related methods 2605 */ 2606 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp) 2607 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp)) 2608 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp)) 2609 2610 static uint32_t 2611 compute_bw_meter_flags(struct bw_upcall *req) 2612 { 2613 uint32_t flags = 0; 2614 2615 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2616 flags |= BW_METER_UNIT_PACKETS; 2617 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2618 flags |= BW_METER_UNIT_BYTES; 2619 if (req->bu_flags & BW_UPCALL_GEQ) 2620 flags |= BW_METER_GEQ; 2621 if (req->bu_flags & BW_UPCALL_LEQ) 2622 flags |= BW_METER_LEQ; 2623 2624 return flags; 2625 } 2626 2627 /* 2628 * Add a bw_meter entry 2629 */ 2630 static int 2631 add_bw_upcall(struct mbuf *m) 2632 { 2633 int s; 2634 struct mfc *mfc; 2635 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2636 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2637 struct timeval now; 2638 struct bw_meter *x; 2639 uint32_t flags; 2640 struct bw_upcall *req; 2641 2642 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2643 return EINVAL; 2644 2645 req = mtod(m, struct bw_upcall *); 2646 2647 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2648 return EOPNOTSUPP; 2649 2650 /* Test if the flags are valid */ 2651 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2652 return EINVAL; 2653 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2654 return EINVAL; 2655 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2656 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2657 return EINVAL; 2658 2659 /* Test if the threshold time interval is valid */ 2660 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2661 return EINVAL; 2662 2663 flags = compute_bw_meter_flags(req); 2664 2665 /* 2666 * Find if we have already same bw_meter entry 2667 */ 2668 s = splsoftnet(); 2669 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2670 if (mfc == NULL) { 2671 splx(s); 2672 return EADDRNOTAVAIL; 2673 } 2674 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2675 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2676 &req->bu_threshold.b_time, ==)) && 2677 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2678 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2679 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2680 splx(s); 2681 return 0; /* XXX Already installed */ 2682 } 2683 } 2684 2685 /* Allocate the new bw_meter entry */ 2686 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2687 if (x == NULL) { 2688 splx(s); 2689 return ENOBUFS; 2690 } 2691 2692 /* Set the new bw_meter entry */ 2693 x->bm_threshold.b_time = req->bu_threshold.b_time; 2694 microtime(&now); 2695 x->bm_start_time = now; 2696 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2697 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2698 x->bm_measured.b_packets = 0; 2699 x->bm_measured.b_bytes = 0; 2700 x->bm_flags = flags; 2701 x->bm_time_next = NULL; 2702 x->bm_time_hash = BW_METER_BUCKETS; 2703 2704 /* Add the new bw_meter entry to the front of entries for this MFC */ 2705 x->bm_mfc = mfc; 2706 x->bm_mfc_next = mfc->mfc_bw_meter; 2707 mfc->mfc_bw_meter = x; 2708 schedule_bw_meter(x, &now); 2709 splx(s); 2710 2711 return 0; 2712 } 2713 2714 static void 2715 free_bw_list(struct bw_meter *list) 2716 { 2717 while (list != NULL) { 2718 struct bw_meter *x = list; 2719 2720 list = list->bm_mfc_next; 2721 unschedule_bw_meter(x); 2722 free(x, M_BWMETER); 2723 } 2724 } 2725 2726 /* 2727 * Delete one or multiple bw_meter entries 2728 */ 2729 static int 2730 del_bw_upcall(struct mbuf *m) 2731 { 2732 int s; 2733 struct mfc *mfc; 2734 struct bw_meter *x; 2735 struct bw_upcall *req; 2736 2737 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2738 return EINVAL; 2739 2740 req = mtod(m, struct bw_upcall *); 2741 2742 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2743 return EOPNOTSUPP; 2744 2745 s = splsoftnet(); 2746 /* Find the corresponding MFC entry */ 2747 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2748 if (mfc == NULL) { 2749 splx(s); 2750 return EADDRNOTAVAIL; 2751 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2752 /* 2753 * Delete all bw_meter entries for this mfc 2754 */ 2755 struct bw_meter *list; 2756 2757 list = mfc->mfc_bw_meter; 2758 mfc->mfc_bw_meter = NULL; 2759 free_bw_list(list); 2760 splx(s); 2761 return 0; 2762 } else { /* Delete a single bw_meter entry */ 2763 struct bw_meter *prev; 2764 uint32_t flags = 0; 2765 2766 flags = compute_bw_meter_flags(req); 2767 2768 /* Find the bw_meter entry to delete */ 2769 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2770 prev = x, x = x->bm_mfc_next) { 2771 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2772 &req->bu_threshold.b_time, ==)) && 2773 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2774 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2775 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2776 break; 2777 } 2778 if (x != NULL) { /* Delete entry from the list for this MFC */ 2779 if (prev != NULL) 2780 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2781 else 2782 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2783 2784 unschedule_bw_meter(x); 2785 splx(s); 2786 /* Free the bw_meter entry */ 2787 free(x, M_BWMETER); 2788 return 0; 2789 } else { 2790 splx(s); 2791 return EINVAL; 2792 } 2793 } 2794 /* NOTREACHED */ 2795 } 2796 2797 /* 2798 * Perform bandwidth measurement processing that may result in an upcall 2799 */ 2800 static void 2801 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2802 { 2803 struct timeval delta; 2804 2805 delta = *nowp; 2806 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2807 2808 if (x->bm_flags & BW_METER_GEQ) { 2809 /* 2810 * Processing for ">=" type of bw_meter entry 2811 */ 2812 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2813 /* Reset the bw_meter entry */ 2814 x->bm_start_time = *nowp; 2815 x->bm_measured.b_packets = 0; 2816 x->bm_measured.b_bytes = 0; 2817 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2818 } 2819 2820 /* Record that a packet is received */ 2821 x->bm_measured.b_packets++; 2822 x->bm_measured.b_bytes += plen; 2823 2824 /* 2825 * Test if we should deliver an upcall 2826 */ 2827 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2828 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2829 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2830 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2831 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2832 /* Prepare an upcall for delivery */ 2833 bw_meter_prepare_upcall(x, nowp); 2834 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2835 } 2836 } 2837 } else if (x->bm_flags & BW_METER_LEQ) { 2838 /* 2839 * Processing for "<=" type of bw_meter entry 2840 */ 2841 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2842 /* 2843 * We are behind time with the multicast forwarding table 2844 * scanning for "<=" type of bw_meter entries, so test now 2845 * if we should deliver an upcall. 2846 */ 2847 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2848 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2849 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2850 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2851 /* Prepare an upcall for delivery */ 2852 bw_meter_prepare_upcall(x, nowp); 2853 } 2854 /* Reschedule the bw_meter entry */ 2855 unschedule_bw_meter(x); 2856 schedule_bw_meter(x, nowp); 2857 } 2858 2859 /* Record that a packet is received */ 2860 x->bm_measured.b_packets++; 2861 x->bm_measured.b_bytes += plen; 2862 2863 /* 2864 * Test if we should restart the measuring interval 2865 */ 2866 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2867 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2868 (x->bm_flags & BW_METER_UNIT_BYTES && 2869 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2870 /* Don't restart the measuring interval */ 2871 } else { 2872 /* Do restart the measuring interval */ 2873 /* 2874 * XXX: note that we don't unschedule and schedule, because this 2875 * might be too much overhead per packet. Instead, when we process 2876 * all entries for a given timer hash bin, we check whether it is 2877 * really a timeout. If not, we reschedule at that time. 2878 */ 2879 x->bm_start_time = *nowp; 2880 x->bm_measured.b_packets = 0; 2881 x->bm_measured.b_bytes = 0; 2882 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2883 } 2884 } 2885 } 2886 2887 /* 2888 * Prepare a bandwidth-related upcall 2889 */ 2890 static void 2891 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2892 { 2893 struct timeval delta; 2894 struct bw_upcall *u; 2895 2896 /* 2897 * Compute the measured time interval 2898 */ 2899 delta = *nowp; 2900 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2901 2902 /* 2903 * If there are too many pending upcalls, deliver them now 2904 */ 2905 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2906 bw_upcalls_send(); 2907 2908 /* 2909 * Set the bw_upcall entry 2910 */ 2911 u = &bw_upcalls[bw_upcalls_n++]; 2912 u->bu_src = x->bm_mfc->mfc_origin; 2913 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2914 u->bu_threshold.b_time = x->bm_threshold.b_time; 2915 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2916 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2917 u->bu_measured.b_time = delta; 2918 u->bu_measured.b_packets = x->bm_measured.b_packets; 2919 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2920 u->bu_flags = 0; 2921 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2922 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2923 if (x->bm_flags & BW_METER_UNIT_BYTES) 2924 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2925 if (x->bm_flags & BW_METER_GEQ) 2926 u->bu_flags |= BW_UPCALL_GEQ; 2927 if (x->bm_flags & BW_METER_LEQ) 2928 u->bu_flags |= BW_UPCALL_LEQ; 2929 } 2930 2931 /* 2932 * Send the pending bandwidth-related upcalls 2933 */ 2934 static void 2935 bw_upcalls_send(void) 2936 { 2937 struct mbuf *m; 2938 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2939 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2940 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2941 0, /* unused2 */ 2942 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2943 0, /* im_mbz */ 2944 0, /* im_vif */ 2945 0, /* unused3 */ 2946 { 0 }, /* im_src */ 2947 { 0 } }; /* im_dst */ 2948 2949 if (bw_upcalls_n == 0) 2950 return; /* No pending upcalls */ 2951 2952 bw_upcalls_n = 0; 2953 2954 /* 2955 * Allocate a new mbuf, initialize it with the header and 2956 * the payload for the pending calls. 2957 */ 2958 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2959 if (m == NULL) { 2960 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2961 return; 2962 } 2963 2964 m->m_len = m->m_pkthdr.len = 0; 2965 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2966 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2967 2968 /* 2969 * Send the upcalls 2970 * XXX do we need to set the address in k_igmpsrc ? 2971 */ 2972 mrtstat.mrts_upcalls++; 2973 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2974 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2975 ++mrtstat.mrts_upq_sockfull; 2976 } 2977 } 2978 2979 /* 2980 * Compute the timeout hash value for the bw_meter entries 2981 */ 2982 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2983 do { \ 2984 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2985 \ 2986 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2987 (hash) = next_timeval.tv_sec; \ 2988 if (next_timeval.tv_usec) \ 2989 (hash)++; /* XXX: make sure we don't timeout early */ \ 2990 (hash) %= BW_METER_BUCKETS; \ 2991 } while (/*CONSTCOND*/ 0) 2992 2993 /* 2994 * Schedule a timer to process periodically bw_meter entry of type "<=" 2995 * by linking the entry in the proper hash bucket. 2996 */ 2997 static void 2998 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2999 { 3000 int time_hash; 3001 3002 if (!(x->bm_flags & BW_METER_LEQ)) 3003 return; /* XXX: we schedule timers only for "<=" entries */ 3004 3005 /* 3006 * Reset the bw_meter entry 3007 */ 3008 x->bm_start_time = *nowp; 3009 x->bm_measured.b_packets = 0; 3010 x->bm_measured.b_bytes = 0; 3011 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 3012 3013 /* 3014 * Compute the timeout hash value and insert the entry 3015 */ 3016 BW_METER_TIMEHASH(x, time_hash); 3017 x->bm_time_next = bw_meter_timers[time_hash]; 3018 bw_meter_timers[time_hash] = x; 3019 x->bm_time_hash = time_hash; 3020 } 3021 3022 /* 3023 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 3024 * by removing the entry from the proper hash bucket. 3025 */ 3026 static void 3027 unschedule_bw_meter(struct bw_meter *x) 3028 { 3029 int time_hash; 3030 struct bw_meter *prev, *tmp; 3031 3032 if (!(x->bm_flags & BW_METER_LEQ)) 3033 return; /* XXX: we schedule timers only for "<=" entries */ 3034 3035 /* 3036 * Compute the timeout hash value and delete the entry 3037 */ 3038 time_hash = x->bm_time_hash; 3039 if (time_hash >= BW_METER_BUCKETS) 3040 return; /* Entry was not scheduled */ 3041 3042 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 3043 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 3044 if (tmp == x) 3045 break; 3046 3047 if (tmp == NULL) 3048 panic("unschedule_bw_meter: bw_meter entry not found"); 3049 3050 if (prev != NULL) 3051 prev->bm_time_next = x->bm_time_next; 3052 else 3053 bw_meter_timers[time_hash] = x->bm_time_next; 3054 3055 x->bm_time_next = NULL; 3056 x->bm_time_hash = BW_METER_BUCKETS; 3057 } 3058 3059 /* 3060 * Process all "<=" type of bw_meter that should be processed now, 3061 * and for each entry prepare an upcall if necessary. Each processed 3062 * entry is rescheduled again for the (periodic) processing. 3063 * 3064 * This is run periodically (once per second normally). On each round, 3065 * all the potentially matching entries are in the hash slot that we are 3066 * looking at. 3067 */ 3068 static void 3069 bw_meter_process() 3070 { 3071 int s; 3072 static uint32_t last_tv_sec; /* last time we processed this */ 3073 3074 uint32_t loops; 3075 int i; 3076 struct timeval now, process_endtime; 3077 3078 microtime(&now); 3079 if (last_tv_sec == now.tv_sec) 3080 return; /* nothing to do */ 3081 3082 loops = now.tv_sec - last_tv_sec; 3083 last_tv_sec = now.tv_sec; 3084 if (loops > BW_METER_BUCKETS) 3085 loops = BW_METER_BUCKETS; 3086 3087 s = splsoftnet(); 3088 /* 3089 * Process all bins of bw_meter entries from the one after the last 3090 * processed to the current one. On entry, i points to the last bucket 3091 * visited, so we need to increment i at the beginning of the loop. 3092 */ 3093 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 3094 struct bw_meter *x, *tmp_list; 3095 3096 if (++i >= BW_METER_BUCKETS) 3097 i = 0; 3098 3099 /* Disconnect the list of bw_meter entries from the bin */ 3100 tmp_list = bw_meter_timers[i]; 3101 bw_meter_timers[i] = NULL; 3102 3103 /* Process the list of bw_meter entries */ 3104 while (tmp_list != NULL) { 3105 x = tmp_list; 3106 tmp_list = tmp_list->bm_time_next; 3107 3108 /* Test if the time interval is over */ 3109 process_endtime = x->bm_start_time; 3110 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 3111 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 3112 /* Not yet: reschedule, but don't reset */ 3113 int time_hash; 3114 3115 BW_METER_TIMEHASH(x, time_hash); 3116 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 3117 /* 3118 * XXX: somehow the bin processing is a bit ahead of time. 3119 * Put the entry in the next bin. 3120 */ 3121 if (++time_hash >= BW_METER_BUCKETS) 3122 time_hash = 0; 3123 } 3124 x->bm_time_next = bw_meter_timers[time_hash]; 3125 bw_meter_timers[time_hash] = x; 3126 x->bm_time_hash = time_hash; 3127 3128 continue; 3129 } 3130 3131 /* 3132 * Test if we should deliver an upcall 3133 */ 3134 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 3135 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 3136 ((x->bm_flags & BW_METER_UNIT_BYTES) && 3137 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 3138 /* Prepare an upcall for delivery */ 3139 bw_meter_prepare_upcall(x, &now); 3140 } 3141 3142 /* 3143 * Reschedule for next processing 3144 */ 3145 schedule_bw_meter(x, &now); 3146 } 3147 } 3148 3149 /* Send all upcalls that are pending delivery */ 3150 bw_upcalls_send(); 3151 3152 splx(s); 3153 } 3154 3155 /* 3156 * A periodic function for sending all upcalls that are pending delivery 3157 */ 3158 static void 3159 expire_bw_upcalls_send(void *unused) 3160 { 3161 int s; 3162 3163 s = splsoftnet(); 3164 bw_upcalls_send(); 3165 splx(s); 3166 3167 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 3168 expire_bw_upcalls_send, NULL); 3169 } 3170 3171 /* 3172 * A periodic function for periodic scanning of the multicast forwarding 3173 * table for processing all "<=" bw_meter entries. 3174 */ 3175 static void 3176 expire_bw_meter_process(void *unused) 3177 { 3178 if (mrt_api_config & MRT_MFC_BW_UPCALL) 3179 bw_meter_process(); 3180 3181 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 3182 expire_bw_meter_process, NULL); 3183 } 3184 3185 /* 3186 * End of bandwidth monitoring code 3187 */ 3188 3189 #ifdef PIM 3190 /* 3191 * Send the packet up to the user daemon, or eventually do kernel encapsulation 3192 */ 3193 static int 3194 pim_register_send(struct ip *ip, struct vif *vifp, 3195 struct mbuf *m, struct mfc *rt) 3196 { 3197 struct mbuf *mb_copy, *mm; 3198 3199 if (mrtdebug & DEBUG_PIM) 3200 log(LOG_DEBUG, "pim_register_send: "); 3201 3202 mb_copy = pim_register_prepare(ip, m); 3203 if (mb_copy == NULL) 3204 return ENOBUFS; 3205 3206 /* 3207 * Send all the fragments. Note that the mbuf for each fragment 3208 * is freed by the sending machinery. 3209 */ 3210 for (mm = mb_copy; mm; mm = mb_copy) { 3211 mb_copy = mm->m_nextpkt; 3212 mm->m_nextpkt = NULL; 3213 mm = m_pullup(mm, sizeof(struct ip)); 3214 if (mm != NULL) { 3215 ip = mtod(mm, struct ip *); 3216 if ((mrt_api_config & MRT_MFC_RP) && 3217 !in_nullhost(rt->mfc_rp)) { 3218 pim_register_send_rp(ip, vifp, mm, rt); 3219 } else { 3220 pim_register_send_upcall(ip, vifp, mm, rt); 3221 } 3222 } 3223 } 3224 3225 return 0; 3226 } 3227 3228 /* 3229 * Return a copy of the data packet that is ready for PIM Register 3230 * encapsulation. 3231 * XXX: Note that in the returned copy the IP header is a valid one. 3232 */ 3233 static struct mbuf * 3234 pim_register_prepare(struct ip *ip, struct mbuf *m) 3235 { 3236 struct mbuf *mb_copy = NULL; 3237 int mtu; 3238 3239 /* Take care of delayed checksums */ 3240 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 3241 in_delayed_cksum(m); 3242 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 3243 } 3244 3245 /* 3246 * Copy the old packet & pullup its IP header into the 3247 * new mbuf so we can modify it. 3248 */ 3249 mb_copy = m_copy(m, 0, M_COPYALL); 3250 if (mb_copy == NULL) 3251 return NULL; 3252 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 3253 if (mb_copy == NULL) 3254 return NULL; 3255 3256 /* take care of the TTL */ 3257 ip = mtod(mb_copy, struct ip *); 3258 --ip->ip_ttl; 3259 3260 /* Compute the MTU after the PIM Register encapsulation */ 3261 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 3262 3263 if (ntohs(ip->ip_len) <= mtu) { 3264 /* Turn the IP header into a valid one */ 3265 ip->ip_sum = 0; 3266 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 3267 } else { 3268 /* Fragment the packet */ 3269 if (ip_fragment(mb_copy, NULL, mtu) != 0) { 3270 /* XXX: mb_copy was freed by ip_fragment() */ 3271 return NULL; 3272 } 3273 } 3274 return mb_copy; 3275 } 3276 3277 /* 3278 * Send an upcall with the data packet to the user-level process. 3279 */ 3280 static int 3281 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 3282 struct mbuf *mb_copy, struct mfc *rt) 3283 { 3284 struct mbuf *mb_first; 3285 int len = ntohs(ip->ip_len); 3286 struct igmpmsg *im; 3287 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 3288 3289 /* 3290 * Add a new mbuf with an upcall header 3291 */ 3292 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3293 if (mb_first == NULL) { 3294 m_freem(mb_copy); 3295 return ENOBUFS; 3296 } 3297 mb_first->m_data += max_linkhdr; 3298 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3299 mb_first->m_len = sizeof(struct igmpmsg); 3300 mb_first->m_next = mb_copy; 3301 3302 /* Send message to routing daemon */ 3303 im = mtod(mb_first, struct igmpmsg *); 3304 im->im_msgtype = IGMPMSG_WHOLEPKT; 3305 im->im_mbz = 0; 3306 im->im_vif = vifp - viftable; 3307 im->im_src = ip->ip_src; 3308 im->im_dst = ip->ip_dst; 3309 3310 k_igmpsrc.sin_addr = ip->ip_src; 3311 3312 mrtstat.mrts_upcalls++; 3313 3314 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3315 if (mrtdebug & DEBUG_PIM) 3316 log(LOG_WARNING, 3317 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3318 ++mrtstat.mrts_upq_sockfull; 3319 return ENOBUFS; 3320 } 3321 3322 /* Keep statistics */ 3323 pimstat.pims_snd_registers_msgs++; 3324 pimstat.pims_snd_registers_bytes += len; 3325 3326 return 0; 3327 } 3328 3329 /* 3330 * Encapsulate the data packet in PIM Register message and send it to the RP. 3331 */ 3332 static int 3333 pim_register_send_rp(struct ip *ip, struct vif *vifp, 3334 struct mbuf *mb_copy, struct mfc *rt) 3335 { 3336 struct mbuf *mb_first; 3337 struct ip *ip_outer; 3338 struct pim_encap_pimhdr *pimhdr; 3339 int len = ntohs(ip->ip_len); 3340 vifi_t vifi = rt->mfc_parent; 3341 3342 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) { 3343 m_freem(mb_copy); 3344 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3345 } 3346 3347 /* 3348 * Add a new mbuf with the encapsulating header 3349 */ 3350 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3351 if (mb_first == NULL) { 3352 m_freem(mb_copy); 3353 return ENOBUFS; 3354 } 3355 mb_first->m_data += max_linkhdr; 3356 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3357 mb_first->m_next = mb_copy; 3358 3359 mb_first->m_pkthdr.len = len + mb_first->m_len; 3360 3361 /* 3362 * Fill in the encapsulating IP and PIM header 3363 */ 3364 ip_outer = mtod(mb_first, struct ip *); 3365 *ip_outer = pim_encap_iphdr; 3366 ip_outer->ip_id = ip_newid(); 3367 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + 3368 sizeof(pim_encap_pimhdr)); 3369 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3370 ip_outer->ip_dst = rt->mfc_rp; 3371 /* 3372 * Copy the inner header TOS to the outer header, and take care of the 3373 * IP_DF bit. 3374 */ 3375 ip_outer->ip_tos = ip->ip_tos; 3376 if (ntohs(ip->ip_off) & IP_DF) 3377 ip_outer->ip_off |= IP_DF; 3378 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 3379 + sizeof(pim_encap_iphdr)); 3380 *pimhdr = pim_encap_pimhdr; 3381 /* If the iif crosses a border, set the Border-bit */ 3382 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3383 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3384 3385 mb_first->m_data += sizeof(pim_encap_iphdr); 3386 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3387 mb_first->m_data -= sizeof(pim_encap_iphdr); 3388 3389 if (vifp->v_rate_limit == 0) 3390 tbf_send_packet(vifp, mb_first); 3391 else 3392 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len)); 3393 3394 /* Keep statistics */ 3395 pimstat.pims_snd_registers_msgs++; 3396 pimstat.pims_snd_registers_bytes += len; 3397 3398 return 0; 3399 } 3400 3401 /* 3402 * PIM-SMv2 and PIM-DM messages processing. 3403 * Receives and verifies the PIM control messages, and passes them 3404 * up to the listening socket, using rip_input(). 3405 * The only message with special processing is the PIM_REGISTER message 3406 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3407 * is passed to if_simloop(). 3408 */ 3409 void 3410 pim_input(struct mbuf *m, ...) 3411 { 3412 struct ip *ip = mtod(m, struct ip *); 3413 struct pim *pim; 3414 int minlen; 3415 int datalen; 3416 int ip_tos; 3417 int proto; 3418 int iphlen; 3419 va_list ap; 3420 3421 va_start(ap, m); 3422 iphlen = va_arg(ap, int); 3423 proto = va_arg(ap, int); 3424 va_end(ap); 3425 3426 datalen = ntohs(ip->ip_len) - iphlen; 3427 3428 /* Keep statistics */ 3429 pimstat.pims_rcv_total_msgs++; 3430 pimstat.pims_rcv_total_bytes += datalen; 3431 3432 /* 3433 * Validate lengths 3434 */ 3435 if (datalen < PIM_MINLEN) { 3436 pimstat.pims_rcv_tooshort++; 3437 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3438 datalen, (u_long)ip->ip_src.s_addr); 3439 m_freem(m); 3440 return; 3441 } 3442 3443 /* 3444 * If the packet is at least as big as a REGISTER, go agead 3445 * and grab the PIM REGISTER header size, to avoid another 3446 * possible m_pullup() later. 3447 * 3448 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3449 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3450 */ 3451 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3452 /* 3453 * Get the IP and PIM headers in contiguous memory, and 3454 * possibly the PIM REGISTER header. 3455 */ 3456 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3457 (m = m_pullup(m, minlen)) == NULL) { 3458 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3459 return; 3460 } 3461 /* m_pullup() may have given us a new mbuf so reset ip. */ 3462 ip = mtod(m, struct ip *); 3463 ip_tos = ip->ip_tos; 3464 3465 /* adjust mbuf to point to the PIM header */ 3466 m->m_data += iphlen; 3467 m->m_len -= iphlen; 3468 pim = mtod(m, struct pim *); 3469 3470 /* 3471 * Validate checksum. If PIM REGISTER, exclude the data packet. 3472 * 3473 * XXX: some older PIMv2 implementations don't make this distinction, 3474 * so for compatibility reason perform the checksum over part of the 3475 * message, and if error, then over the whole message. 3476 */ 3477 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3478 /* do nothing, checksum okay */ 3479 } else if (in_cksum(m, datalen)) { 3480 pimstat.pims_rcv_badsum++; 3481 if (mrtdebug & DEBUG_PIM) 3482 log(LOG_DEBUG, "pim_input: invalid checksum"); 3483 m_freem(m); 3484 return; 3485 } 3486 3487 /* PIM version check */ 3488 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3489 pimstat.pims_rcv_badversion++; 3490 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3491 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3492 m_freem(m); 3493 return; 3494 } 3495 3496 /* restore mbuf back to the outer IP */ 3497 m->m_data -= iphlen; 3498 m->m_len += iphlen; 3499 3500 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3501 /* 3502 * Since this is a REGISTER, we'll make a copy of the register 3503 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3504 * routing daemon. 3505 */ 3506 int s; 3507 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 3508 struct mbuf *mcp; 3509 struct ip *encap_ip; 3510 u_int32_t *reghdr; 3511 struct ifnet *vifp; 3512 3513 s = splsoftnet(); 3514 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3515 splx(s); 3516 if (mrtdebug & DEBUG_PIM) 3517 log(LOG_DEBUG, 3518 "pim_input: register vif not set: %d\n", reg_vif_num); 3519 m_freem(m); 3520 return; 3521 } 3522 /* XXX need refcnt? */ 3523 vifp = viftable[reg_vif_num].v_ifp; 3524 splx(s); 3525 3526 /* 3527 * Validate length 3528 */ 3529 if (datalen < PIM_REG_MINLEN) { 3530 pimstat.pims_rcv_tooshort++; 3531 pimstat.pims_rcv_badregisters++; 3532 log(LOG_ERR, 3533 "pim_input: register packet size too small %d from %lx\n", 3534 datalen, (u_long)ip->ip_src.s_addr); 3535 m_freem(m); 3536 return; 3537 } 3538 3539 reghdr = (u_int32_t *)(pim + 1); 3540 encap_ip = (struct ip *)(reghdr + 1); 3541 3542 if (mrtdebug & DEBUG_PIM) { 3543 log(LOG_DEBUG, 3544 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3545 (u_long)ntohl(encap_ip->ip_src.s_addr), 3546 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3547 ntohs(encap_ip->ip_len)); 3548 } 3549 3550 /* verify the version number of the inner packet */ 3551 if (encap_ip->ip_v != IPVERSION) { 3552 pimstat.pims_rcv_badregisters++; 3553 if (mrtdebug & DEBUG_PIM) { 3554 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3555 "of the inner packet\n", encap_ip->ip_v); 3556 } 3557 m_freem(m); 3558 return; 3559 } 3560 3561 /* verify the inner packet is destined to a mcast group */ 3562 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) { 3563 pimstat.pims_rcv_badregisters++; 3564 if (mrtdebug & DEBUG_PIM) 3565 log(LOG_DEBUG, 3566 "pim_input: inner packet of register is not " 3567 "multicast %lx\n", 3568 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3569 m_freem(m); 3570 return; 3571 } 3572 3573 /* If a NULL_REGISTER, pass it to the daemon */ 3574 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3575 goto pim_input_to_daemon; 3576 3577 /* 3578 * Copy the TOS from the outer IP header to the inner IP header. 3579 */ 3580 if (encap_ip->ip_tos != ip_tos) { 3581 /* Outer TOS -> inner TOS */ 3582 encap_ip->ip_tos = ip_tos; 3583 /* Recompute the inner header checksum. Sigh... */ 3584 3585 /* adjust mbuf to point to the inner IP header */ 3586 m->m_data += (iphlen + PIM_MINLEN); 3587 m->m_len -= (iphlen + PIM_MINLEN); 3588 3589 encap_ip->ip_sum = 0; 3590 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3591 3592 /* restore mbuf to point back to the outer IP header */ 3593 m->m_data -= (iphlen + PIM_MINLEN); 3594 m->m_len += (iphlen + PIM_MINLEN); 3595 } 3596 3597 /* 3598 * Decapsulate the inner IP packet and loopback to forward it 3599 * as a normal multicast packet. Also, make a copy of the 3600 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3601 * to pass to the daemon later, so it can take the appropriate 3602 * actions (e.g., send back PIM_REGISTER_STOP). 3603 * XXX: here m->m_data points to the outer IP header. 3604 */ 3605 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3606 if (mcp == NULL) { 3607 log(LOG_ERR, 3608 "pim_input: pim register: could not copy register head\n"); 3609 m_freem(m); 3610 return; 3611 } 3612 3613 /* Keep statistics */ 3614 /* XXX: registers_bytes include only the encap. mcast pkt */ 3615 pimstat.pims_rcv_registers_msgs++; 3616 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3617 3618 /* 3619 * forward the inner ip packet; point m_data at the inner ip. 3620 */ 3621 m_adj(m, iphlen + PIM_MINLEN); 3622 3623 if (mrtdebug & DEBUG_PIM) { 3624 log(LOG_DEBUG, 3625 "pim_input: forwarding decapsulated register: " 3626 "src %lx, dst %lx, vif %d\n", 3627 (u_long)ntohl(encap_ip->ip_src.s_addr), 3628 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3629 reg_vif_num); 3630 } 3631 /* NB: vifp was collected above; can it change on us? */ 3632 looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL); 3633 3634 /* prepare the register head to send to the mrouting daemon */ 3635 m = mcp; 3636 } 3637 3638 pim_input_to_daemon: 3639 /* 3640 * Pass the PIM message up to the daemon; if it is a Register message, 3641 * pass the 'head' only up to the daemon. This includes the 3642 * outer IP header, PIM header, PIM-Register header and the 3643 * inner IP header. 3644 * XXX: the outer IP header pkt size of a Register is not adjust to 3645 * reflect the fact that the inner multicast data is truncated. 3646 */ 3647 rip_input(m, iphlen, proto); 3648 3649 return; 3650 } 3651 #endif /* PIM */ 3652
Indexes created Thu Sep 25 16:09:42 GMT 2025