1 /* $NetBSD: ip_mroute.c,v 1.99 2006/10/12 01:32:38 christos 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 1.99 2006/10/12 01:32:38 christos Exp $"); 97 98 #include "opt_inet.h" 99 #include "opt_ipsec.h" 100 #include "opt_pim.h" 101 102 #ifdef PIM 103 #define _PIM_VT 1 104 #endif 105 106 #include <sys/param.h> 107 #include <sys/systm.h> 108 #include <sys/callout.h> 109 #include <sys/mbuf.h> 110 #include <sys/socket.h> 111 #include <sys/socketvar.h> 112 #include <sys/protosw.h> 113 #include <sys/errno.h> 114 #include <sys/time.h> 115 #include <sys/kernel.h> 116 #include <sys/ioctl.h> 117 #include <sys/syslog.h> 118 119 #include <net/if.h> 120 #include <net/route.h> 121 #include <net/raw_cb.h> 122 123 #include <netinet/in.h> 124 #include <netinet/in_var.h> 125 #include <netinet/in_systm.h> 126 #include <netinet/ip.h> 127 #include <netinet/ip_var.h> 128 #include <netinet/in_pcb.h> 129 #include <netinet/udp.h> 130 #include <netinet/igmp.h> 131 #include <netinet/igmp_var.h> 132 #include <netinet/ip_mroute.h> 133 #ifdef PIM 134 #include <netinet/pim.h> 135 #include <netinet/pim_var.h> 136 #endif 137 #include <netinet/ip_encap.h> 138 139 #ifdef IPSEC 140 #include <netinet6/ipsec.h> 141 #include <netkey/key.h> 142 #endif 143 144 #ifdef FAST_IPSEC 145 #include <netipsec/ipsec.h> 146 #include <netipsec/key.h> 147 #endif 148 149 #include <machine/stdarg.h> 150 151 #define IP_MULTICASTOPTS 0 152 #define M_PULLUP(m, len) \ 153 do { \ 154 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \ 155 (m) = m_pullup((m), (len)); \ 156 } while (/*CONSTCOND*/ 0) 157 158 /* 159 * Globals. All but ip_mrouter and ip_mrtproto could be static, 160 * except for netstat or debugging purposes. 161 */ 162 struct socket *ip_mrouter = NULL; 163 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 164 165 #define NO_RTE_FOUND 0x1 166 #define RTE_FOUND 0x2 167 168 #define MFCHASH(a, g) \ 169 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ 170 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash) 171 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl; 172 u_long mfchash; 173 174 u_char nexpire[MFCTBLSIZ]; 175 struct vif viftable[MAXVIFS]; 176 struct mrtstat mrtstat; 177 u_int mrtdebug = 0; /* debug level */ 178 #define DEBUG_MFC 0x02 179 #define DEBUG_FORWARD 0x04 180 #define DEBUG_EXPIRE 0x08 181 #define DEBUG_XMIT 0x10 182 #define DEBUG_PIM 0x20 183 184 #define VIFI_INVALID ((vifi_t) -1) 185 186 u_int tbfdebug = 0; /* tbf debug level */ 187 #ifdef RSVP_ISI 188 u_int rsvpdebug = 0; /* rsvp debug level */ 189 extern struct socket *ip_rsvpd; 190 extern int rsvp_on; 191 #endif /* RSVP_ISI */ 192 193 /* vif attachment using sys/netinet/ip_encap.c */ 194 static void vif_input(struct mbuf *, ...); 195 static int vif_encapcheck(struct mbuf *, int, int, void *); 196 197 static const struct protosw vif_protosw = 198 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 199 vif_input, rip_output, 0, rip_ctloutput, 200 rip_usrreq, 201 0, 0, 0, 0, 202 }; 203 204 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 205 #define UPCALL_EXPIRE 6 /* number of timeouts */ 206 207 /* 208 * Define the token bucket filter structures 209 */ 210 211 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 212 213 static int get_sg_cnt(struct sioc_sg_req *); 214 static int get_vif_cnt(struct sioc_vif_req *); 215 static int ip_mrouter_init(struct socket *, struct mbuf *); 216 static int get_version(struct mbuf *); 217 static int set_assert(struct mbuf *); 218 static int get_assert(struct mbuf *); 219 static int add_vif(struct mbuf *); 220 static int del_vif(struct mbuf *); 221 static void update_mfc_params(struct mfc *, struct mfcctl2 *); 222 static void init_mfc_params(struct mfc *, struct mfcctl2 *); 223 static void expire_mfc(struct mfc *); 224 static int add_mfc(struct mbuf *); 225 #ifdef UPCALL_TIMING 226 static void collate(struct timeval *); 227 #endif 228 static int del_mfc(struct mbuf *); 229 static int set_api_config(struct mbuf *); /* chose API capabilities */ 230 static int get_api_support(struct mbuf *); 231 static int get_api_config(struct mbuf *); 232 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 233 static void expire_upcalls(void *); 234 #ifdef RSVP_ISI 235 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 236 #else 237 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *); 238 #endif 239 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 240 static void encap_send(struct ip *, struct vif *, struct mbuf *); 241 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t); 242 static void tbf_queue(struct vif *, struct mbuf *); 243 static void tbf_process_q(struct vif *); 244 static void tbf_reprocess_q(void *); 245 static int tbf_dq_sel(struct vif *, struct ip *); 246 static void tbf_send_packet(struct vif *, struct mbuf *); 247 static void tbf_update_tokens(struct vif *); 248 static int priority(struct vif *, struct ip *); 249 250 /* 251 * Bandwidth monitoring 252 */ 253 static void free_bw_list(struct bw_meter *); 254 static int add_bw_upcall(struct mbuf *); 255 static int del_bw_upcall(struct mbuf *); 256 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *); 257 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *); 258 static void bw_upcalls_send(void); 259 static void schedule_bw_meter(struct bw_meter *, struct timeval *); 260 static void unschedule_bw_meter(struct bw_meter *); 261 static void bw_meter_process(void); 262 static void expire_bw_upcalls_send(void *); 263 static void expire_bw_meter_process(void *); 264 265 #ifdef PIM 266 static int pim_register_send(struct ip *, struct vif *, 267 struct mbuf *, struct mfc *); 268 static int pim_register_send_rp(struct ip *, struct vif *, 269 struct mbuf *, struct mfc *); 270 static int pim_register_send_upcall(struct ip *, struct vif *, 271 struct mbuf *, struct mfc *); 272 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 273 #endif 274 275 /* 276 * 'Interfaces' associated with decapsulator (so we can tell 277 * packets that went through it from ones that get reflected 278 * by a broken gateway). These interfaces are never linked into 279 * the system ifnet list & no routes point to them. I.e., packets 280 * can't be sent this way. They only exist as a placeholder for 281 * multicast source verification. 282 */ 283 #if 0 284 struct ifnet multicast_decap_if[MAXVIFS]; 285 #endif 286 287 #define ENCAP_TTL 64 288 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 289 290 /* prototype IP hdr for encapsulated packets */ 291 struct ip multicast_encap_iphdr = { 292 .ip_hl = sizeof(struct ip) >> 2, 293 .ip_v = IPVERSION, 294 .ip_len = sizeof(struct ip), 295 .ip_ttl = ENCAP_TTL, 296 .ip_p = ENCAP_PROTO, 297 }; 298 299 /* 300 * Bandwidth meter variables and constants 301 */ 302 303 /* 304 * Pending timeouts are stored in a hash table, the key being the 305 * expiration time. Periodically, the entries are analysed and processed. 306 */ 307 #define BW_METER_BUCKETS 1024 308 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 309 struct callout bw_meter_ch; 310 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 311 312 /* 313 * Pending upcalls are stored in a vector which is flushed when 314 * full, or periodically 315 */ 316 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 317 static u_int bw_upcalls_n; /* # of pending upcalls */ 318 struct callout bw_upcalls_ch; 319 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 320 321 #ifdef PIM 322 struct pimstat pimstat; 323 324 /* 325 * Note: the PIM Register encapsulation adds the following in front of a 326 * data packet: 327 * 328 * struct pim_encap_hdr { 329 * struct ip ip; 330 * struct pim_encap_pimhdr pim; 331 * } 332 * 333 */ 334 335 struct pim_encap_pimhdr { 336 struct pim pim; 337 uint32_t flags; 338 }; 339 340 static struct ip pim_encap_iphdr = { 341 .ip_v = IPVERSION, 342 .ip_hl = sizeof(struct ip) >> 2, 343 .ip_len = sizeof(struct ip), 344 .ip_ttl = ENCAP_TTL, 345 .ip_p = IPPROTO_PIM, 346 }; 347 348 static struct pim_encap_pimhdr pim_encap_pimhdr = { 349 { 350 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 351 0, /* reserved */ 352 0, /* checksum */ 353 }, 354 0 /* flags */ 355 }; 356 357 static struct ifnet multicast_register_if; 358 static vifi_t reg_vif_num = VIFI_INVALID; 359 #endif /* PIM */ 360 361 362 /* 363 * Private variables. 364 */ 365 static vifi_t numvifs = 0; 366 367 static struct callout expire_upcalls_ch; 368 369 /* 370 * whether or not special PIM assert processing is enabled. 371 */ 372 static int pim_assert; 373 /* 374 * Rate limit for assert notification messages, in usec 375 */ 376 #define ASSERT_MSG_TIME 3000000 377 378 /* 379 * Kernel multicast routing API capabilities and setup. 380 * If more API capabilities are added to the kernel, they should be 381 * recorded in `mrt_api_support'. 382 */ 383 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 384 MRT_MFC_FLAGS_BORDER_VIF | 385 MRT_MFC_RP | 386 MRT_MFC_BW_UPCALL); 387 static u_int32_t mrt_api_config = 0; 388 389 /* 390 * Find a route for a given origin IP address and Multicast group address 391 * Type of service parameter to be added in the future!!! 392 * Statistics are updated by the caller if needed 393 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 394 */ 395 static struct mfc * 396 mfc_find(struct in_addr *o, struct in_addr *g) 397 { 398 struct mfc *rt; 399 400 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { 401 if (in_hosteq(rt->mfc_origin, *o) && 402 in_hosteq(rt->mfc_mcastgrp, *g) && 403 (rt->mfc_stall == NULL)) 404 break; 405 } 406 407 return (rt); 408 } 409 410 /* 411 * Macros to compute elapsed time efficiently 412 * Borrowed from Van Jacobson's scheduling code 413 */ 414 #define TV_DELTA(a, b, delta) do { \ 415 int xxs; \ 416 delta = (a).tv_usec - (b).tv_usec; \ 417 xxs = (a).tv_sec - (b).tv_sec; \ 418 switch (xxs) { \ 419 case 2: \ 420 delta += 1000000; \ 421 /* fall through */ \ 422 case 1: \ 423 delta += 1000000; \ 424 /* fall through */ \ 425 case 0: \ 426 break; \ 427 default: \ 428 delta += (1000000 * xxs); \ 429 break; \ 430 } \ 431 } while (/*CONSTCOND*/ 0) 432 433 #ifdef UPCALL_TIMING 434 u_int32_t upcall_data[51]; 435 #endif /* UPCALL_TIMING */ 436 437 /* 438 * Handle MRT setsockopt commands to modify the multicast routing tables. 439 */ 440 int 441 ip_mrouter_set(struct socket *so, int optname, struct mbuf **m) 442 { 443 int error; 444 445 if (optname != MRT_INIT && so != ip_mrouter) 446 error = ENOPROTOOPT; 447 else 448 switch (optname) { 449 case MRT_INIT: 450 error = ip_mrouter_init(so, *m); 451 break; 452 case MRT_DONE: 453 error = ip_mrouter_done(); 454 break; 455 case MRT_ADD_VIF: 456 error = add_vif(*m); 457 break; 458 case MRT_DEL_VIF: 459 error = del_vif(*m); 460 break; 461 case MRT_ADD_MFC: 462 error = add_mfc(*m); 463 break; 464 case MRT_DEL_MFC: 465 error = del_mfc(*m); 466 break; 467 case MRT_ASSERT: 468 error = set_assert(*m); 469 break; 470 case MRT_API_CONFIG: 471 error = set_api_config(*m); 472 break; 473 case MRT_ADD_BW_UPCALL: 474 error = add_bw_upcall(*m); 475 break; 476 case MRT_DEL_BW_UPCALL: 477 error = del_bw_upcall(*m); 478 break; 479 default: 480 error = ENOPROTOOPT; 481 break; 482 } 483 484 if (*m) 485 m_free(*m); 486 return (error); 487 } 488 489 /* 490 * Handle MRT getsockopt commands 491 */ 492 int 493 ip_mrouter_get(struct socket *so, int optname, struct mbuf **m) 494 { 495 int error; 496 497 if (so != ip_mrouter) 498 error = ENOPROTOOPT; 499 else { 500 *m = m_get(M_WAIT, MT_SOOPTS); 501 MCLAIM(*m, so->so_mowner); 502 503 switch (optname) { 504 case MRT_VERSION: 505 error = get_version(*m); 506 break; 507 case MRT_ASSERT: 508 error = get_assert(*m); 509 break; 510 case MRT_API_SUPPORT: 511 error = get_api_support(*m); 512 break; 513 case MRT_API_CONFIG: 514 error = get_api_config(*m); 515 break; 516 default: 517 error = ENOPROTOOPT; 518 break; 519 } 520 521 if (error) 522 m_free(*m); 523 } 524 525 return (error); 526 } 527 528 /* 529 * Handle ioctl commands to obtain information from the cache 530 */ 531 int 532 mrt_ioctl(struct socket *so, u_long cmd, caddr_t data) 533 { 534 int error; 535 536 if (so != ip_mrouter) 537 error = EINVAL; 538 else 539 switch (cmd) { 540 case SIOCGETVIFCNT: 541 error = get_vif_cnt((struct sioc_vif_req *)data); 542 break; 543 case SIOCGETSGCNT: 544 error = get_sg_cnt((struct sioc_sg_req *)data); 545 break; 546 default: 547 error = EINVAL; 548 break; 549 } 550 551 return (error); 552 } 553 554 /* 555 * returns the packet, byte, rpf-failure count for the source group provided 556 */ 557 static int 558 get_sg_cnt(struct sioc_sg_req *req) 559 { 560 int s; 561 struct mfc *rt; 562 563 s = splsoftnet(); 564 rt = mfc_find(&req->src, &req->grp); 565 if (rt == NULL) { 566 splx(s); 567 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 568 return (EADDRNOTAVAIL); 569 } 570 req->pktcnt = rt->mfc_pkt_cnt; 571 req->bytecnt = rt->mfc_byte_cnt; 572 req->wrong_if = rt->mfc_wrong_if; 573 splx(s); 574 575 return (0); 576 } 577 578 /* 579 * returns the input and output packet and byte counts on the vif provided 580 */ 581 static int 582 get_vif_cnt(struct sioc_vif_req *req) 583 { 584 vifi_t vifi = req->vifi; 585 586 if (vifi >= numvifs) 587 return (EINVAL); 588 589 req->icount = viftable[vifi].v_pkt_in; 590 req->ocount = viftable[vifi].v_pkt_out; 591 req->ibytes = viftable[vifi].v_bytes_in; 592 req->obytes = viftable[vifi].v_bytes_out; 593 594 return (0); 595 } 596 597 /* 598 * Enable multicast routing 599 */ 600 static int 601 ip_mrouter_init(struct socket *so, struct mbuf *m) 602 { 603 int *v; 604 605 if (mrtdebug) 606 log(LOG_DEBUG, 607 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 608 so->so_type, so->so_proto->pr_protocol); 609 610 if (so->so_type != SOCK_RAW || 611 so->so_proto->pr_protocol != IPPROTO_IGMP) 612 return (EOPNOTSUPP); 613 614 if (m == NULL || m->m_len < sizeof(int)) 615 return (EINVAL); 616 617 v = mtod(m, int *); 618 if (*v != 1) 619 return (EINVAL); 620 621 if (ip_mrouter != NULL) 622 return (EADDRINUSE); 623 624 ip_mrouter = so; 625 626 mfchashtbl = 627 hashinit(MFCTBLSIZ, HASH_LIST, M_MRTABLE, M_WAITOK, &mfchash); 628 bzero((caddr_t)nexpire, sizeof(nexpire)); 629 630 pim_assert = 0; 631 632 callout_init(&expire_upcalls_ch); 633 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 634 expire_upcalls, NULL); 635 636 callout_init(&bw_upcalls_ch); 637 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 638 expire_bw_upcalls_send, NULL); 639 640 callout_init(&bw_meter_ch); 641 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 642 expire_bw_meter_process, NULL); 643 644 if (mrtdebug) 645 log(LOG_DEBUG, "ip_mrouter_init\n"); 646 647 return (0); 648 } 649 650 /* 651 * Disable multicast routing 652 */ 653 int 654 ip_mrouter_done(void) 655 { 656 vifi_t vifi; 657 struct vif *vifp; 658 int i; 659 int s; 660 661 s = splsoftnet(); 662 663 /* Clear out all the vifs currently in use. */ 664 for (vifi = 0; vifi < numvifs; vifi++) { 665 vifp = &viftable[vifi]; 666 if (!in_nullhost(vifp->v_lcl_addr)) 667 reset_vif(vifp); 668 } 669 670 numvifs = 0; 671 pim_assert = 0; 672 mrt_api_config = 0; 673 674 callout_stop(&expire_upcalls_ch); 675 callout_stop(&bw_upcalls_ch); 676 callout_stop(&bw_meter_ch); 677 678 /* 679 * Free all multicast forwarding cache entries. 680 */ 681 for (i = 0; i < MFCTBLSIZ; i++) { 682 struct mfc *rt, *nrt; 683 684 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 685 nrt = LIST_NEXT(rt, mfc_hash); 686 687 expire_mfc(rt); 688 } 689 } 690 691 bzero((caddr_t)nexpire, sizeof(nexpire)); 692 free(mfchashtbl, M_MRTABLE); 693 mfchashtbl = NULL; 694 695 bw_upcalls_n = 0; 696 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 697 698 /* Reset de-encapsulation cache. */ 699 700 ip_mrouter = NULL; 701 702 splx(s); 703 704 if (mrtdebug) 705 log(LOG_DEBUG, "ip_mrouter_done\n"); 706 707 return (0); 708 } 709 710 void 711 ip_mrouter_detach(struct ifnet *ifp) 712 { 713 int vifi, i; 714 struct vif *vifp; 715 struct mfc *rt; 716 struct rtdetq *rte; 717 718 /* XXX not sure about side effect to userland routing daemon */ 719 for (vifi = 0; vifi < numvifs; vifi++) { 720 vifp = &viftable[vifi]; 721 if (vifp->v_ifp == ifp) 722 reset_vif(vifp); 723 } 724 for (i = 0; i < MFCTBLSIZ; i++) { 725 if (nexpire[i] == 0) 726 continue; 727 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) { 728 for (rte = rt->mfc_stall; rte; rte = rte->next) { 729 if (rte->ifp == ifp) 730 rte->ifp = NULL; 731 } 732 } 733 } 734 } 735 736 static int 737 get_version(struct mbuf *m) 738 { 739 int *v = mtod(m, int *); 740 741 *v = 0x0305; /* XXX !!!! */ 742 m->m_len = sizeof(int); 743 return (0); 744 } 745 746 /* 747 * Set PIM assert processing global 748 */ 749 static int 750 set_assert(struct mbuf *m) 751 { 752 int *i; 753 754 if (m == NULL || m->m_len < sizeof(int)) 755 return (EINVAL); 756 757 i = mtod(m, int *); 758 pim_assert = !!*i; 759 return (0); 760 } 761 762 /* 763 * Get PIM assert processing global 764 */ 765 static int 766 get_assert(struct mbuf *m) 767 { 768 int *i = mtod(m, int *); 769 770 *i = pim_assert; 771 m->m_len = sizeof(int); 772 return (0); 773 } 774 775 /* 776 * Configure API capabilities 777 */ 778 static int 779 set_api_config(struct mbuf *m) 780 { 781 int i; 782 u_int32_t *apival; 783 784 if (m == NULL || m->m_len < sizeof(u_int32_t)) 785 return (EINVAL); 786 787 apival = mtod(m, u_int32_t *); 788 789 /* 790 * We can set the API capabilities only if it is the first operation 791 * after MRT_INIT. I.e.: 792 * - there are no vifs installed 793 * - pim_assert is not enabled 794 * - the MFC table is empty 795 */ 796 if (numvifs > 0) { 797 *apival = 0; 798 return (EPERM); 799 } 800 if (pim_assert) { 801 *apival = 0; 802 return (EPERM); 803 } 804 for (i = 0; i < MFCTBLSIZ; i++) { 805 if (LIST_FIRST(&mfchashtbl[i]) != NULL) { 806 *apival = 0; 807 return (EPERM); 808 } 809 } 810 811 mrt_api_config = *apival & mrt_api_support; 812 *apival = mrt_api_config; 813 814 return (0); 815 } 816 817 /* 818 * Get API capabilities 819 */ 820 static int 821 get_api_support(struct mbuf *m) 822 { 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 *apival = mrt_api_support; 831 832 return (0); 833 } 834 835 /* 836 * Get API configured capabilities 837 */ 838 static int 839 get_api_config(struct mbuf *m) 840 { 841 u_int32_t *apival; 842 843 if (m == NULL || m->m_len < sizeof(u_int32_t)) 844 return (EINVAL); 845 846 apival = mtod(m, u_int32_t *); 847 848 *apival = mrt_api_config; 849 850 return (0); 851 } 852 853 static struct sockaddr_in sin = { 854 .sin_len = sizeof(sin), 855 .sin_family = AF_INET 856 }; 857 858 /* 859 * Add a vif to the vif table 860 */ 861 static int 862 add_vif(struct mbuf *m) 863 { 864 struct vifctl *vifcp; 865 struct vif *vifp; 866 struct ifaddr *ifa; 867 struct ifnet *ifp; 868 struct ifreq ifr; 869 int error, s; 870 871 if (m == NULL || m->m_len < sizeof(struct vifctl)) 872 return (EINVAL); 873 874 vifcp = mtod(m, struct vifctl *); 875 if (vifcp->vifc_vifi >= MAXVIFS) 876 return (EINVAL); 877 if (in_nullhost(vifcp->vifc_lcl_addr)) 878 return (EADDRNOTAVAIL); 879 880 vifp = &viftable[vifcp->vifc_vifi]; 881 if (!in_nullhost(vifp->v_lcl_addr)) 882 return (EADDRINUSE); 883 884 /* Find the interface with an address in AF_INET family. */ 885 #ifdef PIM 886 if (vifcp->vifc_flags & VIFF_REGISTER) { 887 /* 888 * XXX: Because VIFF_REGISTER does not really need a valid 889 * local interface (e.g. it could be 127.0.0.2), we don't 890 * check its address. 891 */ 892 ifp = NULL; 893 } else 894 #endif 895 { 896 sin.sin_addr = vifcp->vifc_lcl_addr; 897 ifa = ifa_ifwithaddr(sintosa(&sin)); 898 if (ifa == NULL) 899 return (EADDRNOTAVAIL); 900 ifp = ifa->ifa_ifp; 901 } 902 903 if (vifcp->vifc_flags & VIFF_TUNNEL) { 904 if (vifcp->vifc_flags & VIFF_SRCRT) { 905 log(LOG_ERR, "source routed tunnels not supported\n"); 906 return (EOPNOTSUPP); 907 } 908 909 /* attach this vif to decapsulator dispatch table */ 910 /* 911 * XXX Use addresses in registration so that matching 912 * can be done with radix tree in decapsulator. But, 913 * we need to check inner header for multicast, so 914 * this requires both radix tree lookup and then a 915 * function to check, and this is not supported yet. 916 */ 917 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 918 vif_encapcheck, &vif_protosw, vifp); 919 if (!vifp->v_encap_cookie) 920 return (EINVAL); 921 922 /* Create a fake encapsulation interface. */ 923 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 924 bzero(ifp, sizeof(*ifp)); 925 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 926 "mdecap%d", vifcp->vifc_vifi); 927 928 /* Prepare cached route entry. */ 929 bzero(&vifp->v_route, sizeof(vifp->v_route)); 930 #ifdef PIM 931 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 932 ifp = &multicast_register_if; 933 if (mrtdebug) 934 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 935 (void *)ifp); 936 if (reg_vif_num == VIFI_INVALID) { 937 bzero(ifp, sizeof(*ifp)); 938 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 939 "register_vif"); 940 ifp->if_flags = IFF_LOOPBACK; 941 bzero(&vifp->v_route, sizeof(vifp->v_route)); 942 reg_vif_num = vifcp->vifc_vifi; 943 } 944 #endif 945 } else { 946 /* Make sure the interface supports multicast. */ 947 if ((ifp->if_flags & IFF_MULTICAST) == 0) 948 return (EOPNOTSUPP); 949 950 /* Enable promiscuous reception of all IP multicasts. */ 951 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 952 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 953 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 954 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (caddr_t)&ifr); 955 if (error) 956 return (error); 957 } 958 959 s = splsoftnet(); 960 961 /* Define parameters for the tbf structure. */ 962 vifp->tbf_q = NULL; 963 vifp->tbf_t = &vifp->tbf_q; 964 microtime(&vifp->tbf_last_pkt_t); 965 vifp->tbf_n_tok = 0; 966 vifp->tbf_q_len = 0; 967 vifp->tbf_max_q_len = MAXQSIZE; 968 969 vifp->v_flags = vifcp->vifc_flags; 970 vifp->v_threshold = vifcp->vifc_threshold; 971 /* scaling up here allows division by 1024 in critical code */ 972 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 973 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 974 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 975 vifp->v_ifp = ifp; 976 /* Initialize per vif pkt counters. */ 977 vifp->v_pkt_in = 0; 978 vifp->v_pkt_out = 0; 979 vifp->v_bytes_in = 0; 980 vifp->v_bytes_out = 0; 981 982 callout_init(&vifp->v_repq_ch); 983 984 #ifdef RSVP_ISI 985 vifp->v_rsvp_on = 0; 986 vifp->v_rsvpd = NULL; 987 #endif /* RSVP_ISI */ 988 989 splx(s); 990 991 /* Adjust numvifs up if the vifi is higher than numvifs. */ 992 if (numvifs <= vifcp->vifc_vifi) 993 numvifs = vifcp->vifc_vifi + 1; 994 995 if (mrtdebug) 996 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 997 vifcp->vifc_vifi, 998 ntohl(vifcp->vifc_lcl_addr.s_addr), 999 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 1000 ntohl(vifcp->vifc_rmt_addr.s_addr), 1001 vifcp->vifc_threshold, 1002 vifcp->vifc_rate_limit); 1003 1004 return (0); 1005 } 1006 1007 void 1008 reset_vif(struct vif *vifp) 1009 { 1010 struct mbuf *m, *n; 1011 struct ifnet *ifp; 1012 struct ifreq ifr; 1013 1014 callout_stop(&vifp->v_repq_ch); 1015 1016 /* detach this vif from decapsulator dispatch table */ 1017 encap_detach(vifp->v_encap_cookie); 1018 vifp->v_encap_cookie = NULL; 1019 1020 /* 1021 * Free packets queued at the interface 1022 */ 1023 for (m = vifp->tbf_q; m != NULL; m = n) { 1024 n = m->m_nextpkt; 1025 m_freem(m); 1026 } 1027 1028 if (vifp->v_flags & VIFF_TUNNEL) 1029 free(vifp->v_ifp, M_MRTABLE); 1030 else if (vifp->v_flags & VIFF_REGISTER) { 1031 #ifdef PIM 1032 reg_vif_num = VIFI_INVALID; 1033 #endif 1034 } else { 1035 satosin(&ifr.ifr_addr)->sin_len = sizeof(struct sockaddr_in); 1036 satosin(&ifr.ifr_addr)->sin_family = AF_INET; 1037 satosin(&ifr.ifr_addr)->sin_addr = zeroin_addr; 1038 ifp = vifp->v_ifp; 1039 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (caddr_t)&ifr); 1040 } 1041 bzero((caddr_t)vifp, sizeof(*vifp)); 1042 } 1043 1044 /* 1045 * Delete a vif from the vif table 1046 */ 1047 static int 1048 del_vif(struct mbuf *m) 1049 { 1050 vifi_t *vifip; 1051 struct vif *vifp; 1052 vifi_t vifi; 1053 int s; 1054 1055 if (m == NULL || m->m_len < sizeof(vifi_t)) 1056 return (EINVAL); 1057 1058 vifip = mtod(m, vifi_t *); 1059 if (*vifip >= numvifs) 1060 return (EINVAL); 1061 1062 vifp = &viftable[*vifip]; 1063 if (in_nullhost(vifp->v_lcl_addr)) 1064 return (EADDRNOTAVAIL); 1065 1066 s = splsoftnet(); 1067 1068 reset_vif(vifp); 1069 1070 /* Adjust numvifs down */ 1071 for (vifi = numvifs; vifi > 0; vifi--) 1072 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr)) 1073 break; 1074 numvifs = vifi; 1075 1076 splx(s); 1077 1078 if (mrtdebug) 1079 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 1080 1081 return (0); 1082 } 1083 1084 /* 1085 * update an mfc entry without resetting counters and S,G addresses. 1086 */ 1087 static void 1088 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1089 { 1090 int i; 1091 1092 rt->mfc_parent = mfccp->mfcc_parent; 1093 for (i = 0; i < numvifs; i++) { 1094 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1095 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1096 MRT_MFC_FLAGS_ALL; 1097 } 1098 /* set the RP address */ 1099 if (mrt_api_config & MRT_MFC_RP) 1100 rt->mfc_rp = mfccp->mfcc_rp; 1101 else 1102 rt->mfc_rp = zeroin_addr; 1103 } 1104 1105 /* 1106 * fully initialize an mfc entry from the parameter. 1107 */ 1108 static void 1109 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1110 { 1111 rt->mfc_origin = mfccp->mfcc_origin; 1112 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1113 1114 update_mfc_params(rt, mfccp); 1115 1116 /* initialize pkt counters per src-grp */ 1117 rt->mfc_pkt_cnt = 0; 1118 rt->mfc_byte_cnt = 0; 1119 rt->mfc_wrong_if = 0; 1120 timerclear(&rt->mfc_last_assert); 1121 } 1122 1123 static void 1124 expire_mfc(struct mfc *rt) 1125 { 1126 struct rtdetq *rte, *nrte; 1127 1128 free_bw_list(rt->mfc_bw_meter); 1129 1130 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 1131 nrte = rte->next; 1132 m_freem(rte->m); 1133 free(rte, M_MRTABLE); 1134 } 1135 1136 LIST_REMOVE(rt, mfc_hash); 1137 free(rt, M_MRTABLE); 1138 } 1139 1140 /* 1141 * Add an mfc entry 1142 */ 1143 static int 1144 add_mfc(struct mbuf *m) 1145 { 1146 struct mfcctl2 mfcctl2; 1147 struct mfcctl2 *mfccp; 1148 struct mfc *rt; 1149 u_int32_t hash = 0; 1150 struct rtdetq *rte, *nrte; 1151 u_short nstl; 1152 int s; 1153 int mfcctl_size = sizeof(struct mfcctl); 1154 1155 if (mrt_api_config & MRT_API_FLAGS_ALL) 1156 mfcctl_size = sizeof(struct mfcctl2); 1157 1158 if (m == NULL || m->m_len < mfcctl_size) 1159 return (EINVAL); 1160 1161 /* 1162 * select data size depending on API version. 1163 */ 1164 if (mrt_api_config & MRT_API_FLAGS_ALL) { 1165 struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *); 1166 bcopy(mp2, (caddr_t)&mfcctl2, sizeof(*mp2)); 1167 } else { 1168 struct mfcctl *mp = mtod(m, struct mfcctl *); 1169 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp)); 1170 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl), 1171 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1172 } 1173 mfccp = &mfcctl2; 1174 1175 s = splsoftnet(); 1176 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1177 1178 /* If an entry already exists, just update the fields */ 1179 if (rt) { 1180 if (mrtdebug & DEBUG_MFC) 1181 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n", 1182 ntohl(mfccp->mfcc_origin.s_addr), 1183 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1184 mfccp->mfcc_parent); 1185 1186 update_mfc_params(rt, mfccp); 1187 1188 splx(s); 1189 return (0); 1190 } 1191 1192 /* 1193 * Find the entry for which the upcall was made and update 1194 */ 1195 nstl = 0; 1196 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1197 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1198 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1199 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1200 rt->mfc_stall != NULL) { 1201 if (nstl++) 1202 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 1203 "multiple kernel entries", 1204 ntohl(mfccp->mfcc_origin.s_addr), 1205 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1206 mfccp->mfcc_parent, rt->mfc_stall); 1207 1208 if (mrtdebug & DEBUG_MFC) 1209 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n", 1210 ntohl(mfccp->mfcc_origin.s_addr), 1211 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1212 mfccp->mfcc_parent, rt->mfc_stall); 1213 1214 rte = rt->mfc_stall; 1215 init_mfc_params(rt, mfccp); 1216 rt->mfc_stall = NULL; 1217 1218 rt->mfc_expire = 0; /* Don't clean this guy up */ 1219 nexpire[hash]--; 1220 1221 /* free packets Qed at the end of this entry */ 1222 for (; rte != NULL; rte = nrte) { 1223 nrte = rte->next; 1224 if (rte->ifp) { 1225 #ifdef RSVP_ISI 1226 ip_mdq(rte->m, rte->ifp, rt, -1); 1227 #else 1228 ip_mdq(rte->m, rte->ifp, rt); 1229 #endif /* RSVP_ISI */ 1230 } 1231 m_freem(rte->m); 1232 #ifdef UPCALL_TIMING 1233 collate(&rte->t); 1234 #endif /* UPCALL_TIMING */ 1235 free(rte, M_MRTABLE); 1236 } 1237 } 1238 } 1239 1240 /* 1241 * It is possible that an entry is being inserted without an upcall 1242 */ 1243 if (nstl == 0) { 1244 /* 1245 * No mfc; make a new one 1246 */ 1247 if (mrtdebug & DEBUG_MFC) 1248 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n", 1249 ntohl(mfccp->mfcc_origin.s_addr), 1250 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1251 mfccp->mfcc_parent); 1252 1253 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1254 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1255 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1256 init_mfc_params(rt, mfccp); 1257 if (rt->mfc_expire) 1258 nexpire[hash]--; 1259 rt->mfc_expire = 0; 1260 break; /* XXX */ 1261 } 1262 } 1263 if (rt == NULL) { /* no upcall, so make a new entry */ 1264 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1265 M_NOWAIT); 1266 if (rt == NULL) { 1267 splx(s); 1268 return (ENOBUFS); 1269 } 1270 1271 init_mfc_params(rt, mfccp); 1272 rt->mfc_expire = 0; 1273 rt->mfc_stall = NULL; 1274 rt->mfc_bw_meter = NULL; 1275 1276 /* insert new entry at head of hash chain */ 1277 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1278 } 1279 } 1280 1281 splx(s); 1282 return (0); 1283 } 1284 1285 #ifdef UPCALL_TIMING 1286 /* 1287 * collect delay statistics on the upcalls 1288 */ 1289 static void 1290 collate(struct timeval *t) 1291 { 1292 u_int32_t d; 1293 struct timeval tp; 1294 u_int32_t delta; 1295 1296 microtime(&tp); 1297 1298 if (timercmp(t, &tp, <)) { 1299 TV_DELTA(tp, *t, delta); 1300 1301 d = delta >> 10; 1302 if (d > 50) 1303 d = 50; 1304 1305 ++upcall_data[d]; 1306 } 1307 } 1308 #endif /* UPCALL_TIMING */ 1309 1310 /* 1311 * Delete an mfc entry 1312 */ 1313 static int 1314 del_mfc(struct mbuf *m) 1315 { 1316 struct mfcctl2 mfcctl2; 1317 struct mfcctl2 *mfccp; 1318 struct mfc *rt; 1319 int s; 1320 int mfcctl_size = sizeof(struct mfcctl); 1321 struct mfcctl *mp = mtod(m, struct mfcctl *); 1322 1323 /* 1324 * XXX: for deleting MFC entries the information in entries 1325 * of size "struct mfcctl" is sufficient. 1326 */ 1327 1328 if (m == NULL || m->m_len < mfcctl_size) 1329 return (EINVAL); 1330 1331 bcopy(mp, (caddr_t)&mfcctl2, sizeof(*mp)); 1332 bzero((caddr_t)&mfcctl2 + sizeof(struct mfcctl), 1333 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1334 1335 mfccp = &mfcctl2; 1336 1337 if (mrtdebug & DEBUG_MFC) 1338 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 1339 ntohl(mfccp->mfcc_origin.s_addr), 1340 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 1341 1342 s = splsoftnet(); 1343 1344 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1345 if (rt == NULL) { 1346 splx(s); 1347 return (EADDRNOTAVAIL); 1348 } 1349 1350 /* 1351 * free the bw_meter entries 1352 */ 1353 free_bw_list(rt->mfc_bw_meter); 1354 rt->mfc_bw_meter = NULL; 1355 1356 LIST_REMOVE(rt, mfc_hash); 1357 free(rt, M_MRTABLE); 1358 1359 splx(s); 1360 return (0); 1361 } 1362 1363 static int 1364 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1365 { 1366 if (s) { 1367 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, 1368 (struct mbuf *)NULL) != 0) { 1369 sorwakeup(s); 1370 return (0); 1371 } 1372 } 1373 m_freem(mm); 1374 return (-1); 1375 } 1376 1377 /* 1378 * IP multicast forwarding function. This function assumes that the packet 1379 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1380 * pointed to by "ifp", and the packet is to be relayed to other networks 1381 * that have members of the packet's destination IP multicast group. 1382 * 1383 * The packet is returned unscathed to the caller, unless it is 1384 * erroneous, in which case a non-zero return value tells the caller to 1385 * discard it. 1386 */ 1387 1388 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1389 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1390 1391 int 1392 #ifdef RSVP_ISI 1393 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo) 1394 #else 1395 ip_mforward(struct mbuf *m, struct ifnet *ifp) 1396 #endif /* RSVP_ISI */ 1397 { 1398 struct ip *ip = mtod(m, struct ip *); 1399 struct mfc *rt; 1400 static int srctun = 0; 1401 struct mbuf *mm; 1402 int s; 1403 vifi_t vifi; 1404 1405 if (mrtdebug & DEBUG_FORWARD) 1406 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1407 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1408 1409 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1410 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1411 /* 1412 * Packet arrived via a physical interface or 1413 * an encapsulated tunnel or a register_vif. 1414 */ 1415 } else { 1416 /* 1417 * Packet arrived through a source-route tunnel. 1418 * Source-route tunnels are no longer supported. 1419 */ 1420 if ((srctun++ % 1000) == 0) 1421 log(LOG_ERR, 1422 "ip_mforward: received source-routed packet from %x\n", 1423 ntohl(ip->ip_src.s_addr)); 1424 1425 return (1); 1426 } 1427 1428 #ifdef RSVP_ISI 1429 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1430 if (ip->ip_ttl < MAXTTL) 1431 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1432 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1433 struct vif *vifp = viftable + vifi; 1434 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1435 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1436 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1437 vifp->v_ifp->if_xname); 1438 } 1439 return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi)); 1440 } 1441 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1442 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1443 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1444 } 1445 #endif /* RSVP_ISI */ 1446 1447 /* 1448 * Don't forward a packet with time-to-live of zero or one, 1449 * or a packet destined to a local-only group. 1450 */ 1451 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1452 return (0); 1453 1454 /* 1455 * Determine forwarding vifs from the forwarding cache table 1456 */ 1457 s = splsoftnet(); 1458 ++mrtstat.mrts_mfc_lookups; 1459 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1460 1461 /* Entry exists, so forward if necessary */ 1462 if (rt != NULL) { 1463 splx(s); 1464 #ifdef RSVP_ISI 1465 return (ip_mdq(m, ifp, rt, -1)); 1466 #else 1467 return (ip_mdq(m, ifp, rt)); 1468 #endif /* RSVP_ISI */ 1469 } else { 1470 /* 1471 * If we don't have a route for packet's origin, 1472 * Make a copy of the packet & send message to routing daemon 1473 */ 1474 1475 struct mbuf *mb0; 1476 struct rtdetq *rte; 1477 u_int32_t hash; 1478 int hlen = ip->ip_hl << 2; 1479 #ifdef UPCALL_TIMING 1480 struct timeval tp; 1481 1482 microtime(&tp); 1483 #endif /* UPCALL_TIMING */ 1484 1485 ++mrtstat.mrts_mfc_misses; 1486 1487 mrtstat.mrts_no_route++; 1488 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1489 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1490 ntohl(ip->ip_src.s_addr), 1491 ntohl(ip->ip_dst.s_addr)); 1492 1493 /* 1494 * Allocate mbufs early so that we don't do extra work if we are 1495 * just going to fail anyway. Make sure to pullup the header so 1496 * that other people can't step on it. 1497 */ 1498 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1499 M_NOWAIT); 1500 if (rte == NULL) { 1501 splx(s); 1502 return (ENOBUFS); 1503 } 1504 mb0 = m_copy(m, 0, M_COPYALL); 1505 M_PULLUP(mb0, hlen); 1506 if (mb0 == NULL) { 1507 free(rte, M_MRTABLE); 1508 splx(s); 1509 return (ENOBUFS); 1510 } 1511 1512 /* is there an upcall waiting for this flow? */ 1513 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1514 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1515 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1516 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1517 rt->mfc_stall != NULL) 1518 break; 1519 } 1520 1521 if (rt == NULL) { 1522 int i; 1523 struct igmpmsg *im; 1524 1525 /* 1526 * Locate the vifi for the incoming interface for 1527 * this packet. 1528 * If none found, drop packet. 1529 */ 1530 for (vifi = 0; vifi < numvifs && 1531 viftable[vifi].v_ifp != ifp; vifi++) 1532 ; 1533 if (vifi >= numvifs) /* vif not found, drop packet */ 1534 goto non_fatal; 1535 1536 /* no upcall, so make a new entry */ 1537 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1538 M_NOWAIT); 1539 if (rt == NULL) 1540 goto fail; 1541 1542 /* 1543 * Make a copy of the header to send to the user level 1544 * process 1545 */ 1546 mm = m_copy(m, 0, hlen); 1547 M_PULLUP(mm, hlen); 1548 if (mm == NULL) 1549 goto fail1; 1550 1551 /* 1552 * Send message to routing daemon to install 1553 * a route into the kernel table 1554 */ 1555 1556 im = mtod(mm, struct igmpmsg *); 1557 im->im_msgtype = IGMPMSG_NOCACHE; 1558 im->im_mbz = 0; 1559 im->im_vif = vifi; 1560 1561 mrtstat.mrts_upcalls++; 1562 1563 sin.sin_addr = ip->ip_src; 1564 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1565 log(LOG_WARNING, 1566 "ip_mforward: ip_mrouter socket queue full\n"); 1567 ++mrtstat.mrts_upq_sockfull; 1568 fail1: 1569 free(rt, M_MRTABLE); 1570 fail: 1571 free(rte, M_MRTABLE); 1572 m_freem(mb0); 1573 splx(s); 1574 return (ENOBUFS); 1575 } 1576 1577 /* insert new entry at head of hash chain */ 1578 rt->mfc_origin = ip->ip_src; 1579 rt->mfc_mcastgrp = ip->ip_dst; 1580 rt->mfc_pkt_cnt = 0; 1581 rt->mfc_byte_cnt = 0; 1582 rt->mfc_wrong_if = 0; 1583 rt->mfc_expire = UPCALL_EXPIRE; 1584 nexpire[hash]++; 1585 for (i = 0; i < numvifs; i++) { 1586 rt->mfc_ttls[i] = 0; 1587 rt->mfc_flags[i] = 0; 1588 } 1589 rt->mfc_parent = -1; 1590 1591 /* clear the RP address */ 1592 rt->mfc_rp = zeroin_addr; 1593 1594 rt->mfc_bw_meter = NULL; 1595 1596 /* link into table */ 1597 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1598 /* Add this entry to the end of the queue */ 1599 rt->mfc_stall = rte; 1600 } else { 1601 /* determine if q has overflowed */ 1602 struct rtdetq **p; 1603 int npkts = 0; 1604 1605 /* 1606 * XXX ouch! we need to append to the list, but we 1607 * only have a pointer to the front, so we have to 1608 * scan the entire list every time. 1609 */ 1610 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1611 if (++npkts > MAX_UPQ) { 1612 mrtstat.mrts_upq_ovflw++; 1613 non_fatal: 1614 free(rte, M_MRTABLE); 1615 m_freem(mb0); 1616 splx(s); 1617 return (0); 1618 } 1619 1620 /* Add this entry to the end of the queue */ 1621 *p = rte; 1622 } 1623 1624 rte->next = NULL; 1625 rte->m = mb0; 1626 rte->ifp = ifp; 1627 #ifdef UPCALL_TIMING 1628 rte->t = tp; 1629 #endif /* UPCALL_TIMING */ 1630 1631 splx(s); 1632 1633 return (0); 1634 } 1635 } 1636 1637 1638 /*ARGSUSED*/ 1639 static void 1640 expire_upcalls(void *v __unused) 1641 { 1642 int i; 1643 int s; 1644 1645 s = splsoftnet(); 1646 1647 for (i = 0; i < MFCTBLSIZ; i++) { 1648 struct mfc *rt, *nrt; 1649 1650 if (nexpire[i] == 0) 1651 continue; 1652 1653 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1654 nrt = LIST_NEXT(rt, mfc_hash); 1655 1656 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1657 continue; 1658 nexpire[i]--; 1659 1660 /* 1661 * free the bw_meter entries 1662 */ 1663 while (rt->mfc_bw_meter != NULL) { 1664 struct bw_meter *x = rt->mfc_bw_meter; 1665 1666 rt->mfc_bw_meter = x->bm_mfc_next; 1667 free(x, M_BWMETER); 1668 } 1669 1670 ++mrtstat.mrts_cache_cleanups; 1671 if (mrtdebug & DEBUG_EXPIRE) 1672 log(LOG_DEBUG, 1673 "expire_upcalls: expiring (%x %x)\n", 1674 ntohl(rt->mfc_origin.s_addr), 1675 ntohl(rt->mfc_mcastgrp.s_addr)); 1676 1677 expire_mfc(rt); 1678 } 1679 } 1680 1681 splx(s); 1682 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1683 expire_upcalls, NULL); 1684 } 1685 1686 /* 1687 * Packet forwarding routine once entry in the cache is made 1688 */ 1689 static int 1690 #ifdef RSVP_ISI 1691 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1692 #else 1693 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt) 1694 #endif /* RSVP_ISI */ 1695 { 1696 struct ip *ip = mtod(m, struct ip *); 1697 vifi_t vifi; 1698 struct vif *vifp; 1699 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2); 1700 1701 /* 1702 * Macro to send packet on vif. Since RSVP packets don't get counted on 1703 * input, they shouldn't get counted on output, so statistics keeping is 1704 * separate. 1705 */ 1706 #define MC_SEND(ip, vifp, m) do { \ 1707 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1708 encap_send((ip), (vifp), (m)); \ 1709 else \ 1710 phyint_send((ip), (vifp), (m)); \ 1711 } while (/*CONSTCOND*/ 0) 1712 1713 #ifdef RSVP_ISI 1714 /* 1715 * If xmt_vif is not -1, send on only the requested vif. 1716 * 1717 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1718 */ 1719 if (xmt_vif < numvifs) { 1720 #ifdef PIM 1721 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1722 pim_register_send(ip, viftable + xmt_vif, m, rt); 1723 else 1724 #endif 1725 MC_SEND(ip, viftable + xmt_vif, m); 1726 return (1); 1727 } 1728 #endif /* RSVP_ISI */ 1729 1730 /* 1731 * Don't forward if it didn't arrive from the parent vif for its origin. 1732 */ 1733 vifi = rt->mfc_parent; 1734 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1735 /* came in the wrong interface */ 1736 if (mrtdebug & DEBUG_FORWARD) 1737 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1738 ifp, vifi, 1739 vifi >= numvifs ? 0 : viftable[vifi].v_ifp); 1740 ++mrtstat.mrts_wrong_if; 1741 ++rt->mfc_wrong_if; 1742 /* 1743 * If we are doing PIM assert processing, send a message 1744 * to the routing daemon. 1745 * 1746 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1747 * can complete the SPT switch, regardless of the type 1748 * of the iif (broadcast media, GRE tunnel, etc). 1749 */ 1750 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1751 struct timeval now; 1752 u_int32_t delta; 1753 1754 #ifdef PIM 1755 if (ifp == &multicast_register_if) 1756 pimstat.pims_rcv_registers_wrongiif++; 1757 #endif 1758 1759 /* Get vifi for the incoming packet */ 1760 for (vifi = 0; 1761 vifi < numvifs && viftable[vifi].v_ifp != ifp; 1762 vifi++) 1763 ; 1764 if (vifi >= numvifs) { 1765 /* The iif is not found: ignore the packet. */ 1766 return (0); 1767 } 1768 1769 if (rt->mfc_flags[vifi] & 1770 MRT_MFC_FLAGS_DISABLE_WRONGVIF) { 1771 /* WRONGVIF disabled: ignore the packet */ 1772 return (0); 1773 } 1774 1775 microtime(&now); 1776 1777 TV_DELTA(rt->mfc_last_assert, now, delta); 1778 1779 if (delta > ASSERT_MSG_TIME) { 1780 struct igmpmsg *im; 1781 int hlen = ip->ip_hl << 2; 1782 struct mbuf *mm = m_copy(m, 0, hlen); 1783 1784 M_PULLUP(mm, hlen); 1785 if (mm == NULL) 1786 return (ENOBUFS); 1787 1788 rt->mfc_last_assert = now; 1789 1790 im = mtod(mm, struct igmpmsg *); 1791 im->im_msgtype = IGMPMSG_WRONGVIF; 1792 im->im_mbz = 0; 1793 im->im_vif = vifi; 1794 1795 mrtstat.mrts_upcalls++; 1796 1797 sin.sin_addr = im->im_src; 1798 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1799 log(LOG_WARNING, 1800 "ip_mforward: ip_mrouter socket queue full\n"); 1801 ++mrtstat.mrts_upq_sockfull; 1802 return (ENOBUFS); 1803 } 1804 } 1805 } 1806 return (0); 1807 } 1808 1809 /* If I sourced this packet, it counts as output, else it was input. */ 1810 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1811 viftable[vifi].v_pkt_out++; 1812 viftable[vifi].v_bytes_out += plen; 1813 } else { 1814 viftable[vifi].v_pkt_in++; 1815 viftable[vifi].v_bytes_in += plen; 1816 } 1817 rt->mfc_pkt_cnt++; 1818 rt->mfc_byte_cnt += plen; 1819 1820 /* 1821 * For each vif, decide if a copy of the packet should be forwarded. 1822 * Forward if: 1823 * - the ttl exceeds the vif's threshold 1824 * - there are group members downstream on interface 1825 */ 1826 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1827 if ((rt->mfc_ttls[vifi] > 0) && 1828 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1829 vifp->v_pkt_out++; 1830 vifp->v_bytes_out += plen; 1831 #ifdef PIM 1832 if (vifp->v_flags & VIFF_REGISTER) 1833 pim_register_send(ip, vifp, m, rt); 1834 else 1835 #endif 1836 MC_SEND(ip, vifp, m); 1837 } 1838 1839 /* 1840 * Perform upcall-related bw measuring. 1841 */ 1842 if (rt->mfc_bw_meter != NULL) { 1843 struct bw_meter *x; 1844 struct timeval now; 1845 1846 microtime(&now); 1847 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1848 bw_meter_receive_packet(x, plen, &now); 1849 } 1850 1851 return (0); 1852 } 1853 1854 #ifdef RSVP_ISI 1855 /* 1856 * check if a vif number is legal/ok. This is used by ip_output. 1857 */ 1858 int 1859 legal_vif_num(int vif) 1860 { 1861 if (vif >= 0 && vif < numvifs) 1862 return (1); 1863 else 1864 return (0); 1865 } 1866 #endif /* RSVP_ISI */ 1867 1868 static void 1869 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1870 { 1871 struct mbuf *mb_copy; 1872 int hlen = ip->ip_hl << 2; 1873 1874 /* 1875 * Make a new reference to the packet; make sure that 1876 * the IP header is actually copied, not just referenced, 1877 * so that ip_output() only scribbles on the copy. 1878 */ 1879 mb_copy = m_copy(m, 0, M_COPYALL); 1880 M_PULLUP(mb_copy, hlen); 1881 if (mb_copy == NULL) 1882 return; 1883 1884 if (vifp->v_rate_limit <= 0) 1885 tbf_send_packet(vifp, mb_copy); 1886 else 1887 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1888 ntohs(ip->ip_len)); 1889 } 1890 1891 static void 1892 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1893 { 1894 struct mbuf *mb_copy; 1895 struct ip *ip_copy; 1896 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1897 1898 /* Take care of delayed checksums */ 1899 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1900 in_delayed_cksum(m); 1901 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1902 } 1903 1904 /* 1905 * copy the old packet & pullup it's IP header into the 1906 * new mbuf so we can modify it. Try to fill the new 1907 * mbuf since if we don't the ethernet driver will. 1908 */ 1909 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1910 if (mb_copy == NULL) 1911 return; 1912 mb_copy->m_data += max_linkhdr; 1913 mb_copy->m_pkthdr.len = len; 1914 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1915 1916 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1917 m_freem(mb_copy); 1918 return; 1919 } 1920 i = MHLEN - max_linkhdr; 1921 if (i > len) 1922 i = len; 1923 mb_copy = m_pullup(mb_copy, i); 1924 if (mb_copy == NULL) 1925 return; 1926 1927 /* 1928 * fill in the encapsulating IP header. 1929 */ 1930 ip_copy = mtod(mb_copy, struct ip *); 1931 *ip_copy = multicast_encap_iphdr; 1932 ip_copy->ip_id = ip_newid(); 1933 ip_copy->ip_len = htons(len); 1934 ip_copy->ip_src = vifp->v_lcl_addr; 1935 ip_copy->ip_dst = vifp->v_rmt_addr; 1936 1937 /* 1938 * turn the encapsulated IP header back into a valid one. 1939 */ 1940 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1941 --ip->ip_ttl; 1942 ip->ip_sum = 0; 1943 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1944 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1945 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1946 1947 if (vifp->v_rate_limit <= 0) 1948 tbf_send_packet(vifp, mb_copy); 1949 else 1950 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len)); 1951 } 1952 1953 /* 1954 * De-encapsulate a packet and feed it back through ip input. 1955 */ 1956 static void 1957 vif_input(struct mbuf *m, ...) 1958 { 1959 int off, proto; 1960 va_list ap; 1961 struct vif *vifp; 1962 int s; 1963 struct ifqueue *ifq; 1964 1965 va_start(ap, m); 1966 off = va_arg(ap, int); 1967 proto = va_arg(ap, int); 1968 va_end(ap); 1969 1970 vifp = (struct vif *)encap_getarg(m); 1971 if (!vifp || proto != ENCAP_PROTO) { 1972 m_freem(m); 1973 mrtstat.mrts_bad_tunnel++; 1974 return; 1975 } 1976 1977 m_adj(m, off); 1978 m->m_pkthdr.rcvif = vifp->v_ifp; 1979 ifq = &ipintrq; 1980 s = splnet(); 1981 if (IF_QFULL(ifq)) { 1982 IF_DROP(ifq); 1983 m_freem(m); 1984 } else { 1985 IF_ENQUEUE(ifq, m); 1986 /* 1987 * normally we would need a "schednetisr(NETISR_IP)" 1988 * here but we were called by ip_input and it is going 1989 * to loop back & try to dequeue the packet we just 1990 * queued as soon as we return so we avoid the 1991 * unnecessary software interrrupt. 1992 */ 1993 } 1994 splx(s); 1995 } 1996 1997 /* 1998 * Check if the packet should be received on the vif denoted by arg. 1999 * (The encap selection code will call this once per vif since each is 2000 * registered separately.) 2001 */ 2002 static int 2003 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg) 2004 { 2005 struct vif *vifp; 2006 struct ip ip; 2007 2008 #ifdef DIAGNOSTIC 2009 if (!arg || proto != IPPROTO_IPV4) 2010 panic("unexpected arg in vif_encapcheck"); 2011 #endif 2012 2013 /* 2014 * Accept the packet only if the inner heaader is multicast 2015 * and the outer header matches a tunnel-mode vif. Order 2016 * checks in the hope that common non-matching packets will be 2017 * rejected quickly. Assume that unicast IPv4 traffic in a 2018 * parallel tunnel (e.g. gif(4)) is unlikely. 2019 */ 2020 2021 /* Obtain the outer IP header and the vif pointer. */ 2022 m_copydata((struct mbuf *)m, 0, sizeof(ip), (caddr_t)&ip); 2023 vifp = (struct vif *)arg; 2024 2025 /* 2026 * The outer source must match the vif's remote peer address. 2027 * For a multicast router with several tunnels, this is the 2028 * only check that will fail on packets in other tunnels, 2029 * assuming the local address is the same. 2030 */ 2031 if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 2032 return 0; 2033 2034 /* The outer destination must match the vif's local address. */ 2035 if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst)) 2036 return 0; 2037 2038 /* The vif must be of tunnel type. */ 2039 if ((vifp->v_flags & VIFF_TUNNEL) == 0) 2040 return 0; 2041 2042 /* Check that the inner destination is multicast. */ 2043 m_copydata((struct mbuf *)m, off, sizeof(ip), (caddr_t)&ip); 2044 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 2045 return 0; 2046 2047 /* 2048 * We have checked that both the outer src and dst addresses 2049 * match the vif, and that the inner destination is multicast 2050 * (224/5). By claiming more than 64, we intend to 2051 * preferentially take packets that also match a parallel 2052 * gif(4). 2053 */ 2054 return 32 + 32 + 5; 2055 } 2056 2057 /* 2058 * Token bucket filter module 2059 */ 2060 static void 2061 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len) 2062 { 2063 2064 if (len > MAX_BKT_SIZE) { 2065 /* drop if packet is too large */ 2066 mrtstat.mrts_pkt2large++; 2067 m_freem(m); 2068 return; 2069 } 2070 2071 tbf_update_tokens(vifp); 2072 2073 /* 2074 * If there are enough tokens, and the queue is empty, send this packet 2075 * out immediately. Otherwise, try to insert it on this vif's queue. 2076 */ 2077 if (vifp->tbf_q_len == 0) { 2078 if (len <= vifp->tbf_n_tok) { 2079 vifp->tbf_n_tok -= len; 2080 tbf_send_packet(vifp, m); 2081 } else { 2082 /* queue packet and timeout till later */ 2083 tbf_queue(vifp, m); 2084 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2085 tbf_reprocess_q, vifp); 2086 } 2087 } else { 2088 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 2089 !tbf_dq_sel(vifp, ip)) { 2090 /* queue full, and couldn't make room */ 2091 mrtstat.mrts_q_overflow++; 2092 m_freem(m); 2093 } else { 2094 /* queue length low enough, or made room */ 2095 tbf_queue(vifp, m); 2096 tbf_process_q(vifp); 2097 } 2098 } 2099 } 2100 2101 /* 2102 * adds a packet to the queue at the interface 2103 */ 2104 static void 2105 tbf_queue(struct vif *vifp, struct mbuf *m) 2106 { 2107 int s = splsoftnet(); 2108 2109 /* insert at tail */ 2110 *vifp->tbf_t = m; 2111 vifp->tbf_t = &m->m_nextpkt; 2112 vifp->tbf_q_len++; 2113 2114 splx(s); 2115 } 2116 2117 2118 /* 2119 * processes the queue at the interface 2120 */ 2121 static void 2122 tbf_process_q(struct vif *vifp) 2123 { 2124 struct mbuf *m; 2125 int len; 2126 int s = splsoftnet(); 2127 2128 /* 2129 * Loop through the queue at the interface and send as many packets 2130 * as possible. 2131 */ 2132 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) { 2133 len = ntohs(mtod(m, struct ip *)->ip_len); 2134 2135 /* determine if the packet can be sent */ 2136 if (len <= vifp->tbf_n_tok) { 2137 /* if so, 2138 * reduce no of tokens, dequeue the packet, 2139 * send the packet. 2140 */ 2141 if ((vifp->tbf_q = m->m_nextpkt) == NULL) 2142 vifp->tbf_t = &vifp->tbf_q; 2143 --vifp->tbf_q_len; 2144 2145 m->m_nextpkt = NULL; 2146 vifp->tbf_n_tok -= len; 2147 tbf_send_packet(vifp, m); 2148 } else 2149 break; 2150 } 2151 splx(s); 2152 } 2153 2154 static void 2155 tbf_reprocess_q(void *arg) 2156 { 2157 struct vif *vifp = arg; 2158 2159 if (ip_mrouter == NULL) 2160 return; 2161 2162 tbf_update_tokens(vifp); 2163 tbf_process_q(vifp); 2164 2165 if (vifp->tbf_q_len != 0) 2166 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2167 tbf_reprocess_q, vifp); 2168 } 2169 2170 /* function that will selectively discard a member of the queue 2171 * based on the precedence value and the priority 2172 */ 2173 static int 2174 tbf_dq_sel(struct vif *vifp, struct ip *ip) 2175 { 2176 u_int p; 2177 struct mbuf **mp, *m; 2178 int s = splsoftnet(); 2179 2180 p = priority(vifp, ip); 2181 2182 for (mp = &vifp->tbf_q, m = *mp; 2183 m != NULL; 2184 mp = &m->m_nextpkt, m = *mp) { 2185 if (p > priority(vifp, mtod(m, struct ip *))) { 2186 if ((*mp = m->m_nextpkt) == NULL) 2187 vifp->tbf_t = mp; 2188 --vifp->tbf_q_len; 2189 2190 m_freem(m); 2191 mrtstat.mrts_drop_sel++; 2192 splx(s); 2193 return (1); 2194 } 2195 } 2196 splx(s); 2197 return (0); 2198 } 2199 2200 static void 2201 tbf_send_packet(struct vif *vifp, struct mbuf *m) 2202 { 2203 int error; 2204 int s = splsoftnet(); 2205 2206 if (vifp->v_flags & VIFF_TUNNEL) { 2207 /* If tunnel options */ 2208 ip_output(m, (struct mbuf *)NULL, &vifp->v_route, 2209 IP_FORWARDING, (struct ip_moptions *)NULL, 2210 (struct socket *)NULL); 2211 } else { 2212 /* if physical interface option, extract the options and then send */ 2213 struct ip_moptions imo; 2214 2215 imo.imo_multicast_ifp = vifp->v_ifp; 2216 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 2217 imo.imo_multicast_loop = 1; 2218 #ifdef RSVP_ISI 2219 imo.imo_multicast_vif = -1; 2220 #endif 2221 2222 error = ip_output(m, (struct mbuf *)NULL, (struct route *)NULL, 2223 IP_FORWARDING|IP_MULTICASTOPTS, &imo, 2224 (struct socket *)NULL); 2225 2226 if (mrtdebug & DEBUG_XMIT) 2227 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 2228 (long)(vifp - viftable), error); 2229 } 2230 splx(s); 2231 } 2232 2233 /* determine the current time and then 2234 * the elapsed time (between the last time and time now) 2235 * in milliseconds & update the no. of tokens in the bucket 2236 */ 2237 static void 2238 tbf_update_tokens(struct vif *vifp) 2239 { 2240 struct timeval tp; 2241 u_int32_t tm; 2242 int s = splsoftnet(); 2243 2244 microtime(&tp); 2245 2246 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 2247 2248 /* 2249 * This formula is actually 2250 * "time in seconds" * "bytes/second". 2251 * 2252 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2253 * 2254 * The (1000/1024) was introduced in add_vif to optimize 2255 * this divide into a shift. 2256 */ 2257 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 2258 vifp->tbf_last_pkt_t = tp; 2259 2260 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 2261 vifp->tbf_n_tok = MAX_BKT_SIZE; 2262 2263 splx(s); 2264 } 2265 2266 static int 2267 priority(struct vif *vifp __unused, struct ip *ip) 2268 { 2269 int prio = 50; /* the lowest priority -- default case */ 2270 2271 /* temporary hack; may add general packet classifier some day */ 2272 2273 /* 2274 * The UDP port space is divided up into four priority ranges: 2275 * [0, 16384) : unclassified - lowest priority 2276 * [16384, 32768) : audio - highest priority 2277 * [32768, 49152) : whiteboard - medium priority 2278 * [49152, 65536) : video - low priority 2279 */ 2280 if (ip->ip_p == IPPROTO_UDP) { 2281 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2282 2283 switch (ntohs(udp->uh_dport) & 0xc000) { 2284 case 0x4000: 2285 prio = 70; 2286 break; 2287 case 0x8000: 2288 prio = 60; 2289 break; 2290 case 0xc000: 2291 prio = 55; 2292 break; 2293 } 2294 2295 if (tbfdebug > 1) 2296 log(LOG_DEBUG, "port %x prio %d\n", 2297 ntohs(udp->uh_dport), prio); 2298 } 2299 2300 return (prio); 2301 } 2302 2303 /* 2304 * End of token bucket filter modifications 2305 */ 2306 #ifdef RSVP_ISI 2307 int 2308 ip_rsvp_vif_init(struct socket *so, struct mbuf *m) 2309 { 2310 int vifi, s; 2311 2312 if (rsvpdebug) 2313 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2314 so->so_type, so->so_proto->pr_protocol); 2315 2316 if (so->so_type != SOCK_RAW || 2317 so->so_proto->pr_protocol != IPPROTO_RSVP) 2318 return (EOPNOTSUPP); 2319 2320 /* Check mbuf. */ 2321 if (m == NULL || m->m_len != sizeof(int)) { 2322 return (EINVAL); 2323 } 2324 vifi = *(mtod(m, int *)); 2325 2326 if (rsvpdebug) 2327 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", 2328 vifi, rsvp_on); 2329 2330 s = splsoftnet(); 2331 2332 /* Check vif. */ 2333 if (!legal_vif_num(vifi)) { 2334 splx(s); 2335 return (EADDRNOTAVAIL); 2336 } 2337 2338 /* Check if socket is available. */ 2339 if (viftable[vifi].v_rsvpd != NULL) { 2340 splx(s); 2341 return (EADDRINUSE); 2342 } 2343 2344 viftable[vifi].v_rsvpd = so; 2345 /* 2346 * This may seem silly, but we need to be sure we don't over-increment 2347 * the RSVP counter, in case something slips up. 2348 */ 2349 if (!viftable[vifi].v_rsvp_on) { 2350 viftable[vifi].v_rsvp_on = 1; 2351 rsvp_on++; 2352 } 2353 2354 splx(s); 2355 return (0); 2356 } 2357 2358 int 2359 ip_rsvp_vif_done(struct socket *so, struct mbuf *m) 2360 { 2361 int vifi, s; 2362 2363 if (rsvpdebug) 2364 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2365 so->so_type, so->so_proto->pr_protocol); 2366 2367 if (so->so_type != SOCK_RAW || 2368 so->so_proto->pr_protocol != IPPROTO_RSVP) 2369 return (EOPNOTSUPP); 2370 2371 /* Check mbuf. */ 2372 if (m == NULL || m->m_len != sizeof(int)) { 2373 return (EINVAL); 2374 } 2375 vifi = *(mtod(m, int *)); 2376 2377 s = splsoftnet(); 2378 2379 /* Check vif. */ 2380 if (!legal_vif_num(vifi)) { 2381 splx(s); 2382 return (EADDRNOTAVAIL); 2383 } 2384 2385 if (rsvpdebug) 2386 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 2387 viftable[vifi].v_rsvpd, so); 2388 2389 viftable[vifi].v_rsvpd = NULL; 2390 /* 2391 * This may seem silly, but we need to be sure we don't over-decrement 2392 * the RSVP counter, in case something slips up. 2393 */ 2394 if (viftable[vifi].v_rsvp_on) { 2395 viftable[vifi].v_rsvp_on = 0; 2396 rsvp_on--; 2397 } 2398 2399 splx(s); 2400 return (0); 2401 } 2402 2403 void 2404 ip_rsvp_force_done(struct socket *so) 2405 { 2406 int vifi, s; 2407 2408 /* Don't bother if it is not the right type of socket. */ 2409 if (so->so_type != SOCK_RAW || 2410 so->so_proto->pr_protocol != IPPROTO_RSVP) 2411 return; 2412 2413 s = splsoftnet(); 2414 2415 /* 2416 * The socket may be attached to more than one vif...this 2417 * is perfectly legal. 2418 */ 2419 for (vifi = 0; vifi < numvifs; vifi++) { 2420 if (viftable[vifi].v_rsvpd == so) { 2421 viftable[vifi].v_rsvpd = NULL; 2422 /* 2423 * This may seem silly, but we need to be sure we don't 2424 * over-decrement the RSVP counter, in case something 2425 * slips up. 2426 */ 2427 if (viftable[vifi].v_rsvp_on) { 2428 viftable[vifi].v_rsvp_on = 0; 2429 rsvp_on--; 2430 } 2431 } 2432 } 2433 2434 splx(s); 2435 return; 2436 } 2437 2438 void 2439 rsvp_input(struct mbuf *m, struct ifnet *ifp) 2440 { 2441 int vifi, s; 2442 struct ip *ip = mtod(m, struct ip *); 2443 static struct sockaddr_in rsvp_src = { sizeof(sin), AF_INET }; 2444 2445 if (rsvpdebug) 2446 printf("rsvp_input: rsvp_on %d\n", rsvp_on); 2447 2448 /* 2449 * Can still get packets with rsvp_on = 0 if there is a local member 2450 * of the group to which the RSVP packet is addressed. But in this 2451 * case we want to throw the packet away. 2452 */ 2453 if (!rsvp_on) { 2454 m_freem(m); 2455 return; 2456 } 2457 2458 /* 2459 * If the old-style non-vif-associated socket is set, then use 2460 * it and ignore the new ones. 2461 */ 2462 if (ip_rsvpd != NULL) { 2463 if (rsvpdebug) 2464 printf("rsvp_input: " 2465 "Sending packet up old-style socket\n"); 2466 rip_input(m); /*XXX*/ 2467 return; 2468 } 2469 2470 s = splsoftnet(); 2471 2472 if (rsvpdebug) 2473 printf("rsvp_input: check vifs\n"); 2474 2475 /* Find which vif the packet arrived on. */ 2476 for (vifi = 0; vifi < numvifs; vifi++) { 2477 if (viftable[vifi].v_ifp == ifp) 2478 break; 2479 } 2480 2481 if (vifi == numvifs) { 2482 /* Can't find vif packet arrived on. Drop packet. */ 2483 if (rsvpdebug) 2484 printf("rsvp_input: " 2485 "Can't find vif for packet...dropping it.\n"); 2486 m_freem(m); 2487 splx(s); 2488 return; 2489 } 2490 2491 if (rsvpdebug) 2492 printf("rsvp_input: check socket\n"); 2493 2494 if (viftable[vifi].v_rsvpd == NULL) { 2495 /* 2496 * drop packet, since there is no specific socket for this 2497 * interface 2498 */ 2499 if (rsvpdebug) 2500 printf("rsvp_input: No socket defined for vif %d\n", 2501 vifi); 2502 m_freem(m); 2503 splx(s); 2504 return; 2505 } 2506 2507 rsvp_src.sin_addr = ip->ip_src; 2508 2509 if (rsvpdebug && m) 2510 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2511 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2512 2513 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2514 if (rsvpdebug) 2515 printf("rsvp_input: Failed to append to socket\n"); 2516 else 2517 if (rsvpdebug) 2518 printf("rsvp_input: send packet up\n"); 2519 2520 splx(s); 2521 } 2522 #endif /* RSVP_ISI */ 2523 2524 /* 2525 * Code for bandwidth monitors 2526 */ 2527 2528 /* 2529 * Define common interface for timeval-related methods 2530 */ 2531 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp) 2532 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp)) 2533 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp)) 2534 2535 static uint32_t 2536 compute_bw_meter_flags(struct bw_upcall *req) 2537 { 2538 uint32_t flags = 0; 2539 2540 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2541 flags |= BW_METER_UNIT_PACKETS; 2542 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2543 flags |= BW_METER_UNIT_BYTES; 2544 if (req->bu_flags & BW_UPCALL_GEQ) 2545 flags |= BW_METER_GEQ; 2546 if (req->bu_flags & BW_UPCALL_LEQ) 2547 flags |= BW_METER_LEQ; 2548 2549 return flags; 2550 } 2551 2552 /* 2553 * Add a bw_meter entry 2554 */ 2555 static int 2556 add_bw_upcall(struct mbuf *m) 2557 { 2558 int s; 2559 struct mfc *mfc; 2560 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2561 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2562 struct timeval now; 2563 struct bw_meter *x; 2564 uint32_t flags; 2565 struct bw_upcall *req; 2566 2567 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2568 return EINVAL; 2569 2570 req = mtod(m, struct bw_upcall *); 2571 2572 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2573 return EOPNOTSUPP; 2574 2575 /* Test if the flags are valid */ 2576 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2577 return EINVAL; 2578 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2579 return EINVAL; 2580 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2581 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2582 return EINVAL; 2583 2584 /* Test if the threshold time interval is valid */ 2585 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2586 return EINVAL; 2587 2588 flags = compute_bw_meter_flags(req); 2589 2590 /* 2591 * Find if we have already same bw_meter entry 2592 */ 2593 s = splsoftnet(); 2594 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2595 if (mfc == NULL) { 2596 splx(s); 2597 return EADDRNOTAVAIL; 2598 } 2599 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2600 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2601 &req->bu_threshold.b_time, ==)) && 2602 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2603 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2604 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2605 splx(s); 2606 return 0; /* XXX Already installed */ 2607 } 2608 } 2609 2610 /* Allocate the new bw_meter entry */ 2611 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2612 if (x == NULL) { 2613 splx(s); 2614 return ENOBUFS; 2615 } 2616 2617 /* Set the new bw_meter entry */ 2618 x->bm_threshold.b_time = req->bu_threshold.b_time; 2619 microtime(&now); 2620 x->bm_start_time = now; 2621 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2622 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2623 x->bm_measured.b_packets = 0; 2624 x->bm_measured.b_bytes = 0; 2625 x->bm_flags = flags; 2626 x->bm_time_next = NULL; 2627 x->bm_time_hash = BW_METER_BUCKETS; 2628 2629 /* Add the new bw_meter entry to the front of entries for this MFC */ 2630 x->bm_mfc = mfc; 2631 x->bm_mfc_next = mfc->mfc_bw_meter; 2632 mfc->mfc_bw_meter = x; 2633 schedule_bw_meter(x, &now); 2634 splx(s); 2635 2636 return 0; 2637 } 2638 2639 static void 2640 free_bw_list(struct bw_meter *list) 2641 { 2642 while (list != NULL) { 2643 struct bw_meter *x = list; 2644 2645 list = list->bm_mfc_next; 2646 unschedule_bw_meter(x); 2647 free(x, M_BWMETER); 2648 } 2649 } 2650 2651 /* 2652 * Delete one or multiple bw_meter entries 2653 */ 2654 static int 2655 del_bw_upcall(struct mbuf *m) 2656 { 2657 int s; 2658 struct mfc *mfc; 2659 struct bw_meter *x; 2660 struct bw_upcall *req; 2661 2662 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2663 return EINVAL; 2664 2665 req = mtod(m, struct bw_upcall *); 2666 2667 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2668 return EOPNOTSUPP; 2669 2670 s = splsoftnet(); 2671 /* Find the corresponding MFC entry */ 2672 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2673 if (mfc == NULL) { 2674 splx(s); 2675 return EADDRNOTAVAIL; 2676 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2677 /* 2678 * Delete all bw_meter entries for this mfc 2679 */ 2680 struct bw_meter *list; 2681 2682 list = mfc->mfc_bw_meter; 2683 mfc->mfc_bw_meter = NULL; 2684 free_bw_list(list); 2685 splx(s); 2686 return 0; 2687 } else { /* Delete a single bw_meter entry */ 2688 struct bw_meter *prev; 2689 uint32_t flags = 0; 2690 2691 flags = compute_bw_meter_flags(req); 2692 2693 /* Find the bw_meter entry to delete */ 2694 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2695 prev = x, x = x->bm_mfc_next) { 2696 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2697 &req->bu_threshold.b_time, ==)) && 2698 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2699 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2700 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2701 break; 2702 } 2703 if (x != NULL) { /* Delete entry from the list for this MFC */ 2704 if (prev != NULL) 2705 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2706 else 2707 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2708 2709 unschedule_bw_meter(x); 2710 splx(s); 2711 /* Free the bw_meter entry */ 2712 free(x, M_BWMETER); 2713 return 0; 2714 } else { 2715 splx(s); 2716 return EINVAL; 2717 } 2718 } 2719 /* NOTREACHED */ 2720 } 2721 2722 /* 2723 * Perform bandwidth measurement processing that may result in an upcall 2724 */ 2725 static void 2726 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2727 { 2728 struct timeval delta; 2729 2730 delta = *nowp; 2731 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2732 2733 if (x->bm_flags & BW_METER_GEQ) { 2734 /* 2735 * Processing for ">=" type of bw_meter entry 2736 */ 2737 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2738 /* Reset the bw_meter entry */ 2739 x->bm_start_time = *nowp; 2740 x->bm_measured.b_packets = 0; 2741 x->bm_measured.b_bytes = 0; 2742 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2743 } 2744 2745 /* Record that a packet is received */ 2746 x->bm_measured.b_packets++; 2747 x->bm_measured.b_bytes += plen; 2748 2749 /* 2750 * Test if we should deliver an upcall 2751 */ 2752 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2753 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2754 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2755 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2756 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2757 /* Prepare an upcall for delivery */ 2758 bw_meter_prepare_upcall(x, nowp); 2759 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2760 } 2761 } 2762 } else if (x->bm_flags & BW_METER_LEQ) { 2763 /* 2764 * Processing for "<=" type of bw_meter entry 2765 */ 2766 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2767 /* 2768 * We are behind time with the multicast forwarding table 2769 * scanning for "<=" type of bw_meter entries, so test now 2770 * if we should deliver an upcall. 2771 */ 2772 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2773 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2774 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2775 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2776 /* Prepare an upcall for delivery */ 2777 bw_meter_prepare_upcall(x, nowp); 2778 } 2779 /* Reschedule the bw_meter entry */ 2780 unschedule_bw_meter(x); 2781 schedule_bw_meter(x, nowp); 2782 } 2783 2784 /* Record that a packet is received */ 2785 x->bm_measured.b_packets++; 2786 x->bm_measured.b_bytes += plen; 2787 2788 /* 2789 * Test if we should restart the measuring interval 2790 */ 2791 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2792 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2793 (x->bm_flags & BW_METER_UNIT_BYTES && 2794 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2795 /* Don't restart the measuring interval */ 2796 } else { 2797 /* Do restart the measuring interval */ 2798 /* 2799 * XXX: note that we don't unschedule and schedule, because this 2800 * might be too much overhead per packet. Instead, when we process 2801 * all entries for a given timer hash bin, we check whether it is 2802 * really a timeout. If not, we reschedule at that time. 2803 */ 2804 x->bm_start_time = *nowp; 2805 x->bm_measured.b_packets = 0; 2806 x->bm_measured.b_bytes = 0; 2807 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2808 } 2809 } 2810 } 2811 2812 /* 2813 * Prepare a bandwidth-related upcall 2814 */ 2815 static void 2816 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2817 { 2818 struct timeval delta; 2819 struct bw_upcall *u; 2820 2821 /* 2822 * Compute the measured time interval 2823 */ 2824 delta = *nowp; 2825 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2826 2827 /* 2828 * If there are too many pending upcalls, deliver them now 2829 */ 2830 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2831 bw_upcalls_send(); 2832 2833 /* 2834 * Set the bw_upcall entry 2835 */ 2836 u = &bw_upcalls[bw_upcalls_n++]; 2837 u->bu_src = x->bm_mfc->mfc_origin; 2838 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2839 u->bu_threshold.b_time = x->bm_threshold.b_time; 2840 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2841 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2842 u->bu_measured.b_time = delta; 2843 u->bu_measured.b_packets = x->bm_measured.b_packets; 2844 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2845 u->bu_flags = 0; 2846 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2847 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2848 if (x->bm_flags & BW_METER_UNIT_BYTES) 2849 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2850 if (x->bm_flags & BW_METER_GEQ) 2851 u->bu_flags |= BW_UPCALL_GEQ; 2852 if (x->bm_flags & BW_METER_LEQ) 2853 u->bu_flags |= BW_UPCALL_LEQ; 2854 } 2855 2856 /* 2857 * Send the pending bandwidth-related upcalls 2858 */ 2859 static void 2860 bw_upcalls_send(void) 2861 { 2862 struct mbuf *m; 2863 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2864 struct sockaddr_in k_igmpsrc = { 2865 .sin_len = sizeof(k_igmpsrc), 2866 .sin_family = AF_INET, 2867 }; 2868 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2869 0, /* unused2 */ 2870 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2871 0, /* im_mbz */ 2872 0, /* im_vif */ 2873 0, /* unused3 */ 2874 { 0 }, /* im_src */ 2875 { 0 } }; /* im_dst */ 2876 2877 if (bw_upcalls_n == 0) 2878 return; /* No pending upcalls */ 2879 2880 bw_upcalls_n = 0; 2881 2882 /* 2883 * Allocate a new mbuf, initialize it with the header and 2884 * the payload for the pending calls. 2885 */ 2886 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2887 if (m == NULL) { 2888 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2889 return; 2890 } 2891 2892 m->m_len = m->m_pkthdr.len = 0; 2893 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2894 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2895 2896 /* 2897 * Send the upcalls 2898 * XXX do we need to set the address in k_igmpsrc ? 2899 */ 2900 mrtstat.mrts_upcalls++; 2901 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2902 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2903 ++mrtstat.mrts_upq_sockfull; 2904 } 2905 } 2906 2907 /* 2908 * Compute the timeout hash value for the bw_meter entries 2909 */ 2910 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2911 do { \ 2912 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2913 \ 2914 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2915 (hash) = next_timeval.tv_sec; \ 2916 if (next_timeval.tv_usec) \ 2917 (hash)++; /* XXX: make sure we don't timeout early */ \ 2918 (hash) %= BW_METER_BUCKETS; \ 2919 } while (/*CONSTCOND*/ 0) 2920 2921 /* 2922 * Schedule a timer to process periodically bw_meter entry of type "<=" 2923 * by linking the entry in the proper hash bucket. 2924 */ 2925 static void 2926 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2927 { 2928 int time_hash; 2929 2930 if (!(x->bm_flags & BW_METER_LEQ)) 2931 return; /* XXX: we schedule timers only for "<=" entries */ 2932 2933 /* 2934 * Reset the bw_meter entry 2935 */ 2936 x->bm_start_time = *nowp; 2937 x->bm_measured.b_packets = 0; 2938 x->bm_measured.b_bytes = 0; 2939 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2940 2941 /* 2942 * Compute the timeout hash value and insert the entry 2943 */ 2944 BW_METER_TIMEHASH(x, time_hash); 2945 x->bm_time_next = bw_meter_timers[time_hash]; 2946 bw_meter_timers[time_hash] = x; 2947 x->bm_time_hash = time_hash; 2948 } 2949 2950 /* 2951 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2952 * by removing the entry from the proper hash bucket. 2953 */ 2954 static void 2955 unschedule_bw_meter(struct bw_meter *x) 2956 { 2957 int time_hash; 2958 struct bw_meter *prev, *tmp; 2959 2960 if (!(x->bm_flags & BW_METER_LEQ)) 2961 return; /* XXX: we schedule timers only for "<=" entries */ 2962 2963 /* 2964 * Compute the timeout hash value and delete the entry 2965 */ 2966 time_hash = x->bm_time_hash; 2967 if (time_hash >= BW_METER_BUCKETS) 2968 return; /* Entry was not scheduled */ 2969 2970 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2971 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2972 if (tmp == x) 2973 break; 2974 2975 if (tmp == NULL) 2976 panic("unschedule_bw_meter: bw_meter entry not found"); 2977 2978 if (prev != NULL) 2979 prev->bm_time_next = x->bm_time_next; 2980 else 2981 bw_meter_timers[time_hash] = x->bm_time_next; 2982 2983 x->bm_time_next = NULL; 2984 x->bm_time_hash = BW_METER_BUCKETS; 2985 } 2986 2987 /* 2988 * Process all "<=" type of bw_meter that should be processed now, 2989 * and for each entry prepare an upcall if necessary. Each processed 2990 * entry is rescheduled again for the (periodic) processing. 2991 * 2992 * This is run periodically (once per second normally). On each round, 2993 * all the potentially matching entries are in the hash slot that we are 2994 * looking at. 2995 */ 2996 static void 2997 bw_meter_process(void) 2998 { 2999 int s; 3000 static uint32_t last_tv_sec; /* last time we processed this */ 3001 3002 uint32_t loops; 3003 int i; 3004 struct timeval now, process_endtime; 3005 3006 microtime(&now); 3007 if (last_tv_sec == now.tv_sec) 3008 return; /* nothing to do */ 3009 3010 loops = now.tv_sec - last_tv_sec; 3011 last_tv_sec = now.tv_sec; 3012 if (loops > BW_METER_BUCKETS) 3013 loops = BW_METER_BUCKETS; 3014 3015 s = splsoftnet(); 3016 /* 3017 * Process all bins of bw_meter entries from the one after the last 3018 * processed to the current one. On entry, i points to the last bucket 3019 * visited, so we need to increment i at the beginning of the loop. 3020 */ 3021 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 3022 struct bw_meter *x, *tmp_list; 3023 3024 if (++i >= BW_METER_BUCKETS) 3025 i = 0; 3026 3027 /* Disconnect the list of bw_meter entries from the bin */ 3028 tmp_list = bw_meter_timers[i]; 3029 bw_meter_timers[i] = NULL; 3030 3031 /* Process the list of bw_meter entries */ 3032 while (tmp_list != NULL) { 3033 x = tmp_list; 3034 tmp_list = tmp_list->bm_time_next; 3035 3036 /* Test if the time interval is over */ 3037 process_endtime = x->bm_start_time; 3038 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 3039 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 3040 /* Not yet: reschedule, but don't reset */ 3041 int time_hash; 3042 3043 BW_METER_TIMEHASH(x, time_hash); 3044 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 3045 /* 3046 * XXX: somehow the bin processing is a bit ahead of time. 3047 * Put the entry in the next bin. 3048 */ 3049 if (++time_hash >= BW_METER_BUCKETS) 3050 time_hash = 0; 3051 } 3052 x->bm_time_next = bw_meter_timers[time_hash]; 3053 bw_meter_timers[time_hash] = x; 3054 x->bm_time_hash = time_hash; 3055 3056 continue; 3057 } 3058 3059 /* 3060 * Test if we should deliver an upcall 3061 */ 3062 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 3063 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 3064 ((x->bm_flags & BW_METER_UNIT_BYTES) && 3065 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 3066 /* Prepare an upcall for delivery */ 3067 bw_meter_prepare_upcall(x, &now); 3068 } 3069 3070 /* 3071 * Reschedule for next processing 3072 */ 3073 schedule_bw_meter(x, &now); 3074 } 3075 } 3076 3077 /* Send all upcalls that are pending delivery */ 3078 bw_upcalls_send(); 3079 3080 splx(s); 3081 } 3082 3083 /* 3084 * A periodic function for sending all upcalls that are pending delivery 3085 */ 3086 static void 3087 expire_bw_upcalls_send(void *unused __unused) 3088 { 3089 int s; 3090 3091 s = splsoftnet(); 3092 bw_upcalls_send(); 3093 splx(s); 3094 3095 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 3096 expire_bw_upcalls_send, NULL); 3097 } 3098 3099 /* 3100 * A periodic function for periodic scanning of the multicast forwarding 3101 * table for processing all "<=" bw_meter entries. 3102 */ 3103 static void 3104 expire_bw_meter_process(void *unused __unused) 3105 { 3106 if (mrt_api_config & MRT_MFC_BW_UPCALL) 3107 bw_meter_process(); 3108 3109 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 3110 expire_bw_meter_process, NULL); 3111 } 3112 3113 /* 3114 * End of bandwidth monitoring code 3115 */ 3116 3117 #ifdef PIM 3118 /* 3119 * Send the packet up to the user daemon, or eventually do kernel encapsulation 3120 */ 3121 static int 3122 pim_register_send(struct ip *ip, struct vif *vifp, 3123 struct mbuf *m, struct mfc *rt) 3124 { 3125 struct mbuf *mb_copy, *mm; 3126 3127 if (mrtdebug & DEBUG_PIM) 3128 log(LOG_DEBUG, "pim_register_send: "); 3129 3130 mb_copy = pim_register_prepare(ip, m); 3131 if (mb_copy == NULL) 3132 return ENOBUFS; 3133 3134 /* 3135 * Send all the fragments. Note that the mbuf for each fragment 3136 * is freed by the sending machinery. 3137 */ 3138 for (mm = mb_copy; mm; mm = mb_copy) { 3139 mb_copy = mm->m_nextpkt; 3140 mm->m_nextpkt = NULL; 3141 mm = m_pullup(mm, sizeof(struct ip)); 3142 if (mm != NULL) { 3143 ip = mtod(mm, struct ip *); 3144 if ((mrt_api_config & MRT_MFC_RP) && 3145 !in_nullhost(rt->mfc_rp)) { 3146 pim_register_send_rp(ip, vifp, mm, rt); 3147 } else { 3148 pim_register_send_upcall(ip, vifp, mm, rt); 3149 } 3150 } 3151 } 3152 3153 return 0; 3154 } 3155 3156 /* 3157 * Return a copy of the data packet that is ready for PIM Register 3158 * encapsulation. 3159 * XXX: Note that in the returned copy the IP header is a valid one. 3160 */ 3161 static struct mbuf * 3162 pim_register_prepare(struct ip *ip, struct mbuf *m) 3163 { 3164 struct mbuf *mb_copy = NULL; 3165 int mtu; 3166 3167 /* Take care of delayed checksums */ 3168 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 3169 in_delayed_cksum(m); 3170 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 3171 } 3172 3173 /* 3174 * Copy the old packet & pullup its IP header into the 3175 * new mbuf so we can modify it. 3176 */ 3177 mb_copy = m_copy(m, 0, M_COPYALL); 3178 if (mb_copy == NULL) 3179 return NULL; 3180 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 3181 if (mb_copy == NULL) 3182 return NULL; 3183 3184 /* take care of the TTL */ 3185 ip = mtod(mb_copy, struct ip *); 3186 --ip->ip_ttl; 3187 3188 /* Compute the MTU after the PIM Register encapsulation */ 3189 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 3190 3191 if (ntohs(ip->ip_len) <= mtu) { 3192 /* Turn the IP header into a valid one */ 3193 ip->ip_sum = 0; 3194 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 3195 } else { 3196 /* Fragment the packet */ 3197 if (ip_fragment(mb_copy, NULL, mtu) != 0) { 3198 /* XXX: mb_copy was freed by ip_fragment() */ 3199 return NULL; 3200 } 3201 } 3202 return mb_copy; 3203 } 3204 3205 /* 3206 * Send an upcall with the data packet to the user-level process. 3207 */ 3208 static int 3209 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 3210 struct mbuf *mb_copy, struct mfc *rt __unused) 3211 { 3212 struct mbuf *mb_first; 3213 int len = ntohs(ip->ip_len); 3214 struct igmpmsg *im; 3215 struct sockaddr_in k_igmpsrc = { 3216 .sin_len = sizeof(k_igmpsrc), 3217 .sin_family = AF_INET, 3218 }; 3219 3220 /* 3221 * Add a new mbuf with an upcall header 3222 */ 3223 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3224 if (mb_first == NULL) { 3225 m_freem(mb_copy); 3226 return ENOBUFS; 3227 } 3228 mb_first->m_data += max_linkhdr; 3229 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3230 mb_first->m_len = sizeof(struct igmpmsg); 3231 mb_first->m_next = mb_copy; 3232 3233 /* Send message to routing daemon */ 3234 im = mtod(mb_first, struct igmpmsg *); 3235 im->im_msgtype = IGMPMSG_WHOLEPKT; 3236 im->im_mbz = 0; 3237 im->im_vif = vifp - viftable; 3238 im->im_src = ip->ip_src; 3239 im->im_dst = ip->ip_dst; 3240 3241 k_igmpsrc.sin_addr = ip->ip_src; 3242 3243 mrtstat.mrts_upcalls++; 3244 3245 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3246 if (mrtdebug & DEBUG_PIM) 3247 log(LOG_WARNING, 3248 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3249 ++mrtstat.mrts_upq_sockfull; 3250 return ENOBUFS; 3251 } 3252 3253 /* Keep statistics */ 3254 pimstat.pims_snd_registers_msgs++; 3255 pimstat.pims_snd_registers_bytes += len; 3256 3257 return 0; 3258 } 3259 3260 /* 3261 * Encapsulate the data packet in PIM Register message and send it to the RP. 3262 */ 3263 static int 3264 pim_register_send_rp(struct ip *ip, struct vif *vifp, 3265 struct mbuf *mb_copy, struct mfc *rt) 3266 { 3267 struct mbuf *mb_first; 3268 struct ip *ip_outer; 3269 struct pim_encap_pimhdr *pimhdr; 3270 int len = ntohs(ip->ip_len); 3271 vifi_t vifi = rt->mfc_parent; 3272 3273 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) { 3274 m_freem(mb_copy); 3275 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3276 } 3277 3278 /* 3279 * Add a new mbuf with the encapsulating header 3280 */ 3281 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3282 if (mb_first == NULL) { 3283 m_freem(mb_copy); 3284 return ENOBUFS; 3285 } 3286 mb_first->m_data += max_linkhdr; 3287 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3288 mb_first->m_next = mb_copy; 3289 3290 mb_first->m_pkthdr.len = len + mb_first->m_len; 3291 3292 /* 3293 * Fill in the encapsulating IP and PIM header 3294 */ 3295 ip_outer = mtod(mb_first, struct ip *); 3296 *ip_outer = pim_encap_iphdr; 3297 ip_outer->ip_id = ip_newid(); 3298 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + 3299 sizeof(pim_encap_pimhdr)); 3300 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3301 ip_outer->ip_dst = rt->mfc_rp; 3302 /* 3303 * Copy the inner header TOS to the outer header, and take care of the 3304 * IP_DF bit. 3305 */ 3306 ip_outer->ip_tos = ip->ip_tos; 3307 if (ntohs(ip->ip_off) & IP_DF) 3308 ip_outer->ip_off |= IP_DF; 3309 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 3310 + sizeof(pim_encap_iphdr)); 3311 *pimhdr = pim_encap_pimhdr; 3312 /* If the iif crosses a border, set the Border-bit */ 3313 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3314 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3315 3316 mb_first->m_data += sizeof(pim_encap_iphdr); 3317 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3318 mb_first->m_data -= sizeof(pim_encap_iphdr); 3319 3320 if (vifp->v_rate_limit == 0) 3321 tbf_send_packet(vifp, mb_first); 3322 else 3323 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len)); 3324 3325 /* Keep statistics */ 3326 pimstat.pims_snd_registers_msgs++; 3327 pimstat.pims_snd_registers_bytes += len; 3328 3329 return 0; 3330 } 3331 3332 /* 3333 * PIM-SMv2 and PIM-DM messages processing. 3334 * Receives and verifies the PIM control messages, and passes them 3335 * up to the listening socket, using rip_input(). 3336 * The only message with special processing is the PIM_REGISTER message 3337 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3338 * is passed to if_simloop(). 3339 */ 3340 void 3341 pim_input(struct mbuf *m, ...) 3342 { 3343 struct ip *ip = mtod(m, struct ip *); 3344 struct pim *pim; 3345 int minlen; 3346 int datalen; 3347 int ip_tos; 3348 int proto; 3349 int iphlen; 3350 va_list ap; 3351 3352 va_start(ap, m); 3353 iphlen = va_arg(ap, int); 3354 proto = va_arg(ap, int); 3355 va_end(ap); 3356 3357 datalen = ntohs(ip->ip_len) - iphlen; 3358 3359 /* Keep statistics */ 3360 pimstat.pims_rcv_total_msgs++; 3361 pimstat.pims_rcv_total_bytes += datalen; 3362 3363 /* 3364 * Validate lengths 3365 */ 3366 if (datalen < PIM_MINLEN) { 3367 pimstat.pims_rcv_tooshort++; 3368 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3369 datalen, (u_long)ip->ip_src.s_addr); 3370 m_freem(m); 3371 return; 3372 } 3373 3374 /* 3375 * If the packet is at least as big as a REGISTER, go agead 3376 * and grab the PIM REGISTER header size, to avoid another 3377 * possible m_pullup() later. 3378 * 3379 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3380 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3381 */ 3382 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3383 /* 3384 * Get the IP and PIM headers in contiguous memory, and 3385 * possibly the PIM REGISTER header. 3386 */ 3387 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3388 (m = m_pullup(m, minlen)) == NULL) { 3389 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3390 return; 3391 } 3392 /* m_pullup() may have given us a new mbuf so reset ip. */ 3393 ip = mtod(m, struct ip *); 3394 ip_tos = ip->ip_tos; 3395 3396 /* adjust mbuf to point to the PIM header */ 3397 m->m_data += iphlen; 3398 m->m_len -= iphlen; 3399 pim = mtod(m, struct pim *); 3400 3401 /* 3402 * Validate checksum. If PIM REGISTER, exclude the data packet. 3403 * 3404 * XXX: some older PIMv2 implementations don't make this distinction, 3405 * so for compatibility reason perform the checksum over part of the 3406 * message, and if error, then over the whole message. 3407 */ 3408 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3409 /* do nothing, checksum okay */ 3410 } else if (in_cksum(m, datalen)) { 3411 pimstat.pims_rcv_badsum++; 3412 if (mrtdebug & DEBUG_PIM) 3413 log(LOG_DEBUG, "pim_input: invalid checksum"); 3414 m_freem(m); 3415 return; 3416 } 3417 3418 /* PIM version check */ 3419 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3420 pimstat.pims_rcv_badversion++; 3421 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3422 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3423 m_freem(m); 3424 return; 3425 } 3426 3427 /* restore mbuf back to the outer IP */ 3428 m->m_data -= iphlen; 3429 m->m_len += iphlen; 3430 3431 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3432 /* 3433 * Since this is a REGISTER, we'll make a copy of the register 3434 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3435 * routing daemon. 3436 */ 3437 int s; 3438 struct sockaddr_in dst = { 3439 .sin_len = sizeof(dst), 3440 .sin_family = AF_INET, 3441 }; 3442 struct mbuf *mcp; 3443 struct ip *encap_ip; 3444 u_int32_t *reghdr; 3445 struct ifnet *vifp; 3446 3447 s = splsoftnet(); 3448 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3449 splx(s); 3450 if (mrtdebug & DEBUG_PIM) 3451 log(LOG_DEBUG, 3452 "pim_input: register vif not set: %d\n", reg_vif_num); 3453 m_freem(m); 3454 return; 3455 } 3456 /* XXX need refcnt? */ 3457 vifp = viftable[reg_vif_num].v_ifp; 3458 splx(s); 3459 3460 /* 3461 * Validate length 3462 */ 3463 if (datalen < PIM_REG_MINLEN) { 3464 pimstat.pims_rcv_tooshort++; 3465 pimstat.pims_rcv_badregisters++; 3466 log(LOG_ERR, 3467 "pim_input: register packet size too small %d from %lx\n", 3468 datalen, (u_long)ip->ip_src.s_addr); 3469 m_freem(m); 3470 return; 3471 } 3472 3473 reghdr = (u_int32_t *)(pim + 1); 3474 encap_ip = (struct ip *)(reghdr + 1); 3475 3476 if (mrtdebug & DEBUG_PIM) { 3477 log(LOG_DEBUG, 3478 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3479 (u_long)ntohl(encap_ip->ip_src.s_addr), 3480 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3481 ntohs(encap_ip->ip_len)); 3482 } 3483 3484 /* verify the version number of the inner packet */ 3485 if (encap_ip->ip_v != IPVERSION) { 3486 pimstat.pims_rcv_badregisters++; 3487 if (mrtdebug & DEBUG_PIM) { 3488 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3489 "of the inner packet\n", encap_ip->ip_v); 3490 } 3491 m_freem(m); 3492 return; 3493 } 3494 3495 /* verify the inner packet is destined to a mcast group */ 3496 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) { 3497 pimstat.pims_rcv_badregisters++; 3498 if (mrtdebug & DEBUG_PIM) 3499 log(LOG_DEBUG, 3500 "pim_input: inner packet of register is not " 3501 "multicast %lx\n", 3502 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3503 m_freem(m); 3504 return; 3505 } 3506 3507 /* If a NULL_REGISTER, pass it to the daemon */ 3508 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3509 goto pim_input_to_daemon; 3510 3511 /* 3512 * Copy the TOS from the outer IP header to the inner IP header. 3513 */ 3514 if (encap_ip->ip_tos != ip_tos) { 3515 /* Outer TOS -> inner TOS */ 3516 encap_ip->ip_tos = ip_tos; 3517 /* Recompute the inner header checksum. Sigh... */ 3518 3519 /* adjust mbuf to point to the inner IP header */ 3520 m->m_data += (iphlen + PIM_MINLEN); 3521 m->m_len -= (iphlen + PIM_MINLEN); 3522 3523 encap_ip->ip_sum = 0; 3524 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3525 3526 /* restore mbuf to point back to the outer IP header */ 3527 m->m_data -= (iphlen + PIM_MINLEN); 3528 m->m_len += (iphlen + PIM_MINLEN); 3529 } 3530 3531 /* 3532 * Decapsulate the inner IP packet and loopback to forward it 3533 * as a normal multicast packet. Also, make a copy of the 3534 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3535 * to pass to the daemon later, so it can take the appropriate 3536 * actions (e.g., send back PIM_REGISTER_STOP). 3537 * XXX: here m->m_data points to the outer IP header. 3538 */ 3539 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 3540 if (mcp == NULL) { 3541 log(LOG_ERR, 3542 "pim_input: pim register: could not copy register head\n"); 3543 m_freem(m); 3544 return; 3545 } 3546 3547 /* Keep statistics */ 3548 /* XXX: registers_bytes include only the encap. mcast pkt */ 3549 pimstat.pims_rcv_registers_msgs++; 3550 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3551 3552 /* 3553 * forward the inner ip packet; point m_data at the inner ip. 3554 */ 3555 m_adj(m, iphlen + PIM_MINLEN); 3556 3557 if (mrtdebug & DEBUG_PIM) { 3558 log(LOG_DEBUG, 3559 "pim_input: forwarding decapsulated register: " 3560 "src %lx, dst %lx, vif %d\n", 3561 (u_long)ntohl(encap_ip->ip_src.s_addr), 3562 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3563 reg_vif_num); 3564 } 3565 /* NB: vifp was collected above; can it change on us? */ 3566 looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL); 3567 3568 /* prepare the register head to send to the mrouting daemon */ 3569 m = mcp; 3570 } 3571 3572 pim_input_to_daemon: 3573 /* 3574 * Pass the PIM message up to the daemon; if it is a Register message, 3575 * pass the 'head' only up to the daemon. This includes the 3576 * outer IP header, PIM header, PIM-Register header and the 3577 * inner IP header. 3578 * XXX: the outer IP header pkt size of a Register is not adjust to 3579 * reflect the fact that the inner multicast data is truncated. 3580 */ 3581 rip_input(m, iphlen, proto); 3582 3583 return; 3584 } 3585 #endif /* PIM */ 3586