ip_mroute.c revision 1.112
1 /* $NetBSD: ip_mroute.c,v 1.112 2008/05/05 17:11:17 ad 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 1.112 2008/05/05 17:11:17 ad 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, void *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 = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash); 627 bzero((void *)nexpire, sizeof(nexpire)); 628 629 pim_assert = 0; 630 631 callout_init(&expire_upcalls_ch, 0); 632 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 633 expire_upcalls, NULL); 634 635 callout_init(&bw_upcalls_ch, 0); 636 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 637 expire_bw_upcalls_send, NULL); 638 639 callout_init(&bw_meter_ch, 0); 640 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 641 expire_bw_meter_process, NULL); 642 643 if (mrtdebug) 644 log(LOG_DEBUG, "ip_mrouter_init\n"); 645 646 return (0); 647 } 648 649 /* 650 * Disable multicast routing 651 */ 652 int 653 ip_mrouter_done(void) 654 { 655 vifi_t vifi; 656 struct vif *vifp; 657 int i; 658 int s; 659 660 s = splsoftnet(); 661 662 /* Clear out all the vifs currently in use. */ 663 for (vifi = 0; vifi < numvifs; vifi++) { 664 vifp = &viftable[vifi]; 665 if (!in_nullhost(vifp->v_lcl_addr)) 666 reset_vif(vifp); 667 } 668 669 numvifs = 0; 670 pim_assert = 0; 671 mrt_api_config = 0; 672 673 callout_stop(&expire_upcalls_ch); 674 callout_stop(&bw_upcalls_ch); 675 callout_stop(&bw_meter_ch); 676 677 /* 678 * Free all multicast forwarding cache entries. 679 */ 680 for (i = 0; i < MFCTBLSIZ; i++) { 681 struct mfc *rt, *nrt; 682 683 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 684 nrt = LIST_NEXT(rt, mfc_hash); 685 686 expire_mfc(rt); 687 } 688 } 689 690 bzero((void *)nexpire, sizeof(nexpire)); 691 free(mfchashtbl, M_MRTABLE); 692 mfchashtbl = NULL; 693 694 bw_upcalls_n = 0; 695 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 696 697 /* Reset de-encapsulation cache. */ 698 699 ip_mrouter = NULL; 700 701 splx(s); 702 703 if (mrtdebug) 704 log(LOG_DEBUG, "ip_mrouter_done\n"); 705 706 return (0); 707 } 708 709 void 710 ip_mrouter_detach(struct ifnet *ifp) 711 { 712 int vifi, i; 713 struct vif *vifp; 714 struct mfc *rt; 715 struct rtdetq *rte; 716 717 /* XXX not sure about side effect to userland routing daemon */ 718 for (vifi = 0; vifi < numvifs; vifi++) { 719 vifp = &viftable[vifi]; 720 if (vifp->v_ifp == ifp) 721 reset_vif(vifp); 722 } 723 for (i = 0; i < MFCTBLSIZ; i++) { 724 if (nexpire[i] == 0) 725 continue; 726 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) { 727 for (rte = rt->mfc_stall; rte; rte = rte->next) { 728 if (rte->ifp == ifp) 729 rte->ifp = NULL; 730 } 731 } 732 } 733 } 734 735 static int 736 get_version(struct mbuf *m) 737 { 738 int *v = mtod(m, int *); 739 740 *v = 0x0305; /* XXX !!!! */ 741 m->m_len = sizeof(int); 742 return (0); 743 } 744 745 /* 746 * Set PIM assert processing global 747 */ 748 static int 749 set_assert(struct mbuf *m) 750 { 751 int *i; 752 753 if (m == NULL || m->m_len != sizeof(int)) 754 return (EINVAL); 755 756 i = mtod(m, int *); 757 pim_assert = !!*i; 758 return (0); 759 } 760 761 /* 762 * Get PIM assert processing global 763 */ 764 static int 765 get_assert(struct mbuf *m) 766 { 767 int *i = mtod(m, int *); 768 769 *i = pim_assert; 770 m->m_len = sizeof(int); 771 return (0); 772 } 773 774 /* 775 * Configure API capabilities 776 */ 777 static int 778 set_api_config(struct mbuf *m) 779 { 780 int i; 781 u_int32_t *apival; 782 783 if (m == NULL || m->m_len < sizeof(u_int32_t)) 784 return (EINVAL); 785 786 apival = mtod(m, u_int32_t *); 787 788 /* 789 * We can set the API capabilities only if it is the first operation 790 * after MRT_INIT. I.e.: 791 * - there are no vifs installed 792 * - pim_assert is not enabled 793 * - the MFC table is empty 794 */ 795 if (numvifs > 0) { 796 *apival = 0; 797 return (EPERM); 798 } 799 if (pim_assert) { 800 *apival = 0; 801 return (EPERM); 802 } 803 for (i = 0; i < MFCTBLSIZ; i++) { 804 if (LIST_FIRST(&mfchashtbl[i]) != NULL) { 805 *apival = 0; 806 return (EPERM); 807 } 808 } 809 810 mrt_api_config = *apival & mrt_api_support; 811 *apival = mrt_api_config; 812 813 return (0); 814 } 815 816 /* 817 * Get API capabilities 818 */ 819 static int 820 get_api_support(struct mbuf *m) 821 { 822 u_int32_t *apival; 823 824 if (m == NULL || m->m_len < sizeof(u_int32_t)) 825 return (EINVAL); 826 827 apival = mtod(m, u_int32_t *); 828 829 *apival = mrt_api_support; 830 831 return (0); 832 } 833 834 /* 835 * Get API configured capabilities 836 */ 837 static int 838 get_api_config(struct mbuf *m) 839 { 840 u_int32_t *apival; 841 842 if (m == NULL || m->m_len < sizeof(u_int32_t)) 843 return (EINVAL); 844 845 apival = mtod(m, u_int32_t *); 846 847 *apival = mrt_api_config; 848 849 return (0); 850 } 851 852 /* 853 * Add a vif to the vif table 854 */ 855 static int 856 add_vif(struct mbuf *m) 857 { 858 struct vifctl *vifcp; 859 struct vif *vifp; 860 struct ifaddr *ifa; 861 struct ifnet *ifp; 862 struct ifreq ifr; 863 int error, s; 864 struct sockaddr_in sin; 865 866 if (m == NULL || m->m_len < sizeof(struct vifctl)) 867 return (EINVAL); 868 869 vifcp = mtod(m, struct vifctl *); 870 if (vifcp->vifc_vifi >= MAXVIFS) 871 return (EINVAL); 872 if (in_nullhost(vifcp->vifc_lcl_addr)) 873 return (EADDRNOTAVAIL); 874 875 vifp = &viftable[vifcp->vifc_vifi]; 876 if (!in_nullhost(vifp->v_lcl_addr)) 877 return (EADDRINUSE); 878 879 /* Find the interface with an address in AF_INET family. */ 880 #ifdef PIM 881 if (vifcp->vifc_flags & VIFF_REGISTER) { 882 /* 883 * XXX: Because VIFF_REGISTER does not really need a valid 884 * local interface (e.g. it could be 127.0.0.2), we don't 885 * check its address. 886 */ 887 ifp = NULL; 888 } else 889 #endif 890 { 891 sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0); 892 ifa = ifa_ifwithaddr(sintosa(&sin)); 893 if (ifa == NULL) 894 return (EADDRNOTAVAIL); 895 ifp = ifa->ifa_ifp; 896 } 897 898 if (vifcp->vifc_flags & VIFF_TUNNEL) { 899 if (vifcp->vifc_flags & VIFF_SRCRT) { 900 log(LOG_ERR, "source routed tunnels not supported\n"); 901 return (EOPNOTSUPP); 902 } 903 904 /* attach this vif to decapsulator dispatch table */ 905 /* 906 * XXX Use addresses in registration so that matching 907 * can be done with radix tree in decapsulator. But, 908 * we need to check inner header for multicast, so 909 * this requires both radix tree lookup and then a 910 * function to check, and this is not supported yet. 911 */ 912 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 913 vif_encapcheck, &vif_protosw, vifp); 914 if (!vifp->v_encap_cookie) 915 return (EINVAL); 916 917 /* Create a fake encapsulation interface. */ 918 ifp = (struct ifnet *)malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK); 919 bzero(ifp, sizeof(*ifp)); 920 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 921 "mdecap%d", vifcp->vifc_vifi); 922 923 /* Prepare cached route entry. */ 924 bzero(&vifp->v_route, sizeof(vifp->v_route)); 925 #ifdef PIM 926 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 927 ifp = &multicast_register_if; 928 if (mrtdebug) 929 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 930 (void *)ifp); 931 if (reg_vif_num == VIFI_INVALID) { 932 bzero(ifp, sizeof(*ifp)); 933 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 934 "register_vif"); 935 ifp->if_flags = IFF_LOOPBACK; 936 bzero(&vifp->v_route, sizeof(vifp->v_route)); 937 reg_vif_num = vifcp->vifc_vifi; 938 } 939 #endif 940 } else { 941 /* Make sure the interface supports multicast. */ 942 if ((ifp->if_flags & IFF_MULTICAST) == 0) 943 return (EOPNOTSUPP); 944 945 /* Enable promiscuous reception of all IP multicasts. */ 946 sockaddr_in_init(&sin, &zeroin_addr, 0); 947 ifreq_setaddr(SIOCADDMULTI, &ifr, sintosa(&sin)); 948 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, (void *)&ifr); 949 if (error) 950 return (error); 951 } 952 953 s = splsoftnet(); 954 955 /* Define parameters for the tbf structure. */ 956 vifp->tbf_q = NULL; 957 vifp->tbf_t = &vifp->tbf_q; 958 microtime(&vifp->tbf_last_pkt_t); 959 vifp->tbf_n_tok = 0; 960 vifp->tbf_q_len = 0; 961 vifp->tbf_max_q_len = MAXQSIZE; 962 963 vifp->v_flags = vifcp->vifc_flags; 964 vifp->v_threshold = vifcp->vifc_threshold; 965 /* scaling up here allows division by 1024 in critical code */ 966 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 967 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 968 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 969 vifp->v_ifp = ifp; 970 /* Initialize per vif pkt counters. */ 971 vifp->v_pkt_in = 0; 972 vifp->v_pkt_out = 0; 973 vifp->v_bytes_in = 0; 974 vifp->v_bytes_out = 0; 975 976 callout_init(&vifp->v_repq_ch, 0); 977 978 #ifdef RSVP_ISI 979 vifp->v_rsvp_on = 0; 980 vifp->v_rsvpd = NULL; 981 #endif /* RSVP_ISI */ 982 983 splx(s); 984 985 /* Adjust numvifs up if the vifi is higher than numvifs. */ 986 if (numvifs <= vifcp->vifc_vifi) 987 numvifs = vifcp->vifc_vifi + 1; 988 989 if (mrtdebug) 990 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 991 vifcp->vifc_vifi, 992 ntohl(vifcp->vifc_lcl_addr.s_addr), 993 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 994 ntohl(vifcp->vifc_rmt_addr.s_addr), 995 vifcp->vifc_threshold, 996 vifcp->vifc_rate_limit); 997 998 return (0); 999 } 1000 1001 void 1002 reset_vif(struct vif *vifp) 1003 { 1004 struct mbuf *m, *n; 1005 struct ifnet *ifp; 1006 struct ifreq ifr; 1007 struct sockaddr_in sin; 1008 1009 callout_stop(&vifp->v_repq_ch); 1010 1011 /* detach this vif from decapsulator dispatch table */ 1012 encap_detach(vifp->v_encap_cookie); 1013 vifp->v_encap_cookie = NULL; 1014 1015 /* 1016 * Free packets queued at the interface 1017 */ 1018 for (m = vifp->tbf_q; m != NULL; m = n) { 1019 n = m->m_nextpkt; 1020 m_freem(m); 1021 } 1022 1023 if (vifp->v_flags & VIFF_TUNNEL) 1024 free(vifp->v_ifp, M_MRTABLE); 1025 else if (vifp->v_flags & VIFF_REGISTER) { 1026 #ifdef PIM 1027 reg_vif_num = VIFI_INVALID; 1028 #endif 1029 } else { 1030 sockaddr_in_init(&sin, &zeroin_addr, 0); 1031 ifreq_setaddr(SIOCDELMULTI, &ifr, sintosa(&sin)); 1032 ifp = vifp->v_ifp; 1033 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, (void *)&ifr); 1034 } 1035 bzero((void *)vifp, sizeof(*vifp)); 1036 } 1037 1038 /* 1039 * Delete a vif from the vif table 1040 */ 1041 static int 1042 del_vif(struct mbuf *m) 1043 { 1044 vifi_t *vifip; 1045 struct vif *vifp; 1046 vifi_t vifi; 1047 int s; 1048 1049 if (m == NULL || m->m_len < sizeof(vifi_t)) 1050 return (EINVAL); 1051 1052 vifip = mtod(m, vifi_t *); 1053 if (*vifip >= numvifs) 1054 return (EINVAL); 1055 1056 vifp = &viftable[*vifip]; 1057 if (in_nullhost(vifp->v_lcl_addr)) 1058 return (EADDRNOTAVAIL); 1059 1060 s = splsoftnet(); 1061 1062 reset_vif(vifp); 1063 1064 /* Adjust numvifs down */ 1065 for (vifi = numvifs; vifi > 0; vifi--) 1066 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr)) 1067 break; 1068 numvifs = vifi; 1069 1070 splx(s); 1071 1072 if (mrtdebug) 1073 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 1074 1075 return (0); 1076 } 1077 1078 /* 1079 * update an mfc entry without resetting counters and S,G addresses. 1080 */ 1081 static void 1082 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1083 { 1084 int i; 1085 1086 rt->mfc_parent = mfccp->mfcc_parent; 1087 for (i = 0; i < numvifs; i++) { 1088 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1089 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1090 MRT_MFC_FLAGS_ALL; 1091 } 1092 /* set the RP address */ 1093 if (mrt_api_config & MRT_MFC_RP) 1094 rt->mfc_rp = mfccp->mfcc_rp; 1095 else 1096 rt->mfc_rp = zeroin_addr; 1097 } 1098 1099 /* 1100 * fully initialize an mfc entry from the parameter. 1101 */ 1102 static void 1103 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1104 { 1105 rt->mfc_origin = mfccp->mfcc_origin; 1106 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1107 1108 update_mfc_params(rt, mfccp); 1109 1110 /* initialize pkt counters per src-grp */ 1111 rt->mfc_pkt_cnt = 0; 1112 rt->mfc_byte_cnt = 0; 1113 rt->mfc_wrong_if = 0; 1114 timerclear(&rt->mfc_last_assert); 1115 } 1116 1117 static void 1118 expire_mfc(struct mfc *rt) 1119 { 1120 struct rtdetq *rte, *nrte; 1121 1122 free_bw_list(rt->mfc_bw_meter); 1123 1124 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 1125 nrte = rte->next; 1126 m_freem(rte->m); 1127 free(rte, M_MRTABLE); 1128 } 1129 1130 LIST_REMOVE(rt, mfc_hash); 1131 free(rt, M_MRTABLE); 1132 } 1133 1134 /* 1135 * Add an mfc entry 1136 */ 1137 static int 1138 add_mfc(struct mbuf *m) 1139 { 1140 struct mfcctl2 mfcctl2; 1141 struct mfcctl2 *mfccp; 1142 struct mfc *rt; 1143 u_int32_t hash = 0; 1144 struct rtdetq *rte, *nrte; 1145 u_short nstl; 1146 int s; 1147 int mfcctl_size = sizeof(struct mfcctl); 1148 1149 if (mrt_api_config & MRT_API_FLAGS_ALL) 1150 mfcctl_size = sizeof(struct mfcctl2); 1151 1152 if (m == NULL || m->m_len < mfcctl_size) 1153 return (EINVAL); 1154 1155 /* 1156 * select data size depending on API version. 1157 */ 1158 if (mrt_api_config & MRT_API_FLAGS_ALL) { 1159 struct mfcctl2 *mp2 = mtod(m, struct mfcctl2 *); 1160 bcopy(mp2, (void *)&mfcctl2, sizeof(*mp2)); 1161 } else { 1162 struct mfcctl *mp = mtod(m, struct mfcctl *); 1163 memcpy(&mfcctl2, mp, sizeof(*mp)); 1164 memset((char *)&mfcctl2 + sizeof(struct mfcctl), 0, 1165 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1166 } 1167 mfccp = &mfcctl2; 1168 1169 s = splsoftnet(); 1170 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1171 1172 /* If an entry already exists, just update the fields */ 1173 if (rt) { 1174 if (mrtdebug & DEBUG_MFC) 1175 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n", 1176 ntohl(mfccp->mfcc_origin.s_addr), 1177 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1178 mfccp->mfcc_parent); 1179 1180 update_mfc_params(rt, mfccp); 1181 1182 splx(s); 1183 return (0); 1184 } 1185 1186 /* 1187 * Find the entry for which the upcall was made and update 1188 */ 1189 nstl = 0; 1190 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1191 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1192 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1193 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1194 rt->mfc_stall != NULL) { 1195 if (nstl++) 1196 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 1197 "multiple kernel entries", 1198 ntohl(mfccp->mfcc_origin.s_addr), 1199 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1200 mfccp->mfcc_parent, rt->mfc_stall); 1201 1202 if (mrtdebug & DEBUG_MFC) 1203 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n", 1204 ntohl(mfccp->mfcc_origin.s_addr), 1205 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1206 mfccp->mfcc_parent, rt->mfc_stall); 1207 1208 rte = rt->mfc_stall; 1209 init_mfc_params(rt, mfccp); 1210 rt->mfc_stall = NULL; 1211 1212 rt->mfc_expire = 0; /* Don't clean this guy up */ 1213 nexpire[hash]--; 1214 1215 /* free packets Qed at the end of this entry */ 1216 for (; rte != NULL; rte = nrte) { 1217 nrte = rte->next; 1218 if (rte->ifp) { 1219 #ifdef RSVP_ISI 1220 ip_mdq(rte->m, rte->ifp, rt, -1); 1221 #else 1222 ip_mdq(rte->m, rte->ifp, rt); 1223 #endif /* RSVP_ISI */ 1224 } 1225 m_freem(rte->m); 1226 #ifdef UPCALL_TIMING 1227 collate(&rte->t); 1228 #endif /* UPCALL_TIMING */ 1229 free(rte, M_MRTABLE); 1230 } 1231 } 1232 } 1233 1234 /* 1235 * It is possible that an entry is being inserted without an upcall 1236 */ 1237 if (nstl == 0) { 1238 /* 1239 * No mfc; make a new one 1240 */ 1241 if (mrtdebug & DEBUG_MFC) 1242 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n", 1243 ntohl(mfccp->mfcc_origin.s_addr), 1244 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1245 mfccp->mfcc_parent); 1246 1247 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1248 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1249 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1250 init_mfc_params(rt, mfccp); 1251 if (rt->mfc_expire) 1252 nexpire[hash]--; 1253 rt->mfc_expire = 0; 1254 break; /* XXX */ 1255 } 1256 } 1257 if (rt == NULL) { /* no upcall, so make a new entry */ 1258 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1259 M_NOWAIT); 1260 if (rt == NULL) { 1261 splx(s); 1262 return (ENOBUFS); 1263 } 1264 1265 init_mfc_params(rt, mfccp); 1266 rt->mfc_expire = 0; 1267 rt->mfc_stall = NULL; 1268 rt->mfc_bw_meter = NULL; 1269 1270 /* insert new entry at head of hash chain */ 1271 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1272 } 1273 } 1274 1275 splx(s); 1276 return (0); 1277 } 1278 1279 #ifdef UPCALL_TIMING 1280 /* 1281 * collect delay statistics on the upcalls 1282 */ 1283 static void 1284 collate(struct timeval *t) 1285 { 1286 u_int32_t d; 1287 struct timeval tp; 1288 u_int32_t delta; 1289 1290 microtime(&tp); 1291 1292 if (timercmp(t, &tp, <)) { 1293 TV_DELTA(tp, *t, delta); 1294 1295 d = delta >> 10; 1296 if (d > 50) 1297 d = 50; 1298 1299 ++upcall_data[d]; 1300 } 1301 } 1302 #endif /* UPCALL_TIMING */ 1303 1304 /* 1305 * Delete an mfc entry 1306 */ 1307 static int 1308 del_mfc(struct mbuf *m) 1309 { 1310 struct mfcctl2 mfcctl2; 1311 struct mfcctl2 *mfccp; 1312 struct mfc *rt; 1313 int s; 1314 int mfcctl_size = sizeof(struct mfcctl); 1315 struct mfcctl *mp = mtod(m, struct mfcctl *); 1316 1317 /* 1318 * XXX: for deleting MFC entries the information in entries 1319 * of size "struct mfcctl" is sufficient. 1320 */ 1321 1322 if (m == NULL || m->m_len < mfcctl_size) 1323 return (EINVAL); 1324 1325 memcpy(&mfcctl2, mp, sizeof(*mp)); 1326 memset((char *)&mfcctl2 + sizeof(struct mfcctl), 0, 1327 sizeof(mfcctl2) - sizeof(struct mfcctl)); 1328 1329 mfccp = &mfcctl2; 1330 1331 if (mrtdebug & DEBUG_MFC) 1332 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 1333 ntohl(mfccp->mfcc_origin.s_addr), 1334 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 1335 1336 s = splsoftnet(); 1337 1338 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1339 if (rt == NULL) { 1340 splx(s); 1341 return (EADDRNOTAVAIL); 1342 } 1343 1344 /* 1345 * free the bw_meter entries 1346 */ 1347 free_bw_list(rt->mfc_bw_meter); 1348 rt->mfc_bw_meter = NULL; 1349 1350 LIST_REMOVE(rt, mfc_hash); 1351 free(rt, M_MRTABLE); 1352 1353 splx(s); 1354 return (0); 1355 } 1356 1357 static int 1358 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1359 { 1360 if (s) { 1361 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, 1362 (struct mbuf *)NULL) != 0) { 1363 sorwakeup(s); 1364 return (0); 1365 } 1366 } 1367 m_freem(mm); 1368 return (-1); 1369 } 1370 1371 /* 1372 * IP multicast forwarding function. This function assumes that the packet 1373 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1374 * pointed to by "ifp", and the packet is to be relayed to other networks 1375 * that have members of the packet's destination IP multicast group. 1376 * 1377 * The packet is returned unscathed to the caller, unless it is 1378 * erroneous, in which case a non-zero return value tells the caller to 1379 * discard it. 1380 */ 1381 1382 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1383 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1384 1385 int 1386 #ifdef RSVP_ISI 1387 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo) 1388 #else 1389 ip_mforward(struct mbuf *m, struct ifnet *ifp) 1390 #endif /* RSVP_ISI */ 1391 { 1392 struct ip *ip = mtod(m, struct ip *); 1393 struct mfc *rt; 1394 static int srctun = 0; 1395 struct mbuf *mm; 1396 struct sockaddr_in sin; 1397 int s; 1398 vifi_t vifi; 1399 1400 if (mrtdebug & DEBUG_FORWARD) 1401 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1402 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1403 1404 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1405 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1406 /* 1407 * Packet arrived via a physical interface or 1408 * an encapsulated tunnel or a register_vif. 1409 */ 1410 } else { 1411 /* 1412 * Packet arrived through a source-route tunnel. 1413 * Source-route tunnels are no longer supported. 1414 */ 1415 if ((srctun++ % 1000) == 0) 1416 log(LOG_ERR, 1417 "ip_mforward: received source-routed packet from %x\n", 1418 ntohl(ip->ip_src.s_addr)); 1419 1420 return (1); 1421 } 1422 1423 #ifdef RSVP_ISI 1424 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1425 if (ip->ip_ttl < MAXTTL) 1426 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1427 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1428 struct vif *vifp = viftable + vifi; 1429 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1430 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1431 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1432 vifp->v_ifp->if_xname); 1433 } 1434 return (ip_mdq(m, ifp, (struct mfc *)NULL, vifi)); 1435 } 1436 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1437 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1438 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1439 } 1440 #endif /* RSVP_ISI */ 1441 1442 /* 1443 * Don't forward a packet with time-to-live of zero or one, 1444 * or a packet destined to a local-only group. 1445 */ 1446 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1447 return (0); 1448 1449 /* 1450 * Determine forwarding vifs from the forwarding cache table 1451 */ 1452 s = splsoftnet(); 1453 ++mrtstat.mrts_mfc_lookups; 1454 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1455 1456 /* Entry exists, so forward if necessary */ 1457 if (rt != NULL) { 1458 splx(s); 1459 #ifdef RSVP_ISI 1460 return (ip_mdq(m, ifp, rt, -1)); 1461 #else 1462 return (ip_mdq(m, ifp, rt)); 1463 #endif /* RSVP_ISI */ 1464 } else { 1465 /* 1466 * If we don't have a route for packet's origin, 1467 * Make a copy of the packet & send message to routing daemon 1468 */ 1469 1470 struct mbuf *mb0; 1471 struct rtdetq *rte; 1472 u_int32_t hash; 1473 int hlen = ip->ip_hl << 2; 1474 #ifdef UPCALL_TIMING 1475 struct timeval tp; 1476 1477 microtime(&tp); 1478 #endif /* UPCALL_TIMING */ 1479 1480 ++mrtstat.mrts_mfc_misses; 1481 1482 mrtstat.mrts_no_route++; 1483 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1484 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1485 ntohl(ip->ip_src.s_addr), 1486 ntohl(ip->ip_dst.s_addr)); 1487 1488 /* 1489 * Allocate mbufs early so that we don't do extra work if we are 1490 * just going to fail anyway. Make sure to pullup the header so 1491 * that other people can't step on it. 1492 */ 1493 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1494 M_NOWAIT); 1495 if (rte == NULL) { 1496 splx(s); 1497 return (ENOBUFS); 1498 } 1499 mb0 = m_copypacket(m, M_DONTWAIT); 1500 M_PULLUP(mb0, hlen); 1501 if (mb0 == NULL) { 1502 free(rte, M_MRTABLE); 1503 splx(s); 1504 return (ENOBUFS); 1505 } 1506 1507 /* is there an upcall waiting for this flow? */ 1508 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1509 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1510 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1511 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1512 rt->mfc_stall != NULL) 1513 break; 1514 } 1515 1516 if (rt == NULL) { 1517 int i; 1518 struct igmpmsg *im; 1519 1520 /* 1521 * Locate the vifi for the incoming interface for 1522 * this packet. 1523 * If none found, drop packet. 1524 */ 1525 for (vifi = 0; vifi < numvifs && 1526 viftable[vifi].v_ifp != ifp; vifi++) 1527 ; 1528 if (vifi >= numvifs) /* vif not found, drop packet */ 1529 goto non_fatal; 1530 1531 /* no upcall, so make a new entry */ 1532 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1533 M_NOWAIT); 1534 if (rt == NULL) 1535 goto fail; 1536 1537 /* 1538 * Make a copy of the header to send to the user level 1539 * process 1540 */ 1541 mm = m_copym(m, 0, hlen, M_DONTWAIT); 1542 M_PULLUP(mm, hlen); 1543 if (mm == NULL) 1544 goto fail1; 1545 1546 /* 1547 * Send message to routing daemon to install 1548 * a route into the kernel table 1549 */ 1550 1551 im = mtod(mm, struct igmpmsg *); 1552 im->im_msgtype = IGMPMSG_NOCACHE; 1553 im->im_mbz = 0; 1554 im->im_vif = vifi; 1555 1556 mrtstat.mrts_upcalls++; 1557 1558 sockaddr_in_init(&sin, &ip->ip_src, 0); 1559 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1560 log(LOG_WARNING, 1561 "ip_mforward: ip_mrouter socket queue full\n"); 1562 ++mrtstat.mrts_upq_sockfull; 1563 fail1: 1564 free(rt, M_MRTABLE); 1565 fail: 1566 free(rte, M_MRTABLE); 1567 m_freem(mb0); 1568 splx(s); 1569 return (ENOBUFS); 1570 } 1571 1572 /* insert new entry at head of hash chain */ 1573 rt->mfc_origin = ip->ip_src; 1574 rt->mfc_mcastgrp = ip->ip_dst; 1575 rt->mfc_pkt_cnt = 0; 1576 rt->mfc_byte_cnt = 0; 1577 rt->mfc_wrong_if = 0; 1578 rt->mfc_expire = UPCALL_EXPIRE; 1579 nexpire[hash]++; 1580 for (i = 0; i < numvifs; i++) { 1581 rt->mfc_ttls[i] = 0; 1582 rt->mfc_flags[i] = 0; 1583 } 1584 rt->mfc_parent = -1; 1585 1586 /* clear the RP address */ 1587 rt->mfc_rp = zeroin_addr; 1588 1589 rt->mfc_bw_meter = NULL; 1590 1591 /* link into table */ 1592 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1593 /* Add this entry to the end of the queue */ 1594 rt->mfc_stall = rte; 1595 } else { 1596 /* determine if q has overflowed */ 1597 struct rtdetq **p; 1598 int npkts = 0; 1599 1600 /* 1601 * XXX ouch! we need to append to the list, but we 1602 * only have a pointer to the front, so we have to 1603 * scan the entire list every time. 1604 */ 1605 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1606 if (++npkts > MAX_UPQ) { 1607 mrtstat.mrts_upq_ovflw++; 1608 non_fatal: 1609 free(rte, M_MRTABLE); 1610 m_freem(mb0); 1611 splx(s); 1612 return (0); 1613 } 1614 1615 /* Add this entry to the end of the queue */ 1616 *p = rte; 1617 } 1618 1619 rte->next = NULL; 1620 rte->m = mb0; 1621 rte->ifp = ifp; 1622 #ifdef UPCALL_TIMING 1623 rte->t = tp; 1624 #endif /* UPCALL_TIMING */ 1625 1626 splx(s); 1627 1628 return (0); 1629 } 1630 } 1631 1632 1633 /*ARGSUSED*/ 1634 static void 1635 expire_upcalls(void *v) 1636 { 1637 int i; 1638 int s; 1639 1640 s = splsoftnet(); 1641 1642 for (i = 0; i < MFCTBLSIZ; i++) { 1643 struct mfc *rt, *nrt; 1644 1645 if (nexpire[i] == 0) 1646 continue; 1647 1648 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1649 nrt = LIST_NEXT(rt, mfc_hash); 1650 1651 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1652 continue; 1653 nexpire[i]--; 1654 1655 /* 1656 * free the bw_meter entries 1657 */ 1658 while (rt->mfc_bw_meter != NULL) { 1659 struct bw_meter *x = rt->mfc_bw_meter; 1660 1661 rt->mfc_bw_meter = x->bm_mfc_next; 1662 free(x, M_BWMETER); 1663 } 1664 1665 ++mrtstat.mrts_cache_cleanups; 1666 if (mrtdebug & DEBUG_EXPIRE) 1667 log(LOG_DEBUG, 1668 "expire_upcalls: expiring (%x %x)\n", 1669 ntohl(rt->mfc_origin.s_addr), 1670 ntohl(rt->mfc_mcastgrp.s_addr)); 1671 1672 expire_mfc(rt); 1673 } 1674 } 1675 1676 splx(s); 1677 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1678 expire_upcalls, NULL); 1679 } 1680 1681 /* 1682 * Packet forwarding routine once entry in the cache is made 1683 */ 1684 static int 1685 #ifdef RSVP_ISI 1686 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1687 #else 1688 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt) 1689 #endif /* RSVP_ISI */ 1690 { 1691 struct ip *ip = mtod(m, struct ip *); 1692 vifi_t vifi; 1693 struct vif *vifp; 1694 struct sockaddr_in sin; 1695 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2); 1696 1697 /* 1698 * Macro to send packet on vif. Since RSVP packets don't get counted on 1699 * input, they shouldn't get counted on output, so statistics keeping is 1700 * separate. 1701 */ 1702 #define MC_SEND(ip, vifp, m) do { \ 1703 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1704 encap_send((ip), (vifp), (m)); \ 1705 else \ 1706 phyint_send((ip), (vifp), (m)); \ 1707 } while (/*CONSTCOND*/ 0) 1708 1709 #ifdef RSVP_ISI 1710 /* 1711 * If xmt_vif is not -1, send on only the requested vif. 1712 * 1713 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1714 */ 1715 if (xmt_vif < numvifs) { 1716 #ifdef PIM 1717 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1718 pim_register_send(ip, viftable + xmt_vif, m, rt); 1719 else 1720 #endif 1721 MC_SEND(ip, viftable + xmt_vif, m); 1722 return (1); 1723 } 1724 #endif /* RSVP_ISI */ 1725 1726 /* 1727 * Don't forward if it didn't arrive from the parent vif for its origin. 1728 */ 1729 vifi = rt->mfc_parent; 1730 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1731 /* came in the wrong interface */ 1732 if (mrtdebug & DEBUG_FORWARD) 1733 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1734 ifp, vifi, 1735 vifi >= numvifs ? 0 : viftable[vifi].v_ifp); 1736 ++mrtstat.mrts_wrong_if; 1737 ++rt->mfc_wrong_if; 1738 /* 1739 * If we are doing PIM assert processing, send a message 1740 * to the routing daemon. 1741 * 1742 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1743 * can complete the SPT switch, regardless of the type 1744 * of the iif (broadcast media, GRE tunnel, etc). 1745 */ 1746 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1747 struct timeval now; 1748 u_int32_t delta; 1749 1750 #ifdef PIM 1751 if (ifp == &multicast_register_if) 1752 pimstat.pims_rcv_registers_wrongiif++; 1753 #endif 1754 1755 /* Get vifi for the incoming packet */ 1756 for (vifi = 0; 1757 vifi < numvifs && viftable[vifi].v_ifp != ifp; 1758 vifi++) 1759 ; 1760 if (vifi >= numvifs) { 1761 /* The iif is not found: ignore the packet. */ 1762 return (0); 1763 } 1764 1765 if (rt->mfc_flags[vifi] & 1766 MRT_MFC_FLAGS_DISABLE_WRONGVIF) { 1767 /* WRONGVIF disabled: ignore the packet */ 1768 return (0); 1769 } 1770 1771 microtime(&now); 1772 1773 TV_DELTA(rt->mfc_last_assert, now, delta); 1774 1775 if (delta > ASSERT_MSG_TIME) { 1776 struct igmpmsg *im; 1777 int hlen = ip->ip_hl << 2; 1778 struct mbuf *mm = 1779 m_copym(m, 0, hlen, M_DONTWAIT); 1780 1781 M_PULLUP(mm, hlen); 1782 if (mm == NULL) 1783 return (ENOBUFS); 1784 1785 rt->mfc_last_assert = now; 1786 1787 im = mtod(mm, struct igmpmsg *); 1788 im->im_msgtype = IGMPMSG_WRONGVIF; 1789 im->im_mbz = 0; 1790 im->im_vif = vifi; 1791 1792 mrtstat.mrts_upcalls++; 1793 1794 sockaddr_in_init(&sin, &im->im_src, 0); 1795 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1796 log(LOG_WARNING, 1797 "ip_mforward: ip_mrouter socket queue full\n"); 1798 ++mrtstat.mrts_upq_sockfull; 1799 return (ENOBUFS); 1800 } 1801 } 1802 } 1803 return (0); 1804 } 1805 1806 /* If I sourced this packet, it counts as output, else it was input. */ 1807 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1808 viftable[vifi].v_pkt_out++; 1809 viftable[vifi].v_bytes_out += plen; 1810 } else { 1811 viftable[vifi].v_pkt_in++; 1812 viftable[vifi].v_bytes_in += plen; 1813 } 1814 rt->mfc_pkt_cnt++; 1815 rt->mfc_byte_cnt += plen; 1816 1817 /* 1818 * For each vif, decide if a copy of the packet should be forwarded. 1819 * Forward if: 1820 * - the ttl exceeds the vif's threshold 1821 * - there are group members downstream on interface 1822 */ 1823 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1824 if ((rt->mfc_ttls[vifi] > 0) && 1825 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1826 vifp->v_pkt_out++; 1827 vifp->v_bytes_out += plen; 1828 #ifdef PIM 1829 if (vifp->v_flags & VIFF_REGISTER) 1830 pim_register_send(ip, vifp, m, rt); 1831 else 1832 #endif 1833 MC_SEND(ip, vifp, m); 1834 } 1835 1836 /* 1837 * Perform upcall-related bw measuring. 1838 */ 1839 if (rt->mfc_bw_meter != NULL) { 1840 struct bw_meter *x; 1841 struct timeval now; 1842 1843 microtime(&now); 1844 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1845 bw_meter_receive_packet(x, plen, &now); 1846 } 1847 1848 return (0); 1849 } 1850 1851 #ifdef RSVP_ISI 1852 /* 1853 * check if a vif number is legal/ok. This is used by ip_output. 1854 */ 1855 int 1856 legal_vif_num(int vif) 1857 { 1858 if (vif >= 0 && vif < numvifs) 1859 return (1); 1860 else 1861 return (0); 1862 } 1863 #endif /* RSVP_ISI */ 1864 1865 static void 1866 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1867 { 1868 struct mbuf *mb_copy; 1869 int hlen = ip->ip_hl << 2; 1870 1871 /* 1872 * Make a new reference to the packet; make sure that 1873 * the IP header is actually copied, not just referenced, 1874 * so that ip_output() only scribbles on the copy. 1875 */ 1876 mb_copy = m_copypacket(m, M_DONTWAIT); 1877 M_PULLUP(mb_copy, hlen); 1878 if (mb_copy == NULL) 1879 return; 1880 1881 if (vifp->v_rate_limit <= 0) 1882 tbf_send_packet(vifp, mb_copy); 1883 else 1884 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1885 ntohs(ip->ip_len)); 1886 } 1887 1888 static void 1889 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1890 { 1891 struct mbuf *mb_copy; 1892 struct ip *ip_copy; 1893 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1894 1895 /* Take care of delayed checksums */ 1896 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1897 in_delayed_cksum(m); 1898 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1899 } 1900 1901 /* 1902 * copy the old packet & pullup it's IP header into the 1903 * new mbuf so we can modify it. Try to fill the new 1904 * mbuf since if we don't the ethernet driver will. 1905 */ 1906 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1907 if (mb_copy == NULL) 1908 return; 1909 mb_copy->m_data += max_linkhdr; 1910 mb_copy->m_pkthdr.len = len; 1911 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1912 1913 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) { 1914 m_freem(mb_copy); 1915 return; 1916 } 1917 i = MHLEN - max_linkhdr; 1918 if (i > len) 1919 i = len; 1920 mb_copy = m_pullup(mb_copy, i); 1921 if (mb_copy == NULL) 1922 return; 1923 1924 /* 1925 * fill in the encapsulating IP header. 1926 */ 1927 ip_copy = mtod(mb_copy, struct ip *); 1928 *ip_copy = multicast_encap_iphdr; 1929 if (len < IP_MINFRAGSIZE) 1930 ip_copy->ip_id = 0; 1931 else 1932 ip_copy->ip_id = ip_newid(NULL); 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 *)((char *)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), (void *)&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), (void *)&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, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS, 2223 &imo, NULL); 2224 2225 if (mrtdebug & DEBUG_XMIT) 2226 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 2227 (long)(vifp - viftable), error); 2228 } 2229 splx(s); 2230 } 2231 2232 /* determine the current time and then 2233 * the elapsed time (between the last time and time now) 2234 * in milliseconds & update the no. of tokens in the bucket 2235 */ 2236 static void 2237 tbf_update_tokens(struct vif *vifp) 2238 { 2239 struct timeval tp; 2240 u_int32_t tm; 2241 int s = splsoftnet(); 2242 2243 microtime(&tp); 2244 2245 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 2246 2247 /* 2248 * This formula is actually 2249 * "time in seconds" * "bytes/second". 2250 * 2251 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2252 * 2253 * The (1000/1024) was introduced in add_vif to optimize 2254 * this divide into a shift. 2255 */ 2256 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 2257 vifp->tbf_last_pkt_t = tp; 2258 2259 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 2260 vifp->tbf_n_tok = MAX_BKT_SIZE; 2261 2262 splx(s); 2263 } 2264 2265 static int 2266 priority(struct vif *vifp, struct ip *ip) 2267 { 2268 int prio = 50; /* the lowest priority -- default case */ 2269 2270 /* temporary hack; may add general packet classifier some day */ 2271 2272 /* 2273 * The UDP port space is divided up into four priority ranges: 2274 * [0, 16384) : unclassified - lowest priority 2275 * [16384, 32768) : audio - highest priority 2276 * [32768, 49152) : whiteboard - medium priority 2277 * [49152, 65536) : video - low priority 2278 */ 2279 if (ip->ip_p == IPPROTO_UDP) { 2280 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2281 2282 switch (ntohs(udp->uh_dport) & 0xc000) { 2283 case 0x4000: 2284 prio = 70; 2285 break; 2286 case 0x8000: 2287 prio = 60; 2288 break; 2289 case 0xc000: 2290 prio = 55; 2291 break; 2292 } 2293 2294 if (tbfdebug > 1) 2295 log(LOG_DEBUG, "port %x prio %d\n", 2296 ntohs(udp->uh_dport), prio); 2297 } 2298 2299 return (prio); 2300 } 2301 2302 /* 2303 * End of token bucket filter modifications 2304 */ 2305 #ifdef RSVP_ISI 2306 int 2307 ip_rsvp_vif_init(struct socket *so, struct mbuf *m) 2308 { 2309 int vifi, s; 2310 2311 if (rsvpdebug) 2312 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2313 so->so_type, so->so_proto->pr_protocol); 2314 2315 if (so->so_type != SOCK_RAW || 2316 so->so_proto->pr_protocol != IPPROTO_RSVP) 2317 return (EOPNOTSUPP); 2318 2319 /* Check mbuf. */ 2320 if (m == NULL || m->m_len != sizeof(int)) { 2321 return (EINVAL); 2322 } 2323 vifi = *(mtod(m, int *)); 2324 2325 if (rsvpdebug) 2326 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", 2327 vifi, rsvp_on); 2328 2329 s = splsoftnet(); 2330 2331 /* Check vif. */ 2332 if (!legal_vif_num(vifi)) { 2333 splx(s); 2334 return (EADDRNOTAVAIL); 2335 } 2336 2337 /* Check if socket is available. */ 2338 if (viftable[vifi].v_rsvpd != NULL) { 2339 splx(s); 2340 return (EADDRINUSE); 2341 } 2342 2343 viftable[vifi].v_rsvpd = so; 2344 /* 2345 * This may seem silly, but we need to be sure we don't over-increment 2346 * the RSVP counter, in case something slips up. 2347 */ 2348 if (!viftable[vifi].v_rsvp_on) { 2349 viftable[vifi].v_rsvp_on = 1; 2350 rsvp_on++; 2351 } 2352 2353 splx(s); 2354 return (0); 2355 } 2356 2357 int 2358 ip_rsvp_vif_done(struct socket *so, struct mbuf *m) 2359 { 2360 int vifi, s; 2361 2362 if (rsvpdebug) 2363 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2364 so->so_type, so->so_proto->pr_protocol); 2365 2366 if (so->so_type != SOCK_RAW || 2367 so->so_proto->pr_protocol != IPPROTO_RSVP) 2368 return (EOPNOTSUPP); 2369 2370 /* Check mbuf. */ 2371 if (m == NULL || m->m_len != sizeof(int)) { 2372 return (EINVAL); 2373 } 2374 vifi = *(mtod(m, int *)); 2375 2376 s = splsoftnet(); 2377 2378 /* Check vif. */ 2379 if (!legal_vif_num(vifi)) { 2380 splx(s); 2381 return (EADDRNOTAVAIL); 2382 } 2383 2384 if (rsvpdebug) 2385 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 2386 viftable[vifi].v_rsvpd, so); 2387 2388 viftable[vifi].v_rsvpd = NULL; 2389 /* 2390 * This may seem silly, but we need to be sure we don't over-decrement 2391 * the RSVP counter, in case something slips up. 2392 */ 2393 if (viftable[vifi].v_rsvp_on) { 2394 viftable[vifi].v_rsvp_on = 0; 2395 rsvp_on--; 2396 } 2397 2398 splx(s); 2399 return (0); 2400 } 2401 2402 void 2403 ip_rsvp_force_done(struct socket *so) 2404 { 2405 int vifi, s; 2406 2407 /* Don't bother if it is not the right type of socket. */ 2408 if (so->so_type != SOCK_RAW || 2409 so->so_proto->pr_protocol != IPPROTO_RSVP) 2410 return; 2411 2412 s = splsoftnet(); 2413 2414 /* 2415 * The socket may be attached to more than one vif...this 2416 * is perfectly legal. 2417 */ 2418 for (vifi = 0; vifi < numvifs; vifi++) { 2419 if (viftable[vifi].v_rsvpd == so) { 2420 viftable[vifi].v_rsvpd = NULL; 2421 /* 2422 * This may seem silly, but we need to be sure we don't 2423 * over-decrement the RSVP counter, in case something 2424 * slips up. 2425 */ 2426 if (viftable[vifi].v_rsvp_on) { 2427 viftable[vifi].v_rsvp_on = 0; 2428 rsvp_on--; 2429 } 2430 } 2431 } 2432 2433 splx(s); 2434 return; 2435 } 2436 2437 void 2438 rsvp_input(struct mbuf *m, struct ifnet *ifp) 2439 { 2440 int vifi, s; 2441 struct ip *ip = mtod(m, struct ip *); 2442 struct sockaddr_in rsvp_src; 2443 2444 if (rsvpdebug) 2445 printf("rsvp_input: rsvp_on %d\n", rsvp_on); 2446 2447 /* 2448 * Can still get packets with rsvp_on = 0 if there is a local member 2449 * of the group to which the RSVP packet is addressed. But in this 2450 * case we want to throw the packet away. 2451 */ 2452 if (!rsvp_on) { 2453 m_freem(m); 2454 return; 2455 } 2456 2457 /* 2458 * If the old-style non-vif-associated socket is set, then use 2459 * it and ignore the new ones. 2460 */ 2461 if (ip_rsvpd != NULL) { 2462 if (rsvpdebug) 2463 printf("rsvp_input: " 2464 "Sending packet up old-style socket\n"); 2465 rip_input(m); /*XXX*/ 2466 return; 2467 } 2468 2469 s = splsoftnet(); 2470 2471 if (rsvpdebug) 2472 printf("rsvp_input: check vifs\n"); 2473 2474 /* Find which vif the packet arrived on. */ 2475 for (vifi = 0; vifi < numvifs; vifi++) { 2476 if (viftable[vifi].v_ifp == ifp) 2477 break; 2478 } 2479 2480 if (vifi == numvifs) { 2481 /* Can't find vif packet arrived on. Drop packet. */ 2482 if (rsvpdebug) 2483 printf("rsvp_input: " 2484 "Can't find vif for packet...dropping it.\n"); 2485 m_freem(m); 2486 splx(s); 2487 return; 2488 } 2489 2490 if (rsvpdebug) 2491 printf("rsvp_input: check socket\n"); 2492 2493 if (viftable[vifi].v_rsvpd == NULL) { 2494 /* 2495 * drop packet, since there is no specific socket for this 2496 * interface 2497 */ 2498 if (rsvpdebug) 2499 printf("rsvp_input: No socket defined for vif %d\n", 2500 vifi); 2501 m_freem(m); 2502 splx(s); 2503 return; 2504 } 2505 2506 sockaddr_in_init(&rsvp_src, &ip->ip_src, 0); 2507 2508 if (rsvpdebug && m) 2509 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2510 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2511 2512 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2513 if (rsvpdebug) 2514 printf("rsvp_input: Failed to append to socket\n"); 2515 else 2516 if (rsvpdebug) 2517 printf("rsvp_input: send packet up\n"); 2518 2519 splx(s); 2520 } 2521 #endif /* RSVP_ISI */ 2522 2523 /* 2524 * Code for bandwidth monitors 2525 */ 2526 2527 /* 2528 * Define common interface for timeval-related methods 2529 */ 2530 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp) 2531 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp)) 2532 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp)) 2533 2534 static uint32_t 2535 compute_bw_meter_flags(struct bw_upcall *req) 2536 { 2537 uint32_t flags = 0; 2538 2539 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2540 flags |= BW_METER_UNIT_PACKETS; 2541 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2542 flags |= BW_METER_UNIT_BYTES; 2543 if (req->bu_flags & BW_UPCALL_GEQ) 2544 flags |= BW_METER_GEQ; 2545 if (req->bu_flags & BW_UPCALL_LEQ) 2546 flags |= BW_METER_LEQ; 2547 2548 return flags; 2549 } 2550 2551 /* 2552 * Add a bw_meter entry 2553 */ 2554 static int 2555 add_bw_upcall(struct mbuf *m) 2556 { 2557 int s; 2558 struct mfc *mfc; 2559 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2560 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2561 struct timeval now; 2562 struct bw_meter *x; 2563 uint32_t flags; 2564 struct bw_upcall *req; 2565 2566 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2567 return EINVAL; 2568 2569 req = mtod(m, struct bw_upcall *); 2570 2571 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2572 return EOPNOTSUPP; 2573 2574 /* Test if the flags are valid */ 2575 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2576 return EINVAL; 2577 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2578 return EINVAL; 2579 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2580 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2581 return EINVAL; 2582 2583 /* Test if the threshold time interval is valid */ 2584 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2585 return EINVAL; 2586 2587 flags = compute_bw_meter_flags(req); 2588 2589 /* 2590 * Find if we have already same bw_meter entry 2591 */ 2592 s = splsoftnet(); 2593 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2594 if (mfc == NULL) { 2595 splx(s); 2596 return EADDRNOTAVAIL; 2597 } 2598 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2599 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2600 &req->bu_threshold.b_time, ==)) && 2601 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2602 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2603 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2604 splx(s); 2605 return 0; /* XXX Already installed */ 2606 } 2607 } 2608 2609 /* Allocate the new bw_meter entry */ 2610 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2611 if (x == NULL) { 2612 splx(s); 2613 return ENOBUFS; 2614 } 2615 2616 /* Set the new bw_meter entry */ 2617 x->bm_threshold.b_time = req->bu_threshold.b_time; 2618 microtime(&now); 2619 x->bm_start_time = now; 2620 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2621 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2622 x->bm_measured.b_packets = 0; 2623 x->bm_measured.b_bytes = 0; 2624 x->bm_flags = flags; 2625 x->bm_time_next = NULL; 2626 x->bm_time_hash = BW_METER_BUCKETS; 2627 2628 /* Add the new bw_meter entry to the front of entries for this MFC */ 2629 x->bm_mfc = mfc; 2630 x->bm_mfc_next = mfc->mfc_bw_meter; 2631 mfc->mfc_bw_meter = x; 2632 schedule_bw_meter(x, &now); 2633 splx(s); 2634 2635 return 0; 2636 } 2637 2638 static void 2639 free_bw_list(struct bw_meter *list) 2640 { 2641 while (list != NULL) { 2642 struct bw_meter *x = list; 2643 2644 list = list->bm_mfc_next; 2645 unschedule_bw_meter(x); 2646 free(x, M_BWMETER); 2647 } 2648 } 2649 2650 /* 2651 * Delete one or multiple bw_meter entries 2652 */ 2653 static int 2654 del_bw_upcall(struct mbuf *m) 2655 { 2656 int s; 2657 struct mfc *mfc; 2658 struct bw_meter *x; 2659 struct bw_upcall *req; 2660 2661 if (m == NULL || m->m_len < sizeof(struct bw_upcall)) 2662 return EINVAL; 2663 2664 req = mtod(m, struct bw_upcall *); 2665 2666 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2667 return EOPNOTSUPP; 2668 2669 s = splsoftnet(); 2670 /* Find the corresponding MFC entry */ 2671 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2672 if (mfc == NULL) { 2673 splx(s); 2674 return EADDRNOTAVAIL; 2675 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2676 /* 2677 * Delete all bw_meter entries for this mfc 2678 */ 2679 struct bw_meter *list; 2680 2681 list = mfc->mfc_bw_meter; 2682 mfc->mfc_bw_meter = NULL; 2683 free_bw_list(list); 2684 splx(s); 2685 return 0; 2686 } else { /* Delete a single bw_meter entry */ 2687 struct bw_meter *prev; 2688 uint32_t flags = 0; 2689 2690 flags = compute_bw_meter_flags(req); 2691 2692 /* Find the bw_meter entry to delete */ 2693 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2694 prev = x, x = x->bm_mfc_next) { 2695 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2696 &req->bu_threshold.b_time, ==)) && 2697 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2698 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2699 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2700 break; 2701 } 2702 if (x != NULL) { /* Delete entry from the list for this MFC */ 2703 if (prev != NULL) 2704 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2705 else 2706 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2707 2708 unschedule_bw_meter(x); 2709 splx(s); 2710 /* Free the bw_meter entry */ 2711 free(x, M_BWMETER); 2712 return 0; 2713 } else { 2714 splx(s); 2715 return EINVAL; 2716 } 2717 } 2718 /* NOTREACHED */ 2719 } 2720 2721 /* 2722 * Perform bandwidth measurement processing that may result in an upcall 2723 */ 2724 static void 2725 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2726 { 2727 struct timeval delta; 2728 2729 delta = *nowp; 2730 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2731 2732 if (x->bm_flags & BW_METER_GEQ) { 2733 /* 2734 * Processing for ">=" type of bw_meter entry 2735 */ 2736 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2737 /* Reset the bw_meter entry */ 2738 x->bm_start_time = *nowp; 2739 x->bm_measured.b_packets = 0; 2740 x->bm_measured.b_bytes = 0; 2741 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2742 } 2743 2744 /* Record that a packet is received */ 2745 x->bm_measured.b_packets++; 2746 x->bm_measured.b_bytes += plen; 2747 2748 /* 2749 * Test if we should deliver an upcall 2750 */ 2751 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2752 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2753 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2754 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2755 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2756 /* Prepare an upcall for delivery */ 2757 bw_meter_prepare_upcall(x, nowp); 2758 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2759 } 2760 } 2761 } else if (x->bm_flags & BW_METER_LEQ) { 2762 /* 2763 * Processing for "<=" type of bw_meter entry 2764 */ 2765 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2766 /* 2767 * We are behind time with the multicast forwarding table 2768 * scanning for "<=" type of bw_meter entries, so test now 2769 * if we should deliver an upcall. 2770 */ 2771 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2772 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2773 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2774 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2775 /* Prepare an upcall for delivery */ 2776 bw_meter_prepare_upcall(x, nowp); 2777 } 2778 /* Reschedule the bw_meter entry */ 2779 unschedule_bw_meter(x); 2780 schedule_bw_meter(x, nowp); 2781 } 2782 2783 /* Record that a packet is received */ 2784 x->bm_measured.b_packets++; 2785 x->bm_measured.b_bytes += plen; 2786 2787 /* 2788 * Test if we should restart the measuring interval 2789 */ 2790 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2791 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2792 (x->bm_flags & BW_METER_UNIT_BYTES && 2793 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2794 /* Don't restart the measuring interval */ 2795 } else { 2796 /* Do restart the measuring interval */ 2797 /* 2798 * XXX: note that we don't unschedule and schedule, because this 2799 * might be too much overhead per packet. Instead, when we process 2800 * all entries for a given timer hash bin, we check whether it is 2801 * really a timeout. If not, we reschedule at that time. 2802 */ 2803 x->bm_start_time = *nowp; 2804 x->bm_measured.b_packets = 0; 2805 x->bm_measured.b_bytes = 0; 2806 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2807 } 2808 } 2809 } 2810 2811 /* 2812 * Prepare a bandwidth-related upcall 2813 */ 2814 static void 2815 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2816 { 2817 struct timeval delta; 2818 struct bw_upcall *u; 2819 2820 /* 2821 * Compute the measured time interval 2822 */ 2823 delta = *nowp; 2824 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2825 2826 /* 2827 * If there are too many pending upcalls, deliver them now 2828 */ 2829 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2830 bw_upcalls_send(); 2831 2832 /* 2833 * Set the bw_upcall entry 2834 */ 2835 u = &bw_upcalls[bw_upcalls_n++]; 2836 u->bu_src = x->bm_mfc->mfc_origin; 2837 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2838 u->bu_threshold.b_time = x->bm_threshold.b_time; 2839 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2840 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2841 u->bu_measured.b_time = delta; 2842 u->bu_measured.b_packets = x->bm_measured.b_packets; 2843 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2844 u->bu_flags = 0; 2845 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2846 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2847 if (x->bm_flags & BW_METER_UNIT_BYTES) 2848 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2849 if (x->bm_flags & BW_METER_GEQ) 2850 u->bu_flags |= BW_UPCALL_GEQ; 2851 if (x->bm_flags & BW_METER_LEQ) 2852 u->bu_flags |= BW_UPCALL_LEQ; 2853 } 2854 2855 /* 2856 * Send the pending bandwidth-related upcalls 2857 */ 2858 static void 2859 bw_upcalls_send(void) 2860 { 2861 struct mbuf *m; 2862 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2863 struct sockaddr_in k_igmpsrc = { 2864 .sin_len = sizeof(k_igmpsrc), 2865 .sin_family = AF_INET, 2866 }; 2867 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2868 0, /* unused2 */ 2869 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2870 0, /* im_mbz */ 2871 0, /* im_vif */ 2872 0, /* unused3 */ 2873 { 0 }, /* im_src */ 2874 { 0 } }; /* im_dst */ 2875 2876 if (bw_upcalls_n == 0) 2877 return; /* No pending upcalls */ 2878 2879 bw_upcalls_n = 0; 2880 2881 /* 2882 * Allocate a new mbuf, initialize it with the header and 2883 * the payload for the pending calls. 2884 */ 2885 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2886 if (m == NULL) { 2887 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2888 return; 2889 } 2890 2891 m->m_len = m->m_pkthdr.len = 0; 2892 m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg); 2893 m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]); 2894 2895 /* 2896 * Send the upcalls 2897 * XXX do we need to set the address in k_igmpsrc ? 2898 */ 2899 mrtstat.mrts_upcalls++; 2900 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2901 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2902 ++mrtstat.mrts_upq_sockfull; 2903 } 2904 } 2905 2906 /* 2907 * Compute the timeout hash value for the bw_meter entries 2908 */ 2909 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2910 do { \ 2911 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2912 \ 2913 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2914 (hash) = next_timeval.tv_sec; \ 2915 if (next_timeval.tv_usec) \ 2916 (hash)++; /* XXX: make sure we don't timeout early */ \ 2917 (hash) %= BW_METER_BUCKETS; \ 2918 } while (/*CONSTCOND*/ 0) 2919 2920 /* 2921 * Schedule a timer to process periodically bw_meter entry of type "<=" 2922 * by linking the entry in the proper hash bucket. 2923 */ 2924 static void 2925 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2926 { 2927 int time_hash; 2928 2929 if (!(x->bm_flags & BW_METER_LEQ)) 2930 return; /* XXX: we schedule timers only for "<=" entries */ 2931 2932 /* 2933 * Reset the bw_meter entry 2934 */ 2935 x->bm_start_time = *nowp; 2936 x->bm_measured.b_packets = 0; 2937 x->bm_measured.b_bytes = 0; 2938 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2939 2940 /* 2941 * Compute the timeout hash value and insert the entry 2942 */ 2943 BW_METER_TIMEHASH(x, time_hash); 2944 x->bm_time_next = bw_meter_timers[time_hash]; 2945 bw_meter_timers[time_hash] = x; 2946 x->bm_time_hash = time_hash; 2947 } 2948 2949 /* 2950 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2951 * by removing the entry from the proper hash bucket. 2952 */ 2953 static void 2954 unschedule_bw_meter(struct bw_meter *x) 2955 { 2956 int time_hash; 2957 struct bw_meter *prev, *tmp; 2958 2959 if (!(x->bm_flags & BW_METER_LEQ)) 2960 return; /* XXX: we schedule timers only for "<=" entries */ 2961 2962 /* 2963 * Compute the timeout hash value and delete the entry 2964 */ 2965 time_hash = x->bm_time_hash; 2966 if (time_hash >= BW_METER_BUCKETS) 2967 return; /* Entry was not scheduled */ 2968 2969 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2970 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2971 if (tmp == x) 2972 break; 2973 2974 if (tmp == NULL) 2975 panic("unschedule_bw_meter: bw_meter entry not found"); 2976 2977 if (prev != NULL) 2978 prev->bm_time_next = x->bm_time_next; 2979 else 2980 bw_meter_timers[time_hash] = x->bm_time_next; 2981 2982 x->bm_time_next = NULL; 2983 x->bm_time_hash = BW_METER_BUCKETS; 2984 } 2985 2986 /* 2987 * Process all "<=" type of bw_meter that should be processed now, 2988 * and for each entry prepare an upcall if necessary. Each processed 2989 * entry is rescheduled again for the (periodic) processing. 2990 * 2991 * This is run periodically (once per second normally). On each round, 2992 * all the potentially matching entries are in the hash slot that we are 2993 * looking at. 2994 */ 2995 static void 2996 bw_meter_process(void) 2997 { 2998 int s; 2999 static uint32_t last_tv_sec; /* last time we processed this */ 3000 3001 uint32_t loops; 3002 int i; 3003 struct timeval now, process_endtime; 3004 3005 microtime(&now); 3006 if (last_tv_sec == now.tv_sec) 3007 return; /* nothing to do */ 3008 3009 loops = now.tv_sec - last_tv_sec; 3010 last_tv_sec = now.tv_sec; 3011 if (loops > BW_METER_BUCKETS) 3012 loops = BW_METER_BUCKETS; 3013 3014 s = splsoftnet(); 3015 /* 3016 * Process all bins of bw_meter entries from the one after the last 3017 * processed to the current one. On entry, i points to the last bucket 3018 * visited, so we need to increment i at the beginning of the loop. 3019 */ 3020 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 3021 struct bw_meter *x, *tmp_list; 3022 3023 if (++i >= BW_METER_BUCKETS) 3024 i = 0; 3025 3026 /* Disconnect the list of bw_meter entries from the bin */ 3027 tmp_list = bw_meter_timers[i]; 3028 bw_meter_timers[i] = NULL; 3029 3030 /* Process the list of bw_meter entries */ 3031 while (tmp_list != NULL) { 3032 x = tmp_list; 3033 tmp_list = tmp_list->bm_time_next; 3034 3035 /* Test if the time interval is over */ 3036 process_endtime = x->bm_start_time; 3037 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 3038 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 3039 /* Not yet: reschedule, but don't reset */ 3040 int time_hash; 3041 3042 BW_METER_TIMEHASH(x, time_hash); 3043 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 3044 /* 3045 * XXX: somehow the bin processing is a bit ahead of time. 3046 * Put the entry in the next bin. 3047 */ 3048 if (++time_hash >= BW_METER_BUCKETS) 3049 time_hash = 0; 3050 } 3051 x->bm_time_next = bw_meter_timers[time_hash]; 3052 bw_meter_timers[time_hash] = x; 3053 x->bm_time_hash = time_hash; 3054 3055 continue; 3056 } 3057 3058 /* 3059 * Test if we should deliver an upcall 3060 */ 3061 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 3062 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 3063 ((x->bm_flags & BW_METER_UNIT_BYTES) && 3064 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 3065 /* Prepare an upcall for delivery */ 3066 bw_meter_prepare_upcall(x, &now); 3067 } 3068 3069 /* 3070 * Reschedule for next processing 3071 */ 3072 schedule_bw_meter(x, &now); 3073 } 3074 } 3075 3076 /* Send all upcalls that are pending delivery */ 3077 bw_upcalls_send(); 3078 3079 splx(s); 3080 } 3081 3082 /* 3083 * A periodic function for sending all upcalls that are pending delivery 3084 */ 3085 static void 3086 expire_bw_upcalls_send(void *unused) 3087 { 3088 int s; 3089 3090 s = splsoftnet(); 3091 bw_upcalls_send(); 3092 splx(s); 3093 3094 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 3095 expire_bw_upcalls_send, NULL); 3096 } 3097 3098 /* 3099 * A periodic function for periodic scanning of the multicast forwarding 3100 * table for processing all "<=" bw_meter entries. 3101 */ 3102 static void 3103 expire_bw_meter_process(void *unused) 3104 { 3105 if (mrt_api_config & MRT_MFC_BW_UPCALL) 3106 bw_meter_process(); 3107 3108 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 3109 expire_bw_meter_process, NULL); 3110 } 3111 3112 /* 3113 * End of bandwidth monitoring code 3114 */ 3115 3116 #ifdef PIM 3117 /* 3118 * Send the packet up to the user daemon, or eventually do kernel encapsulation 3119 */ 3120 static int 3121 pim_register_send(struct ip *ip, struct vif *vifp, 3122 struct mbuf *m, struct mfc *rt) 3123 { 3124 struct mbuf *mb_copy, *mm; 3125 3126 if (mrtdebug & DEBUG_PIM) 3127 log(LOG_DEBUG, "pim_register_send: "); 3128 3129 mb_copy = pim_register_prepare(ip, m); 3130 if (mb_copy == NULL) 3131 return ENOBUFS; 3132 3133 /* 3134 * Send all the fragments. Note that the mbuf for each fragment 3135 * is freed by the sending machinery. 3136 */ 3137 for (mm = mb_copy; mm; mm = mb_copy) { 3138 mb_copy = mm->m_nextpkt; 3139 mm->m_nextpkt = NULL; 3140 mm = m_pullup(mm, sizeof(struct ip)); 3141 if (mm != NULL) { 3142 ip = mtod(mm, struct ip *); 3143 if ((mrt_api_config & MRT_MFC_RP) && 3144 !in_nullhost(rt->mfc_rp)) { 3145 pim_register_send_rp(ip, vifp, mm, rt); 3146 } else { 3147 pim_register_send_upcall(ip, vifp, mm, rt); 3148 } 3149 } 3150 } 3151 3152 return 0; 3153 } 3154 3155 /* 3156 * Return a copy of the data packet that is ready for PIM Register 3157 * encapsulation. 3158 * XXX: Note that in the returned copy the IP header is a valid one. 3159 */ 3160 static struct mbuf * 3161 pim_register_prepare(struct ip *ip, struct mbuf *m) 3162 { 3163 struct mbuf *mb_copy = NULL; 3164 int mtu; 3165 3166 /* Take care of delayed checksums */ 3167 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 3168 in_delayed_cksum(m); 3169 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 3170 } 3171 3172 /* 3173 * Copy the old packet & pullup its IP header into the 3174 * new mbuf so we can modify it. 3175 */ 3176 mb_copy = m_copypacket(m, M_DONTWAIT); 3177 if (mb_copy == NULL) 3178 return NULL; 3179 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 3180 if (mb_copy == NULL) 3181 return NULL; 3182 3183 /* take care of the TTL */ 3184 ip = mtod(mb_copy, struct ip *); 3185 --ip->ip_ttl; 3186 3187 /* Compute the MTU after the PIM Register encapsulation */ 3188 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 3189 3190 if (ntohs(ip->ip_len) <= mtu) { 3191 /* Turn the IP header into a valid one */ 3192 ip->ip_sum = 0; 3193 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 3194 } else { 3195 /* Fragment the packet */ 3196 if (ip_fragment(mb_copy, NULL, mtu) != 0) { 3197 /* XXX: mb_copy was freed by ip_fragment() */ 3198 return NULL; 3199 } 3200 } 3201 return mb_copy; 3202 } 3203 3204 /* 3205 * Send an upcall with the data packet to the user-level process. 3206 */ 3207 static int 3208 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 3209 struct mbuf *mb_copy, struct mfc *rt) 3210 { 3211 struct mbuf *mb_first; 3212 int len = ntohs(ip->ip_len); 3213 struct igmpmsg *im; 3214 struct sockaddr_in k_igmpsrc = { 3215 .sin_len = sizeof(k_igmpsrc), 3216 .sin_family = AF_INET, 3217 }; 3218 3219 /* 3220 * Add a new mbuf with an upcall header 3221 */ 3222 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3223 if (mb_first == NULL) { 3224 m_freem(mb_copy); 3225 return ENOBUFS; 3226 } 3227 mb_first->m_data += max_linkhdr; 3228 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3229 mb_first->m_len = sizeof(struct igmpmsg); 3230 mb_first->m_next = mb_copy; 3231 3232 /* Send message to routing daemon */ 3233 im = mtod(mb_first, struct igmpmsg *); 3234 im->im_msgtype = IGMPMSG_WHOLEPKT; 3235 im->im_mbz = 0; 3236 im->im_vif = vifp - viftable; 3237 im->im_src = ip->ip_src; 3238 im->im_dst = ip->ip_dst; 3239 3240 k_igmpsrc.sin_addr = ip->ip_src; 3241 3242 mrtstat.mrts_upcalls++; 3243 3244 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3245 if (mrtdebug & DEBUG_PIM) 3246 log(LOG_WARNING, 3247 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3248 ++mrtstat.mrts_upq_sockfull; 3249 return ENOBUFS; 3250 } 3251 3252 /* Keep statistics */ 3253 pimstat.pims_snd_registers_msgs++; 3254 pimstat.pims_snd_registers_bytes += len; 3255 3256 return 0; 3257 } 3258 3259 /* 3260 * Encapsulate the data packet in PIM Register message and send it to the RP. 3261 */ 3262 static int 3263 pim_register_send_rp(struct ip *ip, struct vif *vifp, 3264 struct mbuf *mb_copy, struct mfc *rt) 3265 { 3266 struct mbuf *mb_first; 3267 struct ip *ip_outer; 3268 struct pim_encap_pimhdr *pimhdr; 3269 int len = ntohs(ip->ip_len); 3270 vifi_t vifi = rt->mfc_parent; 3271 3272 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) { 3273 m_freem(mb_copy); 3274 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3275 } 3276 3277 /* 3278 * Add a new mbuf with the encapsulating header 3279 */ 3280 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3281 if (mb_first == NULL) { 3282 m_freem(mb_copy); 3283 return ENOBUFS; 3284 } 3285 mb_first->m_data += max_linkhdr; 3286 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3287 mb_first->m_next = mb_copy; 3288 3289 mb_first->m_pkthdr.len = len + mb_first->m_len; 3290 3291 /* 3292 * Fill in the encapsulating IP and PIM header 3293 */ 3294 ip_outer = mtod(mb_first, struct ip *); 3295 *ip_outer = pim_encap_iphdr; 3296 if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE) 3297 ip_outer->ip_id = 0; 3298 else 3299 ip_outer->ip_id = ip_newid(NULL); 3300 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + 3301 sizeof(pim_encap_pimhdr)); 3302 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3303 ip_outer->ip_dst = rt->mfc_rp; 3304 /* 3305 * Copy the inner header TOS to the outer header, and take care of the 3306 * IP_DF bit. 3307 */ 3308 ip_outer->ip_tos = ip->ip_tos; 3309 if (ntohs(ip->ip_off) & IP_DF) 3310 ip_outer->ip_off |= htons(IP_DF); 3311 pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer 3312 + sizeof(pim_encap_iphdr)); 3313 *pimhdr = pim_encap_pimhdr; 3314 /* If the iif crosses a border, set the Border-bit */ 3315 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3316 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3317 3318 mb_first->m_data += sizeof(pim_encap_iphdr); 3319 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3320 mb_first->m_data -= sizeof(pim_encap_iphdr); 3321 3322 if (vifp->v_rate_limit == 0) 3323 tbf_send_packet(vifp, mb_first); 3324 else 3325 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len)); 3326 3327 /* Keep statistics */ 3328 pimstat.pims_snd_registers_msgs++; 3329 pimstat.pims_snd_registers_bytes += len; 3330 3331 return 0; 3332 } 3333 3334 /* 3335 * PIM-SMv2 and PIM-DM messages processing. 3336 * Receives and verifies the PIM control messages, and passes them 3337 * up to the listening socket, using rip_input(). 3338 * The only message with special processing is the PIM_REGISTER message 3339 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3340 * is passed to if_simloop(). 3341 */ 3342 void 3343 pim_input(struct mbuf *m, ...) 3344 { 3345 struct ip *ip = mtod(m, struct ip *); 3346 struct pim *pim; 3347 int minlen; 3348 int datalen; 3349 int ip_tos; 3350 int proto; 3351 int iphlen; 3352 va_list ap; 3353 3354 va_start(ap, m); 3355 iphlen = va_arg(ap, int); 3356 proto = va_arg(ap, int); 3357 va_end(ap); 3358 3359 datalen = ntohs(ip->ip_len) - iphlen; 3360 3361 /* Keep statistics */ 3362 pimstat.pims_rcv_total_msgs++; 3363 pimstat.pims_rcv_total_bytes += datalen; 3364 3365 /* 3366 * Validate lengths 3367 */ 3368 if (datalen < PIM_MINLEN) { 3369 pimstat.pims_rcv_tooshort++; 3370 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3371 datalen, (u_long)ip->ip_src.s_addr); 3372 m_freem(m); 3373 return; 3374 } 3375 3376 /* 3377 * If the packet is at least as big as a REGISTER, go agead 3378 * and grab the PIM REGISTER header size, to avoid another 3379 * possible m_pullup() later. 3380 * 3381 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3382 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3383 */ 3384 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3385 /* 3386 * Get the IP and PIM headers in contiguous memory, and 3387 * possibly the PIM REGISTER header. 3388 */ 3389 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3390 (m = m_pullup(m, minlen)) == NULL) { 3391 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3392 return; 3393 } 3394 /* m_pullup() may have given us a new mbuf so reset ip. */ 3395 ip = mtod(m, struct ip *); 3396 ip_tos = ip->ip_tos; 3397 3398 /* adjust mbuf to point to the PIM header */ 3399 m->m_data += iphlen; 3400 m->m_len -= iphlen; 3401 pim = mtod(m, struct pim *); 3402 3403 /* 3404 * Validate checksum. If PIM REGISTER, exclude the data packet. 3405 * 3406 * XXX: some older PIMv2 implementations don't make this distinction, 3407 * so for compatibility reason perform the checksum over part of the 3408 * message, and if error, then over the whole message. 3409 */ 3410 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3411 /* do nothing, checksum okay */ 3412 } else if (in_cksum(m, datalen)) { 3413 pimstat.pims_rcv_badsum++; 3414 if (mrtdebug & DEBUG_PIM) 3415 log(LOG_DEBUG, "pim_input: invalid checksum"); 3416 m_freem(m); 3417 return; 3418 } 3419 3420 /* PIM version check */ 3421 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3422 pimstat.pims_rcv_badversion++; 3423 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3424 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3425 m_freem(m); 3426 return; 3427 } 3428 3429 /* restore mbuf back to the outer IP */ 3430 m->m_data -= iphlen; 3431 m->m_len += iphlen; 3432 3433 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3434 /* 3435 * Since this is a REGISTER, we'll make a copy of the register 3436 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3437 * routing daemon. 3438 */ 3439 int s; 3440 struct sockaddr_in dst = { 3441 .sin_len = sizeof(dst), 3442 .sin_family = AF_INET, 3443 }; 3444 struct mbuf *mcp; 3445 struct ip *encap_ip; 3446 u_int32_t *reghdr; 3447 struct ifnet *vifp; 3448 3449 s = splsoftnet(); 3450 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3451 splx(s); 3452 if (mrtdebug & DEBUG_PIM) 3453 log(LOG_DEBUG, 3454 "pim_input: register vif not set: %d\n", reg_vif_num); 3455 m_freem(m); 3456 return; 3457 } 3458 /* XXX need refcnt? */ 3459 vifp = viftable[reg_vif_num].v_ifp; 3460 splx(s); 3461 3462 /* 3463 * Validate length 3464 */ 3465 if (datalen < PIM_REG_MINLEN) { 3466 pimstat.pims_rcv_tooshort++; 3467 pimstat.pims_rcv_badregisters++; 3468 log(LOG_ERR, 3469 "pim_input: register packet size too small %d from %lx\n", 3470 datalen, (u_long)ip->ip_src.s_addr); 3471 m_freem(m); 3472 return; 3473 } 3474 3475 reghdr = (u_int32_t *)(pim + 1); 3476 encap_ip = (struct ip *)(reghdr + 1); 3477 3478 if (mrtdebug & DEBUG_PIM) { 3479 log(LOG_DEBUG, 3480 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3481 (u_long)ntohl(encap_ip->ip_src.s_addr), 3482 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3483 ntohs(encap_ip->ip_len)); 3484 } 3485 3486 /* verify the version number of the inner packet */ 3487 if (encap_ip->ip_v != IPVERSION) { 3488 pimstat.pims_rcv_badregisters++; 3489 if (mrtdebug & DEBUG_PIM) { 3490 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3491 "of the inner packet\n", encap_ip->ip_v); 3492 } 3493 m_freem(m); 3494 return; 3495 } 3496 3497 /* verify the inner packet is destined to a mcast group */ 3498 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) { 3499 pimstat.pims_rcv_badregisters++; 3500 if (mrtdebug & DEBUG_PIM) 3501 log(LOG_DEBUG, 3502 "pim_input: inner packet of register is not " 3503 "multicast %lx\n", 3504 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3505 m_freem(m); 3506 return; 3507 } 3508 3509 /* If a NULL_REGISTER, pass it to the daemon */ 3510 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3511 goto pim_input_to_daemon; 3512 3513 /* 3514 * Copy the TOS from the outer IP header to the inner IP header. 3515 */ 3516 if (encap_ip->ip_tos != ip_tos) { 3517 /* Outer TOS -> inner TOS */ 3518 encap_ip->ip_tos = ip_tos; 3519 /* Recompute the inner header checksum. Sigh... */ 3520 3521 /* adjust mbuf to point to the inner IP header */ 3522 m->m_data += (iphlen + PIM_MINLEN); 3523 m->m_len -= (iphlen + PIM_MINLEN); 3524 3525 encap_ip->ip_sum = 0; 3526 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3527 3528 /* restore mbuf to point back to the outer IP header */ 3529 m->m_data -= (iphlen + PIM_MINLEN); 3530 m->m_len += (iphlen + PIM_MINLEN); 3531 } 3532 3533 /* 3534 * Decapsulate the inner IP packet and loopback to forward it 3535 * as a normal multicast packet. Also, make a copy of the 3536 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3537 * to pass to the daemon later, so it can take the appropriate 3538 * actions (e.g., send back PIM_REGISTER_STOP). 3539 * XXX: here m->m_data points to the outer IP header. 3540 */ 3541 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT); 3542 if (mcp == NULL) { 3543 log(LOG_ERR, 3544 "pim_input: pim register: could not copy register head\n"); 3545 m_freem(m); 3546 return; 3547 } 3548 3549 /* Keep statistics */ 3550 /* XXX: registers_bytes include only the encap. mcast pkt */ 3551 pimstat.pims_rcv_registers_msgs++; 3552 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3553 3554 /* 3555 * forward the inner ip packet; point m_data at the inner ip. 3556 */ 3557 m_adj(m, iphlen + PIM_MINLEN); 3558 3559 if (mrtdebug & DEBUG_PIM) { 3560 log(LOG_DEBUG, 3561 "pim_input: forwarding decapsulated register: " 3562 "src %lx, dst %lx, vif %d\n", 3563 (u_long)ntohl(encap_ip->ip_src.s_addr), 3564 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3565 reg_vif_num); 3566 } 3567 /* NB: vifp was collected above; can it change on us? */ 3568 looutput(vifp, m, (struct sockaddr *)&dst, (struct rtentry *)NULL); 3569 3570 /* prepare the register head to send to the mrouting daemon */ 3571 m = mcp; 3572 } 3573 3574 pim_input_to_daemon: 3575 /* 3576 * Pass the PIM message up to the daemon; if it is a Register message, 3577 * pass the 'head' only up to the daemon. This includes the 3578 * outer IP header, PIM header, PIM-Register header and the 3579 * inner IP header. 3580 * XXX: the outer IP header pkt size of a Register is not adjust to 3581 * reflect the fact that the inner multicast data is truncated. 3582 */ 3583 rip_input(m, iphlen, proto); 3584 3585 return; 3586 } 3587 #endif /* PIM */ 3588
Indexes created Tue Sep 23 07:09:52 GMT 2025