ip_mroute.c revision 1.120
1 /* $NetBSD: ip_mroute.c,v 1.120 2011/08/31 18:31:03 plunky 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 1.120 2011/08/31 18:31:03 plunky 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 #define IP_MULTICASTOPTS 0 150 #define M_PULLUP(m, len) \ 151 do { \ 152 if ((m) && ((m)->m_flags & M_EXT || (m)->m_len < (len))) \ 153 (m) = m_pullup((m), (len)); \ 154 } while (/*CONSTCOND*/ 0) 155 156 /* 157 * Globals. All but ip_mrouter and ip_mrtproto could be static, 158 * except for netstat or debugging purposes. 159 */ 160 struct socket *ip_mrouter = NULL; 161 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 162 163 #define NO_RTE_FOUND 0x1 164 #define RTE_FOUND 0x2 165 166 #define MFCHASH(a, g) \ 167 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ 168 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash) 169 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl; 170 u_long mfchash; 171 172 u_char nexpire[MFCTBLSIZ]; 173 struct vif viftable[MAXVIFS]; 174 struct mrtstat mrtstat; 175 u_int mrtdebug = 0; /* debug level */ 176 #define DEBUG_MFC 0x02 177 #define DEBUG_FORWARD 0x04 178 #define DEBUG_EXPIRE 0x08 179 #define DEBUG_XMIT 0x10 180 #define DEBUG_PIM 0x20 181 182 #define VIFI_INVALID ((vifi_t) -1) 183 184 u_int tbfdebug = 0; /* tbf debug level */ 185 #ifdef RSVP_ISI 186 u_int rsvpdebug = 0; /* rsvp debug level */ 187 extern struct socket *ip_rsvpd; 188 extern int rsvp_on; 189 #endif /* RSVP_ISI */ 190 191 /* vif attachment using sys/netinet/ip_encap.c */ 192 static void vif_input(struct mbuf *, ...); 193 static int vif_encapcheck(struct mbuf *, int, int, void *); 194 195 static const struct protosw vif_protosw = 196 { SOCK_RAW, &inetdomain, IPPROTO_IPV4, PR_ATOMIC|PR_ADDR, 197 vif_input, rip_output, 0, rip_ctloutput, 198 rip_usrreq, 199 0, 0, 0, 0, 200 }; 201 202 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 203 #define UPCALL_EXPIRE 6 /* number of timeouts */ 204 205 /* 206 * Define the token bucket filter structures 207 */ 208 209 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 210 211 static int get_sg_cnt(struct sioc_sg_req *); 212 static int get_vif_cnt(struct sioc_vif_req *); 213 static int ip_mrouter_init(struct socket *, int); 214 static int set_assert(int); 215 static int add_vif(struct vifctl *); 216 static int del_vif(vifi_t *); 217 static void update_mfc_params(struct mfc *, struct mfcctl2 *); 218 static void init_mfc_params(struct mfc *, struct mfcctl2 *); 219 static void expire_mfc(struct mfc *); 220 static int add_mfc(struct sockopt *); 221 #ifdef UPCALL_TIMING 222 static void collate(struct timeval *); 223 #endif 224 static int del_mfc(struct sockopt *); 225 static int set_api_config(struct sockopt *); /* chose API capabilities */ 226 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 227 static void expire_upcalls(void *); 228 #ifdef RSVP_ISI 229 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 230 #else 231 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *); 232 #endif 233 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 234 static void encap_send(struct ip *, struct vif *, struct mbuf *); 235 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_int32_t); 236 static void tbf_queue(struct vif *, struct mbuf *); 237 static void tbf_process_q(struct vif *); 238 static void tbf_reprocess_q(void *); 239 static int tbf_dq_sel(struct vif *, struct ip *); 240 static void tbf_send_packet(struct vif *, struct mbuf *); 241 static void tbf_update_tokens(struct vif *); 242 static int priority(struct vif *, struct ip *); 243 244 /* 245 * Bandwidth monitoring 246 */ 247 static void free_bw_list(struct bw_meter *); 248 static int add_bw_upcall(struct bw_upcall *); 249 static int del_bw_upcall(struct bw_upcall *); 250 static void bw_meter_receive_packet(struct bw_meter *, int , struct timeval *); 251 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *); 252 static void bw_upcalls_send(void); 253 static void schedule_bw_meter(struct bw_meter *, struct timeval *); 254 static void unschedule_bw_meter(struct bw_meter *); 255 static void bw_meter_process(void); 256 static void expire_bw_upcalls_send(void *); 257 static void expire_bw_meter_process(void *); 258 259 #ifdef PIM 260 static int pim_register_send(struct ip *, struct vif *, 261 struct mbuf *, struct mfc *); 262 static int pim_register_send_rp(struct ip *, struct vif *, 263 struct mbuf *, struct mfc *); 264 static int pim_register_send_upcall(struct ip *, struct vif *, 265 struct mbuf *, struct mfc *); 266 static struct mbuf *pim_register_prepare(struct ip *, struct mbuf *); 267 #endif 268 269 /* 270 * 'Interfaces' associated with decapsulator (so we can tell 271 * packets that went through it from ones that get reflected 272 * by a broken gateway). These interfaces are never linked into 273 * the system ifnet list & no routes point to them. I.e., packets 274 * can't be sent this way. They only exist as a placeholder for 275 * multicast source verification. 276 */ 277 #if 0 278 struct ifnet multicast_decap_if[MAXVIFS]; 279 #endif 280 281 #define ENCAP_TTL 64 282 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 283 284 /* prototype IP hdr for encapsulated packets */ 285 struct ip multicast_encap_iphdr = { 286 .ip_hl = sizeof(struct ip) >> 2, 287 .ip_v = IPVERSION, 288 .ip_len = sizeof(struct ip), 289 .ip_ttl = ENCAP_TTL, 290 .ip_p = ENCAP_PROTO, 291 }; 292 293 /* 294 * Bandwidth meter variables and constants 295 */ 296 297 /* 298 * Pending timeouts are stored in a hash table, the key being the 299 * expiration time. Periodically, the entries are analysed and processed. 300 */ 301 #define BW_METER_BUCKETS 1024 302 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 303 struct callout bw_meter_ch; 304 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 305 306 /* 307 * Pending upcalls are stored in a vector which is flushed when 308 * full, or periodically 309 */ 310 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 311 static u_int bw_upcalls_n; /* # of pending upcalls */ 312 struct callout bw_upcalls_ch; 313 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 314 315 #ifdef PIM 316 struct pimstat pimstat; 317 318 /* 319 * Note: the PIM Register encapsulation adds the following in front of a 320 * data packet: 321 * 322 * struct pim_encap_hdr { 323 * struct ip ip; 324 * struct pim_encap_pimhdr pim; 325 * } 326 * 327 */ 328 329 struct pim_encap_pimhdr { 330 struct pim pim; 331 uint32_t flags; 332 }; 333 334 static struct ip pim_encap_iphdr = { 335 .ip_v = IPVERSION, 336 .ip_hl = sizeof(struct ip) >> 2, 337 .ip_len = sizeof(struct ip), 338 .ip_ttl = ENCAP_TTL, 339 .ip_p = IPPROTO_PIM, 340 }; 341 342 static struct pim_encap_pimhdr pim_encap_pimhdr = { 343 { 344 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 345 0, /* reserved */ 346 0, /* checksum */ 347 }, 348 0 /* flags */ 349 }; 350 351 static struct ifnet multicast_register_if; 352 static vifi_t reg_vif_num = VIFI_INVALID; 353 #endif /* PIM */ 354 355 356 /* 357 * Private variables. 358 */ 359 static vifi_t numvifs = 0; 360 361 static struct callout expire_upcalls_ch; 362 363 /* 364 * whether or not special PIM assert processing is enabled. 365 */ 366 static int pim_assert; 367 /* 368 * Rate limit for assert notification messages, in usec 369 */ 370 #define ASSERT_MSG_TIME 3000000 371 372 /* 373 * Kernel multicast routing API capabilities and setup. 374 * If more API capabilities are added to the kernel, they should be 375 * recorded in `mrt_api_support'. 376 */ 377 static const u_int32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 378 MRT_MFC_FLAGS_BORDER_VIF | 379 MRT_MFC_RP | 380 MRT_MFC_BW_UPCALL); 381 static u_int32_t mrt_api_config = 0; 382 383 /* 384 * Find a route for a given origin IP address and Multicast group address 385 * Type of service parameter to be added in the future!!! 386 * Statistics are updated by the caller if needed 387 * (mrtstat.mrts_mfc_lookups and mrtstat.mrts_mfc_misses) 388 */ 389 static struct mfc * 390 mfc_find(struct in_addr *o, struct in_addr *g) 391 { 392 struct mfc *rt; 393 394 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { 395 if (in_hosteq(rt->mfc_origin, *o) && 396 in_hosteq(rt->mfc_mcastgrp, *g) && 397 (rt->mfc_stall == NULL)) 398 break; 399 } 400 401 return (rt); 402 } 403 404 /* 405 * Macros to compute elapsed time efficiently 406 * Borrowed from Van Jacobson's scheduling code 407 */ 408 #define TV_DELTA(a, b, delta) do { \ 409 int xxs; \ 410 delta = (a).tv_usec - (b).tv_usec; \ 411 xxs = (a).tv_sec - (b).tv_sec; \ 412 switch (xxs) { \ 413 case 2: \ 414 delta += 1000000; \ 415 /* fall through */ \ 416 case 1: \ 417 delta += 1000000; \ 418 /* fall through */ \ 419 case 0: \ 420 break; \ 421 default: \ 422 delta += (1000000 * xxs); \ 423 break; \ 424 } \ 425 } while (/*CONSTCOND*/ 0) 426 427 #ifdef UPCALL_TIMING 428 u_int32_t upcall_data[51]; 429 #endif /* UPCALL_TIMING */ 430 431 /* 432 * Handle MRT setsockopt commands to modify the multicast routing tables. 433 */ 434 int 435 ip_mrouter_set(struct socket *so, struct sockopt *sopt) 436 { 437 int error; 438 int optval; 439 struct vifctl vifc; 440 vifi_t vifi; 441 struct bw_upcall bwuc; 442 443 if (sopt->sopt_name != MRT_INIT && so != ip_mrouter) 444 error = ENOPROTOOPT; 445 else { 446 switch (sopt->sopt_name) { 447 case MRT_INIT: 448 error = sockopt_getint(sopt, &optval); 449 if (error) 450 break; 451 452 error = ip_mrouter_init(so, optval); 453 break; 454 case MRT_DONE: 455 error = ip_mrouter_done(); 456 break; 457 case MRT_ADD_VIF: 458 error = sockopt_get(sopt, &vifc, sizeof(vifc)); 459 if (error) 460 break; 461 error = add_vif(&vifc); 462 break; 463 case MRT_DEL_VIF: 464 error = sockopt_get(sopt, &vifi, sizeof(vifi)); 465 if (error) 466 break; 467 error = del_vif(&vifi); 468 break; 469 case MRT_ADD_MFC: 470 error = add_mfc(sopt); 471 break; 472 case MRT_DEL_MFC: 473 error = del_mfc(sopt); 474 break; 475 case MRT_ASSERT: 476 error = sockopt_getint(sopt, &optval); 477 if (error) 478 break; 479 error = set_assert(optval); 480 break; 481 case MRT_API_CONFIG: 482 error = set_api_config(sopt); 483 break; 484 case MRT_ADD_BW_UPCALL: 485 error = sockopt_get(sopt, &bwuc, sizeof(bwuc)); 486 if (error) 487 break; 488 error = add_bw_upcall(&bwuc); 489 break; 490 case MRT_DEL_BW_UPCALL: 491 error = sockopt_get(sopt, &bwuc, sizeof(bwuc)); 492 if (error) 493 break; 494 error = del_bw_upcall(&bwuc); 495 break; 496 default: 497 error = ENOPROTOOPT; 498 break; 499 } 500 } 501 return (error); 502 } 503 504 /* 505 * Handle MRT getsockopt commands 506 */ 507 int 508 ip_mrouter_get(struct socket *so, struct sockopt *sopt) 509 { 510 int error; 511 512 if (so != ip_mrouter) 513 error = ENOPROTOOPT; 514 else { 515 switch (sopt->sopt_name) { 516 case MRT_VERSION: 517 error = sockopt_setint(sopt, 0x0305); /* XXX !!!! */ 518 break; 519 case MRT_ASSERT: 520 error = sockopt_setint(sopt, pim_assert); 521 break; 522 case MRT_API_SUPPORT: 523 error = sockopt_set(sopt, &mrt_api_support, 524 sizeof(mrt_api_support)); 525 break; 526 case MRT_API_CONFIG: 527 error = sockopt_set(sopt, &mrt_api_config, 528 sizeof(mrt_api_config)); 529 break; 530 default: 531 error = ENOPROTOOPT; 532 break; 533 } 534 } 535 return (error); 536 } 537 538 /* 539 * Handle ioctl commands to obtain information from the cache 540 */ 541 int 542 mrt_ioctl(struct socket *so, u_long cmd, void *data) 543 { 544 int error; 545 546 if (so != ip_mrouter) 547 error = EINVAL; 548 else 549 switch (cmd) { 550 case SIOCGETVIFCNT: 551 error = get_vif_cnt((struct sioc_vif_req *)data); 552 break; 553 case SIOCGETSGCNT: 554 error = get_sg_cnt((struct sioc_sg_req *)data); 555 break; 556 default: 557 error = EINVAL; 558 break; 559 } 560 561 return (error); 562 } 563 564 /* 565 * returns the packet, byte, rpf-failure count for the source group provided 566 */ 567 static int 568 get_sg_cnt(struct sioc_sg_req *req) 569 { 570 int s; 571 struct mfc *rt; 572 573 s = splsoftnet(); 574 rt = mfc_find(&req->src, &req->grp); 575 if (rt == NULL) { 576 splx(s); 577 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 578 return (EADDRNOTAVAIL); 579 } 580 req->pktcnt = rt->mfc_pkt_cnt; 581 req->bytecnt = rt->mfc_byte_cnt; 582 req->wrong_if = rt->mfc_wrong_if; 583 splx(s); 584 585 return (0); 586 } 587 588 /* 589 * returns the input and output packet and byte counts on the vif provided 590 */ 591 static int 592 get_vif_cnt(struct sioc_vif_req *req) 593 { 594 vifi_t vifi = req->vifi; 595 596 if (vifi >= numvifs) 597 return (EINVAL); 598 599 req->icount = viftable[vifi].v_pkt_in; 600 req->ocount = viftable[vifi].v_pkt_out; 601 req->ibytes = viftable[vifi].v_bytes_in; 602 req->obytes = viftable[vifi].v_bytes_out; 603 604 return (0); 605 } 606 607 /* 608 * Enable multicast routing 609 */ 610 static int 611 ip_mrouter_init(struct socket *so, int v) 612 { 613 if (mrtdebug) 614 log(LOG_DEBUG, 615 "ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 616 so->so_type, so->so_proto->pr_protocol); 617 618 if (so->so_type != SOCK_RAW || 619 so->so_proto->pr_protocol != IPPROTO_IGMP) 620 return (EOPNOTSUPP); 621 622 if (v != 1) 623 return (EINVAL); 624 625 if (ip_mrouter != NULL) 626 return (EADDRINUSE); 627 628 ip_mrouter = so; 629 630 mfchashtbl = hashinit(MFCTBLSIZ, HASH_LIST, true, &mfchash); 631 memset((void *)nexpire, 0, sizeof(nexpire)); 632 633 pim_assert = 0; 634 635 callout_init(&expire_upcalls_ch, 0); 636 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 637 expire_upcalls, NULL); 638 639 callout_init(&bw_upcalls_ch, 0); 640 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 641 expire_bw_upcalls_send, NULL); 642 643 callout_init(&bw_meter_ch, 0); 644 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 645 expire_bw_meter_process, NULL); 646 647 if (mrtdebug) 648 log(LOG_DEBUG, "ip_mrouter_init\n"); 649 650 return (0); 651 } 652 653 /* 654 * Disable multicast routing 655 */ 656 int 657 ip_mrouter_done(void) 658 { 659 vifi_t vifi; 660 struct vif *vifp; 661 int i; 662 int s; 663 664 s = splsoftnet(); 665 666 /* Clear out all the vifs currently in use. */ 667 for (vifi = 0; vifi < numvifs; vifi++) { 668 vifp = &viftable[vifi]; 669 if (!in_nullhost(vifp->v_lcl_addr)) 670 reset_vif(vifp); 671 } 672 673 numvifs = 0; 674 pim_assert = 0; 675 mrt_api_config = 0; 676 677 callout_stop(&expire_upcalls_ch); 678 callout_stop(&bw_upcalls_ch); 679 callout_stop(&bw_meter_ch); 680 681 /* 682 * Free all multicast forwarding cache entries. 683 */ 684 for (i = 0; i < MFCTBLSIZ; i++) { 685 struct mfc *rt, *nrt; 686 687 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 688 nrt = LIST_NEXT(rt, mfc_hash); 689 690 expire_mfc(rt); 691 } 692 } 693 694 memset((void *)nexpire, 0, sizeof(nexpire)); 695 hashdone(mfchashtbl, HASH_LIST, mfchash); 696 mfchashtbl = NULL; 697 698 bw_upcalls_n = 0; 699 memset(bw_meter_timers, 0, sizeof(bw_meter_timers)); 700 701 /* Reset de-encapsulation cache. */ 702 703 ip_mrouter = NULL; 704 705 splx(s); 706 707 if (mrtdebug) 708 log(LOG_DEBUG, "ip_mrouter_done\n"); 709 710 return (0); 711 } 712 713 void 714 ip_mrouter_detach(struct ifnet *ifp) 715 { 716 int vifi, i; 717 struct vif *vifp; 718 struct mfc *rt; 719 struct rtdetq *rte; 720 721 /* XXX not sure about side effect to userland routing daemon */ 722 for (vifi = 0; vifi < numvifs; vifi++) { 723 vifp = &viftable[vifi]; 724 if (vifp->v_ifp == ifp) 725 reset_vif(vifp); 726 } 727 for (i = 0; i < MFCTBLSIZ; i++) { 728 if (nexpire[i] == 0) 729 continue; 730 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) { 731 for (rte = rt->mfc_stall; rte; rte = rte->next) { 732 if (rte->ifp == ifp) 733 rte->ifp = NULL; 734 } 735 } 736 } 737 } 738 739 /* 740 * Set PIM assert processing global 741 */ 742 static int 743 set_assert(int i) 744 { 745 pim_assert = !!i; 746 return (0); 747 } 748 749 /* 750 * Configure API capabilities 751 */ 752 static int 753 set_api_config(struct sockopt *sopt) 754 { 755 u_int32_t apival; 756 int i, error; 757 758 /* 759 * We can set the API capabilities only if it is the first operation 760 * after MRT_INIT. I.e.: 761 * - there are no vifs installed 762 * - pim_assert is not enabled 763 * - the MFC table is empty 764 */ 765 error = sockopt_get(sopt, &apival, sizeof(apival)); 766 if (error) 767 return (error); 768 if (numvifs > 0) 769 return (EPERM); 770 if (pim_assert) 771 return (EPERM); 772 for (i = 0; i < MFCTBLSIZ; i++) { 773 if (LIST_FIRST(&mfchashtbl[i]) != NULL) 774 return (EPERM); 775 } 776 777 mrt_api_config = apival & mrt_api_support; 778 return (0); 779 } 780 781 /* 782 * Add a vif to the vif table 783 */ 784 static int 785 add_vif(struct vifctl *vifcp) 786 { 787 struct vif *vifp; 788 struct ifaddr *ifa; 789 struct ifnet *ifp; 790 struct ifreq ifr; 791 int error, s; 792 struct sockaddr_in sin; 793 794 if (vifcp->vifc_vifi >= MAXVIFS) 795 return (EINVAL); 796 if (in_nullhost(vifcp->vifc_lcl_addr)) 797 return (EADDRNOTAVAIL); 798 799 vifp = &viftable[vifcp->vifc_vifi]; 800 if (!in_nullhost(vifp->v_lcl_addr)) 801 return (EADDRINUSE); 802 803 /* Find the interface with an address in AF_INET family. */ 804 #ifdef PIM 805 if (vifcp->vifc_flags & VIFF_REGISTER) { 806 /* 807 * XXX: Because VIFF_REGISTER does not really need a valid 808 * local interface (e.g. it could be 127.0.0.2), we don't 809 * check its address. 810 */ 811 ifp = NULL; 812 } else 813 #endif 814 { 815 sockaddr_in_init(&sin, &vifcp->vifc_lcl_addr, 0); 816 ifa = ifa_ifwithaddr(sintosa(&sin)); 817 if (ifa == NULL) 818 return (EADDRNOTAVAIL); 819 ifp = ifa->ifa_ifp; 820 } 821 822 if (vifcp->vifc_flags & VIFF_TUNNEL) { 823 if (vifcp->vifc_flags & VIFF_SRCRT) { 824 log(LOG_ERR, "source routed tunnels not supported\n"); 825 return (EOPNOTSUPP); 826 } 827 828 /* attach this vif to decapsulator dispatch table */ 829 /* 830 * XXX Use addresses in registration so that matching 831 * can be done with radix tree in decapsulator. But, 832 * we need to check inner header for multicast, so 833 * this requires both radix tree lookup and then a 834 * function to check, and this is not supported yet. 835 */ 836 vifp->v_encap_cookie = encap_attach_func(AF_INET, IPPROTO_IPV4, 837 vif_encapcheck, &vif_protosw, vifp); 838 if (!vifp->v_encap_cookie) 839 return (EINVAL); 840 841 /* Create a fake encapsulation interface. */ 842 ifp = malloc(sizeof(*ifp), M_MRTABLE, M_WAITOK|M_ZERO); 843 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 844 "mdecap%d", vifcp->vifc_vifi); 845 846 /* Prepare cached route entry. */ 847 memset(&vifp->v_route, 0, sizeof(vifp->v_route)); 848 #ifdef PIM 849 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 850 ifp = &multicast_register_if; 851 if (mrtdebug) 852 log(LOG_DEBUG, "Adding a register vif, ifp: %p\n", 853 (void *)ifp); 854 if (reg_vif_num == VIFI_INVALID) { 855 memset(ifp, 0, sizeof(*ifp)); 856 snprintf(ifp->if_xname, sizeof(ifp->if_xname), 857 "register_vif"); 858 ifp->if_flags = IFF_LOOPBACK; 859 memset(&vifp->v_route, 0, sizeof(vifp->v_route)); 860 reg_vif_num = vifcp->vifc_vifi; 861 } 862 #endif 863 } else { 864 /* Make sure the interface supports multicast. */ 865 if ((ifp->if_flags & IFF_MULTICAST) == 0) 866 return (EOPNOTSUPP); 867 868 /* Enable promiscuous reception of all IP multicasts. */ 869 sockaddr_in_init(&sin, &zeroin_addr, 0); 870 ifreq_setaddr(SIOCADDMULTI, &ifr, sintosa(&sin)); 871 error = (*ifp->if_ioctl)(ifp, SIOCADDMULTI, &ifr); 872 if (error) 873 return (error); 874 } 875 876 s = splsoftnet(); 877 878 /* Define parameters for the tbf structure. */ 879 vifp->tbf_q = NULL; 880 vifp->tbf_t = &vifp->tbf_q; 881 microtime(&vifp->tbf_last_pkt_t); 882 vifp->tbf_n_tok = 0; 883 vifp->tbf_q_len = 0; 884 vifp->tbf_max_q_len = MAXQSIZE; 885 886 vifp->v_flags = vifcp->vifc_flags; 887 vifp->v_threshold = vifcp->vifc_threshold; 888 /* scaling up here allows division by 1024 in critical code */ 889 vifp->v_rate_limit = vifcp->vifc_rate_limit * 1024 / 1000; 890 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 891 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 892 vifp->v_ifp = ifp; 893 /* Initialize per vif pkt counters. */ 894 vifp->v_pkt_in = 0; 895 vifp->v_pkt_out = 0; 896 vifp->v_bytes_in = 0; 897 vifp->v_bytes_out = 0; 898 899 callout_init(&vifp->v_repq_ch, 0); 900 901 #ifdef RSVP_ISI 902 vifp->v_rsvp_on = 0; 903 vifp->v_rsvpd = NULL; 904 #endif /* RSVP_ISI */ 905 906 splx(s); 907 908 /* Adjust numvifs up if the vifi is higher than numvifs. */ 909 if (numvifs <= vifcp->vifc_vifi) 910 numvifs = vifcp->vifc_vifi + 1; 911 912 if (mrtdebug) 913 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 914 vifcp->vifc_vifi, 915 ntohl(vifcp->vifc_lcl_addr.s_addr), 916 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 917 ntohl(vifcp->vifc_rmt_addr.s_addr), 918 vifcp->vifc_threshold, 919 vifcp->vifc_rate_limit); 920 921 return (0); 922 } 923 924 void 925 reset_vif(struct vif *vifp) 926 { 927 struct mbuf *m, *n; 928 struct ifnet *ifp; 929 struct ifreq ifr; 930 struct sockaddr_in sin; 931 932 callout_stop(&vifp->v_repq_ch); 933 934 /* detach this vif from decapsulator dispatch table */ 935 encap_detach(vifp->v_encap_cookie); 936 vifp->v_encap_cookie = NULL; 937 938 /* 939 * Free packets queued at the interface 940 */ 941 for (m = vifp->tbf_q; m != NULL; m = n) { 942 n = m->m_nextpkt; 943 m_freem(m); 944 } 945 946 if (vifp->v_flags & VIFF_TUNNEL) 947 free(vifp->v_ifp, M_MRTABLE); 948 else if (vifp->v_flags & VIFF_REGISTER) { 949 #ifdef PIM 950 reg_vif_num = VIFI_INVALID; 951 #endif 952 } else { 953 sockaddr_in_init(&sin, &zeroin_addr, 0); 954 ifreq_setaddr(SIOCDELMULTI, &ifr, sintosa(&sin)); 955 ifp = vifp->v_ifp; 956 (*ifp->if_ioctl)(ifp, SIOCDELMULTI, &ifr); 957 } 958 memset((void *)vifp, 0, sizeof(*vifp)); 959 } 960 961 /* 962 * Delete a vif from the vif table 963 */ 964 static int 965 del_vif(vifi_t *vifip) 966 { 967 struct vif *vifp; 968 vifi_t vifi; 969 int s; 970 971 if (*vifip >= numvifs) 972 return (EINVAL); 973 974 vifp = &viftable[*vifip]; 975 if (in_nullhost(vifp->v_lcl_addr)) 976 return (EADDRNOTAVAIL); 977 978 s = splsoftnet(); 979 980 reset_vif(vifp); 981 982 /* Adjust numvifs down */ 983 for (vifi = numvifs; vifi > 0; vifi--) 984 if (!in_nullhost(viftable[vifi - 1].v_lcl_addr)) 985 break; 986 numvifs = vifi; 987 988 splx(s); 989 990 if (mrtdebug) 991 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 992 993 return (0); 994 } 995 996 /* 997 * update an mfc entry without resetting counters and S,G addresses. 998 */ 999 static void 1000 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1001 { 1002 int i; 1003 1004 rt->mfc_parent = mfccp->mfcc_parent; 1005 for (i = 0; i < numvifs; i++) { 1006 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 1007 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 1008 MRT_MFC_FLAGS_ALL; 1009 } 1010 /* set the RP address */ 1011 if (mrt_api_config & MRT_MFC_RP) 1012 rt->mfc_rp = mfccp->mfcc_rp; 1013 else 1014 rt->mfc_rp = zeroin_addr; 1015 } 1016 1017 /* 1018 * fully initialize an mfc entry from the parameter. 1019 */ 1020 static void 1021 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1022 { 1023 rt->mfc_origin = mfccp->mfcc_origin; 1024 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1025 1026 update_mfc_params(rt, mfccp); 1027 1028 /* initialize pkt counters per src-grp */ 1029 rt->mfc_pkt_cnt = 0; 1030 rt->mfc_byte_cnt = 0; 1031 rt->mfc_wrong_if = 0; 1032 timerclear(&rt->mfc_last_assert); 1033 } 1034 1035 static void 1036 expire_mfc(struct mfc *rt) 1037 { 1038 struct rtdetq *rte, *nrte; 1039 1040 free_bw_list(rt->mfc_bw_meter); 1041 1042 for (rte = rt->mfc_stall; rte != NULL; rte = nrte) { 1043 nrte = rte->next; 1044 m_freem(rte->m); 1045 free(rte, M_MRTABLE); 1046 } 1047 1048 LIST_REMOVE(rt, mfc_hash); 1049 free(rt, M_MRTABLE); 1050 } 1051 1052 /* 1053 * Add an mfc entry 1054 */ 1055 static int 1056 add_mfc(struct sockopt *sopt) 1057 { 1058 struct mfcctl2 mfcctl2; 1059 struct mfcctl2 *mfccp; 1060 struct mfc *rt; 1061 u_int32_t hash = 0; 1062 struct rtdetq *rte, *nrte; 1063 u_short nstl; 1064 int s; 1065 int mfcctl_size = sizeof(struct mfcctl); 1066 int error; 1067 1068 if (mrt_api_config & MRT_API_FLAGS_ALL) 1069 mfcctl_size = sizeof(struct mfcctl2); 1070 1071 /* 1072 * select data size depending on API version. 1073 */ 1074 mfccp = &mfcctl2; 1075 memset(&mfcctl2, 0, sizeof(mfcctl2)); 1076 1077 if (mrt_api_config & MRT_API_FLAGS_ALL) 1078 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2)); 1079 else 1080 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl)); 1081 1082 if (error) 1083 return (error); 1084 1085 s = splsoftnet(); 1086 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1087 1088 /* If an entry already exists, just update the fields */ 1089 if (rt) { 1090 if (mrtdebug & DEBUG_MFC) 1091 log(LOG_DEBUG, "add_mfc update o %x g %x p %x\n", 1092 ntohl(mfccp->mfcc_origin.s_addr), 1093 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1094 mfccp->mfcc_parent); 1095 1096 update_mfc_params(rt, mfccp); 1097 1098 splx(s); 1099 return (0); 1100 } 1101 1102 /* 1103 * Find the entry for which the upcall was made and update 1104 */ 1105 nstl = 0; 1106 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1107 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1108 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1109 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1110 rt->mfc_stall != NULL) { 1111 if (nstl++) 1112 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %p\n", 1113 "multiple kernel entries", 1114 ntohl(mfccp->mfcc_origin.s_addr), 1115 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1116 mfccp->mfcc_parent, rt->mfc_stall); 1117 1118 if (mrtdebug & DEBUG_MFC) 1119 log(LOG_DEBUG, "add_mfc o %x g %x p %x dbg %p\n", 1120 ntohl(mfccp->mfcc_origin.s_addr), 1121 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1122 mfccp->mfcc_parent, rt->mfc_stall); 1123 1124 rte = rt->mfc_stall; 1125 init_mfc_params(rt, mfccp); 1126 rt->mfc_stall = NULL; 1127 1128 rt->mfc_expire = 0; /* Don't clean this guy up */ 1129 nexpire[hash]--; 1130 1131 /* free packets Qed at the end of this entry */ 1132 for (; rte != NULL; rte = nrte) { 1133 nrte = rte->next; 1134 if (rte->ifp) { 1135 #ifdef RSVP_ISI 1136 ip_mdq(rte->m, rte->ifp, rt, -1); 1137 #else 1138 ip_mdq(rte->m, rte->ifp, rt); 1139 #endif /* RSVP_ISI */ 1140 } 1141 m_freem(rte->m); 1142 #ifdef UPCALL_TIMING 1143 collate(&rte->t); 1144 #endif /* UPCALL_TIMING */ 1145 free(rte, M_MRTABLE); 1146 } 1147 } 1148 } 1149 1150 /* 1151 * It is possible that an entry is being inserted without an upcall 1152 */ 1153 if (nstl == 0) { 1154 /* 1155 * No mfc; make a new one 1156 */ 1157 if (mrtdebug & DEBUG_MFC) 1158 log(LOG_DEBUG, "add_mfc no upcall o %x g %x p %x\n", 1159 ntohl(mfccp->mfcc_origin.s_addr), 1160 ntohl(mfccp->mfcc_mcastgrp.s_addr), 1161 mfccp->mfcc_parent); 1162 1163 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1164 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1165 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1166 init_mfc_params(rt, mfccp); 1167 if (rt->mfc_expire) 1168 nexpire[hash]--; 1169 rt->mfc_expire = 0; 1170 break; /* XXX */ 1171 } 1172 } 1173 if (rt == NULL) { /* no upcall, so make a new entry */ 1174 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1175 M_NOWAIT); 1176 if (rt == NULL) { 1177 splx(s); 1178 return (ENOBUFS); 1179 } 1180 1181 init_mfc_params(rt, mfccp); 1182 rt->mfc_expire = 0; 1183 rt->mfc_stall = NULL; 1184 rt->mfc_bw_meter = NULL; 1185 1186 /* insert new entry at head of hash chain */ 1187 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1188 } 1189 } 1190 1191 splx(s); 1192 return (0); 1193 } 1194 1195 #ifdef UPCALL_TIMING 1196 /* 1197 * collect delay statistics on the upcalls 1198 */ 1199 static void 1200 collate(struct timeval *t) 1201 { 1202 u_int32_t d; 1203 struct timeval tp; 1204 u_int32_t delta; 1205 1206 microtime(&tp); 1207 1208 if (timercmp(t, &tp, <)) { 1209 TV_DELTA(tp, *t, delta); 1210 1211 d = delta >> 10; 1212 if (d > 50) 1213 d = 50; 1214 1215 ++upcall_data[d]; 1216 } 1217 } 1218 #endif /* UPCALL_TIMING */ 1219 1220 /* 1221 * Delete an mfc entry 1222 */ 1223 static int 1224 del_mfc(struct sockopt *sopt) 1225 { 1226 struct mfcctl2 mfcctl2; 1227 struct mfcctl2 *mfccp; 1228 struct mfc *rt; 1229 int s; 1230 int error; 1231 1232 /* 1233 * XXX: for deleting MFC entries the information in entries 1234 * of size "struct mfcctl" is sufficient. 1235 */ 1236 1237 mfccp = &mfcctl2; 1238 memset(&mfcctl2, 0, sizeof(mfcctl2)); 1239 1240 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl)); 1241 if (error) { 1242 /* Try with the size of mfcctl2. */ 1243 error = sockopt_get(sopt, mfccp, sizeof(struct mfcctl2)); 1244 if (error) 1245 return (error); 1246 } 1247 1248 if (mrtdebug & DEBUG_MFC) 1249 log(LOG_DEBUG, "del_mfc origin %x mcastgrp %x\n", 1250 ntohl(mfccp->mfcc_origin.s_addr), 1251 ntohl(mfccp->mfcc_mcastgrp.s_addr)); 1252 1253 s = splsoftnet(); 1254 1255 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1256 if (rt == NULL) { 1257 splx(s); 1258 return (EADDRNOTAVAIL); 1259 } 1260 1261 /* 1262 * free the bw_meter entries 1263 */ 1264 free_bw_list(rt->mfc_bw_meter); 1265 rt->mfc_bw_meter = NULL; 1266 1267 LIST_REMOVE(rt, mfc_hash); 1268 free(rt, M_MRTABLE); 1269 1270 splx(s); 1271 return (0); 1272 } 1273 1274 static int 1275 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1276 { 1277 if (s) { 1278 if (sbappendaddr(&s->so_rcv, sintosa(src), mm, NULL) != 0) { 1279 sorwakeup(s); 1280 return (0); 1281 } 1282 } 1283 m_freem(mm); 1284 return (-1); 1285 } 1286 1287 /* 1288 * IP multicast forwarding function. This function assumes that the packet 1289 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1290 * pointed to by "ifp", and the packet is to be relayed to other networks 1291 * that have members of the packet's destination IP multicast group. 1292 * 1293 * The packet is returned unscathed to the caller, unless it is 1294 * erroneous, in which case a non-zero return value tells the caller to 1295 * discard it. 1296 */ 1297 1298 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1299 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1300 1301 int 1302 #ifdef RSVP_ISI 1303 ip_mforward(struct mbuf *m, struct ifnet *ifp, struct ip_moptions *imo) 1304 #else 1305 ip_mforward(struct mbuf *m, struct ifnet *ifp) 1306 #endif /* RSVP_ISI */ 1307 { 1308 struct ip *ip = mtod(m, struct ip *); 1309 struct mfc *rt; 1310 static int srctun = 0; 1311 struct mbuf *mm; 1312 struct sockaddr_in sin; 1313 int s; 1314 vifi_t vifi; 1315 1316 if (mrtdebug & DEBUG_FORWARD) 1317 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %p\n", 1318 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1319 1320 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1321 ((u_char *)(ip + 1))[1] != IPOPT_LSRR) { 1322 /* 1323 * Packet arrived via a physical interface or 1324 * an encapsulated tunnel or a register_vif. 1325 */ 1326 } else { 1327 /* 1328 * Packet arrived through a source-route tunnel. 1329 * Source-route tunnels are no longer supported. 1330 */ 1331 if ((srctun++ % 1000) == 0) 1332 log(LOG_ERR, 1333 "ip_mforward: received source-routed packet from %x\n", 1334 ntohl(ip->ip_src.s_addr)); 1335 1336 return (1); 1337 } 1338 1339 /* 1340 * Clear any in-bound checksum flags for this packet. 1341 */ 1342 m->m_pkthdr.csum_flags = 0; 1343 1344 #ifdef RSVP_ISI 1345 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1346 if (ip->ip_ttl < MAXTTL) 1347 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1348 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1349 struct vif *vifp = viftable + vifi; 1350 printf("Sending IPPROTO_RSVP from %x to %x on vif %d (%s%s)\n", 1351 ntohl(ip->ip_src), ntohl(ip->ip_dst), vifi, 1352 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1353 vifp->v_ifp->if_xname); 1354 } 1355 return (ip_mdq(m, ifp, NULL, vifi)); 1356 } 1357 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1358 printf("Warning: IPPROTO_RSVP from %x to %x without vif option\n", 1359 ntohl(ip->ip_src), ntohl(ip->ip_dst)); 1360 } 1361 #endif /* RSVP_ISI */ 1362 1363 /* 1364 * Don't forward a packet with time-to-live of zero or one, 1365 * or a packet destined to a local-only group. 1366 */ 1367 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ip->ip_dst.s_addr)) 1368 return (0); 1369 1370 /* 1371 * Determine forwarding vifs from the forwarding cache table 1372 */ 1373 s = splsoftnet(); 1374 ++mrtstat.mrts_mfc_lookups; 1375 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1376 1377 /* Entry exists, so forward if necessary */ 1378 if (rt != NULL) { 1379 splx(s); 1380 #ifdef RSVP_ISI 1381 return (ip_mdq(m, ifp, rt, -1)); 1382 #else 1383 return (ip_mdq(m, ifp, rt)); 1384 #endif /* RSVP_ISI */ 1385 } else { 1386 /* 1387 * If we don't have a route for packet's origin, 1388 * Make a copy of the packet & send message to routing daemon 1389 */ 1390 1391 struct mbuf *mb0; 1392 struct rtdetq *rte; 1393 u_int32_t hash; 1394 int hlen = ip->ip_hl << 2; 1395 #ifdef UPCALL_TIMING 1396 struct timeval tp; 1397 1398 microtime(&tp); 1399 #endif /* UPCALL_TIMING */ 1400 1401 ++mrtstat.mrts_mfc_misses; 1402 1403 mrtstat.mrts_no_route++; 1404 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1405 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1406 ntohl(ip->ip_src.s_addr), 1407 ntohl(ip->ip_dst.s_addr)); 1408 1409 /* 1410 * Allocate mbufs early so that we don't do extra work if we are 1411 * just going to fail anyway. Make sure to pullup the header so 1412 * that other people can't step on it. 1413 */ 1414 rte = (struct rtdetq *)malloc(sizeof(*rte), M_MRTABLE, 1415 M_NOWAIT); 1416 if (rte == NULL) { 1417 splx(s); 1418 return (ENOBUFS); 1419 } 1420 mb0 = m_copypacket(m, M_DONTWAIT); 1421 M_PULLUP(mb0, hlen); 1422 if (mb0 == NULL) { 1423 free(rte, M_MRTABLE); 1424 splx(s); 1425 return (ENOBUFS); 1426 } 1427 1428 /* is there an upcall waiting for this flow? */ 1429 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1430 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1431 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1432 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1433 rt->mfc_stall != NULL) 1434 break; 1435 } 1436 1437 if (rt == NULL) { 1438 int i; 1439 struct igmpmsg *im; 1440 1441 /* 1442 * Locate the vifi for the incoming interface for 1443 * this packet. 1444 * If none found, drop packet. 1445 */ 1446 for (vifi = 0; vifi < numvifs && 1447 viftable[vifi].v_ifp != ifp; vifi++) 1448 ; 1449 if (vifi >= numvifs) /* vif not found, drop packet */ 1450 goto non_fatal; 1451 1452 /* no upcall, so make a new entry */ 1453 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, 1454 M_NOWAIT); 1455 if (rt == NULL) 1456 goto fail; 1457 1458 /* 1459 * Make a copy of the header to send to the user level 1460 * process 1461 */ 1462 mm = m_copym(m, 0, hlen, M_DONTWAIT); 1463 M_PULLUP(mm, hlen); 1464 if (mm == NULL) 1465 goto fail1; 1466 1467 /* 1468 * Send message to routing daemon to install 1469 * a route into the kernel table 1470 */ 1471 1472 im = mtod(mm, struct igmpmsg *); 1473 im->im_msgtype = IGMPMSG_NOCACHE; 1474 im->im_mbz = 0; 1475 im->im_vif = vifi; 1476 1477 mrtstat.mrts_upcalls++; 1478 1479 sockaddr_in_init(&sin, &ip->ip_src, 0); 1480 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1481 log(LOG_WARNING, 1482 "ip_mforward: ip_mrouter socket queue full\n"); 1483 ++mrtstat.mrts_upq_sockfull; 1484 fail1: 1485 free(rt, M_MRTABLE); 1486 fail: 1487 free(rte, M_MRTABLE); 1488 m_freem(mb0); 1489 splx(s); 1490 return (ENOBUFS); 1491 } 1492 1493 /* insert new entry at head of hash chain */ 1494 rt->mfc_origin = ip->ip_src; 1495 rt->mfc_mcastgrp = ip->ip_dst; 1496 rt->mfc_pkt_cnt = 0; 1497 rt->mfc_byte_cnt = 0; 1498 rt->mfc_wrong_if = 0; 1499 rt->mfc_expire = UPCALL_EXPIRE; 1500 nexpire[hash]++; 1501 for (i = 0; i < numvifs; i++) { 1502 rt->mfc_ttls[i] = 0; 1503 rt->mfc_flags[i] = 0; 1504 } 1505 rt->mfc_parent = -1; 1506 1507 /* clear the RP address */ 1508 rt->mfc_rp = zeroin_addr; 1509 1510 rt->mfc_bw_meter = NULL; 1511 1512 /* link into table */ 1513 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1514 /* Add this entry to the end of the queue */ 1515 rt->mfc_stall = rte; 1516 } else { 1517 /* determine if q has overflowed */ 1518 struct rtdetq **p; 1519 int npkts = 0; 1520 1521 /* 1522 * XXX ouch! we need to append to the list, but we 1523 * only have a pointer to the front, so we have to 1524 * scan the entire list every time. 1525 */ 1526 for (p = &rt->mfc_stall; *p != NULL; p = &(*p)->next) 1527 if (++npkts > MAX_UPQ) { 1528 mrtstat.mrts_upq_ovflw++; 1529 non_fatal: 1530 free(rte, M_MRTABLE); 1531 m_freem(mb0); 1532 splx(s); 1533 return (0); 1534 } 1535 1536 /* Add this entry to the end of the queue */ 1537 *p = rte; 1538 } 1539 1540 rte->next = NULL; 1541 rte->m = mb0; 1542 rte->ifp = ifp; 1543 #ifdef UPCALL_TIMING 1544 rte->t = tp; 1545 #endif /* UPCALL_TIMING */ 1546 1547 splx(s); 1548 1549 return (0); 1550 } 1551 } 1552 1553 1554 /*ARGSUSED*/ 1555 static void 1556 expire_upcalls(void *v) 1557 { 1558 int i; 1559 int s; 1560 1561 s = splsoftnet(); 1562 1563 for (i = 0; i < MFCTBLSIZ; i++) { 1564 struct mfc *rt, *nrt; 1565 1566 if (nexpire[i] == 0) 1567 continue; 1568 1569 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1570 nrt = LIST_NEXT(rt, mfc_hash); 1571 1572 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1573 continue; 1574 nexpire[i]--; 1575 1576 /* 1577 * free the bw_meter entries 1578 */ 1579 while (rt->mfc_bw_meter != NULL) { 1580 struct bw_meter *x = rt->mfc_bw_meter; 1581 1582 rt->mfc_bw_meter = x->bm_mfc_next; 1583 free(x, M_BWMETER); 1584 } 1585 1586 ++mrtstat.mrts_cache_cleanups; 1587 if (mrtdebug & DEBUG_EXPIRE) 1588 log(LOG_DEBUG, 1589 "expire_upcalls: expiring (%x %x)\n", 1590 ntohl(rt->mfc_origin.s_addr), 1591 ntohl(rt->mfc_mcastgrp.s_addr)); 1592 1593 expire_mfc(rt); 1594 } 1595 } 1596 1597 splx(s); 1598 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, 1599 expire_upcalls, NULL); 1600 } 1601 1602 /* 1603 * Packet forwarding routine once entry in the cache is made 1604 */ 1605 static int 1606 #ifdef RSVP_ISI 1607 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1608 #else 1609 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt) 1610 #endif /* RSVP_ISI */ 1611 { 1612 struct ip *ip = mtod(m, struct ip *); 1613 vifi_t vifi; 1614 struct vif *vifp; 1615 struct sockaddr_in sin; 1616 int plen = ntohs(ip->ip_len) - (ip->ip_hl << 2); 1617 1618 /* 1619 * Macro to send packet on vif. Since RSVP packets don't get counted on 1620 * input, they shouldn't get counted on output, so statistics keeping is 1621 * separate. 1622 */ 1623 #define MC_SEND(ip, vifp, m) do { \ 1624 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1625 encap_send((ip), (vifp), (m)); \ 1626 else \ 1627 phyint_send((ip), (vifp), (m)); \ 1628 } while (/*CONSTCOND*/ 0) 1629 1630 #ifdef RSVP_ISI 1631 /* 1632 * If xmt_vif is not -1, send on only the requested vif. 1633 * 1634 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs. 1635 */ 1636 if (xmt_vif < numvifs) { 1637 #ifdef PIM 1638 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1639 pim_register_send(ip, viftable + xmt_vif, m, rt); 1640 else 1641 #endif 1642 MC_SEND(ip, viftable + xmt_vif, m); 1643 return (1); 1644 } 1645 #endif /* RSVP_ISI */ 1646 1647 /* 1648 * Don't forward if it didn't arrive from the parent vif for its origin. 1649 */ 1650 vifi = rt->mfc_parent; 1651 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1652 /* came in the wrong interface */ 1653 if (mrtdebug & DEBUG_FORWARD) 1654 log(LOG_DEBUG, "wrong if: ifp %p vifi %d vififp %p\n", 1655 ifp, vifi, 1656 vifi >= numvifs ? 0 : viftable[vifi].v_ifp); 1657 ++mrtstat.mrts_wrong_if; 1658 ++rt->mfc_wrong_if; 1659 /* 1660 * If we are doing PIM assert processing, send a message 1661 * to the routing daemon. 1662 * 1663 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1664 * can complete the SPT switch, regardless of the type 1665 * of the iif (broadcast media, GRE tunnel, etc). 1666 */ 1667 if (pim_assert && (vifi < numvifs) && viftable[vifi].v_ifp) { 1668 struct timeval now; 1669 u_int32_t delta; 1670 1671 #ifdef PIM 1672 if (ifp == &multicast_register_if) 1673 pimstat.pims_rcv_registers_wrongiif++; 1674 #endif 1675 1676 /* Get vifi for the incoming packet */ 1677 for (vifi = 0; 1678 vifi < numvifs && viftable[vifi].v_ifp != ifp; 1679 vifi++) 1680 ; 1681 if (vifi >= numvifs) { 1682 /* The iif is not found: ignore the packet. */ 1683 return (0); 1684 } 1685 1686 if (rt->mfc_flags[vifi] & 1687 MRT_MFC_FLAGS_DISABLE_WRONGVIF) { 1688 /* WRONGVIF disabled: ignore the packet */ 1689 return (0); 1690 } 1691 1692 microtime(&now); 1693 1694 TV_DELTA(rt->mfc_last_assert, now, delta); 1695 1696 if (delta > ASSERT_MSG_TIME) { 1697 struct igmpmsg *im; 1698 int hlen = ip->ip_hl << 2; 1699 struct mbuf *mm = 1700 m_copym(m, 0, hlen, M_DONTWAIT); 1701 1702 M_PULLUP(mm, hlen); 1703 if (mm == NULL) 1704 return (ENOBUFS); 1705 1706 rt->mfc_last_assert = now; 1707 1708 im = mtod(mm, struct igmpmsg *); 1709 im->im_msgtype = IGMPMSG_WRONGVIF; 1710 im->im_mbz = 0; 1711 im->im_vif = vifi; 1712 1713 mrtstat.mrts_upcalls++; 1714 1715 sockaddr_in_init(&sin, &im->im_src, 0); 1716 if (socket_send(ip_mrouter, mm, &sin) < 0) { 1717 log(LOG_WARNING, 1718 "ip_mforward: ip_mrouter socket queue full\n"); 1719 ++mrtstat.mrts_upq_sockfull; 1720 return (ENOBUFS); 1721 } 1722 } 1723 } 1724 return (0); 1725 } 1726 1727 /* If I sourced this packet, it counts as output, else it was input. */ 1728 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1729 viftable[vifi].v_pkt_out++; 1730 viftable[vifi].v_bytes_out += plen; 1731 } else { 1732 viftable[vifi].v_pkt_in++; 1733 viftable[vifi].v_bytes_in += plen; 1734 } 1735 rt->mfc_pkt_cnt++; 1736 rt->mfc_byte_cnt += plen; 1737 1738 /* 1739 * For each vif, decide if a copy of the packet should be forwarded. 1740 * Forward if: 1741 * - the ttl exceeds the vif's threshold 1742 * - there are group members downstream on interface 1743 */ 1744 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1745 if ((rt->mfc_ttls[vifi] > 0) && 1746 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1747 vifp->v_pkt_out++; 1748 vifp->v_bytes_out += plen; 1749 #ifdef PIM 1750 if (vifp->v_flags & VIFF_REGISTER) 1751 pim_register_send(ip, vifp, m, rt); 1752 else 1753 #endif 1754 MC_SEND(ip, vifp, m); 1755 } 1756 1757 /* 1758 * Perform upcall-related bw measuring. 1759 */ 1760 if (rt->mfc_bw_meter != NULL) { 1761 struct bw_meter *x; 1762 struct timeval now; 1763 1764 microtime(&now); 1765 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1766 bw_meter_receive_packet(x, plen, &now); 1767 } 1768 1769 return (0); 1770 } 1771 1772 #ifdef RSVP_ISI 1773 /* 1774 * check if a vif number is legal/ok. This is used by ip_output. 1775 */ 1776 int 1777 legal_vif_num(int vif) 1778 { 1779 if (vif >= 0 && vif < numvifs) 1780 return (1); 1781 else 1782 return (0); 1783 } 1784 #endif /* RSVP_ISI */ 1785 1786 static void 1787 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1788 { 1789 struct mbuf *mb_copy; 1790 int hlen = ip->ip_hl << 2; 1791 1792 /* 1793 * Make a new reference to the packet; make sure that 1794 * the IP header is actually copied, not just referenced, 1795 * so that ip_output() only scribbles on the copy. 1796 */ 1797 mb_copy = m_copypacket(m, M_DONTWAIT); 1798 M_PULLUP(mb_copy, hlen); 1799 if (mb_copy == NULL) 1800 return; 1801 1802 if (vifp->v_rate_limit <= 0) 1803 tbf_send_packet(vifp, mb_copy); 1804 else 1805 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), 1806 ntohs(ip->ip_len)); 1807 } 1808 1809 static void 1810 encap_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1811 { 1812 struct mbuf *mb_copy; 1813 struct ip *ip_copy; 1814 int i, len = ntohs(ip->ip_len) + sizeof(multicast_encap_iphdr); 1815 1816 /* Take care of delayed checksums */ 1817 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 1818 in_delayed_cksum(m); 1819 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 1820 } 1821 1822 /* 1823 * copy the old packet & pullup it's IP header into the 1824 * new mbuf so we can modify it. Try to fill the new 1825 * mbuf since if we don't the ethernet driver will. 1826 */ 1827 MGETHDR(mb_copy, M_DONTWAIT, MT_DATA); 1828 if (mb_copy == NULL) 1829 return; 1830 mb_copy->m_data += max_linkhdr; 1831 mb_copy->m_pkthdr.len = len; 1832 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1833 1834 if ((mb_copy->m_next = m_copypacket(m, M_DONTWAIT)) == NULL) { 1835 m_freem(mb_copy); 1836 return; 1837 } 1838 i = MHLEN - max_linkhdr; 1839 if (i > len) 1840 i = len; 1841 mb_copy = m_pullup(mb_copy, i); 1842 if (mb_copy == NULL) 1843 return; 1844 1845 /* 1846 * fill in the encapsulating IP header. 1847 */ 1848 ip_copy = mtod(mb_copy, struct ip *); 1849 *ip_copy = multicast_encap_iphdr; 1850 if (len < IP_MINFRAGSIZE) 1851 ip_copy->ip_id = 0; 1852 else 1853 ip_copy->ip_id = ip_newid(NULL); 1854 ip_copy->ip_len = htons(len); 1855 ip_copy->ip_src = vifp->v_lcl_addr; 1856 ip_copy->ip_dst = vifp->v_rmt_addr; 1857 1858 /* 1859 * turn the encapsulated IP header back into a valid one. 1860 */ 1861 ip = (struct ip *)((char *)ip_copy + sizeof(multicast_encap_iphdr)); 1862 --ip->ip_ttl; 1863 ip->ip_sum = 0; 1864 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1865 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1866 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1867 1868 if (vifp->v_rate_limit <= 0) 1869 tbf_send_packet(vifp, mb_copy); 1870 else 1871 tbf_control(vifp, mb_copy, ip, ntohs(ip_copy->ip_len)); 1872 } 1873 1874 /* 1875 * De-encapsulate a packet and feed it back through ip input. 1876 */ 1877 static void 1878 vif_input(struct mbuf *m, ...) 1879 { 1880 int off, proto; 1881 va_list ap; 1882 struct vif *vifp; 1883 int s; 1884 struct ifqueue *ifq; 1885 1886 va_start(ap, m); 1887 off = va_arg(ap, int); 1888 proto = va_arg(ap, int); 1889 va_end(ap); 1890 1891 vifp = (struct vif *)encap_getarg(m); 1892 if (!vifp || proto != ENCAP_PROTO) { 1893 m_freem(m); 1894 mrtstat.mrts_bad_tunnel++; 1895 return; 1896 } 1897 1898 m_adj(m, off); 1899 m->m_pkthdr.rcvif = vifp->v_ifp; 1900 ifq = &ipintrq; 1901 s = splnet(); 1902 if (IF_QFULL(ifq)) { 1903 IF_DROP(ifq); 1904 m_freem(m); 1905 } else { 1906 IF_ENQUEUE(ifq, m); 1907 /* 1908 * normally we would need a "schednetisr(NETISR_IP)" 1909 * here but we were called by ip_input and it is going 1910 * to loop back & try to dequeue the packet we just 1911 * queued as soon as we return so we avoid the 1912 * unnecessary software interrrupt. 1913 */ 1914 } 1915 splx(s); 1916 } 1917 1918 /* 1919 * Check if the packet should be received on the vif denoted by arg. 1920 * (The encap selection code will call this once per vif since each is 1921 * registered separately.) 1922 */ 1923 static int 1924 vif_encapcheck(struct mbuf *m, int off, int proto, void *arg) 1925 { 1926 struct vif *vifp; 1927 struct ip ip; 1928 1929 #ifdef DIAGNOSTIC 1930 if (!arg || proto != IPPROTO_IPV4) 1931 panic("unexpected arg in vif_encapcheck"); 1932 #endif 1933 1934 /* 1935 * Accept the packet only if the inner heaader is multicast 1936 * and the outer header matches a tunnel-mode vif. Order 1937 * checks in the hope that common non-matching packets will be 1938 * rejected quickly. Assume that unicast IPv4 traffic in a 1939 * parallel tunnel (e.g. gif(4)) is unlikely. 1940 */ 1941 1942 /* Obtain the outer IP header and the vif pointer. */ 1943 m_copydata((struct mbuf *)m, 0, sizeof(ip), (void *)&ip); 1944 vifp = (struct vif *)arg; 1945 1946 /* 1947 * The outer source must match the vif's remote peer address. 1948 * For a multicast router with several tunnels, this is the 1949 * only check that will fail on packets in other tunnels, 1950 * assuming the local address is the same. 1951 */ 1952 if (!in_hosteq(vifp->v_rmt_addr, ip.ip_src)) 1953 return 0; 1954 1955 /* The outer destination must match the vif's local address. */ 1956 if (!in_hosteq(vifp->v_lcl_addr, ip.ip_dst)) 1957 return 0; 1958 1959 /* The vif must be of tunnel type. */ 1960 if ((vifp->v_flags & VIFF_TUNNEL) == 0) 1961 return 0; 1962 1963 /* Check that the inner destination is multicast. */ 1964 m_copydata((struct mbuf *)m, off, sizeof(ip), (void *)&ip); 1965 if (!IN_MULTICAST(ip.ip_dst.s_addr)) 1966 return 0; 1967 1968 /* 1969 * We have checked that both the outer src and dst addresses 1970 * match the vif, and that the inner destination is multicast 1971 * (224/5). By claiming more than 64, we intend to 1972 * preferentially take packets that also match a parallel 1973 * gif(4). 1974 */ 1975 return 32 + 32 + 5; 1976 } 1977 1978 /* 1979 * Token bucket filter module 1980 */ 1981 static void 1982 tbf_control(struct vif *vifp, struct mbuf *m, struct ip *ip, u_int32_t len) 1983 { 1984 1985 if (len > MAX_BKT_SIZE) { 1986 /* drop if packet is too large */ 1987 mrtstat.mrts_pkt2large++; 1988 m_freem(m); 1989 return; 1990 } 1991 1992 tbf_update_tokens(vifp); 1993 1994 /* 1995 * If there are enough tokens, and the queue is empty, send this packet 1996 * out immediately. Otherwise, try to insert it on this vif's queue. 1997 */ 1998 if (vifp->tbf_q_len == 0) { 1999 if (len <= vifp->tbf_n_tok) { 2000 vifp->tbf_n_tok -= len; 2001 tbf_send_packet(vifp, m); 2002 } else { 2003 /* queue packet and timeout till later */ 2004 tbf_queue(vifp, m); 2005 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2006 tbf_reprocess_q, vifp); 2007 } 2008 } else { 2009 if (vifp->tbf_q_len >= vifp->tbf_max_q_len && 2010 !tbf_dq_sel(vifp, ip)) { 2011 /* queue full, and couldn't make room */ 2012 mrtstat.mrts_q_overflow++; 2013 m_freem(m); 2014 } else { 2015 /* queue length low enough, or made room */ 2016 tbf_queue(vifp, m); 2017 tbf_process_q(vifp); 2018 } 2019 } 2020 } 2021 2022 /* 2023 * adds a packet to the queue at the interface 2024 */ 2025 static void 2026 tbf_queue(struct vif *vifp, struct mbuf *m) 2027 { 2028 int s = splsoftnet(); 2029 2030 /* insert at tail */ 2031 *vifp->tbf_t = m; 2032 vifp->tbf_t = &m->m_nextpkt; 2033 vifp->tbf_q_len++; 2034 2035 splx(s); 2036 } 2037 2038 2039 /* 2040 * processes the queue at the interface 2041 */ 2042 static void 2043 tbf_process_q(struct vif *vifp) 2044 { 2045 struct mbuf *m; 2046 int len; 2047 int s = splsoftnet(); 2048 2049 /* 2050 * Loop through the queue at the interface and send as many packets 2051 * as possible. 2052 */ 2053 for (m = vifp->tbf_q; m != NULL; m = vifp->tbf_q) { 2054 len = ntohs(mtod(m, struct ip *)->ip_len); 2055 2056 /* determine if the packet can be sent */ 2057 if (len <= vifp->tbf_n_tok) { 2058 /* if so, 2059 * reduce no of tokens, dequeue the packet, 2060 * send the packet. 2061 */ 2062 if ((vifp->tbf_q = m->m_nextpkt) == NULL) 2063 vifp->tbf_t = &vifp->tbf_q; 2064 --vifp->tbf_q_len; 2065 2066 m->m_nextpkt = NULL; 2067 vifp->tbf_n_tok -= len; 2068 tbf_send_packet(vifp, m); 2069 } else 2070 break; 2071 } 2072 splx(s); 2073 } 2074 2075 static void 2076 tbf_reprocess_q(void *arg) 2077 { 2078 struct vif *vifp = arg; 2079 2080 if (ip_mrouter == NULL) 2081 return; 2082 2083 tbf_update_tokens(vifp); 2084 tbf_process_q(vifp); 2085 2086 if (vifp->tbf_q_len != 0) 2087 callout_reset(&vifp->v_repq_ch, TBF_REPROCESS, 2088 tbf_reprocess_q, vifp); 2089 } 2090 2091 /* function that will selectively discard a member of the queue 2092 * based on the precedence value and the priority 2093 */ 2094 static int 2095 tbf_dq_sel(struct vif *vifp, struct ip *ip) 2096 { 2097 u_int p; 2098 struct mbuf **mp, *m; 2099 int s = splsoftnet(); 2100 2101 p = priority(vifp, ip); 2102 2103 for (mp = &vifp->tbf_q, m = *mp; 2104 m != NULL; 2105 mp = &m->m_nextpkt, m = *mp) { 2106 if (p > priority(vifp, mtod(m, struct ip *))) { 2107 if ((*mp = m->m_nextpkt) == NULL) 2108 vifp->tbf_t = mp; 2109 --vifp->tbf_q_len; 2110 2111 m_freem(m); 2112 mrtstat.mrts_drop_sel++; 2113 splx(s); 2114 return (1); 2115 } 2116 } 2117 splx(s); 2118 return (0); 2119 } 2120 2121 static void 2122 tbf_send_packet(struct vif *vifp, struct mbuf *m) 2123 { 2124 int error; 2125 int s = splsoftnet(); 2126 2127 if (vifp->v_flags & VIFF_TUNNEL) { 2128 /* If tunnel options */ 2129 ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, NULL, NULL); 2130 } else { 2131 /* if physical interface option, extract the options and then send */ 2132 struct ip_moptions imo; 2133 2134 imo.imo_multicast_ifp = vifp->v_ifp; 2135 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 2136 imo.imo_multicast_loop = 1; 2137 #ifdef RSVP_ISI 2138 imo.imo_multicast_vif = -1; 2139 #endif 2140 2141 error = ip_output(m, NULL, NULL, IP_FORWARDING|IP_MULTICASTOPTS, 2142 &imo, NULL); 2143 2144 if (mrtdebug & DEBUG_XMIT) 2145 log(LOG_DEBUG, "phyint_send on vif %ld err %d\n", 2146 (long)(vifp - viftable), error); 2147 } 2148 splx(s); 2149 } 2150 2151 /* determine the current time and then 2152 * the elapsed time (between the last time and time now) 2153 * in milliseconds & update the no. of tokens in the bucket 2154 */ 2155 static void 2156 tbf_update_tokens(struct vif *vifp) 2157 { 2158 struct timeval tp; 2159 u_int32_t tm; 2160 int s = splsoftnet(); 2161 2162 microtime(&tp); 2163 2164 TV_DELTA(tp, vifp->tbf_last_pkt_t, tm); 2165 2166 /* 2167 * This formula is actually 2168 * "time in seconds" * "bytes/second". 2169 * 2170 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 2171 * 2172 * The (1000/1024) was introduced in add_vif to optimize 2173 * this divide into a shift. 2174 */ 2175 vifp->tbf_n_tok += tm * vifp->v_rate_limit / 8192; 2176 vifp->tbf_last_pkt_t = tp; 2177 2178 if (vifp->tbf_n_tok > MAX_BKT_SIZE) 2179 vifp->tbf_n_tok = MAX_BKT_SIZE; 2180 2181 splx(s); 2182 } 2183 2184 static int 2185 priority(struct vif *vifp, struct ip *ip) 2186 { 2187 int prio = 50; /* the lowest priority -- default case */ 2188 2189 /* temporary hack; may add general packet classifier some day */ 2190 2191 /* 2192 * The UDP port space is divided up into four priority ranges: 2193 * [0, 16384) : unclassified - lowest priority 2194 * [16384, 32768) : audio - highest priority 2195 * [32768, 49152) : whiteboard - medium priority 2196 * [49152, 65536) : video - low priority 2197 */ 2198 if (ip->ip_p == IPPROTO_UDP) { 2199 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 2200 2201 switch (ntohs(udp->uh_dport) & 0xc000) { 2202 case 0x4000: 2203 prio = 70; 2204 break; 2205 case 0x8000: 2206 prio = 60; 2207 break; 2208 case 0xc000: 2209 prio = 55; 2210 break; 2211 } 2212 2213 if (tbfdebug > 1) 2214 log(LOG_DEBUG, "port %x prio %d\n", 2215 ntohs(udp->uh_dport), prio); 2216 } 2217 2218 return (prio); 2219 } 2220 2221 /* 2222 * End of token bucket filter modifications 2223 */ 2224 #ifdef RSVP_ISI 2225 int 2226 ip_rsvp_vif_init(struct socket *so, struct mbuf *m) 2227 { 2228 int vifi, s; 2229 2230 if (rsvpdebug) 2231 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2232 so->so_type, so->so_proto->pr_protocol); 2233 2234 if (so->so_type != SOCK_RAW || 2235 so->so_proto->pr_protocol != IPPROTO_RSVP) 2236 return (EOPNOTSUPP); 2237 2238 /* Check mbuf. */ 2239 if (m == NULL || m->m_len != sizeof(int)) { 2240 return (EINVAL); 2241 } 2242 vifi = *(mtod(m, int *)); 2243 2244 if (rsvpdebug) 2245 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n", 2246 vifi, rsvp_on); 2247 2248 s = splsoftnet(); 2249 2250 /* Check vif. */ 2251 if (!legal_vif_num(vifi)) { 2252 splx(s); 2253 return (EADDRNOTAVAIL); 2254 } 2255 2256 /* Check if socket is available. */ 2257 if (viftable[vifi].v_rsvpd != NULL) { 2258 splx(s); 2259 return (EADDRINUSE); 2260 } 2261 2262 viftable[vifi].v_rsvpd = so; 2263 /* 2264 * This may seem silly, but we need to be sure we don't over-increment 2265 * the RSVP counter, in case something slips up. 2266 */ 2267 if (!viftable[vifi].v_rsvp_on) { 2268 viftable[vifi].v_rsvp_on = 1; 2269 rsvp_on++; 2270 } 2271 2272 splx(s); 2273 return (0); 2274 } 2275 2276 int 2277 ip_rsvp_vif_done(struct socket *so, struct mbuf *m) 2278 { 2279 int vifi, s; 2280 2281 if (rsvpdebug) 2282 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2283 so->so_type, so->so_proto->pr_protocol); 2284 2285 if (so->so_type != SOCK_RAW || 2286 so->so_proto->pr_protocol != IPPROTO_RSVP) 2287 return (EOPNOTSUPP); 2288 2289 /* Check mbuf. */ 2290 if (m == NULL || m->m_len != sizeof(int)) { 2291 return (EINVAL); 2292 } 2293 vifi = *(mtod(m, int *)); 2294 2295 s = splsoftnet(); 2296 2297 /* Check vif. */ 2298 if (!legal_vif_num(vifi)) { 2299 splx(s); 2300 return (EADDRNOTAVAIL); 2301 } 2302 2303 if (rsvpdebug) 2304 printf("ip_rsvp_vif_done: v_rsvpd = %x so = %x\n", 2305 viftable[vifi].v_rsvpd, so); 2306 2307 viftable[vifi].v_rsvpd = NULL; 2308 /* 2309 * This may seem silly, but we need to be sure we don't over-decrement 2310 * the RSVP counter, in case something slips up. 2311 */ 2312 if (viftable[vifi].v_rsvp_on) { 2313 viftable[vifi].v_rsvp_on = 0; 2314 rsvp_on--; 2315 } 2316 2317 splx(s); 2318 return (0); 2319 } 2320 2321 void 2322 ip_rsvp_force_done(struct socket *so) 2323 { 2324 int vifi, s; 2325 2326 /* Don't bother if it is not the right type of socket. */ 2327 if (so->so_type != SOCK_RAW || 2328 so->so_proto->pr_protocol != IPPROTO_RSVP) 2329 return; 2330 2331 s = splsoftnet(); 2332 2333 /* 2334 * The socket may be attached to more than one vif...this 2335 * is perfectly legal. 2336 */ 2337 for (vifi = 0; vifi < numvifs; vifi++) { 2338 if (viftable[vifi].v_rsvpd == so) { 2339 viftable[vifi].v_rsvpd = NULL; 2340 /* 2341 * This may seem silly, but we need to be sure we don't 2342 * over-decrement the RSVP counter, in case something 2343 * slips up. 2344 */ 2345 if (viftable[vifi].v_rsvp_on) { 2346 viftable[vifi].v_rsvp_on = 0; 2347 rsvp_on--; 2348 } 2349 } 2350 } 2351 2352 splx(s); 2353 return; 2354 } 2355 2356 void 2357 rsvp_input(struct mbuf *m, struct ifnet *ifp) 2358 { 2359 int vifi, s; 2360 struct ip *ip = mtod(m, struct ip *); 2361 struct sockaddr_in rsvp_src; 2362 2363 if (rsvpdebug) 2364 printf("rsvp_input: rsvp_on %d\n", rsvp_on); 2365 2366 /* 2367 * Can still get packets with rsvp_on = 0 if there is a local member 2368 * of the group to which the RSVP packet is addressed. But in this 2369 * case we want to throw the packet away. 2370 */ 2371 if (!rsvp_on) { 2372 m_freem(m); 2373 return; 2374 } 2375 2376 /* 2377 * If the old-style non-vif-associated socket is set, then use 2378 * it and ignore the new ones. 2379 */ 2380 if (ip_rsvpd != NULL) { 2381 if (rsvpdebug) 2382 printf("rsvp_input: " 2383 "Sending packet up old-style socket\n"); 2384 rip_input(m); /*XXX*/ 2385 return; 2386 } 2387 2388 s = splsoftnet(); 2389 2390 if (rsvpdebug) 2391 printf("rsvp_input: check vifs\n"); 2392 2393 /* Find which vif the packet arrived on. */ 2394 for (vifi = 0; vifi < numvifs; vifi++) { 2395 if (viftable[vifi].v_ifp == ifp) 2396 break; 2397 } 2398 2399 if (vifi == numvifs) { 2400 /* Can't find vif packet arrived on. Drop packet. */ 2401 if (rsvpdebug) 2402 printf("rsvp_input: " 2403 "Can't find vif for packet...dropping it.\n"); 2404 m_freem(m); 2405 splx(s); 2406 return; 2407 } 2408 2409 if (rsvpdebug) 2410 printf("rsvp_input: check socket\n"); 2411 2412 if (viftable[vifi].v_rsvpd == NULL) { 2413 /* 2414 * drop packet, since there is no specific socket for this 2415 * interface 2416 */ 2417 if (rsvpdebug) 2418 printf("rsvp_input: No socket defined for vif %d\n", 2419 vifi); 2420 m_freem(m); 2421 splx(s); 2422 return; 2423 } 2424 2425 sockaddr_in_init(&rsvp_src, &ip->ip_src, 0); 2426 2427 if (rsvpdebug && m) 2428 printf("rsvp_input: m->m_len = %d, sbspace() = %d\n", 2429 m->m_len, sbspace(&viftable[vifi].v_rsvpd->so_rcv)); 2430 2431 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2432 if (rsvpdebug) 2433 printf("rsvp_input: Failed to append to socket\n"); 2434 else 2435 if (rsvpdebug) 2436 printf("rsvp_input: send packet up\n"); 2437 2438 splx(s); 2439 } 2440 #endif /* RSVP_ISI */ 2441 2442 /* 2443 * Code for bandwidth monitors 2444 */ 2445 2446 /* 2447 * Define common interface for timeval-related methods 2448 */ 2449 #define BW_TIMEVALCMP(tvp, uvp, cmp) timercmp((tvp), (uvp), cmp) 2450 #define BW_TIMEVALDECR(vvp, uvp) timersub((vvp), (uvp), (vvp)) 2451 #define BW_TIMEVALADD(vvp, uvp) timeradd((vvp), (uvp), (vvp)) 2452 2453 static uint32_t 2454 compute_bw_meter_flags(struct bw_upcall *req) 2455 { 2456 uint32_t flags = 0; 2457 2458 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 2459 flags |= BW_METER_UNIT_PACKETS; 2460 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 2461 flags |= BW_METER_UNIT_BYTES; 2462 if (req->bu_flags & BW_UPCALL_GEQ) 2463 flags |= BW_METER_GEQ; 2464 if (req->bu_flags & BW_UPCALL_LEQ) 2465 flags |= BW_METER_LEQ; 2466 2467 return flags; 2468 } 2469 2470 /* 2471 * Add a bw_meter entry 2472 */ 2473 static int 2474 add_bw_upcall(struct bw_upcall *req) 2475 { 2476 int s; 2477 struct mfc *mfc; 2478 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 2479 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 2480 struct timeval now; 2481 struct bw_meter *x; 2482 uint32_t flags; 2483 2484 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2485 return EOPNOTSUPP; 2486 2487 /* Test if the flags are valid */ 2488 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 2489 return EINVAL; 2490 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 2491 return EINVAL; 2492 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2493 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 2494 return EINVAL; 2495 2496 /* Test if the threshold time interval is valid */ 2497 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 2498 return EINVAL; 2499 2500 flags = compute_bw_meter_flags(req); 2501 2502 /* 2503 * Find if we have already same bw_meter entry 2504 */ 2505 s = splsoftnet(); 2506 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2507 if (mfc == NULL) { 2508 splx(s); 2509 return EADDRNOTAVAIL; 2510 } 2511 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 2512 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2513 &req->bu_threshold.b_time, ==)) && 2514 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2515 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2516 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 2517 splx(s); 2518 return 0; /* XXX Already installed */ 2519 } 2520 } 2521 2522 /* Allocate the new bw_meter entry */ 2523 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 2524 if (x == NULL) { 2525 splx(s); 2526 return ENOBUFS; 2527 } 2528 2529 /* Set the new bw_meter entry */ 2530 x->bm_threshold.b_time = req->bu_threshold.b_time; 2531 microtime(&now); 2532 x->bm_start_time = now; 2533 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 2534 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 2535 x->bm_measured.b_packets = 0; 2536 x->bm_measured.b_bytes = 0; 2537 x->bm_flags = flags; 2538 x->bm_time_next = NULL; 2539 x->bm_time_hash = BW_METER_BUCKETS; 2540 2541 /* Add the new bw_meter entry to the front of entries for this MFC */ 2542 x->bm_mfc = mfc; 2543 x->bm_mfc_next = mfc->mfc_bw_meter; 2544 mfc->mfc_bw_meter = x; 2545 schedule_bw_meter(x, &now); 2546 splx(s); 2547 2548 return 0; 2549 } 2550 2551 static void 2552 free_bw_list(struct bw_meter *list) 2553 { 2554 while (list != NULL) { 2555 struct bw_meter *x = list; 2556 2557 list = list->bm_mfc_next; 2558 unschedule_bw_meter(x); 2559 free(x, M_BWMETER); 2560 } 2561 } 2562 2563 /* 2564 * Delete one or multiple bw_meter entries 2565 */ 2566 static int 2567 del_bw_upcall(struct bw_upcall *req) 2568 { 2569 int s; 2570 struct mfc *mfc; 2571 struct bw_meter *x; 2572 2573 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 2574 return EOPNOTSUPP; 2575 2576 s = splsoftnet(); 2577 /* Find the corresponding MFC entry */ 2578 mfc = mfc_find(&req->bu_src, &req->bu_dst); 2579 if (mfc == NULL) { 2580 splx(s); 2581 return EADDRNOTAVAIL; 2582 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 2583 /* 2584 * Delete all bw_meter entries for this mfc 2585 */ 2586 struct bw_meter *list; 2587 2588 list = mfc->mfc_bw_meter; 2589 mfc->mfc_bw_meter = NULL; 2590 free_bw_list(list); 2591 splx(s); 2592 return 0; 2593 } else { /* Delete a single bw_meter entry */ 2594 struct bw_meter *prev; 2595 uint32_t flags = 0; 2596 2597 flags = compute_bw_meter_flags(req); 2598 2599 /* Find the bw_meter entry to delete */ 2600 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 2601 prev = x, x = x->bm_mfc_next) { 2602 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 2603 &req->bu_threshold.b_time, ==)) && 2604 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 2605 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 2606 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 2607 break; 2608 } 2609 if (x != NULL) { /* Delete entry from the list for this MFC */ 2610 if (prev != NULL) 2611 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 2612 else 2613 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 2614 2615 unschedule_bw_meter(x); 2616 splx(s); 2617 /* Free the bw_meter entry */ 2618 free(x, M_BWMETER); 2619 return 0; 2620 } else { 2621 splx(s); 2622 return EINVAL; 2623 } 2624 } 2625 /* NOTREACHED */ 2626 } 2627 2628 /* 2629 * Perform bandwidth measurement processing that may result in an upcall 2630 */ 2631 static void 2632 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 2633 { 2634 struct timeval delta; 2635 2636 delta = *nowp; 2637 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2638 2639 if (x->bm_flags & BW_METER_GEQ) { 2640 /* 2641 * Processing for ">=" type of bw_meter entry 2642 */ 2643 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2644 /* Reset the bw_meter entry */ 2645 x->bm_start_time = *nowp; 2646 x->bm_measured.b_packets = 0; 2647 x->bm_measured.b_bytes = 0; 2648 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2649 } 2650 2651 /* Record that a packet is received */ 2652 x->bm_measured.b_packets++; 2653 x->bm_measured.b_bytes += plen; 2654 2655 /* 2656 * Test if we should deliver an upcall 2657 */ 2658 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 2659 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2660 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 2661 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2662 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 2663 /* Prepare an upcall for delivery */ 2664 bw_meter_prepare_upcall(x, nowp); 2665 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 2666 } 2667 } 2668 } else if (x->bm_flags & BW_METER_LEQ) { 2669 /* 2670 * Processing for "<=" type of bw_meter entry 2671 */ 2672 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 2673 /* 2674 * We are behind time with the multicast forwarding table 2675 * scanning for "<=" type of bw_meter entries, so test now 2676 * if we should deliver an upcall. 2677 */ 2678 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2679 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2680 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2681 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2682 /* Prepare an upcall for delivery */ 2683 bw_meter_prepare_upcall(x, nowp); 2684 } 2685 /* Reschedule the bw_meter entry */ 2686 unschedule_bw_meter(x); 2687 schedule_bw_meter(x, nowp); 2688 } 2689 2690 /* Record that a packet is received */ 2691 x->bm_measured.b_packets++; 2692 x->bm_measured.b_bytes += plen; 2693 2694 /* 2695 * Test if we should restart the measuring interval 2696 */ 2697 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 2698 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 2699 (x->bm_flags & BW_METER_UNIT_BYTES && 2700 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 2701 /* Don't restart the measuring interval */ 2702 } else { 2703 /* Do restart the measuring interval */ 2704 /* 2705 * XXX: note that we don't unschedule and schedule, because this 2706 * might be too much overhead per packet. Instead, when we process 2707 * all entries for a given timer hash bin, we check whether it is 2708 * really a timeout. If not, we reschedule at that time. 2709 */ 2710 x->bm_start_time = *nowp; 2711 x->bm_measured.b_packets = 0; 2712 x->bm_measured.b_bytes = 0; 2713 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2714 } 2715 } 2716 } 2717 2718 /* 2719 * Prepare a bandwidth-related upcall 2720 */ 2721 static void 2722 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2723 { 2724 struct timeval delta; 2725 struct bw_upcall *u; 2726 2727 /* 2728 * Compute the measured time interval 2729 */ 2730 delta = *nowp; 2731 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2732 2733 /* 2734 * If there are too many pending upcalls, deliver them now 2735 */ 2736 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2737 bw_upcalls_send(); 2738 2739 /* 2740 * Set the bw_upcall entry 2741 */ 2742 u = &bw_upcalls[bw_upcalls_n++]; 2743 u->bu_src = x->bm_mfc->mfc_origin; 2744 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2745 u->bu_threshold.b_time = x->bm_threshold.b_time; 2746 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2747 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2748 u->bu_measured.b_time = delta; 2749 u->bu_measured.b_packets = x->bm_measured.b_packets; 2750 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2751 u->bu_flags = 0; 2752 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2753 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2754 if (x->bm_flags & BW_METER_UNIT_BYTES) 2755 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2756 if (x->bm_flags & BW_METER_GEQ) 2757 u->bu_flags |= BW_UPCALL_GEQ; 2758 if (x->bm_flags & BW_METER_LEQ) 2759 u->bu_flags |= BW_UPCALL_LEQ; 2760 } 2761 2762 /* 2763 * Send the pending bandwidth-related upcalls 2764 */ 2765 static void 2766 bw_upcalls_send(void) 2767 { 2768 struct mbuf *m; 2769 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2770 struct sockaddr_in k_igmpsrc = { 2771 .sin_len = sizeof(k_igmpsrc), 2772 .sin_family = AF_INET, 2773 }; 2774 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2775 0, /* unused2 */ 2776 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2777 0, /* im_mbz */ 2778 0, /* im_vif */ 2779 0, /* unused3 */ 2780 { 0 }, /* im_src */ 2781 { 0 } }; /* im_dst */ 2782 2783 if (bw_upcalls_n == 0) 2784 return; /* No pending upcalls */ 2785 2786 bw_upcalls_n = 0; 2787 2788 /* 2789 * Allocate a new mbuf, initialize it with the header and 2790 * the payload for the pending calls. 2791 */ 2792 MGETHDR(m, M_DONTWAIT, MT_HEADER); 2793 if (m == NULL) { 2794 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2795 return; 2796 } 2797 2798 m->m_len = m->m_pkthdr.len = 0; 2799 m_copyback(m, 0, sizeof(struct igmpmsg), (void *)&igmpmsg); 2800 m_copyback(m, sizeof(struct igmpmsg), len, (void *)&bw_upcalls[0]); 2801 2802 /* 2803 * Send the upcalls 2804 * XXX do we need to set the address in k_igmpsrc ? 2805 */ 2806 mrtstat.mrts_upcalls++; 2807 if (socket_send(ip_mrouter, m, &k_igmpsrc) < 0) { 2808 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2809 ++mrtstat.mrts_upq_sockfull; 2810 } 2811 } 2812 2813 /* 2814 * Compute the timeout hash value for the bw_meter entries 2815 */ 2816 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2817 do { \ 2818 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2819 \ 2820 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2821 (hash) = next_timeval.tv_sec; \ 2822 if (next_timeval.tv_usec) \ 2823 (hash)++; /* XXX: make sure we don't timeout early */ \ 2824 (hash) %= BW_METER_BUCKETS; \ 2825 } while (/*CONSTCOND*/ 0) 2826 2827 /* 2828 * Schedule a timer to process periodically bw_meter entry of type "<=" 2829 * by linking the entry in the proper hash bucket. 2830 */ 2831 static void 2832 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2833 { 2834 int time_hash; 2835 2836 if (!(x->bm_flags & BW_METER_LEQ)) 2837 return; /* XXX: we schedule timers only for "<=" entries */ 2838 2839 /* 2840 * Reset the bw_meter entry 2841 */ 2842 x->bm_start_time = *nowp; 2843 x->bm_measured.b_packets = 0; 2844 x->bm_measured.b_bytes = 0; 2845 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2846 2847 /* 2848 * Compute the timeout hash value and insert the entry 2849 */ 2850 BW_METER_TIMEHASH(x, time_hash); 2851 x->bm_time_next = bw_meter_timers[time_hash]; 2852 bw_meter_timers[time_hash] = x; 2853 x->bm_time_hash = time_hash; 2854 } 2855 2856 /* 2857 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2858 * by removing the entry from the proper hash bucket. 2859 */ 2860 static void 2861 unschedule_bw_meter(struct bw_meter *x) 2862 { 2863 int time_hash; 2864 struct bw_meter *prev, *tmp; 2865 2866 if (!(x->bm_flags & BW_METER_LEQ)) 2867 return; /* XXX: we schedule timers only for "<=" entries */ 2868 2869 /* 2870 * Compute the timeout hash value and delete the entry 2871 */ 2872 time_hash = x->bm_time_hash; 2873 if (time_hash >= BW_METER_BUCKETS) 2874 return; /* Entry was not scheduled */ 2875 2876 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2877 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2878 if (tmp == x) 2879 break; 2880 2881 if (tmp == NULL) 2882 panic("unschedule_bw_meter: bw_meter entry not found"); 2883 2884 if (prev != NULL) 2885 prev->bm_time_next = x->bm_time_next; 2886 else 2887 bw_meter_timers[time_hash] = x->bm_time_next; 2888 2889 x->bm_time_next = NULL; 2890 x->bm_time_hash = BW_METER_BUCKETS; 2891 } 2892 2893 /* 2894 * Process all "<=" type of bw_meter that should be processed now, 2895 * and for each entry prepare an upcall if necessary. Each processed 2896 * entry is rescheduled again for the (periodic) processing. 2897 * 2898 * This is run periodically (once per second normally). On each round, 2899 * all the potentially matching entries are in the hash slot that we are 2900 * looking at. 2901 */ 2902 static void 2903 bw_meter_process(void) 2904 { 2905 int s; 2906 static uint32_t last_tv_sec; /* last time we processed this */ 2907 2908 uint32_t loops; 2909 int i; 2910 struct timeval now, process_endtime; 2911 2912 microtime(&now); 2913 if (last_tv_sec == now.tv_sec) 2914 return; /* nothing to do */ 2915 2916 loops = now.tv_sec - last_tv_sec; 2917 last_tv_sec = now.tv_sec; 2918 if (loops > BW_METER_BUCKETS) 2919 loops = BW_METER_BUCKETS; 2920 2921 s = splsoftnet(); 2922 /* 2923 * Process all bins of bw_meter entries from the one after the last 2924 * processed to the current one. On entry, i points to the last bucket 2925 * visited, so we need to increment i at the beginning of the loop. 2926 */ 2927 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2928 struct bw_meter *x, *tmp_list; 2929 2930 if (++i >= BW_METER_BUCKETS) 2931 i = 0; 2932 2933 /* Disconnect the list of bw_meter entries from the bin */ 2934 tmp_list = bw_meter_timers[i]; 2935 bw_meter_timers[i] = NULL; 2936 2937 /* Process the list of bw_meter entries */ 2938 while (tmp_list != NULL) { 2939 x = tmp_list; 2940 tmp_list = tmp_list->bm_time_next; 2941 2942 /* Test if the time interval is over */ 2943 process_endtime = x->bm_start_time; 2944 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2945 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2946 /* Not yet: reschedule, but don't reset */ 2947 int time_hash; 2948 2949 BW_METER_TIMEHASH(x, time_hash); 2950 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2951 /* 2952 * XXX: somehow the bin processing is a bit ahead of time. 2953 * Put the entry in the next bin. 2954 */ 2955 if (++time_hash >= BW_METER_BUCKETS) 2956 time_hash = 0; 2957 } 2958 x->bm_time_next = bw_meter_timers[time_hash]; 2959 bw_meter_timers[time_hash] = x; 2960 x->bm_time_hash = time_hash; 2961 2962 continue; 2963 } 2964 2965 /* 2966 * Test if we should deliver an upcall 2967 */ 2968 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2969 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2970 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2971 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2972 /* Prepare an upcall for delivery */ 2973 bw_meter_prepare_upcall(x, &now); 2974 } 2975 2976 /* 2977 * Reschedule for next processing 2978 */ 2979 schedule_bw_meter(x, &now); 2980 } 2981 } 2982 2983 /* Send all upcalls that are pending delivery */ 2984 bw_upcalls_send(); 2985 2986 splx(s); 2987 } 2988 2989 /* 2990 * A periodic function for sending all upcalls that are pending delivery 2991 */ 2992 static void 2993 expire_bw_upcalls_send(void *unused) 2994 { 2995 int s; 2996 2997 s = splsoftnet(); 2998 bw_upcalls_send(); 2999 splx(s); 3000 3001 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 3002 expire_bw_upcalls_send, NULL); 3003 } 3004 3005 /* 3006 * A periodic function for periodic scanning of the multicast forwarding 3007 * table for processing all "<=" bw_meter entries. 3008 */ 3009 static void 3010 expire_bw_meter_process(void *unused) 3011 { 3012 if (mrt_api_config & MRT_MFC_BW_UPCALL) 3013 bw_meter_process(); 3014 3015 callout_reset(&bw_meter_ch, BW_METER_PERIOD, 3016 expire_bw_meter_process, NULL); 3017 } 3018 3019 /* 3020 * End of bandwidth monitoring code 3021 */ 3022 3023 #ifdef PIM 3024 /* 3025 * Send the packet up to the user daemon, or eventually do kernel encapsulation 3026 */ 3027 static int 3028 pim_register_send(struct ip *ip, struct vif *vifp, 3029 struct mbuf *m, struct mfc *rt) 3030 { 3031 struct mbuf *mb_copy, *mm; 3032 3033 if (mrtdebug & DEBUG_PIM) 3034 log(LOG_DEBUG, "pim_register_send: "); 3035 3036 mb_copy = pim_register_prepare(ip, m); 3037 if (mb_copy == NULL) 3038 return ENOBUFS; 3039 3040 /* 3041 * Send all the fragments. Note that the mbuf for each fragment 3042 * is freed by the sending machinery. 3043 */ 3044 for (mm = mb_copy; mm; mm = mb_copy) { 3045 mb_copy = mm->m_nextpkt; 3046 mm->m_nextpkt = NULL; 3047 mm = m_pullup(mm, sizeof(struct ip)); 3048 if (mm != NULL) { 3049 ip = mtod(mm, struct ip *); 3050 if ((mrt_api_config & MRT_MFC_RP) && 3051 !in_nullhost(rt->mfc_rp)) { 3052 pim_register_send_rp(ip, vifp, mm, rt); 3053 } else { 3054 pim_register_send_upcall(ip, vifp, mm, rt); 3055 } 3056 } 3057 } 3058 3059 return 0; 3060 } 3061 3062 /* 3063 * Return a copy of the data packet that is ready for PIM Register 3064 * encapsulation. 3065 * XXX: Note that in the returned copy the IP header is a valid one. 3066 */ 3067 static struct mbuf * 3068 pim_register_prepare(struct ip *ip, struct mbuf *m) 3069 { 3070 struct mbuf *mb_copy = NULL; 3071 int mtu; 3072 3073 /* Take care of delayed checksums */ 3074 if (m->m_pkthdr.csum_flags & (M_CSUM_TCPv4|M_CSUM_UDPv4)) { 3075 in_delayed_cksum(m); 3076 m->m_pkthdr.csum_flags &= ~(M_CSUM_TCPv4|M_CSUM_UDPv4); 3077 } 3078 3079 /* 3080 * Copy the old packet & pullup its IP header into the 3081 * new mbuf so we can modify it. 3082 */ 3083 mb_copy = m_copypacket(m, M_DONTWAIT); 3084 if (mb_copy == NULL) 3085 return NULL; 3086 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 3087 if (mb_copy == NULL) 3088 return NULL; 3089 3090 /* take care of the TTL */ 3091 ip = mtod(mb_copy, struct ip *); 3092 --ip->ip_ttl; 3093 3094 /* Compute the MTU after the PIM Register encapsulation */ 3095 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 3096 3097 if (ntohs(ip->ip_len) <= mtu) { 3098 /* Turn the IP header into a valid one */ 3099 ip->ip_sum = 0; 3100 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 3101 } else { 3102 /* Fragment the packet */ 3103 if (ip_fragment(mb_copy, NULL, mtu) != 0) { 3104 /* XXX: mb_copy was freed by ip_fragment() */ 3105 return NULL; 3106 } 3107 } 3108 return mb_copy; 3109 } 3110 3111 /* 3112 * Send an upcall with the data packet to the user-level process. 3113 */ 3114 static int 3115 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 3116 struct mbuf *mb_copy, struct mfc *rt) 3117 { 3118 struct mbuf *mb_first; 3119 int len = ntohs(ip->ip_len); 3120 struct igmpmsg *im; 3121 struct sockaddr_in k_igmpsrc = { 3122 .sin_len = sizeof(k_igmpsrc), 3123 .sin_family = AF_INET, 3124 }; 3125 3126 /* 3127 * Add a new mbuf with an upcall header 3128 */ 3129 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3130 if (mb_first == NULL) { 3131 m_freem(mb_copy); 3132 return ENOBUFS; 3133 } 3134 mb_first->m_data += max_linkhdr; 3135 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 3136 mb_first->m_len = sizeof(struct igmpmsg); 3137 mb_first->m_next = mb_copy; 3138 3139 /* Send message to routing daemon */ 3140 im = mtod(mb_first, struct igmpmsg *); 3141 im->im_msgtype = IGMPMSG_WHOLEPKT; 3142 im->im_mbz = 0; 3143 im->im_vif = vifp - viftable; 3144 im->im_src = ip->ip_src; 3145 im->im_dst = ip->ip_dst; 3146 3147 k_igmpsrc.sin_addr = ip->ip_src; 3148 3149 mrtstat.mrts_upcalls++; 3150 3151 if (socket_send(ip_mrouter, mb_first, &k_igmpsrc) < 0) { 3152 if (mrtdebug & DEBUG_PIM) 3153 log(LOG_WARNING, 3154 "mcast: pim_register_send_upcall: ip_mrouter socket queue full"); 3155 ++mrtstat.mrts_upq_sockfull; 3156 return ENOBUFS; 3157 } 3158 3159 /* Keep statistics */ 3160 pimstat.pims_snd_registers_msgs++; 3161 pimstat.pims_snd_registers_bytes += len; 3162 3163 return 0; 3164 } 3165 3166 /* 3167 * Encapsulate the data packet in PIM Register message and send it to the RP. 3168 */ 3169 static int 3170 pim_register_send_rp(struct ip *ip, struct vif *vifp, 3171 struct mbuf *mb_copy, struct mfc *rt) 3172 { 3173 struct mbuf *mb_first; 3174 struct ip *ip_outer; 3175 struct pim_encap_pimhdr *pimhdr; 3176 int len = ntohs(ip->ip_len); 3177 vifi_t vifi = rt->mfc_parent; 3178 3179 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) { 3180 m_freem(mb_copy); 3181 return EADDRNOTAVAIL; /* The iif vif is invalid */ 3182 } 3183 3184 /* 3185 * Add a new mbuf with the encapsulating header 3186 */ 3187 MGETHDR(mb_first, M_DONTWAIT, MT_HEADER); 3188 if (mb_first == NULL) { 3189 m_freem(mb_copy); 3190 return ENOBUFS; 3191 } 3192 mb_first->m_data += max_linkhdr; 3193 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 3194 mb_first->m_next = mb_copy; 3195 3196 mb_first->m_pkthdr.len = len + mb_first->m_len; 3197 3198 /* 3199 * Fill in the encapsulating IP and PIM header 3200 */ 3201 ip_outer = mtod(mb_first, struct ip *); 3202 *ip_outer = pim_encap_iphdr; 3203 if (mb_first->m_pkthdr.len < IP_MINFRAGSIZE) 3204 ip_outer->ip_id = 0; 3205 else 3206 ip_outer->ip_id = ip_newid(NULL); 3207 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + 3208 sizeof(pim_encap_pimhdr)); 3209 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 3210 ip_outer->ip_dst = rt->mfc_rp; 3211 /* 3212 * Copy the inner header TOS to the outer header, and take care of the 3213 * IP_DF bit. 3214 */ 3215 ip_outer->ip_tos = ip->ip_tos; 3216 if (ntohs(ip->ip_off) & IP_DF) 3217 ip_outer->ip_off |= htons(IP_DF); 3218 pimhdr = (struct pim_encap_pimhdr *)((char *)ip_outer 3219 + sizeof(pim_encap_iphdr)); 3220 *pimhdr = pim_encap_pimhdr; 3221 /* If the iif crosses a border, set the Border-bit */ 3222 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 3223 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 3224 3225 mb_first->m_data += sizeof(pim_encap_iphdr); 3226 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 3227 mb_first->m_data -= sizeof(pim_encap_iphdr); 3228 3229 if (vifp->v_rate_limit == 0) 3230 tbf_send_packet(vifp, mb_first); 3231 else 3232 tbf_control(vifp, mb_first, ip, ntohs(ip_outer->ip_len)); 3233 3234 /* Keep statistics */ 3235 pimstat.pims_snd_registers_msgs++; 3236 pimstat.pims_snd_registers_bytes += len; 3237 3238 return 0; 3239 } 3240 3241 /* 3242 * PIM-SMv2 and PIM-DM messages processing. 3243 * Receives and verifies the PIM control messages, and passes them 3244 * up to the listening socket, using rip_input(). 3245 * The only message with special processing is the PIM_REGISTER message 3246 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 3247 * is passed to if_simloop(). 3248 */ 3249 void 3250 pim_input(struct mbuf *m, ...) 3251 { 3252 struct ip *ip = mtod(m, struct ip *); 3253 struct pim *pim; 3254 int minlen; 3255 int datalen; 3256 int ip_tos; 3257 int proto; 3258 int iphlen; 3259 va_list ap; 3260 3261 va_start(ap, m); 3262 iphlen = va_arg(ap, int); 3263 proto = va_arg(ap, int); 3264 va_end(ap); 3265 3266 datalen = ntohs(ip->ip_len) - iphlen; 3267 3268 /* Keep statistics */ 3269 pimstat.pims_rcv_total_msgs++; 3270 pimstat.pims_rcv_total_bytes += datalen; 3271 3272 /* 3273 * Validate lengths 3274 */ 3275 if (datalen < PIM_MINLEN) { 3276 pimstat.pims_rcv_tooshort++; 3277 log(LOG_ERR, "pim_input: packet size too small %d from %lx\n", 3278 datalen, (u_long)ip->ip_src.s_addr); 3279 m_freem(m); 3280 return; 3281 } 3282 3283 /* 3284 * If the packet is at least as big as a REGISTER, go agead 3285 * and grab the PIM REGISTER header size, to avoid another 3286 * possible m_pullup() later. 3287 * 3288 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 3289 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 3290 */ 3291 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 3292 /* 3293 * Get the IP and PIM headers in contiguous memory, and 3294 * possibly the PIM REGISTER header. 3295 */ 3296 if ((m->m_flags & M_EXT || m->m_len < minlen) && 3297 (m = m_pullup(m, minlen)) == NULL) { 3298 log(LOG_ERR, "pim_input: m_pullup failure\n"); 3299 return; 3300 } 3301 /* m_pullup() may have given us a new mbuf so reset ip. */ 3302 ip = mtod(m, struct ip *); 3303 ip_tos = ip->ip_tos; 3304 3305 /* adjust mbuf to point to the PIM header */ 3306 m->m_data += iphlen; 3307 m->m_len -= iphlen; 3308 pim = mtod(m, struct pim *); 3309 3310 /* 3311 * Validate checksum. If PIM REGISTER, exclude the data packet. 3312 * 3313 * XXX: some older PIMv2 implementations don't make this distinction, 3314 * so for compatibility reason perform the checksum over part of the 3315 * message, and if error, then over the whole message. 3316 */ 3317 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 3318 /* do nothing, checksum okay */ 3319 } else if (in_cksum(m, datalen)) { 3320 pimstat.pims_rcv_badsum++; 3321 if (mrtdebug & DEBUG_PIM) 3322 log(LOG_DEBUG, "pim_input: invalid checksum"); 3323 m_freem(m); 3324 return; 3325 } 3326 3327 /* PIM version check */ 3328 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 3329 pimstat.pims_rcv_badversion++; 3330 log(LOG_ERR, "pim_input: incorrect version %d, expecting %d\n", 3331 PIM_VT_V(pim->pim_vt), PIM_VERSION); 3332 m_freem(m); 3333 return; 3334 } 3335 3336 /* restore mbuf back to the outer IP */ 3337 m->m_data -= iphlen; 3338 m->m_len += iphlen; 3339 3340 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 3341 /* 3342 * Since this is a REGISTER, we'll make a copy of the register 3343 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 3344 * routing daemon. 3345 */ 3346 int s; 3347 struct sockaddr_in dst = { 3348 .sin_len = sizeof(dst), 3349 .sin_family = AF_INET, 3350 }; 3351 struct mbuf *mcp; 3352 struct ip *encap_ip; 3353 u_int32_t *reghdr; 3354 struct ifnet *vifp; 3355 3356 s = splsoftnet(); 3357 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 3358 splx(s); 3359 if (mrtdebug & DEBUG_PIM) 3360 log(LOG_DEBUG, 3361 "pim_input: register vif not set: %d\n", reg_vif_num); 3362 m_freem(m); 3363 return; 3364 } 3365 /* XXX need refcnt? */ 3366 vifp = viftable[reg_vif_num].v_ifp; 3367 splx(s); 3368 3369 /* 3370 * Validate length 3371 */ 3372 if (datalen < PIM_REG_MINLEN) { 3373 pimstat.pims_rcv_tooshort++; 3374 pimstat.pims_rcv_badregisters++; 3375 log(LOG_ERR, 3376 "pim_input: register packet size too small %d from %lx\n", 3377 datalen, (u_long)ip->ip_src.s_addr); 3378 m_freem(m); 3379 return; 3380 } 3381 3382 reghdr = (u_int32_t *)(pim + 1); 3383 encap_ip = (struct ip *)(reghdr + 1); 3384 3385 if (mrtdebug & DEBUG_PIM) { 3386 log(LOG_DEBUG, 3387 "pim_input[register], encap_ip: %lx -> %lx, encap_ip len %d\n", 3388 (u_long)ntohl(encap_ip->ip_src.s_addr), 3389 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3390 ntohs(encap_ip->ip_len)); 3391 } 3392 3393 /* verify the version number of the inner packet */ 3394 if (encap_ip->ip_v != IPVERSION) { 3395 pimstat.pims_rcv_badregisters++; 3396 if (mrtdebug & DEBUG_PIM) { 3397 log(LOG_DEBUG, "pim_input: invalid IP version (%d) " 3398 "of the inner packet\n", encap_ip->ip_v); 3399 } 3400 m_freem(m); 3401 return; 3402 } 3403 3404 /* verify the inner packet is destined to a mcast group */ 3405 if (!IN_MULTICAST(encap_ip->ip_dst.s_addr)) { 3406 pimstat.pims_rcv_badregisters++; 3407 if (mrtdebug & DEBUG_PIM) 3408 log(LOG_DEBUG, 3409 "pim_input: inner packet of register is not " 3410 "multicast %lx\n", 3411 (u_long)ntohl(encap_ip->ip_dst.s_addr)); 3412 m_freem(m); 3413 return; 3414 } 3415 3416 /* If a NULL_REGISTER, pass it to the daemon */ 3417 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 3418 goto pim_input_to_daemon; 3419 3420 /* 3421 * Copy the TOS from the outer IP header to the inner IP header. 3422 */ 3423 if (encap_ip->ip_tos != ip_tos) { 3424 /* Outer TOS -> inner TOS */ 3425 encap_ip->ip_tos = ip_tos; 3426 /* Recompute the inner header checksum. Sigh... */ 3427 3428 /* adjust mbuf to point to the inner IP header */ 3429 m->m_data += (iphlen + PIM_MINLEN); 3430 m->m_len -= (iphlen + PIM_MINLEN); 3431 3432 encap_ip->ip_sum = 0; 3433 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 3434 3435 /* restore mbuf to point back to the outer IP header */ 3436 m->m_data -= (iphlen + PIM_MINLEN); 3437 m->m_len += (iphlen + PIM_MINLEN); 3438 } 3439 3440 /* 3441 * Decapsulate the inner IP packet and loopback to forward it 3442 * as a normal multicast packet. Also, make a copy of the 3443 * outer_iphdr + pimhdr + reghdr + encap_iphdr 3444 * to pass to the daemon later, so it can take the appropriate 3445 * actions (e.g., send back PIM_REGISTER_STOP). 3446 * XXX: here m->m_data points to the outer IP header. 3447 */ 3448 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_DONTWAIT); 3449 if (mcp == NULL) { 3450 log(LOG_ERR, 3451 "pim_input: pim register: could not copy register head\n"); 3452 m_freem(m); 3453 return; 3454 } 3455 3456 /* Keep statistics */ 3457 /* XXX: registers_bytes include only the encap. mcast pkt */ 3458 pimstat.pims_rcv_registers_msgs++; 3459 pimstat.pims_rcv_registers_bytes += ntohs(encap_ip->ip_len); 3460 3461 /* 3462 * forward the inner ip packet; point m_data at the inner ip. 3463 */ 3464 m_adj(m, iphlen + PIM_MINLEN); 3465 3466 if (mrtdebug & DEBUG_PIM) { 3467 log(LOG_DEBUG, 3468 "pim_input: forwarding decapsulated register: " 3469 "src %lx, dst %lx, vif %d\n", 3470 (u_long)ntohl(encap_ip->ip_src.s_addr), 3471 (u_long)ntohl(encap_ip->ip_dst.s_addr), 3472 reg_vif_num); 3473 } 3474 /* NB: vifp was collected above; can it change on us? */ 3475 looutput(vifp, m, (struct sockaddr *)&dst, NULL); 3476 3477 /* prepare the register head to send to the mrouting daemon */ 3478 m = mcp; 3479 } 3480 3481 pim_input_to_daemon: 3482 /* 3483 * Pass the PIM message up to the daemon; if it is a Register message, 3484 * pass the 'head' only up to the daemon. This includes the 3485 * outer IP header, PIM header, PIM-Register header and the 3486 * inner IP header. 3487 * XXX: the outer IP header pkt size of a Register is not adjust to 3488 * reflect the fact that the inner multicast data is truncated. 3489 */ 3490 rip_input(m, iphlen, proto); 3491 3492 return; 3493 } 3494 #endif /* PIM */ 3495
Indexes created Fri Sep 26 20:09:58 GMT 2025