1 /*- 2 * Copyright (c) 1989 Stephen Deering 3 * Copyright (c) 1992, 1993 4 * The Regents of the University of California. All rights reserved. 5 * 6 * This code is derived from software contributed to Berkeley by 7 * Stephen Deering of Stanford University. 8 * 9 * Redistribution and use in source and binary forms, with or without 10 * modification, are permitted provided that the following conditions 11 * are met: 12 * 1. Redistributions of source code must retain the above copyright 13 * notice, this list of conditions and the following disclaimer. 14 * 2. Redistributions in binary form must reproduce the above copyright 15 * notice, this list of conditions and the following disclaimer in the 16 * documentation and/or other materials provided with the distribution. 17 * 4. Neither the name of the University nor the names of its contributors 18 * may be used to endorse or promote products derived from this software 19 * without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 31 * SUCH DAMAGE. 32 * 33 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 34 */ 35 36 /* 37 * IP multicast forwarding procedures 38 * 39 * Written by David Waitzman, BBN Labs, August 1988. 40 * Modified by Steve Deering, Stanford, February 1989. 41 * Modified by Mark J. Steiglitz, Stanford, May, 1991 42 * Modified by Van Jacobson, LBL, January 1993 43 * Modified by Ajit Thyagarajan, PARC, August 1993 44 * Modified by Bill Fenner, PARC, April 1995 45 * Modified by Ahmed Helmy, SGI, June 1996 46 * Modified by George Edmond Eddy (Rusty), ISI, February 1998 47 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000 48 * Modified by Hitoshi Asaeda, WIDE, August 2000 49 * Modified by Pavlin Radoslavov, ICSI, October 2002 50 * 51 * MROUTING Revision: 3.5 52 * and PIM-SMv2 and PIM-DM support, advanced API support, 53 * bandwidth metering and signaling 54 */ 55 56 /* 57 * TODO: Prefix functions with ipmf_. 58 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol 59 * domain attachment (if_afdata) so we can track consumers of that service. 60 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT, 61 * move it to socket options. 62 * TODO: Cleanup LSRR removal further. 63 * TODO: Push RSVP stubs into raw_ip.c. 64 * TODO: Use bitstring.h for vif set. 65 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded. 66 * TODO: Sync ip6_mroute.c with this file. 67 */ 68 69 #include <sys/cdefs.h> 70 __FBSDID("$FreeBSD$"); 71 72 #include "opt_inet.h" 73 #include "opt_mrouting.h" 74 75 #define _PIM_VT 1 76 77 #include <sys/param.h> 78 #include <sys/kernel.h> 79 #include <sys/stddef.h> 80 #include <sys/lock.h> 81 #include <sys/ktr.h> 82 #include <sys/malloc.h> 83 #include <sys/mbuf.h> 84 #include <sys/module.h> 85 #include <sys/priv.h> 86 #include <sys/protosw.h> 87 #include <sys/signalvar.h> 88 #include <sys/socket.h> 89 #include <sys/socketvar.h> 90 #include <sys/sockio.h> 91 #include <sys/sx.h> 92 #include <sys/sysctl.h> 93 #include <sys/syslog.h> 94 #include <sys/systm.h> 95 #include <sys/time.h> 96 97 #include <net/if.h> 98 #include <net/netisr.h> 99 #include <net/route.h> 100 #include <net/vnet.h> 101 102 #include <netinet/in.h> 103 #include <netinet/igmp.h> 104 #include <netinet/in_systm.h> 105 #include <netinet/in_var.h> 106 #include <netinet/ip.h> 107 #include <netinet/ip_encap.h> 108 #include <netinet/ip_mroute.h> 109 #include <netinet/ip_var.h> 110 #include <netinet/ip_options.h> 111 #include <netinet/pim.h> 112 #include <netinet/pim_var.h> 113 #include <netinet/udp.h> 114 115 #include <machine/in_cksum.h> 116 117 #include <security/mac/mac_framework.h> 118 119 #ifndef KTR_IPMF 120 #define KTR_IPMF KTR_INET 121 #endif 122 123 #define VIFI_INVALID ((vifi_t) -1) 124 #define M_HASCL(m) ((m)->m_flags & M_EXT) 125 126 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache"); 127 128 /* 129 * Locking. We use two locks: one for the virtual interface table and 130 * one for the forwarding table. These locks may be nested in which case 131 * the VIF lock must always be taken first. Note that each lock is used 132 * to cover not only the specific data structure but also related data 133 * structures. 134 */ 135 136 static struct mtx mrouter_mtx; 137 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx) 138 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx) 139 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED) 140 #define MROUTER_LOCK_INIT() \ 141 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF) 142 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx) 143 144 static struct mrtstat mrtstat; 145 SYSCTL_STRUCT(_net_inet_ip, OID_AUTO, mrtstat, CTLFLAG_RW, 146 &mrtstat, mrtstat, 147 "IPv4 Multicast Forwarding Statistics (struct mrtstat, " 148 "netinet/ip_mroute.h)"); 149 150 static u_long mfchash; 151 #define MFCHASH(a, g) \ 152 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ 153 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & mfchash) 154 #define MFCHASHSIZE 256 155 156 static u_char *nexpire; /* 0..mfchashsize-1 */ 157 static u_long mfchashsize; /* Hash size */ 158 LIST_HEAD(mfchashhdr, mfc) *mfchashtbl; 159 160 static struct mtx mfc_mtx; 161 #define MFC_LOCK() mtx_lock(&mfc_mtx) 162 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 163 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED) 164 #define MFC_LOCK_INIT() \ 165 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF) 166 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 167 168 static vifi_t numvifs; 169 static struct vif viftable[MAXVIFS]; 170 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_RD, 171 &viftable, sizeof(viftable), "S,vif[MAXVIFS]", 172 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 173 174 static struct mtx vif_mtx; 175 #define VIF_LOCK() mtx_lock(&vif_mtx) 176 #define VIF_UNLOCK() mtx_unlock(&vif_mtx) 177 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 178 #define VIF_LOCK_INIT() \ 179 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF) 180 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 181 182 static eventhandler_tag if_detach_event_tag = NULL; 183 184 static struct callout expire_upcalls_ch; 185 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 186 #define UPCALL_EXPIRE 6 /* number of timeouts */ 187 188 /* 189 * Bandwidth meter variables and constants 190 */ 191 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 192 /* 193 * Pending timeouts are stored in a hash table, the key being the 194 * expiration time. Periodically, the entries are analysed and processed. 195 */ 196 #define BW_METER_BUCKETS 1024 197 static struct bw_meter *bw_meter_timers[BW_METER_BUCKETS]; 198 static struct callout bw_meter_ch; 199 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 200 201 /* 202 * Pending upcalls are stored in a vector which is flushed when 203 * full, or periodically 204 */ 205 static struct bw_upcall bw_upcalls[BW_UPCALLS_MAX]; 206 static u_int bw_upcalls_n; /* # of pending upcalls */ 207 static struct callout bw_upcalls_ch; 208 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 209 210 static struct pimstat pimstat; 211 212 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); 213 SYSCTL_STRUCT(_net_inet_pim, PIMCTL_STATS, stats, CTLFLAG_RD, 214 &pimstat, pimstat, 215 "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 216 217 static u_long pim_squelch_wholepkt = 0; 218 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, 219 &pim_squelch_wholepkt, 0, 220 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); 221 222 extern struct domain inetdomain; 223 static const struct protosw in_pim_protosw = { 224 .pr_type = SOCK_RAW, 225 .pr_domain = &inetdomain, 226 .pr_protocol = IPPROTO_PIM, 227 .pr_flags = PR_ATOMIC|PR_ADDR|PR_LASTHDR, 228 .pr_input = pim_input, 229 .pr_output = (pr_output_t*)rip_output, 230 .pr_ctloutput = rip_ctloutput, 231 .pr_usrreqs = &rip_usrreqs 232 }; 233 static const struct encaptab *pim_encap_cookie; 234 235 static int pim_encapcheck(const struct mbuf *, int, int, void *); 236 237 /* 238 * Note: the PIM Register encapsulation adds the following in front of a 239 * data packet: 240 * 241 * struct pim_encap_hdr { 242 * struct ip ip; 243 * struct pim_encap_pimhdr pim; 244 * } 245 * 246 */ 247 248 struct pim_encap_pimhdr { 249 struct pim pim; 250 uint32_t flags; 251 }; 252 #define PIM_ENCAP_TTL 64 253 254 static struct ip pim_encap_iphdr = { 255 #if BYTE_ORDER == LITTLE_ENDIAN 256 sizeof(struct ip) >> 2, 257 IPVERSION, 258 #else 259 IPVERSION, 260 sizeof(struct ip) >> 2, 261 #endif 262 0, /* tos */ 263 sizeof(struct ip), /* total length */ 264 0, /* id */ 265 0, /* frag offset */ 266 PIM_ENCAP_TTL, 267 IPPROTO_PIM, 268 0, /* checksum */ 269 }; 270 271 static struct pim_encap_pimhdr pim_encap_pimhdr = { 272 { 273 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 274 0, /* reserved */ 275 0, /* checksum */ 276 }, 277 0 /* flags */ 278 }; 279 280 static struct ifnet multicast_register_if; 281 static vifi_t reg_vif_num = VIFI_INVALID; 282 283 /* 284 * Private variables. 285 */ 286 287 static u_long X_ip_mcast_src(int); 288 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *, 289 struct ip_moptions *); 290 static int X_ip_mrouter_done(void); 291 static int X_ip_mrouter_get(struct socket *, struct sockopt *); 292 static int X_ip_mrouter_set(struct socket *, struct sockopt *); 293 static int X_legal_vif_num(int); 294 static int X_mrt_ioctl(u_long, caddr_t, int); 295 296 static int add_bw_upcall(struct bw_upcall *); 297 static int add_mfc(struct mfcctl2 *); 298 static int add_vif(struct vifctl *); 299 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *); 300 static void bw_meter_process(void); 301 static void bw_meter_receive_packet(struct bw_meter *, int, 302 struct timeval *); 303 static void bw_upcalls_send(void); 304 static int del_bw_upcall(struct bw_upcall *); 305 static int del_mfc(struct mfcctl2 *); 306 static int del_vif(vifi_t); 307 static int del_vif_locked(vifi_t); 308 static void expire_bw_meter_process(void *); 309 static void expire_bw_upcalls_send(void *); 310 static void expire_mfc(struct mfc *); 311 static void expire_upcalls(void *); 312 static void free_bw_list(struct bw_meter *); 313 static int get_sg_cnt(struct sioc_sg_req *); 314 static int get_vif_cnt(struct sioc_vif_req *); 315 static void if_detached_event(void *, struct ifnet *); 316 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 317 static int ip_mrouter_init(struct socket *, int); 318 static __inline struct mfc * 319 mfc_find(struct in_addr *, struct in_addr *); 320 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 321 static struct mbuf * 322 pim_register_prepare(struct ip *, struct mbuf *); 323 static int pim_register_send(struct ip *, struct vif *, 324 struct mbuf *, struct mfc *); 325 static int pim_register_send_rp(struct ip *, struct vif *, 326 struct mbuf *, struct mfc *); 327 static int pim_register_send_upcall(struct ip *, struct vif *, 328 struct mbuf *, struct mfc *); 329 static void schedule_bw_meter(struct bw_meter *, struct timeval *); 330 static void send_packet(struct vif *, struct mbuf *); 331 static int set_api_config(uint32_t *); 332 static int set_assert(int); 333 static int socket_send(struct socket *, struct mbuf *, 334 struct sockaddr_in *); 335 static void unschedule_bw_meter(struct bw_meter *); 336 337 /* 338 * Kernel multicast forwarding API capabilities and setup. 339 * If more API capabilities are added to the kernel, they should be 340 * recorded in `mrt_api_support'. 341 */ 342 #define MRT_API_VERSION 0x0305 343 344 static const int mrt_api_version = MRT_API_VERSION; 345 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 346 MRT_MFC_FLAGS_BORDER_VIF | 347 MRT_MFC_RP | 348 MRT_MFC_BW_UPCALL); 349 static uint32_t mrt_api_config = 0; 350 351 static int pim_assert_enabled; 352 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */ 353 354 /* 355 * Find a route for a given origin IP address and multicast group address. 356 * Statistics must be updated by the caller. 357 */ 358 static __inline struct mfc * 359 mfc_find(struct in_addr *o, struct in_addr *g) 360 { 361 struct mfc *rt; 362 363 MFC_LOCK_ASSERT(); 364 365 LIST_FOREACH(rt, &mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { 366 if (in_hosteq(rt->mfc_origin, *o) && 367 in_hosteq(rt->mfc_mcastgrp, *g) && 368 TAILQ_EMPTY(&rt->mfc_stall)) 369 break; 370 } 371 372 return (rt); 373 } 374 375 /* 376 * Handle MRT setsockopt commands to modify the multicast forwarding tables. 377 */ 378 static int 379 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 380 { 381 int error, optval; 382 vifi_t vifi; 383 struct vifctl vifc; 384 struct mfcctl2 mfc; 385 struct bw_upcall bw_upcall; 386 uint32_t i; 387 388 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT) 389 return EPERM; 390 391 error = 0; 392 switch (sopt->sopt_name) { 393 case MRT_INIT: 394 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 395 if (error) 396 break; 397 error = ip_mrouter_init(so, optval); 398 break; 399 400 case MRT_DONE: 401 error = ip_mrouter_done(); 402 break; 403 404 case MRT_ADD_VIF: 405 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 406 if (error) 407 break; 408 error = add_vif(&vifc); 409 break; 410 411 case MRT_DEL_VIF: 412 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 413 if (error) 414 break; 415 error = del_vif(vifi); 416 break; 417 418 case MRT_ADD_MFC: 419 case MRT_DEL_MFC: 420 /* 421 * select data size depending on API version. 422 */ 423 if (sopt->sopt_name == MRT_ADD_MFC && 424 mrt_api_config & MRT_API_FLAGS_ALL) { 425 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 426 sizeof(struct mfcctl2)); 427 } else { 428 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 429 sizeof(struct mfcctl)); 430 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 431 sizeof(mfc) - sizeof(struct mfcctl)); 432 } 433 if (error) 434 break; 435 if (sopt->sopt_name == MRT_ADD_MFC) 436 error = add_mfc(&mfc); 437 else 438 error = del_mfc(&mfc); 439 break; 440 441 case MRT_ASSERT: 442 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 443 if (error) 444 break; 445 set_assert(optval); 446 break; 447 448 case MRT_API_CONFIG: 449 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 450 if (!error) 451 error = set_api_config(&i); 452 if (!error) 453 error = sooptcopyout(sopt, &i, sizeof i); 454 break; 455 456 case MRT_ADD_BW_UPCALL: 457 case MRT_DEL_BW_UPCALL: 458 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 459 sizeof bw_upcall); 460 if (error) 461 break; 462 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 463 error = add_bw_upcall(&bw_upcall); 464 else 465 error = del_bw_upcall(&bw_upcall); 466 break; 467 468 default: 469 error = EOPNOTSUPP; 470 break; 471 } 472 return error; 473 } 474 475 /* 476 * Handle MRT getsockopt commands 477 */ 478 static int 479 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 480 { 481 int error; 482 483 switch (sopt->sopt_name) { 484 case MRT_VERSION: 485 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version); 486 break; 487 488 case MRT_ASSERT: 489 error = sooptcopyout(sopt, &pim_assert_enabled, 490 sizeof pim_assert_enabled); 491 break; 492 493 case MRT_API_SUPPORT: 494 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 495 break; 496 497 case MRT_API_CONFIG: 498 error = sooptcopyout(sopt, &mrt_api_config, sizeof mrt_api_config); 499 break; 500 501 default: 502 error = EOPNOTSUPP; 503 break; 504 } 505 return error; 506 } 507 508 /* 509 * Handle ioctl commands to obtain information from the cache 510 */ 511 static int 512 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused) 513 { 514 int error = 0; 515 516 /* 517 * Currently the only function calling this ioctl routine is rtioctl(). 518 * Typically, only root can create the raw socket in order to execute 519 * this ioctl method, however the request might be coming from a prison 520 */ 521 error = priv_check(curthread, PRIV_NETINET_MROUTE); 522 if (error) 523 return (error); 524 switch (cmd) { 525 case (SIOCGETVIFCNT): 526 error = get_vif_cnt((struct sioc_vif_req *)data); 527 break; 528 529 case (SIOCGETSGCNT): 530 error = get_sg_cnt((struct sioc_sg_req *)data); 531 break; 532 533 default: 534 error = EINVAL; 535 break; 536 } 537 return error; 538 } 539 540 /* 541 * returns the packet, byte, rpf-failure count for the source group provided 542 */ 543 static int 544 get_sg_cnt(struct sioc_sg_req *req) 545 { 546 struct mfc *rt; 547 548 MFC_LOCK(); 549 rt = mfc_find(&req->src, &req->grp); 550 if (rt == NULL) { 551 MFC_UNLOCK(); 552 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 553 return EADDRNOTAVAIL; 554 } 555 req->pktcnt = rt->mfc_pkt_cnt; 556 req->bytecnt = rt->mfc_byte_cnt; 557 req->wrong_if = rt->mfc_wrong_if; 558 MFC_UNLOCK(); 559 return 0; 560 } 561 562 /* 563 * returns the input and output packet and byte counts on the vif provided 564 */ 565 static int 566 get_vif_cnt(struct sioc_vif_req *req) 567 { 568 vifi_t vifi = req->vifi; 569 570 VIF_LOCK(); 571 if (vifi >= numvifs) { 572 VIF_UNLOCK(); 573 return EINVAL; 574 } 575 576 req->icount = viftable[vifi].v_pkt_in; 577 req->ocount = viftable[vifi].v_pkt_out; 578 req->ibytes = viftable[vifi].v_bytes_in; 579 req->obytes = viftable[vifi].v_bytes_out; 580 VIF_UNLOCK(); 581 582 return 0; 583 } 584 585 static void 586 ip_mrouter_reset(void) 587 { 588 589 pim_assert_enabled = 0; 590 mrt_api_config = 0; 591 592 callout_init(&expire_upcalls_ch, CALLOUT_MPSAFE); 593 594 bw_upcalls_n = 0; 595 bzero((caddr_t)bw_meter_timers, sizeof(bw_meter_timers)); 596 callout_init(&bw_upcalls_ch, CALLOUT_MPSAFE); 597 callout_init(&bw_meter_ch, CALLOUT_MPSAFE); 598 } 599 600 static void 601 if_detached_event(void *arg __unused, struct ifnet *ifp) 602 { 603 vifi_t vifi; 604 int i; 605 606 MROUTER_LOCK(); 607 608 if (V_ip_mrouter == NULL) { 609 MROUTER_UNLOCK(); 610 return; 611 } 612 613 VIF_LOCK(); 614 MFC_LOCK(); 615 616 /* 617 * Tear down multicast forwarder state associated with this ifnet. 618 * 1. Walk the vif list, matching vifs against this ifnet. 619 * 2. Walk the multicast forwarding cache (mfc) looking for 620 * inner matches with this vif's index. 621 * 3. Expire any matching multicast forwarding cache entries. 622 * 4. Free vif state. This should disable ALLMULTI on the interface. 623 */ 624 for (vifi = 0; vifi < numvifs; vifi++) { 625 if (viftable[vifi].v_ifp != ifp) 626 continue; 627 for (i = 0; i < mfchashsize; i++) { 628 struct mfc *rt, *nrt; 629 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 630 nrt = LIST_NEXT(rt, mfc_hash); 631 if (rt->mfc_parent == vifi) { 632 expire_mfc(rt); 633 } 634 } 635 } 636 del_vif_locked(vifi); 637 } 638 639 MFC_UNLOCK(); 640 VIF_UNLOCK(); 641 642 MROUTER_UNLOCK(); 643 } 644 645 /* 646 * Enable multicast forwarding. 647 */ 648 static int 649 ip_mrouter_init(struct socket *so, int version) 650 { 651 652 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__, 653 so->so_type, so->so_proto->pr_protocol); 654 655 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 656 return EOPNOTSUPP; 657 658 if (version != 1) 659 return ENOPROTOOPT; 660 661 MROUTER_LOCK(); 662 663 if (V_ip_mrouter != NULL) { 664 MROUTER_UNLOCK(); 665 return EADDRINUSE; 666 } 667 668 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 669 if_detached_event, NULL, EVENTHANDLER_PRI_ANY); 670 if (if_detach_event_tag == NULL) { 671 MROUTER_UNLOCK(); 672 return (ENOMEM); 673 } 674 675 mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &mfchash, HASH_NOWAIT); 676 677 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 678 679 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 680 expire_bw_upcalls_send, NULL); 681 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 682 683 V_ip_mrouter = so; 684 685 MROUTER_UNLOCK(); 686 687 CTR1(KTR_IPMF, "%s: done", __func__); 688 689 return 0; 690 } 691 692 /* 693 * Disable multicast forwarding. 694 */ 695 static int 696 X_ip_mrouter_done(void) 697 { 698 vifi_t vifi; 699 int i; 700 struct ifnet *ifp; 701 struct ifreq ifr; 702 703 MROUTER_LOCK(); 704 705 if (V_ip_mrouter == NULL) { 706 MROUTER_UNLOCK(); 707 return EINVAL; 708 } 709 710 /* 711 * Detach/disable hooks to the reset of the system. 712 */ 713 V_ip_mrouter = NULL; 714 mrt_api_config = 0; 715 716 VIF_LOCK(); 717 718 /* 719 * For each phyint in use, disable promiscuous reception of all IP 720 * multicasts. 721 */ 722 for (vifi = 0; vifi < numvifs; vifi++) { 723 if (!in_nullhost(viftable[vifi].v_lcl_addr) && 724 !(viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 725 struct sockaddr_in *so = (struct sockaddr_in *)&(ifr.ifr_addr); 726 727 so->sin_len = sizeof(struct sockaddr_in); 728 so->sin_family = AF_INET; 729 so->sin_addr.s_addr = INADDR_ANY; 730 ifp = viftable[vifi].v_ifp; 731 if_allmulti(ifp, 0); 732 } 733 } 734 bzero((caddr_t)viftable, sizeof(viftable)); 735 numvifs = 0; 736 pim_assert_enabled = 0; 737 738 VIF_UNLOCK(); 739 740 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 741 742 callout_stop(&expire_upcalls_ch); 743 callout_stop(&bw_upcalls_ch); 744 callout_stop(&bw_meter_ch); 745 746 MFC_LOCK(); 747 748 /* 749 * Free all multicast forwarding cache entries. 750 * Do not use hashdestroy(), as we must perform other cleanup. 751 */ 752 for (i = 0; i < mfchashsize; i++) { 753 struct mfc *rt, *nrt; 754 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 755 nrt = LIST_NEXT(rt, mfc_hash); 756 expire_mfc(rt); 757 } 758 } 759 free(mfchashtbl, M_MRTABLE); 760 mfchashtbl = NULL; 761 762 bzero(nexpire, sizeof(nexpire[0]) * mfchashsize); 763 764 bw_upcalls_n = 0; 765 bzero(bw_meter_timers, sizeof(bw_meter_timers)); 766 767 MFC_UNLOCK(); 768 769 reg_vif_num = VIFI_INVALID; 770 771 MROUTER_UNLOCK(); 772 773 CTR1(KTR_IPMF, "%s: done", __func__); 774 775 return 0; 776 } 777 778 /* 779 * Set PIM assert processing global 780 */ 781 static int 782 set_assert(int i) 783 { 784 if ((i != 1) && (i != 0)) 785 return EINVAL; 786 787 pim_assert_enabled = i; 788 789 return 0; 790 } 791 792 /* 793 * Configure API capabilities 794 */ 795 int 796 set_api_config(uint32_t *apival) 797 { 798 int i; 799 800 /* 801 * We can set the API capabilities only if it is the first operation 802 * after MRT_INIT. I.e.: 803 * - there are no vifs installed 804 * - pim_assert is not enabled 805 * - the MFC table is empty 806 */ 807 if (numvifs > 0) { 808 *apival = 0; 809 return EPERM; 810 } 811 if (pim_assert_enabled) { 812 *apival = 0; 813 return EPERM; 814 } 815 816 MFC_LOCK(); 817 818 for (i = 0; i < mfchashsize; i++) { 819 if (LIST_FIRST(&mfchashtbl[i]) != NULL) { 820 *apival = 0; 821 return EPERM; 822 } 823 } 824 825 MFC_UNLOCK(); 826 827 mrt_api_config = *apival & mrt_api_support; 828 *apival = mrt_api_config; 829 830 return 0; 831 } 832 833 /* 834 * Add a vif to the vif table 835 */ 836 static int 837 add_vif(struct vifctl *vifcp) 838 { 839 struct vif *vifp = viftable + vifcp->vifc_vifi; 840 struct sockaddr_in sin = {sizeof sin, AF_INET}; 841 struct ifaddr *ifa; 842 struct ifnet *ifp; 843 int error; 844 845 VIF_LOCK(); 846 if (vifcp->vifc_vifi >= MAXVIFS) { 847 VIF_UNLOCK(); 848 return EINVAL; 849 } 850 /* rate limiting is no longer supported by this code */ 851 if (vifcp->vifc_rate_limit != 0) { 852 log(LOG_ERR, "rate limiting is no longer supported\n"); 853 VIF_UNLOCK(); 854 return EINVAL; 855 } 856 if (!in_nullhost(vifp->v_lcl_addr)) { 857 VIF_UNLOCK(); 858 return EADDRINUSE; 859 } 860 if (in_nullhost(vifcp->vifc_lcl_addr)) { 861 VIF_UNLOCK(); 862 return EADDRNOTAVAIL; 863 } 864 865 /* Find the interface with an address in AF_INET family */ 866 if (vifcp->vifc_flags & VIFF_REGISTER) { 867 /* 868 * XXX: Because VIFF_REGISTER does not really need a valid 869 * local interface (e.g. it could be 127.0.0.2), we don't 870 * check its address. 871 */ 872 ifp = NULL; 873 } else { 874 sin.sin_addr = vifcp->vifc_lcl_addr; 875 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 876 if (ifa == NULL) { 877 VIF_UNLOCK(); 878 return EADDRNOTAVAIL; 879 } 880 ifp = ifa->ifa_ifp; 881 ifa_free(ifa); 882 } 883 884 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { 885 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__); 886 VIF_UNLOCK(); 887 return EOPNOTSUPP; 888 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 889 ifp = &multicast_register_if; 890 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp); 891 if (reg_vif_num == VIFI_INVALID) { 892 if_initname(&multicast_register_if, "register_vif", 0); 893 multicast_register_if.if_flags = IFF_LOOPBACK; 894 reg_vif_num = vifcp->vifc_vifi; 895 } 896 } else { /* Make sure the interface supports multicast */ 897 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 898 VIF_UNLOCK(); 899 return EOPNOTSUPP; 900 } 901 902 /* Enable promiscuous reception of all IP multicasts from the if */ 903 error = if_allmulti(ifp, 1); 904 if (error) { 905 VIF_UNLOCK(); 906 return error; 907 } 908 } 909 910 vifp->v_flags = vifcp->vifc_flags; 911 vifp->v_threshold = vifcp->vifc_threshold; 912 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 913 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 914 vifp->v_ifp = ifp; 915 /* initialize per vif pkt counters */ 916 vifp->v_pkt_in = 0; 917 vifp->v_pkt_out = 0; 918 vifp->v_bytes_in = 0; 919 vifp->v_bytes_out = 0; 920 bzero(&vifp->v_route, sizeof(vifp->v_route)); 921 922 /* Adjust numvifs up if the vifi is higher than numvifs */ 923 if (numvifs <= vifcp->vifc_vifi) 924 numvifs = vifcp->vifc_vifi + 1; 925 926 VIF_UNLOCK(); 927 928 CTR4(KTR_IPMF, "%s: add vif %d laddr %s thresh %x", __func__, 929 (int)vifcp->vifc_vifi, inet_ntoa(vifcp->vifc_lcl_addr), 930 (int)vifcp->vifc_threshold); 931 932 return 0; 933 } 934 935 /* 936 * Delete a vif from the vif table 937 */ 938 static int 939 del_vif_locked(vifi_t vifi) 940 { 941 struct vif *vifp; 942 943 VIF_LOCK_ASSERT(); 944 945 if (vifi >= numvifs) { 946 return EINVAL; 947 } 948 vifp = &viftable[vifi]; 949 if (in_nullhost(vifp->v_lcl_addr)) { 950 return EADDRNOTAVAIL; 951 } 952 953 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 954 if_allmulti(vifp->v_ifp, 0); 955 956 if (vifp->v_flags & VIFF_REGISTER) 957 reg_vif_num = VIFI_INVALID; 958 959 bzero((caddr_t)vifp, sizeof (*vifp)); 960 961 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi); 962 963 /* Adjust numvifs down */ 964 for (vifi = numvifs; vifi > 0; vifi--) 965 if (!in_nullhost(viftable[vifi-1].v_lcl_addr)) 966 break; 967 numvifs = vifi; 968 969 return 0; 970 } 971 972 static int 973 del_vif(vifi_t vifi) 974 { 975 int cc; 976 977 VIF_LOCK(); 978 cc = del_vif_locked(vifi); 979 VIF_UNLOCK(); 980 981 return cc; 982 } 983 984 /* 985 * update an mfc entry without resetting counters and S,G addresses. 986 */ 987 static void 988 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 989 { 990 int i; 991 992 rt->mfc_parent = mfccp->mfcc_parent; 993 for (i = 0; i < numvifs; i++) { 994 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 995 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & mrt_api_config & 996 MRT_MFC_FLAGS_ALL; 997 } 998 /* set the RP address */ 999 if (mrt_api_config & MRT_MFC_RP) 1000 rt->mfc_rp = mfccp->mfcc_rp; 1001 else 1002 rt->mfc_rp.s_addr = INADDR_ANY; 1003 } 1004 1005 /* 1006 * fully initialize an mfc entry from the parameter. 1007 */ 1008 static void 1009 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1010 { 1011 rt->mfc_origin = mfccp->mfcc_origin; 1012 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1013 1014 update_mfc_params(rt, mfccp); 1015 1016 /* initialize pkt counters per src-grp */ 1017 rt->mfc_pkt_cnt = 0; 1018 rt->mfc_byte_cnt = 0; 1019 rt->mfc_wrong_if = 0; 1020 timevalclear(&rt->mfc_last_assert); 1021 } 1022 1023 static void 1024 expire_mfc(struct mfc *rt) 1025 { 1026 struct rtdetq *rte, *nrte; 1027 1028 MFC_LOCK_ASSERT(); 1029 1030 free_bw_list(rt->mfc_bw_meter); 1031 1032 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) { 1033 m_freem(rte->m); 1034 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link); 1035 free(rte, M_MRTABLE); 1036 } 1037 1038 LIST_REMOVE(rt, mfc_hash); 1039 free(rt, M_MRTABLE); 1040 } 1041 1042 /* 1043 * Add an mfc entry 1044 */ 1045 static int 1046 add_mfc(struct mfcctl2 *mfccp) 1047 { 1048 struct mfc *rt; 1049 struct rtdetq *rte, *nrte; 1050 u_long hash = 0; 1051 u_short nstl; 1052 1053 VIF_LOCK(); 1054 MFC_LOCK(); 1055 1056 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1057 1058 /* If an entry already exists, just update the fields */ 1059 if (rt) { 1060 CTR4(KTR_IPMF, "%s: update mfc orig %s group %lx parent %x", 1061 __func__, inet_ntoa(mfccp->mfcc_origin), 1062 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1063 mfccp->mfcc_parent); 1064 update_mfc_params(rt, mfccp); 1065 MFC_UNLOCK(); 1066 VIF_UNLOCK(); 1067 return (0); 1068 } 1069 1070 /* 1071 * Find the entry for which the upcall was made and update 1072 */ 1073 nstl = 0; 1074 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1075 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1076 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1077 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1078 !TAILQ_EMPTY(&rt->mfc_stall)) { 1079 CTR5(KTR_IPMF, 1080 "%s: add mfc orig %s group %lx parent %x qh %p", 1081 __func__, inet_ntoa(mfccp->mfcc_origin), 1082 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1083 mfccp->mfcc_parent, 1084 TAILQ_FIRST(&rt->mfc_stall)); 1085 if (nstl++) 1086 CTR1(KTR_IPMF, "%s: multiple matches", __func__); 1087 1088 init_mfc_params(rt, mfccp); 1089 rt->mfc_expire = 0; /* Don't clean this guy up */ 1090 nexpire[hash]--; 1091 1092 /* Free queued packets, but attempt to forward them first. */ 1093 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) { 1094 if (rte->ifp != NULL) 1095 ip_mdq(rte->m, rte->ifp, rt, -1); 1096 m_freem(rte->m); 1097 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link); 1098 rt->mfc_nstall--; 1099 free(rte, M_MRTABLE); 1100 } 1101 } 1102 } 1103 1104 /* 1105 * It is possible that an entry is being inserted without an upcall 1106 */ 1107 if (nstl == 0) { 1108 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__); 1109 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1110 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1111 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1112 init_mfc_params(rt, mfccp); 1113 if (rt->mfc_expire) 1114 nexpire[hash]--; 1115 rt->mfc_expire = 0; 1116 break; /* XXX */ 1117 } 1118 } 1119 1120 if (rt == NULL) { /* no upcall, so make a new entry */ 1121 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1122 if (rt == NULL) { 1123 MFC_UNLOCK(); 1124 VIF_UNLOCK(); 1125 return (ENOBUFS); 1126 } 1127 1128 init_mfc_params(rt, mfccp); 1129 TAILQ_INIT(&rt->mfc_stall); 1130 rt->mfc_nstall = 0; 1131 1132 rt->mfc_expire = 0; 1133 rt->mfc_bw_meter = NULL; 1134 1135 /* insert new entry at head of hash chain */ 1136 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1137 } 1138 } 1139 1140 MFC_UNLOCK(); 1141 VIF_UNLOCK(); 1142 1143 return (0); 1144 } 1145 1146 /* 1147 * Delete an mfc entry 1148 */ 1149 static int 1150 del_mfc(struct mfcctl2 *mfccp) 1151 { 1152 struct in_addr origin; 1153 struct in_addr mcastgrp; 1154 struct mfc *rt; 1155 1156 origin = mfccp->mfcc_origin; 1157 mcastgrp = mfccp->mfcc_mcastgrp; 1158 1159 CTR3(KTR_IPMF, "%s: delete mfc orig %s group %lx", __func__, 1160 inet_ntoa(origin), (u_long)ntohl(mcastgrp.s_addr)); 1161 1162 MFC_LOCK(); 1163 1164 rt = mfc_find(&origin, &mcastgrp); 1165 if (rt == NULL) { 1166 MFC_UNLOCK(); 1167 return EADDRNOTAVAIL; 1168 } 1169 1170 /* 1171 * free the bw_meter entries 1172 */ 1173 free_bw_list(rt->mfc_bw_meter); 1174 rt->mfc_bw_meter = NULL; 1175 1176 LIST_REMOVE(rt, mfc_hash); 1177 free(rt, M_MRTABLE); 1178 1179 MFC_UNLOCK(); 1180 1181 return (0); 1182 } 1183 1184 /* 1185 * Send a message to the routing daemon on the multicast routing socket. 1186 */ 1187 static int 1188 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1189 { 1190 if (s) { 1191 SOCKBUF_LOCK(&s->so_rcv); 1192 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, 1193 NULL) != 0) { 1194 sorwakeup_locked(s); 1195 return 0; 1196 } 1197 SOCKBUF_UNLOCK(&s->so_rcv); 1198 } 1199 m_freem(mm); 1200 return -1; 1201 } 1202 1203 /* 1204 * IP multicast forwarding function. This function assumes that the packet 1205 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1206 * pointed to by "ifp", and the packet is to be relayed to other networks 1207 * that have members of the packet's destination IP multicast group. 1208 * 1209 * The packet is returned unscathed to the caller, unless it is 1210 * erroneous, in which case a non-zero return value tells the caller to 1211 * discard it. 1212 */ 1213 1214 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1215 1216 static int 1217 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1218 struct ip_moptions *imo) 1219 { 1220 struct mfc *rt; 1221 int error; 1222 vifi_t vifi; 1223 1224 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig %s group %lx ifp %p", 1225 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr), ifp); 1226 1227 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1228 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1229 /* 1230 * Packet arrived via a physical interface or 1231 * an encapsulated tunnel or a register_vif. 1232 */ 1233 } else { 1234 /* 1235 * Packet arrived through a source-route tunnel. 1236 * Source-route tunnels are no longer supported. 1237 */ 1238 return (1); 1239 } 1240 1241 VIF_LOCK(); 1242 MFC_LOCK(); 1243 if (imo && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1244 if (ip->ip_ttl < MAXTTL) 1245 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1246 error = ip_mdq(m, ifp, NULL, vifi); 1247 MFC_UNLOCK(); 1248 VIF_UNLOCK(); 1249 return error; 1250 } 1251 1252 /* 1253 * Don't forward a packet with time-to-live of zero or one, 1254 * or a packet destined to a local-only group. 1255 */ 1256 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) { 1257 MFC_UNLOCK(); 1258 VIF_UNLOCK(); 1259 return 0; 1260 } 1261 1262 /* 1263 * Determine forwarding vifs from the forwarding cache table 1264 */ 1265 MRTSTAT_INC(mrts_mfc_lookups); 1266 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1267 1268 /* Entry exists, so forward if necessary */ 1269 if (rt != NULL) { 1270 error = ip_mdq(m, ifp, rt, -1); 1271 MFC_UNLOCK(); 1272 VIF_UNLOCK(); 1273 return error; 1274 } else { 1275 /* 1276 * If we don't have a route for packet's origin, 1277 * Make a copy of the packet & send message to routing daemon 1278 */ 1279 1280 struct mbuf *mb0; 1281 struct rtdetq *rte; 1282 u_long hash; 1283 int hlen = ip->ip_hl << 2; 1284 1285 MRTSTAT_INC(mrts_mfc_misses); 1286 MRTSTAT_INC(mrts_no_route); 1287 CTR2(KTR_IPMF, "ip_mforward: no mfc for (%s,%lx)", 1288 inet_ntoa(ip->ip_src), (u_long)ntohl(ip->ip_dst.s_addr)); 1289 1290 /* 1291 * Allocate mbufs early so that we don't do extra work if we are 1292 * just going to fail anyway. Make sure to pullup the header so 1293 * that other people can't step on it. 1294 */ 1295 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, 1296 M_NOWAIT|M_ZERO); 1297 if (rte == NULL) { 1298 MFC_UNLOCK(); 1299 VIF_UNLOCK(); 1300 return ENOBUFS; 1301 } 1302 1303 mb0 = m_copypacket(m, M_DONTWAIT); 1304 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1305 mb0 = m_pullup(mb0, hlen); 1306 if (mb0 == NULL) { 1307 free(rte, M_MRTABLE); 1308 MFC_UNLOCK(); 1309 VIF_UNLOCK(); 1310 return ENOBUFS; 1311 } 1312 1313 /* is there an upcall waiting for this flow ? */ 1314 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1315 LIST_FOREACH(rt, &mfchashtbl[hash], mfc_hash) { 1316 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1317 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1318 !TAILQ_EMPTY(&rt->mfc_stall)) 1319 break; 1320 } 1321 1322 if (rt == NULL) { 1323 int i; 1324 struct igmpmsg *im; 1325 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1326 struct mbuf *mm; 1327 1328 /* 1329 * Locate the vifi for the incoming interface for this packet. 1330 * If none found, drop packet. 1331 */ 1332 for (vifi = 0; vifi < numvifs && 1333 viftable[vifi].v_ifp != ifp; vifi++) 1334 ; 1335 if (vifi >= numvifs) /* vif not found, drop packet */ 1336 goto non_fatal; 1337 1338 /* no upcall, so make a new entry */ 1339 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1340 if (rt == NULL) 1341 goto fail; 1342 1343 /* Make a copy of the header to send to the user level process */ 1344 mm = m_copy(mb0, 0, hlen); 1345 if (mm == NULL) 1346 goto fail1; 1347 1348 /* 1349 * Send message to routing daemon to install 1350 * a route into the kernel table 1351 */ 1352 1353 im = mtod(mm, struct igmpmsg *); 1354 im->im_msgtype = IGMPMSG_NOCACHE; 1355 im->im_mbz = 0; 1356 im->im_vif = vifi; 1357 1358 MRTSTAT_INC(mrts_upcalls); 1359 1360 k_igmpsrc.sin_addr = ip->ip_src; 1361 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { 1362 CTR0(KTR_IPMF, "ip_mforward: socket queue full"); 1363 MRTSTAT_INC(mrts_upq_sockfull); 1364 fail1: 1365 free(rt, M_MRTABLE); 1366 fail: 1367 free(rte, M_MRTABLE); 1368 m_freem(mb0); 1369 MFC_UNLOCK(); 1370 VIF_UNLOCK(); 1371 return ENOBUFS; 1372 } 1373 1374 /* insert new entry at head of hash chain */ 1375 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1376 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1377 rt->mfc_expire = UPCALL_EXPIRE; 1378 nexpire[hash]++; 1379 for (i = 0; i < numvifs; i++) { 1380 rt->mfc_ttls[i] = 0; 1381 rt->mfc_flags[i] = 0; 1382 } 1383 rt->mfc_parent = -1; 1384 1385 /* clear the RP address */ 1386 rt->mfc_rp.s_addr = INADDR_ANY; 1387 rt->mfc_bw_meter = NULL; 1388 1389 /* initialize pkt counters per src-grp */ 1390 rt->mfc_pkt_cnt = 0; 1391 rt->mfc_byte_cnt = 0; 1392 rt->mfc_wrong_if = 0; 1393 timevalclear(&rt->mfc_last_assert); 1394 1395 TAILQ_INIT(&rt->mfc_stall); 1396 rt->mfc_nstall = 0; 1397 1398 /* link into table */ 1399 LIST_INSERT_HEAD(&mfchashtbl[hash], rt, mfc_hash); 1400 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link); 1401 rt->mfc_nstall++; 1402 1403 } else { 1404 /* determine if queue has overflowed */ 1405 if (rt->mfc_nstall > MAX_UPQ) { 1406 MRTSTAT_INC(mrts_upq_ovflw); 1407 non_fatal: 1408 free(rte, M_MRTABLE); 1409 m_freem(mb0); 1410 MFC_UNLOCK(); 1411 VIF_UNLOCK(); 1412 return (0); 1413 } 1414 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link); 1415 rt->mfc_nstall++; 1416 } 1417 1418 rte->m = mb0; 1419 rte->ifp = ifp; 1420 1421 MFC_UNLOCK(); 1422 VIF_UNLOCK(); 1423 1424 return 0; 1425 } 1426 } 1427 1428 /* 1429 * Clean up the cache entry if upcall is not serviced 1430 */ 1431 static void 1432 expire_upcalls(void *unused) 1433 { 1434 int i; 1435 1436 MFC_LOCK(); 1437 1438 for (i = 0; i < mfchashsize; i++) { 1439 struct mfc *rt, *nrt; 1440 1441 if (nexpire[i] == 0) 1442 continue; 1443 1444 for (rt = LIST_FIRST(&mfchashtbl[i]); rt; rt = nrt) { 1445 nrt = LIST_NEXT(rt, mfc_hash); 1446 1447 if (TAILQ_EMPTY(&rt->mfc_stall)) 1448 continue; 1449 1450 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1451 continue; 1452 1453 /* 1454 * free the bw_meter entries 1455 */ 1456 while (rt->mfc_bw_meter != NULL) { 1457 struct bw_meter *x = rt->mfc_bw_meter; 1458 1459 rt->mfc_bw_meter = x->bm_mfc_next; 1460 free(x, M_BWMETER); 1461 } 1462 1463 MRTSTAT_INC(mrts_cache_cleanups); 1464 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__, 1465 (u_long)ntohl(rt->mfc_origin.s_addr), 1466 (u_long)ntohl(rt->mfc_mcastgrp.s_addr)); 1467 1468 expire_mfc(rt); 1469 } 1470 } 1471 1472 MFC_UNLOCK(); 1473 1474 callout_reset(&expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, NULL); 1475 } 1476 1477 /* 1478 * Packet forwarding routine once entry in the cache is made 1479 */ 1480 static int 1481 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1482 { 1483 struct ip *ip = mtod(m, struct ip *); 1484 vifi_t vifi; 1485 int plen = ip->ip_len; 1486 1487 VIF_LOCK_ASSERT(); 1488 1489 /* 1490 * If xmt_vif is not -1, send on only the requested vif. 1491 * 1492 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1493 */ 1494 if (xmt_vif < numvifs) { 1495 if (viftable[xmt_vif].v_flags & VIFF_REGISTER) 1496 pim_register_send(ip, viftable + xmt_vif, m, rt); 1497 else 1498 phyint_send(ip, viftable + xmt_vif, m); 1499 return 1; 1500 } 1501 1502 /* 1503 * Don't forward if it didn't arrive from the parent vif for its origin. 1504 */ 1505 vifi = rt->mfc_parent; 1506 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1507 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)", 1508 __func__, ifp, (int)vifi, viftable[vifi].v_ifp); 1509 MRTSTAT_INC(mrts_wrong_if); 1510 ++rt->mfc_wrong_if; 1511 /* 1512 * If we are doing PIM assert processing, send a message 1513 * to the routing daemon. 1514 * 1515 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1516 * can complete the SPT switch, regardless of the type 1517 * of the iif (broadcast media, GRE tunnel, etc). 1518 */ 1519 if (pim_assert_enabled && (vifi < numvifs) && viftable[vifi].v_ifp) { 1520 1521 if (ifp == &multicast_register_if) 1522 PIMSTAT_INC(pims_rcv_registers_wrongiif); 1523 1524 /* Get vifi for the incoming packet */ 1525 for (vifi=0; vifi < numvifs && viftable[vifi].v_ifp != ifp; vifi++) 1526 ; 1527 if (vifi >= numvifs) 1528 return 0; /* The iif is not found: ignore the packet. */ 1529 1530 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1531 return 0; /* WRONGVIF disabled: ignore the packet */ 1532 1533 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) { 1534 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1535 struct igmpmsg *im; 1536 int hlen = ip->ip_hl << 2; 1537 struct mbuf *mm = m_copy(m, 0, hlen); 1538 1539 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1540 mm = m_pullup(mm, hlen); 1541 if (mm == NULL) 1542 return ENOBUFS; 1543 1544 im = mtod(mm, struct igmpmsg *); 1545 im->im_msgtype = IGMPMSG_WRONGVIF; 1546 im->im_mbz = 0; 1547 im->im_vif = vifi; 1548 1549 MRTSTAT_INC(mrts_upcalls); 1550 1551 k_igmpsrc.sin_addr = im->im_src; 1552 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { 1553 CTR1(KTR_IPMF, "%s: socket queue full", __func__); 1554 MRTSTAT_INC(mrts_upq_sockfull); 1555 return ENOBUFS; 1556 } 1557 } 1558 } 1559 return 0; 1560 } 1561 1562 1563 /* If I sourced this packet, it counts as output, else it was input. */ 1564 if (in_hosteq(ip->ip_src, viftable[vifi].v_lcl_addr)) { 1565 viftable[vifi].v_pkt_out++; 1566 viftable[vifi].v_bytes_out += plen; 1567 } else { 1568 viftable[vifi].v_pkt_in++; 1569 viftable[vifi].v_bytes_in += plen; 1570 } 1571 rt->mfc_pkt_cnt++; 1572 rt->mfc_byte_cnt += plen; 1573 1574 /* 1575 * For each vif, decide if a copy of the packet should be forwarded. 1576 * Forward if: 1577 * - the ttl exceeds the vif's threshold 1578 * - there are group members downstream on interface 1579 */ 1580 for (vifi = 0; vifi < numvifs; vifi++) 1581 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1582 viftable[vifi].v_pkt_out++; 1583 viftable[vifi].v_bytes_out += plen; 1584 if (viftable[vifi].v_flags & VIFF_REGISTER) 1585 pim_register_send(ip, viftable + vifi, m, rt); 1586 else 1587 phyint_send(ip, viftable + vifi, m); 1588 } 1589 1590 /* 1591 * Perform upcall-related bw measuring. 1592 */ 1593 if (rt->mfc_bw_meter != NULL) { 1594 struct bw_meter *x; 1595 struct timeval now; 1596 1597 microtime(&now); 1598 MFC_LOCK_ASSERT(); 1599 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1600 bw_meter_receive_packet(x, plen, &now); 1601 } 1602 1603 return 0; 1604 } 1605 1606 /* 1607 * Check if a vif number is legal/ok. This is used by in_mcast.c. 1608 */ 1609 static int 1610 X_legal_vif_num(int vif) 1611 { 1612 int ret; 1613 1614 ret = 0; 1615 if (vif < 0) 1616 return (ret); 1617 1618 VIF_LOCK(); 1619 if (vif < numvifs) 1620 ret = 1; 1621 VIF_UNLOCK(); 1622 1623 return (ret); 1624 } 1625 1626 /* 1627 * Return the local address used by this vif 1628 */ 1629 static u_long 1630 X_ip_mcast_src(int vifi) 1631 { 1632 in_addr_t addr; 1633 1634 addr = INADDR_ANY; 1635 if (vifi < 0) 1636 return (addr); 1637 1638 VIF_LOCK(); 1639 if (vifi < numvifs) 1640 addr = viftable[vifi].v_lcl_addr.s_addr; 1641 VIF_UNLOCK(); 1642 1643 return (addr); 1644 } 1645 1646 static void 1647 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1648 { 1649 struct mbuf *mb_copy; 1650 int hlen = ip->ip_hl << 2; 1651 1652 VIF_LOCK_ASSERT(); 1653 1654 /* 1655 * Make a new reference to the packet; make sure that 1656 * the IP header is actually copied, not just referenced, 1657 * so that ip_output() only scribbles on the copy. 1658 */ 1659 mb_copy = m_copypacket(m, M_DONTWAIT); 1660 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1661 mb_copy = m_pullup(mb_copy, hlen); 1662 if (mb_copy == NULL) 1663 return; 1664 1665 send_packet(vifp, mb_copy); 1666 } 1667 1668 static void 1669 send_packet(struct vif *vifp, struct mbuf *m) 1670 { 1671 struct ip_moptions imo; 1672 struct in_multi *imm[2]; 1673 int error; 1674 1675 VIF_LOCK_ASSERT(); 1676 1677 imo.imo_multicast_ifp = vifp->v_ifp; 1678 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1679 imo.imo_multicast_loop = 1; 1680 imo.imo_multicast_vif = -1; 1681 imo.imo_num_memberships = 0; 1682 imo.imo_max_memberships = 2; 1683 imo.imo_membership = &imm[0]; 1684 1685 /* 1686 * Re-entrancy should not be a problem here, because 1687 * the packets that we send out and are looped back at us 1688 * should get rejected because they appear to come from 1689 * the loopback interface, thus preventing looping. 1690 */ 1691 error = ip_output(m, NULL, &vifp->v_route, IP_FORWARDING, &imo, NULL); 1692 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__, 1693 (ptrdiff_t)(vifp - viftable), error); 1694 } 1695 1696 /* 1697 * Stubs for old RSVP socket shim implementation. 1698 */ 1699 1700 static int 1701 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused) 1702 { 1703 1704 return (EOPNOTSUPP); 1705 } 1706 1707 static void 1708 X_ip_rsvp_force_done(struct socket *so __unused) 1709 { 1710 1711 } 1712 1713 static void 1714 X_rsvp_input(struct mbuf *m, int off __unused) 1715 { 1716 1717 if (!V_rsvp_on) 1718 m_freem(m); 1719 } 1720 1721 /* 1722 * Code for bandwidth monitors 1723 */ 1724 1725 /* 1726 * Define common interface for timeval-related methods 1727 */ 1728 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 1729 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 1730 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 1731 1732 static uint32_t 1733 compute_bw_meter_flags(struct bw_upcall *req) 1734 { 1735 uint32_t flags = 0; 1736 1737 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 1738 flags |= BW_METER_UNIT_PACKETS; 1739 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 1740 flags |= BW_METER_UNIT_BYTES; 1741 if (req->bu_flags & BW_UPCALL_GEQ) 1742 flags |= BW_METER_GEQ; 1743 if (req->bu_flags & BW_UPCALL_LEQ) 1744 flags |= BW_METER_LEQ; 1745 1746 return flags; 1747 } 1748 1749 /* 1750 * Add a bw_meter entry 1751 */ 1752 static int 1753 add_bw_upcall(struct bw_upcall *req) 1754 { 1755 struct mfc *mfc; 1756 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 1757 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 1758 struct timeval now; 1759 struct bw_meter *x; 1760 uint32_t flags; 1761 1762 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 1763 return EOPNOTSUPP; 1764 1765 /* Test if the flags are valid */ 1766 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 1767 return EINVAL; 1768 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 1769 return EINVAL; 1770 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 1771 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 1772 return EINVAL; 1773 1774 /* Test if the threshold time interval is valid */ 1775 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 1776 return EINVAL; 1777 1778 flags = compute_bw_meter_flags(req); 1779 1780 /* 1781 * Find if we have already same bw_meter entry 1782 */ 1783 MFC_LOCK(); 1784 mfc = mfc_find(&req->bu_src, &req->bu_dst); 1785 if (mfc == NULL) { 1786 MFC_UNLOCK(); 1787 return EADDRNOTAVAIL; 1788 } 1789 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 1790 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 1791 &req->bu_threshold.b_time, ==)) && 1792 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 1793 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 1794 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 1795 MFC_UNLOCK(); 1796 return 0; /* XXX Already installed */ 1797 } 1798 } 1799 1800 /* Allocate the new bw_meter entry */ 1801 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 1802 if (x == NULL) { 1803 MFC_UNLOCK(); 1804 return ENOBUFS; 1805 } 1806 1807 /* Set the new bw_meter entry */ 1808 x->bm_threshold.b_time = req->bu_threshold.b_time; 1809 microtime(&now); 1810 x->bm_start_time = now; 1811 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 1812 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 1813 x->bm_measured.b_packets = 0; 1814 x->bm_measured.b_bytes = 0; 1815 x->bm_flags = flags; 1816 x->bm_time_next = NULL; 1817 x->bm_time_hash = BW_METER_BUCKETS; 1818 1819 /* Add the new bw_meter entry to the front of entries for this MFC */ 1820 x->bm_mfc = mfc; 1821 x->bm_mfc_next = mfc->mfc_bw_meter; 1822 mfc->mfc_bw_meter = x; 1823 schedule_bw_meter(x, &now); 1824 MFC_UNLOCK(); 1825 1826 return 0; 1827 } 1828 1829 static void 1830 free_bw_list(struct bw_meter *list) 1831 { 1832 while (list != NULL) { 1833 struct bw_meter *x = list; 1834 1835 list = list->bm_mfc_next; 1836 unschedule_bw_meter(x); 1837 free(x, M_BWMETER); 1838 } 1839 } 1840 1841 /* 1842 * Delete one or multiple bw_meter entries 1843 */ 1844 static int 1845 del_bw_upcall(struct bw_upcall *req) 1846 { 1847 struct mfc *mfc; 1848 struct bw_meter *x; 1849 1850 if (!(mrt_api_config & MRT_MFC_BW_UPCALL)) 1851 return EOPNOTSUPP; 1852 1853 MFC_LOCK(); 1854 1855 /* Find the corresponding MFC entry */ 1856 mfc = mfc_find(&req->bu_src, &req->bu_dst); 1857 if (mfc == NULL) { 1858 MFC_UNLOCK(); 1859 return EADDRNOTAVAIL; 1860 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 1861 /* 1862 * Delete all bw_meter entries for this mfc 1863 */ 1864 struct bw_meter *list; 1865 1866 list = mfc->mfc_bw_meter; 1867 mfc->mfc_bw_meter = NULL; 1868 free_bw_list(list); 1869 MFC_UNLOCK(); 1870 return 0; 1871 } else { /* Delete a single bw_meter entry */ 1872 struct bw_meter *prev; 1873 uint32_t flags = 0; 1874 1875 flags = compute_bw_meter_flags(req); 1876 1877 /* Find the bw_meter entry to delete */ 1878 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 1879 prev = x, x = x->bm_mfc_next) { 1880 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 1881 &req->bu_threshold.b_time, ==)) && 1882 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 1883 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 1884 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 1885 break; 1886 } 1887 if (x != NULL) { /* Delete entry from the list for this MFC */ 1888 if (prev != NULL) 1889 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 1890 else 1891 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 1892 1893 unschedule_bw_meter(x); 1894 MFC_UNLOCK(); 1895 /* Free the bw_meter entry */ 1896 free(x, M_BWMETER); 1897 return 0; 1898 } else { 1899 MFC_UNLOCK(); 1900 return EINVAL; 1901 } 1902 } 1903 /* NOTREACHED */ 1904 } 1905 1906 /* 1907 * Perform bandwidth measurement processing that may result in an upcall 1908 */ 1909 static void 1910 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 1911 { 1912 struct timeval delta; 1913 1914 MFC_LOCK_ASSERT(); 1915 1916 delta = *nowp; 1917 BW_TIMEVALDECR(&delta, &x->bm_start_time); 1918 1919 if (x->bm_flags & BW_METER_GEQ) { 1920 /* 1921 * Processing for ">=" type of bw_meter entry 1922 */ 1923 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 1924 /* Reset the bw_meter entry */ 1925 x->bm_start_time = *nowp; 1926 x->bm_measured.b_packets = 0; 1927 x->bm_measured.b_bytes = 0; 1928 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 1929 } 1930 1931 /* Record that a packet is received */ 1932 x->bm_measured.b_packets++; 1933 x->bm_measured.b_bytes += plen; 1934 1935 /* 1936 * Test if we should deliver an upcall 1937 */ 1938 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 1939 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 1940 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 1941 ((x->bm_flags & BW_METER_UNIT_BYTES) && 1942 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 1943 /* Prepare an upcall for delivery */ 1944 bw_meter_prepare_upcall(x, nowp); 1945 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 1946 } 1947 } 1948 } else if (x->bm_flags & BW_METER_LEQ) { 1949 /* 1950 * Processing for "<=" type of bw_meter entry 1951 */ 1952 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 1953 /* 1954 * We are behind time with the multicast forwarding table 1955 * scanning for "<=" type of bw_meter entries, so test now 1956 * if we should deliver an upcall. 1957 */ 1958 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 1959 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 1960 ((x->bm_flags & BW_METER_UNIT_BYTES) && 1961 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 1962 /* Prepare an upcall for delivery */ 1963 bw_meter_prepare_upcall(x, nowp); 1964 } 1965 /* Reschedule the bw_meter entry */ 1966 unschedule_bw_meter(x); 1967 schedule_bw_meter(x, nowp); 1968 } 1969 1970 /* Record that a packet is received */ 1971 x->bm_measured.b_packets++; 1972 x->bm_measured.b_bytes += plen; 1973 1974 /* 1975 * Test if we should restart the measuring interval 1976 */ 1977 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 1978 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 1979 (x->bm_flags & BW_METER_UNIT_BYTES && 1980 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 1981 /* Don't restart the measuring interval */ 1982 } else { 1983 /* Do restart the measuring interval */ 1984 /* 1985 * XXX: note that we don't unschedule and schedule, because this 1986 * might be too much overhead per packet. Instead, when we process 1987 * all entries for a given timer hash bin, we check whether it is 1988 * really a timeout. If not, we reschedule at that time. 1989 */ 1990 x->bm_start_time = *nowp; 1991 x->bm_measured.b_packets = 0; 1992 x->bm_measured.b_bytes = 0; 1993 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 1994 } 1995 } 1996 } 1997 1998 /* 1999 * Prepare a bandwidth-related upcall 2000 */ 2001 static void 2002 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2003 { 2004 struct timeval delta; 2005 struct bw_upcall *u; 2006 2007 MFC_LOCK_ASSERT(); 2008 2009 /* 2010 * Compute the measured time interval 2011 */ 2012 delta = *nowp; 2013 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2014 2015 /* 2016 * If there are too many pending upcalls, deliver them now 2017 */ 2018 if (bw_upcalls_n >= BW_UPCALLS_MAX) 2019 bw_upcalls_send(); 2020 2021 /* 2022 * Set the bw_upcall entry 2023 */ 2024 u = &bw_upcalls[bw_upcalls_n++]; 2025 u->bu_src = x->bm_mfc->mfc_origin; 2026 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2027 u->bu_threshold.b_time = x->bm_threshold.b_time; 2028 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2029 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2030 u->bu_measured.b_time = delta; 2031 u->bu_measured.b_packets = x->bm_measured.b_packets; 2032 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2033 u->bu_flags = 0; 2034 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2035 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2036 if (x->bm_flags & BW_METER_UNIT_BYTES) 2037 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2038 if (x->bm_flags & BW_METER_GEQ) 2039 u->bu_flags |= BW_UPCALL_GEQ; 2040 if (x->bm_flags & BW_METER_LEQ) 2041 u->bu_flags |= BW_UPCALL_LEQ; 2042 } 2043 2044 /* 2045 * Send the pending bandwidth-related upcalls 2046 */ 2047 static void 2048 bw_upcalls_send(void) 2049 { 2050 struct mbuf *m; 2051 int len = bw_upcalls_n * sizeof(bw_upcalls[0]); 2052 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2053 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2054 0, /* unused2 */ 2055 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2056 0, /* im_mbz */ 2057 0, /* im_vif */ 2058 0, /* unused3 */ 2059 { 0 }, /* im_src */ 2060 { 0 } }; /* im_dst */ 2061 2062 MFC_LOCK_ASSERT(); 2063 2064 if (bw_upcalls_n == 0) 2065 return; /* No pending upcalls */ 2066 2067 bw_upcalls_n = 0; 2068 2069 /* 2070 * Allocate a new mbuf, initialize it with the header and 2071 * the payload for the pending calls. 2072 */ 2073 MGETHDR(m, M_DONTWAIT, MT_DATA); 2074 if (m == NULL) { 2075 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2076 return; 2077 } 2078 2079 m->m_len = m->m_pkthdr.len = 0; 2080 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2081 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&bw_upcalls[0]); 2082 2083 /* 2084 * Send the upcalls 2085 * XXX do we need to set the address in k_igmpsrc ? 2086 */ 2087 MRTSTAT_INC(mrts_upcalls); 2088 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) { 2089 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2090 MRTSTAT_INC(mrts_upq_sockfull); 2091 } 2092 } 2093 2094 /* 2095 * Compute the timeout hash value for the bw_meter entries 2096 */ 2097 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2098 do { \ 2099 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2100 \ 2101 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2102 (hash) = next_timeval.tv_sec; \ 2103 if (next_timeval.tv_usec) \ 2104 (hash)++; /* XXX: make sure we don't timeout early */ \ 2105 (hash) %= BW_METER_BUCKETS; \ 2106 } while (0) 2107 2108 /* 2109 * Schedule a timer to process periodically bw_meter entry of type "<=" 2110 * by linking the entry in the proper hash bucket. 2111 */ 2112 static void 2113 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2114 { 2115 int time_hash; 2116 2117 MFC_LOCK_ASSERT(); 2118 2119 if (!(x->bm_flags & BW_METER_LEQ)) 2120 return; /* XXX: we schedule timers only for "<=" entries */ 2121 2122 /* 2123 * Reset the bw_meter entry 2124 */ 2125 x->bm_start_time = *nowp; 2126 x->bm_measured.b_packets = 0; 2127 x->bm_measured.b_bytes = 0; 2128 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2129 2130 /* 2131 * Compute the timeout hash value and insert the entry 2132 */ 2133 BW_METER_TIMEHASH(x, time_hash); 2134 x->bm_time_next = bw_meter_timers[time_hash]; 2135 bw_meter_timers[time_hash] = x; 2136 x->bm_time_hash = time_hash; 2137 } 2138 2139 /* 2140 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2141 * by removing the entry from the proper hash bucket. 2142 */ 2143 static void 2144 unschedule_bw_meter(struct bw_meter *x) 2145 { 2146 int time_hash; 2147 struct bw_meter *prev, *tmp; 2148 2149 MFC_LOCK_ASSERT(); 2150 2151 if (!(x->bm_flags & BW_METER_LEQ)) 2152 return; /* XXX: we schedule timers only for "<=" entries */ 2153 2154 /* 2155 * Compute the timeout hash value and delete the entry 2156 */ 2157 time_hash = x->bm_time_hash; 2158 if (time_hash >= BW_METER_BUCKETS) 2159 return; /* Entry was not scheduled */ 2160 2161 for (prev = NULL, tmp = bw_meter_timers[time_hash]; 2162 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2163 if (tmp == x) 2164 break; 2165 2166 if (tmp == NULL) 2167 panic("unschedule_bw_meter: bw_meter entry not found"); 2168 2169 if (prev != NULL) 2170 prev->bm_time_next = x->bm_time_next; 2171 else 2172 bw_meter_timers[time_hash] = x->bm_time_next; 2173 2174 x->bm_time_next = NULL; 2175 x->bm_time_hash = BW_METER_BUCKETS; 2176 } 2177 2178 2179 /* 2180 * Process all "<=" type of bw_meter that should be processed now, 2181 * and for each entry prepare an upcall if necessary. Each processed 2182 * entry is rescheduled again for the (periodic) processing. 2183 * 2184 * This is run periodically (once per second normally). On each round, 2185 * all the potentially matching entries are in the hash slot that we are 2186 * looking at. 2187 */ 2188 static void 2189 bw_meter_process() 2190 { 2191 static uint32_t last_tv_sec; /* last time we processed this */ 2192 2193 uint32_t loops; 2194 int i; 2195 struct timeval now, process_endtime; 2196 2197 microtime(&now); 2198 if (last_tv_sec == now.tv_sec) 2199 return; /* nothing to do */ 2200 2201 loops = now.tv_sec - last_tv_sec; 2202 last_tv_sec = now.tv_sec; 2203 if (loops > BW_METER_BUCKETS) 2204 loops = BW_METER_BUCKETS; 2205 2206 MFC_LOCK(); 2207 /* 2208 * Process all bins of bw_meter entries from the one after the last 2209 * processed to the current one. On entry, i points to the last bucket 2210 * visited, so we need to increment i at the beginning of the loop. 2211 */ 2212 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2213 struct bw_meter *x, *tmp_list; 2214 2215 if (++i >= BW_METER_BUCKETS) 2216 i = 0; 2217 2218 /* Disconnect the list of bw_meter entries from the bin */ 2219 tmp_list = bw_meter_timers[i]; 2220 bw_meter_timers[i] = NULL; 2221 2222 /* Process the list of bw_meter entries */ 2223 while (tmp_list != NULL) { 2224 x = tmp_list; 2225 tmp_list = tmp_list->bm_time_next; 2226 2227 /* Test if the time interval is over */ 2228 process_endtime = x->bm_start_time; 2229 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2230 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2231 /* Not yet: reschedule, but don't reset */ 2232 int time_hash; 2233 2234 BW_METER_TIMEHASH(x, time_hash); 2235 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2236 /* 2237 * XXX: somehow the bin processing is a bit ahead of time. 2238 * Put the entry in the next bin. 2239 */ 2240 if (++time_hash >= BW_METER_BUCKETS) 2241 time_hash = 0; 2242 } 2243 x->bm_time_next = bw_meter_timers[time_hash]; 2244 bw_meter_timers[time_hash] = x; 2245 x->bm_time_hash = time_hash; 2246 2247 continue; 2248 } 2249 2250 /* 2251 * Test if we should deliver an upcall 2252 */ 2253 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2254 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2255 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2256 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2257 /* Prepare an upcall for delivery */ 2258 bw_meter_prepare_upcall(x, &now); 2259 } 2260 2261 /* 2262 * Reschedule for next processing 2263 */ 2264 schedule_bw_meter(x, &now); 2265 } 2266 } 2267 2268 /* Send all upcalls that are pending delivery */ 2269 bw_upcalls_send(); 2270 2271 MFC_UNLOCK(); 2272 } 2273 2274 /* 2275 * A periodic function for sending all upcalls that are pending delivery 2276 */ 2277 static void 2278 expire_bw_upcalls_send(void *unused) 2279 { 2280 MFC_LOCK(); 2281 bw_upcalls_send(); 2282 MFC_UNLOCK(); 2283 2284 callout_reset(&bw_upcalls_ch, BW_UPCALLS_PERIOD, 2285 expire_bw_upcalls_send, NULL); 2286 } 2287 2288 /* 2289 * A periodic function for periodic scanning of the multicast forwarding 2290 * table for processing all "<=" bw_meter entries. 2291 */ 2292 static void 2293 expire_bw_meter_process(void *unused) 2294 { 2295 if (mrt_api_config & MRT_MFC_BW_UPCALL) 2296 bw_meter_process(); 2297 2298 callout_reset(&bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, NULL); 2299 } 2300 2301 /* 2302 * End of bandwidth monitoring code 2303 */ 2304 2305 /* 2306 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2307 * 2308 */ 2309 static int 2310 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m, 2311 struct mfc *rt) 2312 { 2313 struct mbuf *mb_copy, *mm; 2314 2315 /* 2316 * Do not send IGMP_WHOLEPKT notifications to userland, if the 2317 * rendezvous point was unspecified, and we were told not to. 2318 */ 2319 if (pim_squelch_wholepkt != 0 && (mrt_api_config & MRT_MFC_RP) && 2320 in_nullhost(rt->mfc_rp)) 2321 return 0; 2322 2323 mb_copy = pim_register_prepare(ip, m); 2324 if (mb_copy == NULL) 2325 return ENOBUFS; 2326 2327 /* 2328 * Send all the fragments. Note that the mbuf for each fragment 2329 * is freed by the sending machinery. 2330 */ 2331 for (mm = mb_copy; mm; mm = mb_copy) { 2332 mb_copy = mm->m_nextpkt; 2333 mm->m_nextpkt = 0; 2334 mm = m_pullup(mm, sizeof(struct ip)); 2335 if (mm != NULL) { 2336 ip = mtod(mm, struct ip *); 2337 if ((mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) { 2338 pim_register_send_rp(ip, vifp, mm, rt); 2339 } else { 2340 pim_register_send_upcall(ip, vifp, mm, rt); 2341 } 2342 } 2343 } 2344 2345 return 0; 2346 } 2347 2348 /* 2349 * Return a copy of the data packet that is ready for PIM Register 2350 * encapsulation. 2351 * XXX: Note that in the returned copy the IP header is a valid one. 2352 */ 2353 static struct mbuf * 2354 pim_register_prepare(struct ip *ip, struct mbuf *m) 2355 { 2356 struct mbuf *mb_copy = NULL; 2357 int mtu; 2358 2359 /* Take care of delayed checksums */ 2360 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2361 in_delayed_cksum(m); 2362 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2363 } 2364 2365 /* 2366 * Copy the old packet & pullup its IP header into the 2367 * new mbuf so we can modify it. 2368 */ 2369 mb_copy = m_copypacket(m, M_DONTWAIT); 2370 if (mb_copy == NULL) 2371 return NULL; 2372 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2373 if (mb_copy == NULL) 2374 return NULL; 2375 2376 /* take care of the TTL */ 2377 ip = mtod(mb_copy, struct ip *); 2378 --ip->ip_ttl; 2379 2380 /* Compute the MTU after the PIM Register encapsulation */ 2381 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2382 2383 if (ip->ip_len <= mtu) { 2384 /* Turn the IP header into a valid one */ 2385 ip->ip_len = htons(ip->ip_len); 2386 ip->ip_off = htons(ip->ip_off); 2387 ip->ip_sum = 0; 2388 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2389 } else { 2390 /* Fragment the packet */ 2391 if (ip_fragment(ip, &mb_copy, mtu, 0, CSUM_DELAY_IP) != 0) { 2392 m_freem(mb_copy); 2393 return NULL; 2394 } 2395 } 2396 return mb_copy; 2397 } 2398 2399 /* 2400 * Send an upcall with the data packet to the user-level process. 2401 */ 2402 static int 2403 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2404 struct mbuf *mb_copy, struct mfc *rt) 2405 { 2406 struct mbuf *mb_first; 2407 int len = ntohs(ip->ip_len); 2408 struct igmpmsg *im; 2409 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2410 2411 VIF_LOCK_ASSERT(); 2412 2413 /* 2414 * Add a new mbuf with an upcall header 2415 */ 2416 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2417 if (mb_first == NULL) { 2418 m_freem(mb_copy); 2419 return ENOBUFS; 2420 } 2421 mb_first->m_data += max_linkhdr; 2422 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2423 mb_first->m_len = sizeof(struct igmpmsg); 2424 mb_first->m_next = mb_copy; 2425 2426 /* Send message to routing daemon */ 2427 im = mtod(mb_first, struct igmpmsg *); 2428 im->im_msgtype = IGMPMSG_WHOLEPKT; 2429 im->im_mbz = 0; 2430 im->im_vif = vifp - viftable; 2431 im->im_src = ip->ip_src; 2432 im->im_dst = ip->ip_dst; 2433 2434 k_igmpsrc.sin_addr = ip->ip_src; 2435 2436 MRTSTAT_INC(mrts_upcalls); 2437 2438 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2439 CTR1(KTR_IPMF, "%s: socket queue full", __func__); 2440 MRTSTAT_INC(mrts_upq_sockfull); 2441 return ENOBUFS; 2442 } 2443 2444 /* Keep statistics */ 2445 PIMSTAT_INC(pims_snd_registers_msgs); 2446 PIMSTAT_ADD(pims_snd_registers_bytes, len); 2447 2448 return 0; 2449 } 2450 2451 /* 2452 * Encapsulate the data packet in PIM Register message and send it to the RP. 2453 */ 2454 static int 2455 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy, 2456 struct mfc *rt) 2457 { 2458 struct mbuf *mb_first; 2459 struct ip *ip_outer; 2460 struct pim_encap_pimhdr *pimhdr; 2461 int len = ntohs(ip->ip_len); 2462 vifi_t vifi = rt->mfc_parent; 2463 2464 VIF_LOCK_ASSERT(); 2465 2466 if ((vifi >= numvifs) || in_nullhost(viftable[vifi].v_lcl_addr)) { 2467 m_freem(mb_copy); 2468 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2469 } 2470 2471 /* 2472 * Add a new mbuf with the encapsulating header 2473 */ 2474 MGETHDR(mb_first, M_DONTWAIT, MT_DATA); 2475 if (mb_first == NULL) { 2476 m_freem(mb_copy); 2477 return ENOBUFS; 2478 } 2479 mb_first->m_data += max_linkhdr; 2480 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2481 mb_first->m_next = mb_copy; 2482 2483 mb_first->m_pkthdr.len = len + mb_first->m_len; 2484 2485 /* 2486 * Fill in the encapsulating IP and PIM header 2487 */ 2488 ip_outer = mtod(mb_first, struct ip *); 2489 *ip_outer = pim_encap_iphdr; 2490 ip_outer->ip_id = ip_newid(); 2491 ip_outer->ip_len = len + sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2492 ip_outer->ip_src = viftable[vifi].v_lcl_addr; 2493 ip_outer->ip_dst = rt->mfc_rp; 2494 /* 2495 * Copy the inner header TOS to the outer header, and take care of the 2496 * IP_DF bit. 2497 */ 2498 ip_outer->ip_tos = ip->ip_tos; 2499 if (ntohs(ip->ip_off) & IP_DF) 2500 ip_outer->ip_off |= IP_DF; 2501 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2502 + sizeof(pim_encap_iphdr)); 2503 *pimhdr = pim_encap_pimhdr; 2504 /* If the iif crosses a border, set the Border-bit */ 2505 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & mrt_api_config) 2506 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2507 2508 mb_first->m_data += sizeof(pim_encap_iphdr); 2509 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2510 mb_first->m_data -= sizeof(pim_encap_iphdr); 2511 2512 send_packet(vifp, mb_first); 2513 2514 /* Keep statistics */ 2515 PIMSTAT_INC(pims_snd_registers_msgs); 2516 PIMSTAT_ADD(pims_snd_registers_bytes, len); 2517 2518 return 0; 2519 } 2520 2521 /* 2522 * pim_encapcheck() is called by the encap4_input() path at runtime to 2523 * determine if a packet is for PIM; allowing PIM to be dynamically loaded 2524 * into the kernel. 2525 */ 2526 static int 2527 pim_encapcheck(const struct mbuf *m, int off, int proto, void *arg) 2528 { 2529 2530 #ifdef DIAGNOSTIC 2531 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); 2532 #endif 2533 if (proto != IPPROTO_PIM) 2534 return 0; /* not for us; reject the datagram. */ 2535 2536 return 64; /* claim the datagram. */ 2537 } 2538 2539 /* 2540 * PIM-SMv2 and PIM-DM messages processing. 2541 * Receives and verifies the PIM control messages, and passes them 2542 * up to the listening socket, using rip_input(). 2543 * The only message with special processing is the PIM_REGISTER message 2544 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2545 * is passed to if_simloop(). 2546 */ 2547 void 2548 pim_input(struct mbuf *m, int off) 2549 { 2550 struct ip *ip = mtod(m, struct ip *); 2551 struct pim *pim; 2552 int minlen; 2553 int datalen = ip->ip_len; 2554 int ip_tos; 2555 int iphlen = off; 2556 2557 /* Keep statistics */ 2558 PIMSTAT_INC(pims_rcv_total_msgs); 2559 PIMSTAT_ADD(pims_rcv_total_bytes, datalen); 2560 2561 /* 2562 * Validate lengths 2563 */ 2564 if (datalen < PIM_MINLEN) { 2565 PIMSTAT_INC(pims_rcv_tooshort); 2566 CTR3(KTR_IPMF, "%s: short packet (%d) from %s", 2567 __func__, datalen, inet_ntoa(ip->ip_src)); 2568 m_freem(m); 2569 return; 2570 } 2571 2572 /* 2573 * If the packet is at least as big as a REGISTER, go agead 2574 * and grab the PIM REGISTER header size, to avoid another 2575 * possible m_pullup() later. 2576 * 2577 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 2578 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 2579 */ 2580 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 2581 /* 2582 * Get the IP and PIM headers in contiguous memory, and 2583 * possibly the PIM REGISTER header. 2584 */ 2585 if ((m->m_flags & M_EXT || m->m_len < minlen) && 2586 (m = m_pullup(m, minlen)) == 0) { 2587 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__); 2588 return; 2589 } 2590 2591 /* m_pullup() may have given us a new mbuf so reset ip. */ 2592 ip = mtod(m, struct ip *); 2593 ip_tos = ip->ip_tos; 2594 2595 /* adjust mbuf to point to the PIM header */ 2596 m->m_data += iphlen; 2597 m->m_len -= iphlen; 2598 pim = mtod(m, struct pim *); 2599 2600 /* 2601 * Validate checksum. If PIM REGISTER, exclude the data packet. 2602 * 2603 * XXX: some older PIMv2 implementations don't make this distinction, 2604 * so for compatibility reason perform the checksum over part of the 2605 * message, and if error, then over the whole message. 2606 */ 2607 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 2608 /* do nothing, checksum okay */ 2609 } else if (in_cksum(m, datalen)) { 2610 PIMSTAT_INC(pims_rcv_badsum); 2611 CTR1(KTR_IPMF, "%s: invalid checksum", __func__); 2612 m_freem(m); 2613 return; 2614 } 2615 2616 /* PIM version check */ 2617 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 2618 PIMSTAT_INC(pims_rcv_badversion); 2619 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__, 2620 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION); 2621 m_freem(m); 2622 return; 2623 } 2624 2625 /* restore mbuf back to the outer IP */ 2626 m->m_data -= iphlen; 2627 m->m_len += iphlen; 2628 2629 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 2630 /* 2631 * Since this is a REGISTER, we'll make a copy of the register 2632 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 2633 * routing daemon. 2634 */ 2635 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 2636 struct mbuf *mcp; 2637 struct ip *encap_ip; 2638 u_int32_t *reghdr; 2639 struct ifnet *vifp; 2640 2641 VIF_LOCK(); 2642 if ((reg_vif_num >= numvifs) || (reg_vif_num == VIFI_INVALID)) { 2643 VIF_UNLOCK(); 2644 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__, 2645 (int)reg_vif_num); 2646 m_freem(m); 2647 return; 2648 } 2649 /* XXX need refcnt? */ 2650 vifp = viftable[reg_vif_num].v_ifp; 2651 VIF_UNLOCK(); 2652 2653 /* 2654 * Validate length 2655 */ 2656 if (datalen < PIM_REG_MINLEN) { 2657 PIMSTAT_INC(pims_rcv_tooshort); 2658 PIMSTAT_INC(pims_rcv_badregisters); 2659 CTR1(KTR_IPMF, "%s: register packet size too small", __func__); 2660 m_freem(m); 2661 return; 2662 } 2663 2664 reghdr = (u_int32_t *)(pim + 1); 2665 encap_ip = (struct ip *)(reghdr + 1); 2666 2667 CTR3(KTR_IPMF, "%s: register: encap ip src %s len %d", 2668 __func__, inet_ntoa(encap_ip->ip_src), ntohs(encap_ip->ip_len)); 2669 2670 /* verify the version number of the inner packet */ 2671 if (encap_ip->ip_v != IPVERSION) { 2672 PIMSTAT_INC(pims_rcv_badregisters); 2673 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__); 2674 m_freem(m); 2675 return; 2676 } 2677 2678 /* verify the inner packet is destined to a mcast group */ 2679 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 2680 PIMSTAT_INC(pims_rcv_badregisters); 2681 CTR2(KTR_IPMF, "%s: bad encap ip dest %s", __func__, 2682 inet_ntoa(encap_ip->ip_dst)); 2683 m_freem(m); 2684 return; 2685 } 2686 2687 /* If a NULL_REGISTER, pass it to the daemon */ 2688 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 2689 goto pim_input_to_daemon; 2690 2691 /* 2692 * Copy the TOS from the outer IP header to the inner IP header. 2693 */ 2694 if (encap_ip->ip_tos != ip_tos) { 2695 /* Outer TOS -> inner TOS */ 2696 encap_ip->ip_tos = ip_tos; 2697 /* Recompute the inner header checksum. Sigh... */ 2698 2699 /* adjust mbuf to point to the inner IP header */ 2700 m->m_data += (iphlen + PIM_MINLEN); 2701 m->m_len -= (iphlen + PIM_MINLEN); 2702 2703 encap_ip->ip_sum = 0; 2704 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 2705 2706 /* restore mbuf to point back to the outer IP header */ 2707 m->m_data -= (iphlen + PIM_MINLEN); 2708 m->m_len += (iphlen + PIM_MINLEN); 2709 } 2710 2711 /* 2712 * Decapsulate the inner IP packet and loopback to forward it 2713 * as a normal multicast packet. Also, make a copy of the 2714 * outer_iphdr + pimhdr + reghdr + encap_iphdr 2715 * to pass to the daemon later, so it can take the appropriate 2716 * actions (e.g., send back PIM_REGISTER_STOP). 2717 * XXX: here m->m_data points to the outer IP header. 2718 */ 2719 mcp = m_copy(m, 0, iphlen + PIM_REG_MINLEN); 2720 if (mcp == NULL) { 2721 CTR1(KTR_IPMF, "%s: m_copy() failed", __func__); 2722 m_freem(m); 2723 return; 2724 } 2725 2726 /* Keep statistics */ 2727 /* XXX: registers_bytes include only the encap. mcast pkt */ 2728 PIMSTAT_INC(pims_rcv_registers_msgs); 2729 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len)); 2730 2731 /* 2732 * forward the inner ip packet; point m_data at the inner ip. 2733 */ 2734 m_adj(m, iphlen + PIM_MINLEN); 2735 2736 CTR4(KTR_IPMF, 2737 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d", 2738 __func__, 2739 (u_long)ntohl(encap_ip->ip_src.s_addr), 2740 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2741 (int)reg_vif_num); 2742 2743 /* NB: vifp was collected above; can it change on us? */ 2744 if_simloop(vifp, m, dst.sin_family, 0); 2745 2746 /* prepare the register head to send to the mrouting daemon */ 2747 m = mcp; 2748 } 2749 2750 pim_input_to_daemon: 2751 /* 2752 * Pass the PIM message up to the daemon; if it is a Register message, 2753 * pass the 'head' only up to the daemon. This includes the 2754 * outer IP header, PIM header, PIM-Register header and the 2755 * inner IP header. 2756 * XXX: the outer IP header pkt size of a Register is not adjust to 2757 * reflect the fact that the inner multicast data is truncated. 2758 */ 2759 rip_input(m, iphlen); 2760 2761 return; 2762 } 2763 2764 static int 2765 sysctl_mfctable(SYSCTL_HANDLER_ARGS) 2766 { 2767 struct mfc *rt; 2768 int error, i; 2769 2770 if (req->newptr) 2771 return (EPERM); 2772 if (mfchashtbl == NULL) /* XXX unlocked */ 2773 return (0); 2774 error = sysctl_wire_old_buffer(req, 0); 2775 if (error) 2776 return (error); 2777 2778 MFC_LOCK(); 2779 for (i = 0; i < mfchashsize; i++) { 2780 LIST_FOREACH(rt, &mfchashtbl[i], mfc_hash) { 2781 error = SYSCTL_OUT(req, rt, sizeof(struct mfc)); 2782 if (error) 2783 goto out_locked; 2784 } 2785 } 2786 out_locked: 2787 MFC_UNLOCK(); 2788 return (error); 2789 } 2790 2791 SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, sysctl_mfctable, 2792 "IPv4 Multicast Forwarding Table (struct *mfc[mfchashsize], " 2793 "netinet/ip_mroute.h)"); 2794 2795 static int 2796 ip_mroute_modevent(module_t mod, int type, void *unused) 2797 { 2798 2799 switch (type) { 2800 case MOD_LOAD: 2801 MROUTER_LOCK_INIT(); 2802 MFC_LOCK_INIT(); 2803 VIF_LOCK_INIT(); 2804 2805 mfchashsize = MFCHASHSIZE; 2806 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) && 2807 !powerof2(mfchashsize)) { 2808 printf("WARNING: %s not a power of 2; using default\n", 2809 "net.inet.ip.mfchashsize"); 2810 mfchashsize = MFCHASHSIZE; 2811 } 2812 MALLOC(nexpire, u_char *, mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO); 2813 2814 pim_squelch_wholepkt = 0; 2815 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", 2816 &pim_squelch_wholepkt); 2817 ip_mrouter_reset(); 2818 2819 pim_encap_cookie = encap_attach_func(AF_INET, IPPROTO_PIM, 2820 pim_encapcheck, &in_pim_protosw, NULL); 2821 if (pim_encap_cookie == NULL) { 2822 printf("ip_mroute: unable to attach pim encap\n"); 2823 VIF_LOCK_DESTROY(); 2824 MFC_LOCK_DESTROY(); 2825 MROUTER_LOCK_DESTROY(); 2826 return (EINVAL); 2827 } 2828 2829 ip_mcast_src = X_ip_mcast_src; 2830 ip_mforward = X_ip_mforward; 2831 ip_mrouter_done = X_ip_mrouter_done; 2832 ip_mrouter_get = X_ip_mrouter_get; 2833 ip_mrouter_set = X_ip_mrouter_set; 2834 2835 ip_rsvp_force_done = X_ip_rsvp_force_done; 2836 ip_rsvp_vif = X_ip_rsvp_vif; 2837 2838 legal_vif_num = X_legal_vif_num; 2839 mrt_ioctl = X_mrt_ioctl; 2840 rsvp_input_p = X_rsvp_input; 2841 break; 2842 2843 case MOD_UNLOAD: 2844 /* 2845 * Typically module unload happens after the user-level 2846 * process has shutdown the kernel services (the check 2847 * below insures someone can't just yank the module out 2848 * from under a running process). But if the module is 2849 * just loaded and then unloaded w/o starting up a user 2850 * process we still need to cleanup. 2851 */ 2852 if (V_ip_mrouter != NULL) 2853 return (EINVAL); 2854 2855 if (pim_encap_cookie) { 2856 encap_detach(pim_encap_cookie); 2857 pim_encap_cookie = NULL; 2858 } 2859 X_ip_mrouter_done(); 2860 2861 FREE(nexpire, M_MRTABLE); 2862 nexpire = NULL; 2863 2864 ip_mcast_src = NULL; 2865 ip_mforward = NULL; 2866 ip_mrouter_done = NULL; 2867 ip_mrouter_get = NULL; 2868 ip_mrouter_set = NULL; 2869 2870 ip_rsvp_force_done = NULL; 2871 ip_rsvp_vif = NULL; 2872 2873 legal_vif_num = NULL; 2874 mrt_ioctl = NULL; 2875 rsvp_input_p = NULL; 2876 2877 VIF_LOCK_DESTROY(); 2878 MFC_LOCK_DESTROY(); 2879 MROUTER_LOCK_DESTROY(); 2880 break; 2881 2882 default: 2883 return EOPNOTSUPP; 2884 } 2885 return 0; 2886 } 2887 2888 static moduledata_t ip_mroutemod = { 2889 "ip_mroute", 2890 ip_mroute_modevent, 2891 0 2892 }; 2893 2894 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PSEUDO, SI_ORDER_ANY); 2895