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