1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1989 Stephen Deering 5 * Copyright (c) 1992, 1993 6 * The Regents of the University of California. All rights reserved. 7 * 8 * This code is derived from software contributed to Berkeley by 9 * Stephen Deering of Stanford University. 10 * 11 * Redistribution and use in source and binary forms, with or without 12 * modification, are permitted provided that the following conditions 13 * are met: 14 * 1. Redistributions of source code must retain the above copyright 15 * notice, this list of conditions and the following disclaimer. 16 * 2. Redistributions in binary form must reproduce the above copyright 17 * notice, this list of conditions and the following disclaimer in the 18 * documentation and/or other materials provided with the distribution. 19 * 3. Neither the name of the University nor the names of its contributors 20 * may be used to endorse or promote products derived from this software 21 * without specific prior written permission. 22 * 23 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 24 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 25 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 26 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 27 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 28 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 29 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 30 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 31 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 32 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 33 * SUCH DAMAGE. 34 * 35 * @(#)ip_mroute.c 8.2 (Berkeley) 11/15/93 36 */ 37 38 /* 39 * IP multicast forwarding procedures 40 * 41 * Written by David Waitzman, BBN Labs, August 1988. 42 * Modified by Steve Deering, Stanford, February 1989. 43 * Modified by Mark J. Steiglitz, Stanford, May, 1991 44 * Modified by Van Jacobson, LBL, January 1993 45 * Modified by Ajit Thyagarajan, PARC, August 1993 46 * Modified by Bill Fenner, PARC, April 1995 47 * Modified by Ahmed Helmy, SGI, June 1996 48 * Modified by George Edmond Eddy (Rusty), ISI, February 1998 49 * Modified by Pavlin Radoslavov, USC/ISI, May 1998, August 1999, October 2000 50 * Modified by Hitoshi Asaeda, WIDE, August 2000 51 * Modified by Pavlin Radoslavov, ICSI, October 2002 52 * 53 * MROUTING Revision: 3.5 54 * and PIM-SMv2 and PIM-DM support, advanced API support, 55 * bandwidth metering and signaling 56 */ 57 58 /* 59 * TODO: Prefix functions with ipmf_. 60 * TODO: Maintain a refcount on if_allmulti() in ifnet or in the protocol 61 * domain attachment (if_afdata) so we can track consumers of that service. 62 * TODO: Deprecate routing socket path for SIOCGETSGCNT and SIOCGETVIFCNT, 63 * move it to socket options. 64 * TODO: Cleanup LSRR removal further. 65 * TODO: Push RSVP stubs into raw_ip.c. 66 * TODO: Use bitstring.h for vif set. 67 * TODO: Fix mrt6_ioctl dangling ref when dynamically loaded. 68 * TODO: Sync ip6_mroute.c with this file. 69 */ 70 71 #include <sys/cdefs.h> 72 __FBSDID("$FreeBSD$"); 73 74 #include "opt_inet.h" 75 #include "opt_mrouting.h" 76 77 #define _PIM_VT 1 78 79 #include <sys/param.h> 80 #include <sys/kernel.h> 81 #include <sys/stddef.h> 82 #include <sys/eventhandler.h> 83 #include <sys/lock.h> 84 #include <sys/ktr.h> 85 #include <sys/malloc.h> 86 #include <sys/mbuf.h> 87 #include <sys/module.h> 88 #include <sys/priv.h> 89 #include <sys/protosw.h> 90 #include <sys/signalvar.h> 91 #include <sys/socket.h> 92 #include <sys/socketvar.h> 93 #include <sys/sockio.h> 94 #include <sys/sx.h> 95 #include <sys/sysctl.h> 96 #include <sys/syslog.h> 97 #include <sys/systm.h> 98 #include <sys/time.h> 99 #include <sys/counter.h> 100 101 #include <net/if.h> 102 #include <net/if_var.h> 103 #include <net/netisr.h> 104 #include <net/route.h> 105 #include <net/vnet.h> 106 107 #include <netinet/in.h> 108 #include <netinet/igmp.h> 109 #include <netinet/in_systm.h> 110 #include <netinet/in_var.h> 111 #include <netinet/ip.h> 112 #include <netinet/ip_encap.h> 113 #include <netinet/ip_mroute.h> 114 #include <netinet/ip_var.h> 115 #include <netinet/ip_options.h> 116 #include <netinet/pim.h> 117 #include <netinet/pim_var.h> 118 #include <netinet/udp.h> 119 120 #include <machine/in_cksum.h> 121 122 #ifndef KTR_IPMF 123 #define KTR_IPMF KTR_INET 124 #endif 125 126 #define VIFI_INVALID ((vifi_t) -1) 127 128 VNET_DEFINE_STATIC(uint32_t, last_tv_sec); /* last time we processed this */ 129 #define V_last_tv_sec VNET(last_tv_sec) 130 131 static MALLOC_DEFINE(M_MRTABLE, "mroutetbl", "multicast forwarding cache"); 132 133 /* 134 * Locking. We use two locks: one for the virtual interface table and 135 * one for the forwarding table. These locks may be nested in which case 136 * the VIF lock must always be taken first. Note that each lock is used 137 * to cover not only the specific data structure but also related data 138 * structures. 139 */ 140 141 static struct mtx mrouter_mtx; 142 #define MROUTER_LOCK() mtx_lock(&mrouter_mtx) 143 #define MROUTER_UNLOCK() mtx_unlock(&mrouter_mtx) 144 #define MROUTER_LOCK_ASSERT() mtx_assert(&mrouter_mtx, MA_OWNED) 145 #define MROUTER_LOCK_INIT() \ 146 mtx_init(&mrouter_mtx, "IPv4 multicast forwarding", NULL, MTX_DEF) 147 #define MROUTER_LOCK_DESTROY() mtx_destroy(&mrouter_mtx) 148 149 static int ip_mrouter_cnt; /* # of vnets with active mrouters */ 150 static int ip_mrouter_unloading; /* Allow no more V_ip_mrouter sockets */ 151 152 VNET_PCPUSTAT_DEFINE_STATIC(struct mrtstat, mrtstat); 153 VNET_PCPUSTAT_SYSINIT(mrtstat); 154 VNET_PCPUSTAT_SYSUNINIT(mrtstat); 155 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, OID_AUTO, mrtstat, struct mrtstat, 156 mrtstat, "IPv4 Multicast Forwarding Statistics (struct mrtstat, " 157 "netinet/ip_mroute.h)"); 158 159 VNET_DEFINE_STATIC(u_long, mfchash); 160 #define V_mfchash VNET(mfchash) 161 #define MFCHASH(a, g) \ 162 ((((a).s_addr >> 20) ^ ((a).s_addr >> 10) ^ (a).s_addr ^ \ 163 ((g).s_addr >> 20) ^ ((g).s_addr >> 10) ^ (g).s_addr) & V_mfchash) 164 #define MFCHASHSIZE 256 165 166 static u_long mfchashsize; /* Hash size */ 167 VNET_DEFINE_STATIC(u_char *, nexpire); /* 0..mfchashsize-1 */ 168 #define V_nexpire VNET(nexpire) 169 VNET_DEFINE_STATIC(LIST_HEAD(mfchashhdr, mfc)*, mfchashtbl); 170 #define V_mfchashtbl VNET(mfchashtbl) 171 172 static struct mtx mfc_mtx; 173 #define MFC_LOCK() mtx_lock(&mfc_mtx) 174 #define MFC_UNLOCK() mtx_unlock(&mfc_mtx) 175 #define MFC_LOCK_ASSERT() mtx_assert(&mfc_mtx, MA_OWNED) 176 #define MFC_LOCK_INIT() \ 177 mtx_init(&mfc_mtx, "IPv4 multicast forwarding cache", NULL, MTX_DEF) 178 #define MFC_LOCK_DESTROY() mtx_destroy(&mfc_mtx) 179 180 VNET_DEFINE_STATIC(vifi_t, numvifs); 181 #define V_numvifs VNET(numvifs) 182 VNET_DEFINE_STATIC(struct vif, viftable[MAXVIFS]); 183 #define V_viftable VNET(viftable) 184 SYSCTL_OPAQUE(_net_inet_ip, OID_AUTO, viftable, CTLFLAG_VNET | CTLFLAG_RD, 185 &VNET_NAME(viftable), sizeof(V_viftable), "S,vif[MAXVIFS]", 186 "IPv4 Multicast Interfaces (struct vif[MAXVIFS], netinet/ip_mroute.h)"); 187 188 static struct mtx vif_mtx; 189 #define VIF_LOCK() mtx_lock(&vif_mtx) 190 #define VIF_UNLOCK() mtx_unlock(&vif_mtx) 191 #define VIF_LOCK_ASSERT() mtx_assert(&vif_mtx, MA_OWNED) 192 #define VIF_LOCK_INIT() \ 193 mtx_init(&vif_mtx, "IPv4 multicast interfaces", NULL, MTX_DEF) 194 #define VIF_LOCK_DESTROY() mtx_destroy(&vif_mtx) 195 196 static eventhandler_tag if_detach_event_tag = NULL; 197 198 VNET_DEFINE_STATIC(struct callout, expire_upcalls_ch); 199 #define V_expire_upcalls_ch VNET(expire_upcalls_ch) 200 201 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 202 #define UPCALL_EXPIRE 6 /* number of timeouts */ 203 204 /* 205 * Bandwidth meter variables and constants 206 */ 207 static MALLOC_DEFINE(M_BWMETER, "bwmeter", "multicast upcall bw meters"); 208 /* 209 * Pending timeouts are stored in a hash table, the key being the 210 * expiration time. Periodically, the entries are analysed and processed. 211 */ 212 #define BW_METER_BUCKETS 1024 213 VNET_DEFINE_STATIC(struct bw_meter*, bw_meter_timers[BW_METER_BUCKETS]); 214 #define V_bw_meter_timers VNET(bw_meter_timers) 215 VNET_DEFINE_STATIC(struct callout, bw_meter_ch); 216 #define V_bw_meter_ch VNET(bw_meter_ch) 217 #define BW_METER_PERIOD (hz) /* periodical handling of bw meters */ 218 219 /* 220 * Pending upcalls are stored in a vector which is flushed when 221 * full, or periodically 222 */ 223 VNET_DEFINE_STATIC(struct bw_upcall, bw_upcalls[BW_UPCALLS_MAX]); 224 #define V_bw_upcalls VNET(bw_upcalls) 225 VNET_DEFINE_STATIC(u_int, bw_upcalls_n); /* # of pending upcalls */ 226 #define V_bw_upcalls_n VNET(bw_upcalls_n) 227 VNET_DEFINE_STATIC(struct callout, bw_upcalls_ch); 228 #define V_bw_upcalls_ch VNET(bw_upcalls_ch) 229 230 #define BW_UPCALLS_PERIOD (hz) /* periodical flush of bw upcalls */ 231 232 VNET_PCPUSTAT_DEFINE_STATIC(struct pimstat, pimstat); 233 VNET_PCPUSTAT_SYSINIT(pimstat); 234 VNET_PCPUSTAT_SYSUNINIT(pimstat); 235 236 SYSCTL_NODE(_net_inet, IPPROTO_PIM, pim, CTLFLAG_RW, 0, "PIM"); 237 SYSCTL_VNET_PCPUSTAT(_net_inet_pim, PIMCTL_STATS, stats, struct pimstat, 238 pimstat, "PIM Statistics (struct pimstat, netinet/pim_var.h)"); 239 240 static u_long pim_squelch_wholepkt = 0; 241 SYSCTL_ULONG(_net_inet_pim, OID_AUTO, squelch_wholepkt, CTLFLAG_RW, 242 &pim_squelch_wholepkt, 0, 243 "Disable IGMP_WHOLEPKT notifications if rendezvous point is unspecified"); 244 245 static const struct encaptab *pim_encap_cookie; 246 static int pim_encapcheck(const struct mbuf *, int, int, void *); 247 static int pim_input(struct mbuf *, int, int, void *); 248 249 static const struct encap_config ipv4_encap_cfg = { 250 .proto = IPPROTO_PIM, 251 .min_length = sizeof(struct ip) + PIM_MINLEN, 252 .exact_match = 8, 253 .check = pim_encapcheck, 254 .input = pim_input 255 }; 256 257 /* 258 * Note: the PIM Register encapsulation adds the following in front of a 259 * data packet: 260 * 261 * struct pim_encap_hdr { 262 * struct ip ip; 263 * struct pim_encap_pimhdr pim; 264 * } 265 * 266 */ 267 268 struct pim_encap_pimhdr { 269 struct pim pim; 270 uint32_t flags; 271 }; 272 #define PIM_ENCAP_TTL 64 273 274 static struct ip pim_encap_iphdr = { 275 #if BYTE_ORDER == LITTLE_ENDIAN 276 sizeof(struct ip) >> 2, 277 IPVERSION, 278 #else 279 IPVERSION, 280 sizeof(struct ip) >> 2, 281 #endif 282 0, /* tos */ 283 sizeof(struct ip), /* total length */ 284 0, /* id */ 285 0, /* frag offset */ 286 PIM_ENCAP_TTL, 287 IPPROTO_PIM, 288 0, /* checksum */ 289 }; 290 291 static struct pim_encap_pimhdr pim_encap_pimhdr = { 292 { 293 PIM_MAKE_VT(PIM_VERSION, PIM_REGISTER), /* PIM vers and message type */ 294 0, /* reserved */ 295 0, /* checksum */ 296 }, 297 0 /* flags */ 298 }; 299 300 VNET_DEFINE_STATIC(vifi_t, reg_vif_num) = VIFI_INVALID; 301 #define V_reg_vif_num VNET(reg_vif_num) 302 VNET_DEFINE_STATIC(struct ifnet, multicast_register_if); 303 #define V_multicast_register_if VNET(multicast_register_if) 304 305 /* 306 * Private variables. 307 */ 308 309 static u_long X_ip_mcast_src(int); 310 static int X_ip_mforward(struct ip *, struct ifnet *, struct mbuf *, 311 struct ip_moptions *); 312 static int X_ip_mrouter_done(void); 313 static int X_ip_mrouter_get(struct socket *, struct sockopt *); 314 static int X_ip_mrouter_set(struct socket *, struct sockopt *); 315 static int X_legal_vif_num(int); 316 static int X_mrt_ioctl(u_long, caddr_t, int); 317 318 static int add_bw_upcall(struct bw_upcall *); 319 static int add_mfc(struct mfcctl2 *); 320 static int add_vif(struct vifctl *); 321 static void bw_meter_prepare_upcall(struct bw_meter *, struct timeval *); 322 static void bw_meter_process(void); 323 static void bw_meter_receive_packet(struct bw_meter *, int, 324 struct timeval *); 325 static void bw_upcalls_send(void); 326 static int del_bw_upcall(struct bw_upcall *); 327 static int del_mfc(struct mfcctl2 *); 328 static int del_vif(vifi_t); 329 static int del_vif_locked(vifi_t); 330 static void expire_bw_meter_process(void *); 331 static void expire_bw_upcalls_send(void *); 332 static void expire_mfc(struct mfc *); 333 static void expire_upcalls(void *); 334 static void free_bw_list(struct bw_meter *); 335 static int get_sg_cnt(struct sioc_sg_req *); 336 static int get_vif_cnt(struct sioc_vif_req *); 337 static void if_detached_event(void *, struct ifnet *); 338 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, vifi_t); 339 static int ip_mrouter_init(struct socket *, int); 340 static __inline struct mfc * 341 mfc_find(struct in_addr *, struct in_addr *); 342 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 343 static struct mbuf * 344 pim_register_prepare(struct ip *, struct mbuf *); 345 static int pim_register_send(struct ip *, struct vif *, 346 struct mbuf *, struct mfc *); 347 static int pim_register_send_rp(struct ip *, struct vif *, 348 struct mbuf *, struct mfc *); 349 static int pim_register_send_upcall(struct ip *, struct vif *, 350 struct mbuf *, struct mfc *); 351 static void schedule_bw_meter(struct bw_meter *, struct timeval *); 352 static void send_packet(struct vif *, struct mbuf *); 353 static int set_api_config(uint32_t *); 354 static int set_assert(int); 355 static int socket_send(struct socket *, struct mbuf *, 356 struct sockaddr_in *); 357 static void unschedule_bw_meter(struct bw_meter *); 358 359 /* 360 * Kernel multicast forwarding API capabilities and setup. 361 * If more API capabilities are added to the kernel, they should be 362 * recorded in `mrt_api_support'. 363 */ 364 #define MRT_API_VERSION 0x0305 365 366 static const int mrt_api_version = MRT_API_VERSION; 367 static const uint32_t mrt_api_support = (MRT_MFC_FLAGS_DISABLE_WRONGVIF | 368 MRT_MFC_FLAGS_BORDER_VIF | 369 MRT_MFC_RP | 370 MRT_MFC_BW_UPCALL); 371 VNET_DEFINE_STATIC(uint32_t, mrt_api_config); 372 #define V_mrt_api_config VNET(mrt_api_config) 373 VNET_DEFINE_STATIC(int, pim_assert_enabled); 374 #define V_pim_assert_enabled VNET(pim_assert_enabled) 375 static struct timeval pim_assert_interval = { 3, 0 }; /* Rate limit */ 376 377 /* 378 * Find a route for a given origin IP address and multicast group address. 379 * Statistics must be updated by the caller. 380 */ 381 static __inline struct mfc * 382 mfc_find(struct in_addr *o, struct in_addr *g) 383 { 384 struct mfc *rt; 385 386 MFC_LOCK_ASSERT(); 387 388 LIST_FOREACH(rt, &V_mfchashtbl[MFCHASH(*o, *g)], mfc_hash) { 389 if (in_hosteq(rt->mfc_origin, *o) && 390 in_hosteq(rt->mfc_mcastgrp, *g) && 391 TAILQ_EMPTY(&rt->mfc_stall)) 392 break; 393 } 394 395 return (rt); 396 } 397 398 /* 399 * Handle MRT setsockopt commands to modify the multicast forwarding tables. 400 */ 401 static int 402 X_ip_mrouter_set(struct socket *so, struct sockopt *sopt) 403 { 404 int error, optval; 405 vifi_t vifi; 406 struct vifctl vifc; 407 struct mfcctl2 mfc; 408 struct bw_upcall bw_upcall; 409 uint32_t i; 410 411 if (so != V_ip_mrouter && sopt->sopt_name != MRT_INIT) 412 return EPERM; 413 414 error = 0; 415 switch (sopt->sopt_name) { 416 case MRT_INIT: 417 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 418 if (error) 419 break; 420 error = ip_mrouter_init(so, optval); 421 break; 422 423 case MRT_DONE: 424 error = ip_mrouter_done(); 425 break; 426 427 case MRT_ADD_VIF: 428 error = sooptcopyin(sopt, &vifc, sizeof vifc, sizeof vifc); 429 if (error) 430 break; 431 error = add_vif(&vifc); 432 break; 433 434 case MRT_DEL_VIF: 435 error = sooptcopyin(sopt, &vifi, sizeof vifi, sizeof vifi); 436 if (error) 437 break; 438 error = del_vif(vifi); 439 break; 440 441 case MRT_ADD_MFC: 442 case MRT_DEL_MFC: 443 /* 444 * select data size depending on API version. 445 */ 446 if (sopt->sopt_name == MRT_ADD_MFC && 447 V_mrt_api_config & MRT_API_FLAGS_ALL) { 448 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl2), 449 sizeof(struct mfcctl2)); 450 } else { 451 error = sooptcopyin(sopt, &mfc, sizeof(struct mfcctl), 452 sizeof(struct mfcctl)); 453 bzero((caddr_t)&mfc + sizeof(struct mfcctl), 454 sizeof(mfc) - sizeof(struct mfcctl)); 455 } 456 if (error) 457 break; 458 if (sopt->sopt_name == MRT_ADD_MFC) 459 error = add_mfc(&mfc); 460 else 461 error = del_mfc(&mfc); 462 break; 463 464 case MRT_ASSERT: 465 error = sooptcopyin(sopt, &optval, sizeof optval, sizeof optval); 466 if (error) 467 break; 468 set_assert(optval); 469 break; 470 471 case MRT_API_CONFIG: 472 error = sooptcopyin(sopt, &i, sizeof i, sizeof i); 473 if (!error) 474 error = set_api_config(&i); 475 if (!error) 476 error = sooptcopyout(sopt, &i, sizeof i); 477 break; 478 479 case MRT_ADD_BW_UPCALL: 480 case MRT_DEL_BW_UPCALL: 481 error = sooptcopyin(sopt, &bw_upcall, sizeof bw_upcall, 482 sizeof bw_upcall); 483 if (error) 484 break; 485 if (sopt->sopt_name == MRT_ADD_BW_UPCALL) 486 error = add_bw_upcall(&bw_upcall); 487 else 488 error = del_bw_upcall(&bw_upcall); 489 break; 490 491 default: 492 error = EOPNOTSUPP; 493 break; 494 } 495 return error; 496 } 497 498 /* 499 * Handle MRT getsockopt commands 500 */ 501 static int 502 X_ip_mrouter_get(struct socket *so, struct sockopt *sopt) 503 { 504 int error; 505 506 switch (sopt->sopt_name) { 507 case MRT_VERSION: 508 error = sooptcopyout(sopt, &mrt_api_version, sizeof mrt_api_version); 509 break; 510 511 case MRT_ASSERT: 512 error = sooptcopyout(sopt, &V_pim_assert_enabled, 513 sizeof V_pim_assert_enabled); 514 break; 515 516 case MRT_API_SUPPORT: 517 error = sooptcopyout(sopt, &mrt_api_support, sizeof mrt_api_support); 518 break; 519 520 case MRT_API_CONFIG: 521 error = sooptcopyout(sopt, &V_mrt_api_config, sizeof V_mrt_api_config); 522 break; 523 524 default: 525 error = EOPNOTSUPP; 526 break; 527 } 528 return error; 529 } 530 531 /* 532 * Handle ioctl commands to obtain information from the cache 533 */ 534 static int 535 X_mrt_ioctl(u_long cmd, caddr_t data, int fibnum __unused) 536 { 537 int error = 0; 538 539 /* 540 * Currently the only function calling this ioctl routine is rtioctl_fib(). 541 * Typically, only root can create the raw socket in order to execute 542 * this ioctl method, however the request might be coming from a prison 543 */ 544 error = priv_check(curthread, PRIV_NETINET_MROUTE); 545 if (error) 546 return (error); 547 switch (cmd) { 548 case (SIOCGETVIFCNT): 549 error = get_vif_cnt((struct sioc_vif_req *)data); 550 break; 551 552 case (SIOCGETSGCNT): 553 error = get_sg_cnt((struct sioc_sg_req *)data); 554 break; 555 556 default: 557 error = EINVAL; 558 break; 559 } 560 return error; 561 } 562 563 /* 564 * returns the packet, byte, rpf-failure count for the source group provided 565 */ 566 static int 567 get_sg_cnt(struct sioc_sg_req *req) 568 { 569 struct mfc *rt; 570 571 MFC_LOCK(); 572 rt = mfc_find(&req->src, &req->grp); 573 if (rt == NULL) { 574 MFC_UNLOCK(); 575 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 576 return EADDRNOTAVAIL; 577 } 578 req->pktcnt = rt->mfc_pkt_cnt; 579 req->bytecnt = rt->mfc_byte_cnt; 580 req->wrong_if = rt->mfc_wrong_if; 581 MFC_UNLOCK(); 582 return 0; 583 } 584 585 /* 586 * returns the input and output packet and byte counts on the vif provided 587 */ 588 static int 589 get_vif_cnt(struct sioc_vif_req *req) 590 { 591 vifi_t vifi = req->vifi; 592 593 VIF_LOCK(); 594 if (vifi >= V_numvifs) { 595 VIF_UNLOCK(); 596 return EINVAL; 597 } 598 599 req->icount = V_viftable[vifi].v_pkt_in; 600 req->ocount = V_viftable[vifi].v_pkt_out; 601 req->ibytes = V_viftable[vifi].v_bytes_in; 602 req->obytes = V_viftable[vifi].v_bytes_out; 603 VIF_UNLOCK(); 604 605 return 0; 606 } 607 608 static void 609 if_detached_event(void *arg __unused, struct ifnet *ifp) 610 { 611 vifi_t vifi; 612 u_long i; 613 614 MROUTER_LOCK(); 615 616 if (V_ip_mrouter == NULL) { 617 MROUTER_UNLOCK(); 618 return; 619 } 620 621 VIF_LOCK(); 622 MFC_LOCK(); 623 624 /* 625 * Tear down multicast forwarder state associated with this ifnet. 626 * 1. Walk the vif list, matching vifs against this ifnet. 627 * 2. Walk the multicast forwarding cache (mfc) looking for 628 * inner matches with this vif's index. 629 * 3. Expire any matching multicast forwarding cache entries. 630 * 4. Free vif state. This should disable ALLMULTI on the interface. 631 */ 632 for (vifi = 0; vifi < V_numvifs; vifi++) { 633 if (V_viftable[vifi].v_ifp != ifp) 634 continue; 635 for (i = 0; i < mfchashsize; i++) { 636 struct mfc *rt, *nrt; 637 638 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) { 639 if (rt->mfc_parent == vifi) { 640 expire_mfc(rt); 641 } 642 } 643 } 644 del_vif_locked(vifi); 645 } 646 647 MFC_UNLOCK(); 648 VIF_UNLOCK(); 649 650 MROUTER_UNLOCK(); 651 } 652 653 /* 654 * Enable multicast forwarding. 655 */ 656 static int 657 ip_mrouter_init(struct socket *so, int version) 658 { 659 660 CTR3(KTR_IPMF, "%s: so_type %d, pr_protocol %d", __func__, 661 so->so_type, so->so_proto->pr_protocol); 662 663 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_IGMP) 664 return EOPNOTSUPP; 665 666 if (version != 1) 667 return ENOPROTOOPT; 668 669 MROUTER_LOCK(); 670 671 if (ip_mrouter_unloading) { 672 MROUTER_UNLOCK(); 673 return ENOPROTOOPT; 674 } 675 676 if (V_ip_mrouter != NULL) { 677 MROUTER_UNLOCK(); 678 return EADDRINUSE; 679 } 680 681 V_mfchashtbl = hashinit_flags(mfchashsize, M_MRTABLE, &V_mfchash, 682 HASH_NOWAIT); 683 684 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, 685 curvnet); 686 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send, 687 curvnet); 688 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, 689 curvnet); 690 691 V_ip_mrouter = so; 692 ip_mrouter_cnt++; 693 694 MROUTER_UNLOCK(); 695 696 CTR1(KTR_IPMF, "%s: done", __func__); 697 698 return 0; 699 } 700 701 /* 702 * Disable multicast forwarding. 703 */ 704 static int 705 X_ip_mrouter_done(void) 706 { 707 struct ifnet *ifp; 708 u_long i; 709 vifi_t vifi; 710 711 MROUTER_LOCK(); 712 713 if (V_ip_mrouter == NULL) { 714 MROUTER_UNLOCK(); 715 return EINVAL; 716 } 717 718 /* 719 * Detach/disable hooks to the reset of the system. 720 */ 721 V_ip_mrouter = NULL; 722 ip_mrouter_cnt--; 723 V_mrt_api_config = 0; 724 725 VIF_LOCK(); 726 727 /* 728 * For each phyint in use, disable promiscuous reception of all IP 729 * multicasts. 730 */ 731 for (vifi = 0; vifi < V_numvifs; vifi++) { 732 if (!in_nullhost(V_viftable[vifi].v_lcl_addr) && 733 !(V_viftable[vifi].v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) { 734 ifp = V_viftable[vifi].v_ifp; 735 if_allmulti(ifp, 0); 736 } 737 } 738 bzero((caddr_t)V_viftable, sizeof(V_viftable)); 739 V_numvifs = 0; 740 V_pim_assert_enabled = 0; 741 742 VIF_UNLOCK(); 743 744 callout_stop(&V_expire_upcalls_ch); 745 callout_stop(&V_bw_upcalls_ch); 746 callout_stop(&V_bw_meter_ch); 747 748 MFC_LOCK(); 749 750 /* 751 * Free all multicast forwarding cache entries. 752 * Do not use hashdestroy(), as we must perform other cleanup. 753 */ 754 for (i = 0; i < mfchashsize; i++) { 755 struct mfc *rt, *nrt; 756 757 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) { 758 expire_mfc(rt); 759 } 760 } 761 free(V_mfchashtbl, M_MRTABLE); 762 V_mfchashtbl = NULL; 763 764 bzero(V_nexpire, sizeof(V_nexpire[0]) * mfchashsize); 765 766 V_bw_upcalls_n = 0; 767 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers)); 768 769 MFC_UNLOCK(); 770 771 V_reg_vif_num = VIFI_INVALID; 772 773 MROUTER_UNLOCK(); 774 775 CTR1(KTR_IPMF, "%s: done", __func__); 776 777 return 0; 778 } 779 780 /* 781 * Set PIM assert processing global 782 */ 783 static int 784 set_assert(int i) 785 { 786 if ((i != 1) && (i != 0)) 787 return EINVAL; 788 789 V_pim_assert_enabled = i; 790 791 return 0; 792 } 793 794 /* 795 * Configure API capabilities 796 */ 797 int 798 set_api_config(uint32_t *apival) 799 { 800 u_long i; 801 802 /* 803 * We can set the API capabilities only if it is the first operation 804 * after MRT_INIT. I.e.: 805 * - there are no vifs installed 806 * - pim_assert is not enabled 807 * - the MFC table is empty 808 */ 809 if (V_numvifs > 0) { 810 *apival = 0; 811 return EPERM; 812 } 813 if (V_pim_assert_enabled) { 814 *apival = 0; 815 return EPERM; 816 } 817 818 MFC_LOCK(); 819 820 for (i = 0; i < mfchashsize; i++) { 821 if (LIST_FIRST(&V_mfchashtbl[i]) != NULL) { 822 MFC_UNLOCK(); 823 *apival = 0; 824 return EPERM; 825 } 826 } 827 828 MFC_UNLOCK(); 829 830 V_mrt_api_config = *apival & mrt_api_support; 831 *apival = V_mrt_api_config; 832 833 return 0; 834 } 835 836 /* 837 * Add a vif to the vif table 838 */ 839 static int 840 add_vif(struct vifctl *vifcp) 841 { 842 struct vif *vifp = V_viftable + vifcp->vifc_vifi; 843 struct sockaddr_in sin = {sizeof sin, AF_INET}; 844 struct ifaddr *ifa; 845 struct ifnet *ifp; 846 int error; 847 848 VIF_LOCK(); 849 if (vifcp->vifc_vifi >= MAXVIFS) { 850 VIF_UNLOCK(); 851 return EINVAL; 852 } 853 /* rate limiting is no longer supported by this code */ 854 if (vifcp->vifc_rate_limit != 0) { 855 log(LOG_ERR, "rate limiting is no longer supported\n"); 856 VIF_UNLOCK(); 857 return EINVAL; 858 } 859 if (!in_nullhost(vifp->v_lcl_addr)) { 860 VIF_UNLOCK(); 861 return EADDRINUSE; 862 } 863 if (in_nullhost(vifcp->vifc_lcl_addr)) { 864 VIF_UNLOCK(); 865 return EADDRNOTAVAIL; 866 } 867 868 /* Find the interface with an address in AF_INET family */ 869 if (vifcp->vifc_flags & VIFF_REGISTER) { 870 /* 871 * XXX: Because VIFF_REGISTER does not really need a valid 872 * local interface (e.g. it could be 127.0.0.2), we don't 873 * check its address. 874 */ 875 ifp = NULL; 876 } else { 877 sin.sin_addr = vifcp->vifc_lcl_addr; 878 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 879 if (ifa == NULL) { 880 VIF_UNLOCK(); 881 return EADDRNOTAVAIL; 882 } 883 ifp = ifa->ifa_ifp; 884 } 885 886 if ((vifcp->vifc_flags & VIFF_TUNNEL) != 0) { 887 CTR1(KTR_IPMF, "%s: tunnels are no longer supported", __func__); 888 VIF_UNLOCK(); 889 return EOPNOTSUPP; 890 } else if (vifcp->vifc_flags & VIFF_REGISTER) { 891 ifp = &V_multicast_register_if; 892 CTR2(KTR_IPMF, "%s: add register vif for ifp %p", __func__, ifp); 893 if (V_reg_vif_num == VIFI_INVALID) { 894 if_initname(&V_multicast_register_if, "register_vif", 0); 895 V_multicast_register_if.if_flags = IFF_LOOPBACK; 896 V_reg_vif_num = vifcp->vifc_vifi; 897 } 898 } else { /* Make sure the interface supports multicast */ 899 if ((ifp->if_flags & IFF_MULTICAST) == 0) { 900 VIF_UNLOCK(); 901 return EOPNOTSUPP; 902 } 903 904 /* Enable promiscuous reception of all IP multicasts from the if */ 905 error = if_allmulti(ifp, 1); 906 if (error) { 907 VIF_UNLOCK(); 908 return error; 909 } 910 } 911 912 vifp->v_flags = vifcp->vifc_flags; 913 vifp->v_threshold = vifcp->vifc_threshold; 914 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 915 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 916 vifp->v_ifp = ifp; 917 /* initialize per vif pkt counters */ 918 vifp->v_pkt_in = 0; 919 vifp->v_pkt_out = 0; 920 vifp->v_bytes_in = 0; 921 vifp->v_bytes_out = 0; 922 923 /* Adjust numvifs up if the vifi is higher than numvifs */ 924 if (V_numvifs <= vifcp->vifc_vifi) 925 V_numvifs = vifcp->vifc_vifi + 1; 926 927 VIF_UNLOCK(); 928 929 CTR4(KTR_IPMF, "%s: add vif %d laddr 0x%08x thresh %x", __func__, 930 (int)vifcp->vifc_vifi, ntohl(vifcp->vifc_lcl_addr.s_addr), 931 (int)vifcp->vifc_threshold); 932 933 return 0; 934 } 935 936 /* 937 * Delete a vif from the vif table 938 */ 939 static int 940 del_vif_locked(vifi_t vifi) 941 { 942 struct vif *vifp; 943 944 VIF_LOCK_ASSERT(); 945 946 if (vifi >= V_numvifs) { 947 return EINVAL; 948 } 949 vifp = &V_viftable[vifi]; 950 if (in_nullhost(vifp->v_lcl_addr)) { 951 return EADDRNOTAVAIL; 952 } 953 954 if (!(vifp->v_flags & (VIFF_TUNNEL | VIFF_REGISTER))) 955 if_allmulti(vifp->v_ifp, 0); 956 957 if (vifp->v_flags & VIFF_REGISTER) 958 V_reg_vif_num = VIFI_INVALID; 959 960 bzero((caddr_t)vifp, sizeof (*vifp)); 961 962 CTR2(KTR_IPMF, "%s: delete vif %d", __func__, (int)vifi); 963 964 /* Adjust numvifs down */ 965 for (vifi = V_numvifs; vifi > 0; vifi--) 966 if (!in_nullhost(V_viftable[vifi-1].v_lcl_addr)) 967 break; 968 V_numvifs = vifi; 969 970 return 0; 971 } 972 973 static int 974 del_vif(vifi_t vifi) 975 { 976 int cc; 977 978 VIF_LOCK(); 979 cc = del_vif_locked(vifi); 980 VIF_UNLOCK(); 981 982 return cc; 983 } 984 985 /* 986 * update an mfc entry without resetting counters and S,G addresses. 987 */ 988 static void 989 update_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 990 { 991 int i; 992 993 rt->mfc_parent = mfccp->mfcc_parent; 994 for (i = 0; i < V_numvifs; i++) { 995 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 996 rt->mfc_flags[i] = mfccp->mfcc_flags[i] & V_mrt_api_config & 997 MRT_MFC_FLAGS_ALL; 998 } 999 /* set the RP address */ 1000 if (V_mrt_api_config & MRT_MFC_RP) 1001 rt->mfc_rp = mfccp->mfcc_rp; 1002 else 1003 rt->mfc_rp.s_addr = INADDR_ANY; 1004 } 1005 1006 /* 1007 * fully initialize an mfc entry from the parameter. 1008 */ 1009 static void 1010 init_mfc_params(struct mfc *rt, struct mfcctl2 *mfccp) 1011 { 1012 rt->mfc_origin = mfccp->mfcc_origin; 1013 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 1014 1015 update_mfc_params(rt, mfccp); 1016 1017 /* initialize pkt counters per src-grp */ 1018 rt->mfc_pkt_cnt = 0; 1019 rt->mfc_byte_cnt = 0; 1020 rt->mfc_wrong_if = 0; 1021 timevalclear(&rt->mfc_last_assert); 1022 } 1023 1024 static void 1025 expire_mfc(struct mfc *rt) 1026 { 1027 struct rtdetq *rte, *nrte; 1028 1029 MFC_LOCK_ASSERT(); 1030 1031 free_bw_list(rt->mfc_bw_meter); 1032 1033 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) { 1034 m_freem(rte->m); 1035 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link); 1036 free(rte, M_MRTABLE); 1037 } 1038 1039 LIST_REMOVE(rt, mfc_hash); 1040 free(rt, M_MRTABLE); 1041 } 1042 1043 /* 1044 * Add an mfc entry 1045 */ 1046 static int 1047 add_mfc(struct mfcctl2 *mfccp) 1048 { 1049 struct mfc *rt; 1050 struct rtdetq *rte, *nrte; 1051 u_long hash = 0; 1052 u_short nstl; 1053 1054 VIF_LOCK(); 1055 MFC_LOCK(); 1056 1057 rt = mfc_find(&mfccp->mfcc_origin, &mfccp->mfcc_mcastgrp); 1058 1059 /* If an entry already exists, just update the fields */ 1060 if (rt) { 1061 CTR4(KTR_IPMF, "%s: update mfc orig 0x%08x group %lx parent %x", 1062 __func__, ntohl(mfccp->mfcc_origin.s_addr), 1063 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1064 mfccp->mfcc_parent); 1065 update_mfc_params(rt, mfccp); 1066 MFC_UNLOCK(); 1067 VIF_UNLOCK(); 1068 return (0); 1069 } 1070 1071 /* 1072 * Find the entry for which the upcall was made and update 1073 */ 1074 nstl = 0; 1075 hash = MFCHASH(mfccp->mfcc_origin, mfccp->mfcc_mcastgrp); 1076 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) { 1077 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1078 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp) && 1079 !TAILQ_EMPTY(&rt->mfc_stall)) { 1080 CTR5(KTR_IPMF, 1081 "%s: add mfc orig 0x%08x group %lx parent %x qh %p", 1082 __func__, ntohl(mfccp->mfcc_origin.s_addr), 1083 (u_long)ntohl(mfccp->mfcc_mcastgrp.s_addr), 1084 mfccp->mfcc_parent, 1085 TAILQ_FIRST(&rt->mfc_stall)); 1086 if (nstl++) 1087 CTR1(KTR_IPMF, "%s: multiple matches", __func__); 1088 1089 init_mfc_params(rt, mfccp); 1090 rt->mfc_expire = 0; /* Don't clean this guy up */ 1091 V_nexpire[hash]--; 1092 1093 /* Free queued packets, but attempt to forward them first. */ 1094 TAILQ_FOREACH_SAFE(rte, &rt->mfc_stall, rte_link, nrte) { 1095 if (rte->ifp != NULL) 1096 ip_mdq(rte->m, rte->ifp, rt, -1); 1097 m_freem(rte->m); 1098 TAILQ_REMOVE(&rt->mfc_stall, rte, rte_link); 1099 rt->mfc_nstall--; 1100 free(rte, M_MRTABLE); 1101 } 1102 } 1103 } 1104 1105 /* 1106 * It is possible that an entry is being inserted without an upcall 1107 */ 1108 if (nstl == 0) { 1109 CTR1(KTR_IPMF, "%s: adding mfc w/o upcall", __func__); 1110 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) { 1111 if (in_hosteq(rt->mfc_origin, mfccp->mfcc_origin) && 1112 in_hosteq(rt->mfc_mcastgrp, mfccp->mfcc_mcastgrp)) { 1113 init_mfc_params(rt, mfccp); 1114 if (rt->mfc_expire) 1115 V_nexpire[hash]--; 1116 rt->mfc_expire = 0; 1117 break; /* XXX */ 1118 } 1119 } 1120 1121 if (rt == NULL) { /* no upcall, so make a new entry */ 1122 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1123 if (rt == NULL) { 1124 MFC_UNLOCK(); 1125 VIF_UNLOCK(); 1126 return (ENOBUFS); 1127 } 1128 1129 init_mfc_params(rt, mfccp); 1130 TAILQ_INIT(&rt->mfc_stall); 1131 rt->mfc_nstall = 0; 1132 1133 rt->mfc_expire = 0; 1134 rt->mfc_bw_meter = NULL; 1135 1136 /* insert new entry at head of hash chain */ 1137 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash); 1138 } 1139 } 1140 1141 MFC_UNLOCK(); 1142 VIF_UNLOCK(); 1143 1144 return (0); 1145 } 1146 1147 /* 1148 * Delete an mfc entry 1149 */ 1150 static int 1151 del_mfc(struct mfcctl2 *mfccp) 1152 { 1153 struct in_addr origin; 1154 struct in_addr mcastgrp; 1155 struct mfc *rt; 1156 1157 origin = mfccp->mfcc_origin; 1158 mcastgrp = mfccp->mfcc_mcastgrp; 1159 1160 CTR3(KTR_IPMF, "%s: delete mfc orig 0x%08x group %lx", __func__, 1161 ntohl(origin.s_addr), (u_long)ntohl(mcastgrp.s_addr)); 1162 1163 MFC_LOCK(); 1164 1165 rt = mfc_find(&origin, &mcastgrp); 1166 if (rt == NULL) { 1167 MFC_UNLOCK(); 1168 return EADDRNOTAVAIL; 1169 } 1170 1171 /* 1172 * free the bw_meter entries 1173 */ 1174 free_bw_list(rt->mfc_bw_meter); 1175 rt->mfc_bw_meter = NULL; 1176 1177 LIST_REMOVE(rt, mfc_hash); 1178 free(rt, M_MRTABLE); 1179 1180 MFC_UNLOCK(); 1181 1182 return (0); 1183 } 1184 1185 /* 1186 * Send a message to the routing daemon on the multicast routing socket. 1187 */ 1188 static int 1189 socket_send(struct socket *s, struct mbuf *mm, struct sockaddr_in *src) 1190 { 1191 if (s) { 1192 SOCKBUF_LOCK(&s->so_rcv); 1193 if (sbappendaddr_locked(&s->so_rcv, (struct sockaddr *)src, mm, 1194 NULL) != 0) { 1195 sorwakeup_locked(s); 1196 return 0; 1197 } 1198 SOCKBUF_UNLOCK(&s->so_rcv); 1199 } 1200 m_freem(mm); 1201 return -1; 1202 } 1203 1204 /* 1205 * IP multicast forwarding function. This function assumes that the packet 1206 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1207 * pointed to by "ifp", and the packet is to be relayed to other networks 1208 * that have members of the packet's destination IP multicast group. 1209 * 1210 * The packet is returned unscathed to the caller, unless it is 1211 * erroneous, in which case a non-zero return value tells the caller to 1212 * discard it. 1213 */ 1214 1215 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1216 1217 static int 1218 X_ip_mforward(struct ip *ip, struct ifnet *ifp, struct mbuf *m, 1219 struct ip_moptions *imo) 1220 { 1221 struct mfc *rt; 1222 int error; 1223 vifi_t vifi; 1224 1225 CTR3(KTR_IPMF, "ip_mforward: delete mfc orig 0x%08x group %lx ifp %p", 1226 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr), ifp); 1227 1228 if (ip->ip_hl < (sizeof(struct ip) + TUNNEL_LEN) >> 2 || 1229 ((u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1230 /* 1231 * Packet arrived via a physical interface or 1232 * an encapsulated tunnel or a register_vif. 1233 */ 1234 } else { 1235 /* 1236 * Packet arrived through a source-route tunnel. 1237 * Source-route tunnels are no longer supported. 1238 */ 1239 return (1); 1240 } 1241 1242 VIF_LOCK(); 1243 MFC_LOCK(); 1244 if (imo && ((vifi = imo->imo_multicast_vif) < V_numvifs)) { 1245 if (ip->ip_ttl < MAXTTL) 1246 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1247 error = ip_mdq(m, ifp, NULL, vifi); 1248 MFC_UNLOCK(); 1249 VIF_UNLOCK(); 1250 return error; 1251 } 1252 1253 /* 1254 * Don't forward a packet with time-to-live of zero or one, 1255 * or a packet destined to a local-only group. 1256 */ 1257 if (ip->ip_ttl <= 1 || IN_LOCAL_GROUP(ntohl(ip->ip_dst.s_addr))) { 1258 MFC_UNLOCK(); 1259 VIF_UNLOCK(); 1260 return 0; 1261 } 1262 1263 /* 1264 * Determine forwarding vifs from the forwarding cache table 1265 */ 1266 MRTSTAT_INC(mrts_mfc_lookups); 1267 rt = mfc_find(&ip->ip_src, &ip->ip_dst); 1268 1269 /* Entry exists, so forward if necessary */ 1270 if (rt != NULL) { 1271 error = ip_mdq(m, ifp, rt, -1); 1272 MFC_UNLOCK(); 1273 VIF_UNLOCK(); 1274 return error; 1275 } else { 1276 /* 1277 * If we don't have a route for packet's origin, 1278 * Make a copy of the packet & send message to routing daemon 1279 */ 1280 1281 struct mbuf *mb0; 1282 struct rtdetq *rte; 1283 u_long hash; 1284 int hlen = ip->ip_hl << 2; 1285 1286 MRTSTAT_INC(mrts_mfc_misses); 1287 MRTSTAT_INC(mrts_no_route); 1288 CTR2(KTR_IPMF, "ip_mforward: no mfc for (0x%08x,%lx)", 1289 ntohl(ip->ip_src.s_addr), (u_long)ntohl(ip->ip_dst.s_addr)); 1290 1291 /* 1292 * Allocate mbufs early so that we don't do extra work if we are 1293 * just going to fail anyway. Make sure to pullup the header so 1294 * that other people can't step on it. 1295 */ 1296 rte = (struct rtdetq *)malloc((sizeof *rte), M_MRTABLE, 1297 M_NOWAIT|M_ZERO); 1298 if (rte == NULL) { 1299 MFC_UNLOCK(); 1300 VIF_UNLOCK(); 1301 return ENOBUFS; 1302 } 1303 1304 mb0 = m_copypacket(m, M_NOWAIT); 1305 if (mb0 && (!M_WRITABLE(mb0) || mb0->m_len < hlen)) 1306 mb0 = m_pullup(mb0, hlen); 1307 if (mb0 == NULL) { 1308 free(rte, M_MRTABLE); 1309 MFC_UNLOCK(); 1310 VIF_UNLOCK(); 1311 return ENOBUFS; 1312 } 1313 1314 /* is there an upcall waiting for this flow ? */ 1315 hash = MFCHASH(ip->ip_src, ip->ip_dst); 1316 LIST_FOREACH(rt, &V_mfchashtbl[hash], mfc_hash) { 1317 if (in_hosteq(ip->ip_src, rt->mfc_origin) && 1318 in_hosteq(ip->ip_dst, rt->mfc_mcastgrp) && 1319 !TAILQ_EMPTY(&rt->mfc_stall)) 1320 break; 1321 } 1322 1323 if (rt == NULL) { 1324 int i; 1325 struct igmpmsg *im; 1326 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1327 struct mbuf *mm; 1328 1329 /* 1330 * Locate the vifi for the incoming interface for this packet. 1331 * If none found, drop packet. 1332 */ 1333 for (vifi = 0; vifi < V_numvifs && 1334 V_viftable[vifi].v_ifp != ifp; vifi++) 1335 ; 1336 if (vifi >= V_numvifs) /* vif not found, drop packet */ 1337 goto non_fatal; 1338 1339 /* no upcall, so make a new entry */ 1340 rt = (struct mfc *)malloc(sizeof(*rt), M_MRTABLE, M_NOWAIT); 1341 if (rt == NULL) 1342 goto fail; 1343 1344 /* Make a copy of the header to send to the user level process */ 1345 mm = m_copym(mb0, 0, hlen, M_NOWAIT); 1346 if (mm == NULL) 1347 goto fail1; 1348 1349 /* 1350 * Send message to routing daemon to install 1351 * a route into the kernel table 1352 */ 1353 1354 im = mtod(mm, struct igmpmsg *); 1355 im->im_msgtype = IGMPMSG_NOCACHE; 1356 im->im_mbz = 0; 1357 im->im_vif = vifi; 1358 1359 MRTSTAT_INC(mrts_upcalls); 1360 1361 k_igmpsrc.sin_addr = ip->ip_src; 1362 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { 1363 CTR0(KTR_IPMF, "ip_mforward: socket queue full"); 1364 MRTSTAT_INC(mrts_upq_sockfull); 1365 fail1: 1366 free(rt, M_MRTABLE); 1367 fail: 1368 free(rte, M_MRTABLE); 1369 m_freem(mb0); 1370 MFC_UNLOCK(); 1371 VIF_UNLOCK(); 1372 return ENOBUFS; 1373 } 1374 1375 /* insert new entry at head of hash chain */ 1376 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1377 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1378 rt->mfc_expire = UPCALL_EXPIRE; 1379 V_nexpire[hash]++; 1380 for (i = 0; i < V_numvifs; i++) { 1381 rt->mfc_ttls[i] = 0; 1382 rt->mfc_flags[i] = 0; 1383 } 1384 rt->mfc_parent = -1; 1385 1386 /* clear the RP address */ 1387 rt->mfc_rp.s_addr = INADDR_ANY; 1388 rt->mfc_bw_meter = NULL; 1389 1390 /* initialize pkt counters per src-grp */ 1391 rt->mfc_pkt_cnt = 0; 1392 rt->mfc_byte_cnt = 0; 1393 rt->mfc_wrong_if = 0; 1394 timevalclear(&rt->mfc_last_assert); 1395 1396 TAILQ_INIT(&rt->mfc_stall); 1397 rt->mfc_nstall = 0; 1398 1399 /* link into table */ 1400 LIST_INSERT_HEAD(&V_mfchashtbl[hash], rt, mfc_hash); 1401 TAILQ_INSERT_HEAD(&rt->mfc_stall, rte, rte_link); 1402 rt->mfc_nstall++; 1403 1404 } else { 1405 /* determine if queue has overflowed */ 1406 if (rt->mfc_nstall > MAX_UPQ) { 1407 MRTSTAT_INC(mrts_upq_ovflw); 1408 non_fatal: 1409 free(rte, M_MRTABLE); 1410 m_freem(mb0); 1411 MFC_UNLOCK(); 1412 VIF_UNLOCK(); 1413 return (0); 1414 } 1415 TAILQ_INSERT_TAIL(&rt->mfc_stall, rte, rte_link); 1416 rt->mfc_nstall++; 1417 } 1418 1419 rte->m = mb0; 1420 rte->ifp = ifp; 1421 1422 MFC_UNLOCK(); 1423 VIF_UNLOCK(); 1424 1425 return 0; 1426 } 1427 } 1428 1429 /* 1430 * Clean up the cache entry if upcall is not serviced 1431 */ 1432 static void 1433 expire_upcalls(void *arg) 1434 { 1435 u_long i; 1436 1437 CURVNET_SET((struct vnet *) arg); 1438 1439 MFC_LOCK(); 1440 1441 for (i = 0; i < mfchashsize; i++) { 1442 struct mfc *rt, *nrt; 1443 1444 if (V_nexpire[i] == 0) 1445 continue; 1446 1447 LIST_FOREACH_SAFE(rt, &V_mfchashtbl[i], mfc_hash, nrt) { 1448 if (TAILQ_EMPTY(&rt->mfc_stall)) 1449 continue; 1450 1451 if (rt->mfc_expire == 0 || --rt->mfc_expire > 0) 1452 continue; 1453 1454 /* 1455 * free the bw_meter entries 1456 */ 1457 while (rt->mfc_bw_meter != NULL) { 1458 struct bw_meter *x = rt->mfc_bw_meter; 1459 1460 rt->mfc_bw_meter = x->bm_mfc_next; 1461 free(x, M_BWMETER); 1462 } 1463 1464 MRTSTAT_INC(mrts_cache_cleanups); 1465 CTR3(KTR_IPMF, "%s: expire (%lx, %lx)", __func__, 1466 (u_long)ntohl(rt->mfc_origin.s_addr), 1467 (u_long)ntohl(rt->mfc_mcastgrp.s_addr)); 1468 1469 expire_mfc(rt); 1470 } 1471 } 1472 1473 MFC_UNLOCK(); 1474 1475 callout_reset(&V_expire_upcalls_ch, EXPIRE_TIMEOUT, expire_upcalls, 1476 curvnet); 1477 1478 CURVNET_RESTORE(); 1479 } 1480 1481 /* 1482 * Packet forwarding routine once entry in the cache is made 1483 */ 1484 static int 1485 ip_mdq(struct mbuf *m, struct ifnet *ifp, struct mfc *rt, vifi_t xmt_vif) 1486 { 1487 struct ip *ip = mtod(m, struct ip *); 1488 vifi_t vifi; 1489 int plen = ntohs(ip->ip_len); 1490 1491 VIF_LOCK_ASSERT(); 1492 1493 /* 1494 * If xmt_vif is not -1, send on only the requested vif. 1495 * 1496 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1497 */ 1498 if (xmt_vif < V_numvifs) { 1499 if (V_viftable[xmt_vif].v_flags & VIFF_REGISTER) 1500 pim_register_send(ip, V_viftable + xmt_vif, m, rt); 1501 else 1502 phyint_send(ip, V_viftable + xmt_vif, m); 1503 return 1; 1504 } 1505 1506 /* 1507 * Don't forward if it didn't arrive from the parent vif for its origin. 1508 */ 1509 vifi = rt->mfc_parent; 1510 if ((vifi >= V_numvifs) || (V_viftable[vifi].v_ifp != ifp)) { 1511 CTR4(KTR_IPMF, "%s: rx on wrong ifp %p (vifi %d, v_ifp %p)", 1512 __func__, ifp, (int)vifi, V_viftable[vifi].v_ifp); 1513 MRTSTAT_INC(mrts_wrong_if); 1514 ++rt->mfc_wrong_if; 1515 /* 1516 * If we are doing PIM assert processing, send a message 1517 * to the routing daemon. 1518 * 1519 * XXX: A PIM-SM router needs the WRONGVIF detection so it 1520 * can complete the SPT switch, regardless of the type 1521 * of the iif (broadcast media, GRE tunnel, etc). 1522 */ 1523 if (V_pim_assert_enabled && (vifi < V_numvifs) && 1524 V_viftable[vifi].v_ifp) { 1525 1526 if (ifp == &V_multicast_register_if) 1527 PIMSTAT_INC(pims_rcv_registers_wrongiif); 1528 1529 /* Get vifi for the incoming packet */ 1530 for (vifi = 0; vifi < V_numvifs && V_viftable[vifi].v_ifp != ifp; 1531 vifi++) 1532 ; 1533 if (vifi >= V_numvifs) 1534 return 0; /* The iif is not found: ignore the packet. */ 1535 1536 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_DISABLE_WRONGVIF) 1537 return 0; /* WRONGVIF disabled: ignore the packet */ 1538 1539 if (ratecheck(&rt->mfc_last_assert, &pim_assert_interval)) { 1540 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1541 struct igmpmsg *im; 1542 int hlen = ip->ip_hl << 2; 1543 struct mbuf *mm = m_copym(m, 0, hlen, M_NOWAIT); 1544 1545 if (mm && (!M_WRITABLE(mm) || mm->m_len < hlen)) 1546 mm = m_pullup(mm, hlen); 1547 if (mm == NULL) 1548 return ENOBUFS; 1549 1550 im = mtod(mm, struct igmpmsg *); 1551 im->im_msgtype = IGMPMSG_WRONGVIF; 1552 im->im_mbz = 0; 1553 im->im_vif = vifi; 1554 1555 MRTSTAT_INC(mrts_upcalls); 1556 1557 k_igmpsrc.sin_addr = im->im_src; 1558 if (socket_send(V_ip_mrouter, mm, &k_igmpsrc) < 0) { 1559 CTR1(KTR_IPMF, "%s: socket queue full", __func__); 1560 MRTSTAT_INC(mrts_upq_sockfull); 1561 return ENOBUFS; 1562 } 1563 } 1564 } 1565 return 0; 1566 } 1567 1568 1569 /* If I sourced this packet, it counts as output, else it was input. */ 1570 if (in_hosteq(ip->ip_src, V_viftable[vifi].v_lcl_addr)) { 1571 V_viftable[vifi].v_pkt_out++; 1572 V_viftable[vifi].v_bytes_out += plen; 1573 } else { 1574 V_viftable[vifi].v_pkt_in++; 1575 V_viftable[vifi].v_bytes_in += plen; 1576 } 1577 rt->mfc_pkt_cnt++; 1578 rt->mfc_byte_cnt += plen; 1579 1580 /* 1581 * For each vif, decide if a copy of the packet should be forwarded. 1582 * Forward if: 1583 * - the ttl exceeds the vif's threshold 1584 * - there are group members downstream on interface 1585 */ 1586 for (vifi = 0; vifi < V_numvifs; vifi++) 1587 if ((rt->mfc_ttls[vifi] > 0) && (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1588 V_viftable[vifi].v_pkt_out++; 1589 V_viftable[vifi].v_bytes_out += plen; 1590 if (V_viftable[vifi].v_flags & VIFF_REGISTER) 1591 pim_register_send(ip, V_viftable + vifi, m, rt); 1592 else 1593 phyint_send(ip, V_viftable + vifi, m); 1594 } 1595 1596 /* 1597 * Perform upcall-related bw measuring. 1598 */ 1599 if (rt->mfc_bw_meter != NULL) { 1600 struct bw_meter *x; 1601 struct timeval now; 1602 1603 microtime(&now); 1604 MFC_LOCK_ASSERT(); 1605 for (x = rt->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) 1606 bw_meter_receive_packet(x, plen, &now); 1607 } 1608 1609 return 0; 1610 } 1611 1612 /* 1613 * Check if a vif number is legal/ok. This is used by in_mcast.c. 1614 */ 1615 static int 1616 X_legal_vif_num(int vif) 1617 { 1618 int ret; 1619 1620 ret = 0; 1621 if (vif < 0) 1622 return (ret); 1623 1624 VIF_LOCK(); 1625 if (vif < V_numvifs) 1626 ret = 1; 1627 VIF_UNLOCK(); 1628 1629 return (ret); 1630 } 1631 1632 /* 1633 * Return the local address used by this vif 1634 */ 1635 static u_long 1636 X_ip_mcast_src(int vifi) 1637 { 1638 in_addr_t addr; 1639 1640 addr = INADDR_ANY; 1641 if (vifi < 0) 1642 return (addr); 1643 1644 VIF_LOCK(); 1645 if (vifi < V_numvifs) 1646 addr = V_viftable[vifi].v_lcl_addr.s_addr; 1647 VIF_UNLOCK(); 1648 1649 return (addr); 1650 } 1651 1652 static void 1653 phyint_send(struct ip *ip, struct vif *vifp, struct mbuf *m) 1654 { 1655 struct mbuf *mb_copy; 1656 int hlen = ip->ip_hl << 2; 1657 1658 VIF_LOCK_ASSERT(); 1659 1660 /* 1661 * Make a new reference to the packet; make sure that 1662 * the IP header is actually copied, not just referenced, 1663 * so that ip_output() only scribbles on the copy. 1664 */ 1665 mb_copy = m_copypacket(m, M_NOWAIT); 1666 if (mb_copy && (!M_WRITABLE(mb_copy) || mb_copy->m_len < hlen)) 1667 mb_copy = m_pullup(mb_copy, hlen); 1668 if (mb_copy == NULL) 1669 return; 1670 1671 send_packet(vifp, mb_copy); 1672 } 1673 1674 static void 1675 send_packet(struct vif *vifp, struct mbuf *m) 1676 { 1677 struct ip_moptions imo; 1678 int error __unused; 1679 1680 VIF_LOCK_ASSERT(); 1681 1682 imo.imo_multicast_ifp = vifp->v_ifp; 1683 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1684 imo.imo_multicast_loop = 1; 1685 imo.imo_multicast_vif = -1; 1686 STAILQ_INIT(&imo.imo_head); 1687 1688 /* 1689 * Re-entrancy should not be a problem here, because 1690 * the packets that we send out and are looped back at us 1691 * should get rejected because they appear to come from 1692 * the loopback interface, thus preventing looping. 1693 */ 1694 error = ip_output(m, NULL, NULL, IP_FORWARDING, &imo, NULL); 1695 CTR3(KTR_IPMF, "%s: vif %td err %d", __func__, 1696 (ptrdiff_t)(vifp - V_viftable), error); 1697 } 1698 1699 /* 1700 * Stubs for old RSVP socket shim implementation. 1701 */ 1702 1703 static int 1704 X_ip_rsvp_vif(struct socket *so __unused, struct sockopt *sopt __unused) 1705 { 1706 1707 return (EOPNOTSUPP); 1708 } 1709 1710 static void 1711 X_ip_rsvp_force_done(struct socket *so __unused) 1712 { 1713 1714 } 1715 1716 static int 1717 X_rsvp_input(struct mbuf **mp, int *offp, int proto) 1718 { 1719 struct mbuf *m; 1720 1721 m = *mp; 1722 *mp = NULL; 1723 if (!V_rsvp_on) 1724 m_freem(m); 1725 return (IPPROTO_DONE); 1726 } 1727 1728 /* 1729 * Code for bandwidth monitors 1730 */ 1731 1732 /* 1733 * Define common interface for timeval-related methods 1734 */ 1735 #define BW_TIMEVALCMP(tvp, uvp, cmp) timevalcmp((tvp), (uvp), cmp) 1736 #define BW_TIMEVALDECR(vvp, uvp) timevalsub((vvp), (uvp)) 1737 #define BW_TIMEVALADD(vvp, uvp) timevaladd((vvp), (uvp)) 1738 1739 static uint32_t 1740 compute_bw_meter_flags(struct bw_upcall *req) 1741 { 1742 uint32_t flags = 0; 1743 1744 if (req->bu_flags & BW_UPCALL_UNIT_PACKETS) 1745 flags |= BW_METER_UNIT_PACKETS; 1746 if (req->bu_flags & BW_UPCALL_UNIT_BYTES) 1747 flags |= BW_METER_UNIT_BYTES; 1748 if (req->bu_flags & BW_UPCALL_GEQ) 1749 flags |= BW_METER_GEQ; 1750 if (req->bu_flags & BW_UPCALL_LEQ) 1751 flags |= BW_METER_LEQ; 1752 1753 return flags; 1754 } 1755 1756 /* 1757 * Add a bw_meter entry 1758 */ 1759 static int 1760 add_bw_upcall(struct bw_upcall *req) 1761 { 1762 struct mfc *mfc; 1763 struct timeval delta = { BW_UPCALL_THRESHOLD_INTERVAL_MIN_SEC, 1764 BW_UPCALL_THRESHOLD_INTERVAL_MIN_USEC }; 1765 struct timeval now; 1766 struct bw_meter *x; 1767 uint32_t flags; 1768 1769 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL)) 1770 return EOPNOTSUPP; 1771 1772 /* Test if the flags are valid */ 1773 if (!(req->bu_flags & (BW_UPCALL_UNIT_PACKETS | BW_UPCALL_UNIT_BYTES))) 1774 return EINVAL; 1775 if (!(req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ))) 1776 return EINVAL; 1777 if ((req->bu_flags & (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 1778 == (BW_UPCALL_GEQ | BW_UPCALL_LEQ)) 1779 return EINVAL; 1780 1781 /* Test if the threshold time interval is valid */ 1782 if (BW_TIMEVALCMP(&req->bu_threshold.b_time, &delta, <)) 1783 return EINVAL; 1784 1785 flags = compute_bw_meter_flags(req); 1786 1787 /* 1788 * Find if we have already same bw_meter entry 1789 */ 1790 MFC_LOCK(); 1791 mfc = mfc_find(&req->bu_src, &req->bu_dst); 1792 if (mfc == NULL) { 1793 MFC_UNLOCK(); 1794 return EADDRNOTAVAIL; 1795 } 1796 for (x = mfc->mfc_bw_meter; x != NULL; x = x->bm_mfc_next) { 1797 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 1798 &req->bu_threshold.b_time, ==)) && 1799 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 1800 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 1801 (x->bm_flags & BW_METER_USER_FLAGS) == flags) { 1802 MFC_UNLOCK(); 1803 return 0; /* XXX Already installed */ 1804 } 1805 } 1806 1807 /* Allocate the new bw_meter entry */ 1808 x = (struct bw_meter *)malloc(sizeof(*x), M_BWMETER, M_NOWAIT); 1809 if (x == NULL) { 1810 MFC_UNLOCK(); 1811 return ENOBUFS; 1812 } 1813 1814 /* Set the new bw_meter entry */ 1815 x->bm_threshold.b_time = req->bu_threshold.b_time; 1816 microtime(&now); 1817 x->bm_start_time = now; 1818 x->bm_threshold.b_packets = req->bu_threshold.b_packets; 1819 x->bm_threshold.b_bytes = req->bu_threshold.b_bytes; 1820 x->bm_measured.b_packets = 0; 1821 x->bm_measured.b_bytes = 0; 1822 x->bm_flags = flags; 1823 x->bm_time_next = NULL; 1824 x->bm_time_hash = BW_METER_BUCKETS; 1825 1826 /* Add the new bw_meter entry to the front of entries for this MFC */ 1827 x->bm_mfc = mfc; 1828 x->bm_mfc_next = mfc->mfc_bw_meter; 1829 mfc->mfc_bw_meter = x; 1830 schedule_bw_meter(x, &now); 1831 MFC_UNLOCK(); 1832 1833 return 0; 1834 } 1835 1836 static void 1837 free_bw_list(struct bw_meter *list) 1838 { 1839 while (list != NULL) { 1840 struct bw_meter *x = list; 1841 1842 list = list->bm_mfc_next; 1843 unschedule_bw_meter(x); 1844 free(x, M_BWMETER); 1845 } 1846 } 1847 1848 /* 1849 * Delete one or multiple bw_meter entries 1850 */ 1851 static int 1852 del_bw_upcall(struct bw_upcall *req) 1853 { 1854 struct mfc *mfc; 1855 struct bw_meter *x; 1856 1857 if (!(V_mrt_api_config & MRT_MFC_BW_UPCALL)) 1858 return EOPNOTSUPP; 1859 1860 MFC_LOCK(); 1861 1862 /* Find the corresponding MFC entry */ 1863 mfc = mfc_find(&req->bu_src, &req->bu_dst); 1864 if (mfc == NULL) { 1865 MFC_UNLOCK(); 1866 return EADDRNOTAVAIL; 1867 } else if (req->bu_flags & BW_UPCALL_DELETE_ALL) { 1868 /* 1869 * Delete all bw_meter entries for this mfc 1870 */ 1871 struct bw_meter *list; 1872 1873 list = mfc->mfc_bw_meter; 1874 mfc->mfc_bw_meter = NULL; 1875 free_bw_list(list); 1876 MFC_UNLOCK(); 1877 return 0; 1878 } else { /* Delete a single bw_meter entry */ 1879 struct bw_meter *prev; 1880 uint32_t flags = 0; 1881 1882 flags = compute_bw_meter_flags(req); 1883 1884 /* Find the bw_meter entry to delete */ 1885 for (prev = NULL, x = mfc->mfc_bw_meter; x != NULL; 1886 prev = x, x = x->bm_mfc_next) { 1887 if ((BW_TIMEVALCMP(&x->bm_threshold.b_time, 1888 &req->bu_threshold.b_time, ==)) && 1889 (x->bm_threshold.b_packets == req->bu_threshold.b_packets) && 1890 (x->bm_threshold.b_bytes == req->bu_threshold.b_bytes) && 1891 (x->bm_flags & BW_METER_USER_FLAGS) == flags) 1892 break; 1893 } 1894 if (x != NULL) { /* Delete entry from the list for this MFC */ 1895 if (prev != NULL) 1896 prev->bm_mfc_next = x->bm_mfc_next; /* remove from middle*/ 1897 else 1898 x->bm_mfc->mfc_bw_meter = x->bm_mfc_next;/* new head of list */ 1899 1900 unschedule_bw_meter(x); 1901 MFC_UNLOCK(); 1902 /* Free the bw_meter entry */ 1903 free(x, M_BWMETER); 1904 return 0; 1905 } else { 1906 MFC_UNLOCK(); 1907 return EINVAL; 1908 } 1909 } 1910 /* NOTREACHED */ 1911 } 1912 1913 /* 1914 * Perform bandwidth measurement processing that may result in an upcall 1915 */ 1916 static void 1917 bw_meter_receive_packet(struct bw_meter *x, int plen, struct timeval *nowp) 1918 { 1919 struct timeval delta; 1920 1921 MFC_LOCK_ASSERT(); 1922 1923 delta = *nowp; 1924 BW_TIMEVALDECR(&delta, &x->bm_start_time); 1925 1926 if (x->bm_flags & BW_METER_GEQ) { 1927 /* 1928 * Processing for ">=" type of bw_meter entry 1929 */ 1930 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 1931 /* Reset the bw_meter entry */ 1932 x->bm_start_time = *nowp; 1933 x->bm_measured.b_packets = 0; 1934 x->bm_measured.b_bytes = 0; 1935 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 1936 } 1937 1938 /* Record that a packet is received */ 1939 x->bm_measured.b_packets++; 1940 x->bm_measured.b_bytes += plen; 1941 1942 /* 1943 * Test if we should deliver an upcall 1944 */ 1945 if (!(x->bm_flags & BW_METER_UPCALL_DELIVERED)) { 1946 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 1947 (x->bm_measured.b_packets >= x->bm_threshold.b_packets)) || 1948 ((x->bm_flags & BW_METER_UNIT_BYTES) && 1949 (x->bm_measured.b_bytes >= x->bm_threshold.b_bytes))) { 1950 /* Prepare an upcall for delivery */ 1951 bw_meter_prepare_upcall(x, nowp); 1952 x->bm_flags |= BW_METER_UPCALL_DELIVERED; 1953 } 1954 } 1955 } else if (x->bm_flags & BW_METER_LEQ) { 1956 /* 1957 * Processing for "<=" type of bw_meter entry 1958 */ 1959 if (BW_TIMEVALCMP(&delta, &x->bm_threshold.b_time, >)) { 1960 /* 1961 * We are behind time with the multicast forwarding table 1962 * scanning for "<=" type of bw_meter entries, so test now 1963 * if we should deliver an upcall. 1964 */ 1965 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 1966 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 1967 ((x->bm_flags & BW_METER_UNIT_BYTES) && 1968 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 1969 /* Prepare an upcall for delivery */ 1970 bw_meter_prepare_upcall(x, nowp); 1971 } 1972 /* Reschedule the bw_meter entry */ 1973 unschedule_bw_meter(x); 1974 schedule_bw_meter(x, nowp); 1975 } 1976 1977 /* Record that a packet is received */ 1978 x->bm_measured.b_packets++; 1979 x->bm_measured.b_bytes += plen; 1980 1981 /* 1982 * Test if we should restart the measuring interval 1983 */ 1984 if ((x->bm_flags & BW_METER_UNIT_PACKETS && 1985 x->bm_measured.b_packets <= x->bm_threshold.b_packets) || 1986 (x->bm_flags & BW_METER_UNIT_BYTES && 1987 x->bm_measured.b_bytes <= x->bm_threshold.b_bytes)) { 1988 /* Don't restart the measuring interval */ 1989 } else { 1990 /* Do restart the measuring interval */ 1991 /* 1992 * XXX: note that we don't unschedule and schedule, because this 1993 * might be too much overhead per packet. Instead, when we process 1994 * all entries for a given timer hash bin, we check whether it is 1995 * really a timeout. If not, we reschedule at that time. 1996 */ 1997 x->bm_start_time = *nowp; 1998 x->bm_measured.b_packets = 0; 1999 x->bm_measured.b_bytes = 0; 2000 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2001 } 2002 } 2003 } 2004 2005 /* 2006 * Prepare a bandwidth-related upcall 2007 */ 2008 static void 2009 bw_meter_prepare_upcall(struct bw_meter *x, struct timeval *nowp) 2010 { 2011 struct timeval delta; 2012 struct bw_upcall *u; 2013 2014 MFC_LOCK_ASSERT(); 2015 2016 /* 2017 * Compute the measured time interval 2018 */ 2019 delta = *nowp; 2020 BW_TIMEVALDECR(&delta, &x->bm_start_time); 2021 2022 /* 2023 * If there are too many pending upcalls, deliver them now 2024 */ 2025 if (V_bw_upcalls_n >= BW_UPCALLS_MAX) 2026 bw_upcalls_send(); 2027 2028 /* 2029 * Set the bw_upcall entry 2030 */ 2031 u = &V_bw_upcalls[V_bw_upcalls_n++]; 2032 u->bu_src = x->bm_mfc->mfc_origin; 2033 u->bu_dst = x->bm_mfc->mfc_mcastgrp; 2034 u->bu_threshold.b_time = x->bm_threshold.b_time; 2035 u->bu_threshold.b_packets = x->bm_threshold.b_packets; 2036 u->bu_threshold.b_bytes = x->bm_threshold.b_bytes; 2037 u->bu_measured.b_time = delta; 2038 u->bu_measured.b_packets = x->bm_measured.b_packets; 2039 u->bu_measured.b_bytes = x->bm_measured.b_bytes; 2040 u->bu_flags = 0; 2041 if (x->bm_flags & BW_METER_UNIT_PACKETS) 2042 u->bu_flags |= BW_UPCALL_UNIT_PACKETS; 2043 if (x->bm_flags & BW_METER_UNIT_BYTES) 2044 u->bu_flags |= BW_UPCALL_UNIT_BYTES; 2045 if (x->bm_flags & BW_METER_GEQ) 2046 u->bu_flags |= BW_UPCALL_GEQ; 2047 if (x->bm_flags & BW_METER_LEQ) 2048 u->bu_flags |= BW_UPCALL_LEQ; 2049 } 2050 2051 /* 2052 * Send the pending bandwidth-related upcalls 2053 */ 2054 static void 2055 bw_upcalls_send(void) 2056 { 2057 struct mbuf *m; 2058 int len = V_bw_upcalls_n * sizeof(V_bw_upcalls[0]); 2059 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2060 static struct igmpmsg igmpmsg = { 0, /* unused1 */ 2061 0, /* unused2 */ 2062 IGMPMSG_BW_UPCALL,/* im_msgtype */ 2063 0, /* im_mbz */ 2064 0, /* im_vif */ 2065 0, /* unused3 */ 2066 { 0 }, /* im_src */ 2067 { 0 } }; /* im_dst */ 2068 2069 MFC_LOCK_ASSERT(); 2070 2071 if (V_bw_upcalls_n == 0) 2072 return; /* No pending upcalls */ 2073 2074 V_bw_upcalls_n = 0; 2075 2076 /* 2077 * Allocate a new mbuf, initialize it with the header and 2078 * the payload for the pending calls. 2079 */ 2080 m = m_gethdr(M_NOWAIT, MT_DATA); 2081 if (m == NULL) { 2082 log(LOG_WARNING, "bw_upcalls_send: cannot allocate mbuf\n"); 2083 return; 2084 } 2085 2086 m_copyback(m, 0, sizeof(struct igmpmsg), (caddr_t)&igmpmsg); 2087 m_copyback(m, sizeof(struct igmpmsg), len, (caddr_t)&V_bw_upcalls[0]); 2088 2089 /* 2090 * Send the upcalls 2091 * XXX do we need to set the address in k_igmpsrc ? 2092 */ 2093 MRTSTAT_INC(mrts_upcalls); 2094 if (socket_send(V_ip_mrouter, m, &k_igmpsrc) < 0) { 2095 log(LOG_WARNING, "bw_upcalls_send: ip_mrouter socket queue full\n"); 2096 MRTSTAT_INC(mrts_upq_sockfull); 2097 } 2098 } 2099 2100 /* 2101 * Compute the timeout hash value for the bw_meter entries 2102 */ 2103 #define BW_METER_TIMEHASH(bw_meter, hash) \ 2104 do { \ 2105 struct timeval next_timeval = (bw_meter)->bm_start_time; \ 2106 \ 2107 BW_TIMEVALADD(&next_timeval, &(bw_meter)->bm_threshold.b_time); \ 2108 (hash) = next_timeval.tv_sec; \ 2109 if (next_timeval.tv_usec) \ 2110 (hash)++; /* XXX: make sure we don't timeout early */ \ 2111 (hash) %= BW_METER_BUCKETS; \ 2112 } while (0) 2113 2114 /* 2115 * Schedule a timer to process periodically bw_meter entry of type "<=" 2116 * by linking the entry in the proper hash bucket. 2117 */ 2118 static void 2119 schedule_bw_meter(struct bw_meter *x, struct timeval *nowp) 2120 { 2121 int time_hash; 2122 2123 MFC_LOCK_ASSERT(); 2124 2125 if (!(x->bm_flags & BW_METER_LEQ)) 2126 return; /* XXX: we schedule timers only for "<=" entries */ 2127 2128 /* 2129 * Reset the bw_meter entry 2130 */ 2131 x->bm_start_time = *nowp; 2132 x->bm_measured.b_packets = 0; 2133 x->bm_measured.b_bytes = 0; 2134 x->bm_flags &= ~BW_METER_UPCALL_DELIVERED; 2135 2136 /* 2137 * Compute the timeout hash value and insert the entry 2138 */ 2139 BW_METER_TIMEHASH(x, time_hash); 2140 x->bm_time_next = V_bw_meter_timers[time_hash]; 2141 V_bw_meter_timers[time_hash] = x; 2142 x->bm_time_hash = time_hash; 2143 } 2144 2145 /* 2146 * Unschedule the periodic timer that processes bw_meter entry of type "<=" 2147 * by removing the entry from the proper hash bucket. 2148 */ 2149 static void 2150 unschedule_bw_meter(struct bw_meter *x) 2151 { 2152 int time_hash; 2153 struct bw_meter *prev, *tmp; 2154 2155 MFC_LOCK_ASSERT(); 2156 2157 if (!(x->bm_flags & BW_METER_LEQ)) 2158 return; /* XXX: we schedule timers only for "<=" entries */ 2159 2160 /* 2161 * Compute the timeout hash value and delete the entry 2162 */ 2163 time_hash = x->bm_time_hash; 2164 if (time_hash >= BW_METER_BUCKETS) 2165 return; /* Entry was not scheduled */ 2166 2167 for (prev = NULL, tmp = V_bw_meter_timers[time_hash]; 2168 tmp != NULL; prev = tmp, tmp = tmp->bm_time_next) 2169 if (tmp == x) 2170 break; 2171 2172 if (tmp == NULL) 2173 panic("unschedule_bw_meter: bw_meter entry not found"); 2174 2175 if (prev != NULL) 2176 prev->bm_time_next = x->bm_time_next; 2177 else 2178 V_bw_meter_timers[time_hash] = x->bm_time_next; 2179 2180 x->bm_time_next = NULL; 2181 x->bm_time_hash = BW_METER_BUCKETS; 2182 } 2183 2184 2185 /* 2186 * Process all "<=" type of bw_meter that should be processed now, 2187 * and for each entry prepare an upcall if necessary. Each processed 2188 * entry is rescheduled again for the (periodic) processing. 2189 * 2190 * This is run periodically (once per second normally). On each round, 2191 * all the potentially matching entries are in the hash slot that we are 2192 * looking at. 2193 */ 2194 static void 2195 bw_meter_process() 2196 { 2197 uint32_t loops; 2198 int i; 2199 struct timeval now, process_endtime; 2200 2201 microtime(&now); 2202 if (V_last_tv_sec == now.tv_sec) 2203 return; /* nothing to do */ 2204 2205 loops = now.tv_sec - V_last_tv_sec; 2206 V_last_tv_sec = now.tv_sec; 2207 if (loops > BW_METER_BUCKETS) 2208 loops = BW_METER_BUCKETS; 2209 2210 MFC_LOCK(); 2211 /* 2212 * Process all bins of bw_meter entries from the one after the last 2213 * processed to the current one. On entry, i points to the last bucket 2214 * visited, so we need to increment i at the beginning of the loop. 2215 */ 2216 for (i = (now.tv_sec - loops) % BW_METER_BUCKETS; loops > 0; loops--) { 2217 struct bw_meter *x, *tmp_list; 2218 2219 if (++i >= BW_METER_BUCKETS) 2220 i = 0; 2221 2222 /* Disconnect the list of bw_meter entries from the bin */ 2223 tmp_list = V_bw_meter_timers[i]; 2224 V_bw_meter_timers[i] = NULL; 2225 2226 /* Process the list of bw_meter entries */ 2227 while (tmp_list != NULL) { 2228 x = tmp_list; 2229 tmp_list = tmp_list->bm_time_next; 2230 2231 /* Test if the time interval is over */ 2232 process_endtime = x->bm_start_time; 2233 BW_TIMEVALADD(&process_endtime, &x->bm_threshold.b_time); 2234 if (BW_TIMEVALCMP(&process_endtime, &now, >)) { 2235 /* Not yet: reschedule, but don't reset */ 2236 int time_hash; 2237 2238 BW_METER_TIMEHASH(x, time_hash); 2239 if (time_hash == i && process_endtime.tv_sec == now.tv_sec) { 2240 /* 2241 * XXX: somehow the bin processing is a bit ahead of time. 2242 * Put the entry in the next bin. 2243 */ 2244 if (++time_hash >= BW_METER_BUCKETS) 2245 time_hash = 0; 2246 } 2247 x->bm_time_next = V_bw_meter_timers[time_hash]; 2248 V_bw_meter_timers[time_hash] = x; 2249 x->bm_time_hash = time_hash; 2250 2251 continue; 2252 } 2253 2254 /* 2255 * Test if we should deliver an upcall 2256 */ 2257 if (((x->bm_flags & BW_METER_UNIT_PACKETS) && 2258 (x->bm_measured.b_packets <= x->bm_threshold.b_packets)) || 2259 ((x->bm_flags & BW_METER_UNIT_BYTES) && 2260 (x->bm_measured.b_bytes <= x->bm_threshold.b_bytes))) { 2261 /* Prepare an upcall for delivery */ 2262 bw_meter_prepare_upcall(x, &now); 2263 } 2264 2265 /* 2266 * Reschedule for next processing 2267 */ 2268 schedule_bw_meter(x, &now); 2269 } 2270 } 2271 2272 /* Send all upcalls that are pending delivery */ 2273 bw_upcalls_send(); 2274 2275 MFC_UNLOCK(); 2276 } 2277 2278 /* 2279 * A periodic function for sending all upcalls that are pending delivery 2280 */ 2281 static void 2282 expire_bw_upcalls_send(void *arg) 2283 { 2284 CURVNET_SET((struct vnet *) arg); 2285 2286 MFC_LOCK(); 2287 bw_upcalls_send(); 2288 MFC_UNLOCK(); 2289 2290 callout_reset(&V_bw_upcalls_ch, BW_UPCALLS_PERIOD, expire_bw_upcalls_send, 2291 curvnet); 2292 CURVNET_RESTORE(); 2293 } 2294 2295 /* 2296 * A periodic function for periodic scanning of the multicast forwarding 2297 * table for processing all "<=" bw_meter entries. 2298 */ 2299 static void 2300 expire_bw_meter_process(void *arg) 2301 { 2302 CURVNET_SET((struct vnet *) arg); 2303 2304 if (V_mrt_api_config & MRT_MFC_BW_UPCALL) 2305 bw_meter_process(); 2306 2307 callout_reset(&V_bw_meter_ch, BW_METER_PERIOD, expire_bw_meter_process, 2308 curvnet); 2309 CURVNET_RESTORE(); 2310 } 2311 2312 /* 2313 * End of bandwidth monitoring code 2314 */ 2315 2316 /* 2317 * Send the packet up to the user daemon, or eventually do kernel encapsulation 2318 * 2319 */ 2320 static int 2321 pim_register_send(struct ip *ip, struct vif *vifp, struct mbuf *m, 2322 struct mfc *rt) 2323 { 2324 struct mbuf *mb_copy, *mm; 2325 2326 /* 2327 * Do not send IGMP_WHOLEPKT notifications to userland, if the 2328 * rendezvous point was unspecified, and we were told not to. 2329 */ 2330 if (pim_squelch_wholepkt != 0 && (V_mrt_api_config & MRT_MFC_RP) && 2331 in_nullhost(rt->mfc_rp)) 2332 return 0; 2333 2334 mb_copy = pim_register_prepare(ip, m); 2335 if (mb_copy == NULL) 2336 return ENOBUFS; 2337 2338 /* 2339 * Send all the fragments. Note that the mbuf for each fragment 2340 * is freed by the sending machinery. 2341 */ 2342 for (mm = mb_copy; mm; mm = mb_copy) { 2343 mb_copy = mm->m_nextpkt; 2344 mm->m_nextpkt = 0; 2345 mm = m_pullup(mm, sizeof(struct ip)); 2346 if (mm != NULL) { 2347 ip = mtod(mm, struct ip *); 2348 if ((V_mrt_api_config & MRT_MFC_RP) && !in_nullhost(rt->mfc_rp)) { 2349 pim_register_send_rp(ip, vifp, mm, rt); 2350 } else { 2351 pim_register_send_upcall(ip, vifp, mm, rt); 2352 } 2353 } 2354 } 2355 2356 return 0; 2357 } 2358 2359 /* 2360 * Return a copy of the data packet that is ready for PIM Register 2361 * encapsulation. 2362 * XXX: Note that in the returned copy the IP header is a valid one. 2363 */ 2364 static struct mbuf * 2365 pim_register_prepare(struct ip *ip, struct mbuf *m) 2366 { 2367 struct mbuf *mb_copy = NULL; 2368 int mtu; 2369 2370 /* Take care of delayed checksums */ 2371 if (m->m_pkthdr.csum_flags & CSUM_DELAY_DATA) { 2372 in_delayed_cksum(m); 2373 m->m_pkthdr.csum_flags &= ~CSUM_DELAY_DATA; 2374 } 2375 2376 /* 2377 * Copy the old packet & pullup its IP header into the 2378 * new mbuf so we can modify it. 2379 */ 2380 mb_copy = m_copypacket(m, M_NOWAIT); 2381 if (mb_copy == NULL) 2382 return NULL; 2383 mb_copy = m_pullup(mb_copy, ip->ip_hl << 2); 2384 if (mb_copy == NULL) 2385 return NULL; 2386 2387 /* take care of the TTL */ 2388 ip = mtod(mb_copy, struct ip *); 2389 --ip->ip_ttl; 2390 2391 /* Compute the MTU after the PIM Register encapsulation */ 2392 mtu = 0xffff - sizeof(pim_encap_iphdr) - sizeof(pim_encap_pimhdr); 2393 2394 if (ntohs(ip->ip_len) <= mtu) { 2395 /* Turn the IP header into a valid one */ 2396 ip->ip_sum = 0; 2397 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 2398 } else { 2399 /* Fragment the packet */ 2400 mb_copy->m_pkthdr.csum_flags |= CSUM_IP; 2401 if (ip_fragment(ip, &mb_copy, mtu, 0) != 0) { 2402 m_freem(mb_copy); 2403 return NULL; 2404 } 2405 } 2406 return mb_copy; 2407 } 2408 2409 /* 2410 * Send an upcall with the data packet to the user-level process. 2411 */ 2412 static int 2413 pim_register_send_upcall(struct ip *ip, struct vif *vifp, 2414 struct mbuf *mb_copy, struct mfc *rt) 2415 { 2416 struct mbuf *mb_first; 2417 int len = ntohs(ip->ip_len); 2418 struct igmpmsg *im; 2419 struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 2420 2421 VIF_LOCK_ASSERT(); 2422 2423 /* 2424 * Add a new mbuf with an upcall header 2425 */ 2426 mb_first = m_gethdr(M_NOWAIT, MT_DATA); 2427 if (mb_first == NULL) { 2428 m_freem(mb_copy); 2429 return ENOBUFS; 2430 } 2431 mb_first->m_data += max_linkhdr; 2432 mb_first->m_pkthdr.len = len + sizeof(struct igmpmsg); 2433 mb_first->m_len = sizeof(struct igmpmsg); 2434 mb_first->m_next = mb_copy; 2435 2436 /* Send message to routing daemon */ 2437 im = mtod(mb_first, struct igmpmsg *); 2438 im->im_msgtype = IGMPMSG_WHOLEPKT; 2439 im->im_mbz = 0; 2440 im->im_vif = vifp - V_viftable; 2441 im->im_src = ip->ip_src; 2442 im->im_dst = ip->ip_dst; 2443 2444 k_igmpsrc.sin_addr = ip->ip_src; 2445 2446 MRTSTAT_INC(mrts_upcalls); 2447 2448 if (socket_send(V_ip_mrouter, mb_first, &k_igmpsrc) < 0) { 2449 CTR1(KTR_IPMF, "%s: socket queue full", __func__); 2450 MRTSTAT_INC(mrts_upq_sockfull); 2451 return ENOBUFS; 2452 } 2453 2454 /* Keep statistics */ 2455 PIMSTAT_INC(pims_snd_registers_msgs); 2456 PIMSTAT_ADD(pims_snd_registers_bytes, len); 2457 2458 return 0; 2459 } 2460 2461 /* 2462 * Encapsulate the data packet in PIM Register message and send it to the RP. 2463 */ 2464 static int 2465 pim_register_send_rp(struct ip *ip, struct vif *vifp, struct mbuf *mb_copy, 2466 struct mfc *rt) 2467 { 2468 struct mbuf *mb_first; 2469 struct ip *ip_outer; 2470 struct pim_encap_pimhdr *pimhdr; 2471 int len = ntohs(ip->ip_len); 2472 vifi_t vifi = rt->mfc_parent; 2473 2474 VIF_LOCK_ASSERT(); 2475 2476 if ((vifi >= V_numvifs) || in_nullhost(V_viftable[vifi].v_lcl_addr)) { 2477 m_freem(mb_copy); 2478 return EADDRNOTAVAIL; /* The iif vif is invalid */ 2479 } 2480 2481 /* 2482 * Add a new mbuf with the encapsulating header 2483 */ 2484 mb_first = m_gethdr(M_NOWAIT, MT_DATA); 2485 if (mb_first == NULL) { 2486 m_freem(mb_copy); 2487 return ENOBUFS; 2488 } 2489 mb_first->m_data += max_linkhdr; 2490 mb_first->m_len = sizeof(pim_encap_iphdr) + sizeof(pim_encap_pimhdr); 2491 mb_first->m_next = mb_copy; 2492 2493 mb_first->m_pkthdr.len = len + mb_first->m_len; 2494 2495 /* 2496 * Fill in the encapsulating IP and PIM header 2497 */ 2498 ip_outer = mtod(mb_first, struct ip *); 2499 *ip_outer = pim_encap_iphdr; 2500 ip_outer->ip_len = htons(len + sizeof(pim_encap_iphdr) + 2501 sizeof(pim_encap_pimhdr)); 2502 ip_outer->ip_src = V_viftable[vifi].v_lcl_addr; 2503 ip_outer->ip_dst = rt->mfc_rp; 2504 /* 2505 * Copy the inner header TOS to the outer header, and take care of the 2506 * IP_DF bit. 2507 */ 2508 ip_outer->ip_tos = ip->ip_tos; 2509 if (ip->ip_off & htons(IP_DF)) 2510 ip_outer->ip_off |= htons(IP_DF); 2511 ip_fillid(ip_outer); 2512 pimhdr = (struct pim_encap_pimhdr *)((caddr_t)ip_outer 2513 + sizeof(pim_encap_iphdr)); 2514 *pimhdr = pim_encap_pimhdr; 2515 /* If the iif crosses a border, set the Border-bit */ 2516 if (rt->mfc_flags[vifi] & MRT_MFC_FLAGS_BORDER_VIF & V_mrt_api_config) 2517 pimhdr->flags |= htonl(PIM_BORDER_REGISTER); 2518 2519 mb_first->m_data += sizeof(pim_encap_iphdr); 2520 pimhdr->pim.pim_cksum = in_cksum(mb_first, sizeof(pim_encap_pimhdr)); 2521 mb_first->m_data -= sizeof(pim_encap_iphdr); 2522 2523 send_packet(vifp, mb_first); 2524 2525 /* Keep statistics */ 2526 PIMSTAT_INC(pims_snd_registers_msgs); 2527 PIMSTAT_ADD(pims_snd_registers_bytes, len); 2528 2529 return 0; 2530 } 2531 2532 /* 2533 * pim_encapcheck() is called by the encap4_input() path at runtime to 2534 * determine if a packet is for PIM; allowing PIM to be dynamically loaded 2535 * into the kernel. 2536 */ 2537 static int 2538 pim_encapcheck(const struct mbuf *m __unused, int off __unused, 2539 int proto __unused, void *arg __unused) 2540 { 2541 2542 KASSERT(proto == IPPROTO_PIM, ("not for IPPROTO_PIM")); 2543 return (8); /* claim the datagram. */ 2544 } 2545 2546 /* 2547 * PIM-SMv2 and PIM-DM messages processing. 2548 * Receives and verifies the PIM control messages, and passes them 2549 * up to the listening socket, using rip_input(). 2550 * The only message with special processing is the PIM_REGISTER message 2551 * (used by PIM-SM): the PIM header is stripped off, and the inner packet 2552 * is passed to if_simloop(). 2553 */ 2554 static int 2555 pim_input(struct mbuf *m, int off, int proto, void *arg __unused) 2556 { 2557 struct ip *ip = mtod(m, struct ip *); 2558 struct pim *pim; 2559 int iphlen = off; 2560 int minlen; 2561 int datalen = ntohs(ip->ip_len) - iphlen; 2562 int ip_tos; 2563 2564 /* Keep statistics */ 2565 PIMSTAT_INC(pims_rcv_total_msgs); 2566 PIMSTAT_ADD(pims_rcv_total_bytes, datalen); 2567 2568 /* 2569 * Validate lengths 2570 */ 2571 if (datalen < PIM_MINLEN) { 2572 PIMSTAT_INC(pims_rcv_tooshort); 2573 CTR3(KTR_IPMF, "%s: short packet (%d) from 0x%08x", 2574 __func__, datalen, ntohl(ip->ip_src.s_addr)); 2575 m_freem(m); 2576 return (IPPROTO_DONE); 2577 } 2578 2579 /* 2580 * If the packet is at least as big as a REGISTER, go agead 2581 * and grab the PIM REGISTER header size, to avoid another 2582 * possible m_pullup() later. 2583 * 2584 * PIM_MINLEN == pimhdr + u_int32_t == 4 + 4 = 8 2585 * PIM_REG_MINLEN == pimhdr + reghdr + encap_iphdr == 4 + 4 + 20 = 28 2586 */ 2587 minlen = iphlen + (datalen >= PIM_REG_MINLEN ? PIM_REG_MINLEN : PIM_MINLEN); 2588 /* 2589 * Get the IP and PIM headers in contiguous memory, and 2590 * possibly the PIM REGISTER header. 2591 */ 2592 if (m->m_len < minlen && (m = m_pullup(m, minlen)) == NULL) { 2593 CTR1(KTR_IPMF, "%s: m_pullup() failed", __func__); 2594 return (IPPROTO_DONE); 2595 } 2596 2597 /* m_pullup() may have given us a new mbuf so reset ip. */ 2598 ip = mtod(m, struct ip *); 2599 ip_tos = ip->ip_tos; 2600 2601 /* adjust mbuf to point to the PIM header */ 2602 m->m_data += iphlen; 2603 m->m_len -= iphlen; 2604 pim = mtod(m, struct pim *); 2605 2606 /* 2607 * Validate checksum. If PIM REGISTER, exclude the data packet. 2608 * 2609 * XXX: some older PIMv2 implementations don't make this distinction, 2610 * so for compatibility reason perform the checksum over part of the 2611 * message, and if error, then over the whole message. 2612 */ 2613 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER && in_cksum(m, PIM_MINLEN) == 0) { 2614 /* do nothing, checksum okay */ 2615 } else if (in_cksum(m, datalen)) { 2616 PIMSTAT_INC(pims_rcv_badsum); 2617 CTR1(KTR_IPMF, "%s: invalid checksum", __func__); 2618 m_freem(m); 2619 return (IPPROTO_DONE); 2620 } 2621 2622 /* PIM version check */ 2623 if (PIM_VT_V(pim->pim_vt) < PIM_VERSION) { 2624 PIMSTAT_INC(pims_rcv_badversion); 2625 CTR3(KTR_IPMF, "%s: bad version %d expect %d", __func__, 2626 (int)PIM_VT_V(pim->pim_vt), PIM_VERSION); 2627 m_freem(m); 2628 return (IPPROTO_DONE); 2629 } 2630 2631 /* restore mbuf back to the outer IP */ 2632 m->m_data -= iphlen; 2633 m->m_len += iphlen; 2634 2635 if (PIM_VT_T(pim->pim_vt) == PIM_REGISTER) { 2636 /* 2637 * Since this is a REGISTER, we'll make a copy of the register 2638 * headers ip + pim + u_int32 + encap_ip, to be passed up to the 2639 * routing daemon. 2640 */ 2641 struct sockaddr_in dst = { sizeof(dst), AF_INET }; 2642 struct mbuf *mcp; 2643 struct ip *encap_ip; 2644 u_int32_t *reghdr; 2645 struct ifnet *vifp; 2646 2647 VIF_LOCK(); 2648 if ((V_reg_vif_num >= V_numvifs) || (V_reg_vif_num == VIFI_INVALID)) { 2649 VIF_UNLOCK(); 2650 CTR2(KTR_IPMF, "%s: register vif not set: %d", __func__, 2651 (int)V_reg_vif_num); 2652 m_freem(m); 2653 return (IPPROTO_DONE); 2654 } 2655 /* XXX need refcnt? */ 2656 vifp = V_viftable[V_reg_vif_num].v_ifp; 2657 VIF_UNLOCK(); 2658 2659 /* 2660 * Validate length 2661 */ 2662 if (datalen < PIM_REG_MINLEN) { 2663 PIMSTAT_INC(pims_rcv_tooshort); 2664 PIMSTAT_INC(pims_rcv_badregisters); 2665 CTR1(KTR_IPMF, "%s: register packet size too small", __func__); 2666 m_freem(m); 2667 return (IPPROTO_DONE); 2668 } 2669 2670 reghdr = (u_int32_t *)(pim + 1); 2671 encap_ip = (struct ip *)(reghdr + 1); 2672 2673 CTR3(KTR_IPMF, "%s: register: encap ip src 0x%08x len %d", 2674 __func__, ntohl(encap_ip->ip_src.s_addr), 2675 ntohs(encap_ip->ip_len)); 2676 2677 /* verify the version number of the inner packet */ 2678 if (encap_ip->ip_v != IPVERSION) { 2679 PIMSTAT_INC(pims_rcv_badregisters); 2680 CTR1(KTR_IPMF, "%s: bad encap ip version", __func__); 2681 m_freem(m); 2682 return (IPPROTO_DONE); 2683 } 2684 2685 /* verify the inner packet is destined to a mcast group */ 2686 if (!IN_MULTICAST(ntohl(encap_ip->ip_dst.s_addr))) { 2687 PIMSTAT_INC(pims_rcv_badregisters); 2688 CTR2(KTR_IPMF, "%s: bad encap ip dest 0x%08x", __func__, 2689 ntohl(encap_ip->ip_dst.s_addr)); 2690 m_freem(m); 2691 return (IPPROTO_DONE); 2692 } 2693 2694 /* If a NULL_REGISTER, pass it to the daemon */ 2695 if ((ntohl(*reghdr) & PIM_NULL_REGISTER)) 2696 goto pim_input_to_daemon; 2697 2698 /* 2699 * Copy the TOS from the outer IP header to the inner IP header. 2700 */ 2701 if (encap_ip->ip_tos != ip_tos) { 2702 /* Outer TOS -> inner TOS */ 2703 encap_ip->ip_tos = ip_tos; 2704 /* Recompute the inner header checksum. Sigh... */ 2705 2706 /* adjust mbuf to point to the inner IP header */ 2707 m->m_data += (iphlen + PIM_MINLEN); 2708 m->m_len -= (iphlen + PIM_MINLEN); 2709 2710 encap_ip->ip_sum = 0; 2711 encap_ip->ip_sum = in_cksum(m, encap_ip->ip_hl << 2); 2712 2713 /* restore mbuf to point back to the outer IP header */ 2714 m->m_data -= (iphlen + PIM_MINLEN); 2715 m->m_len += (iphlen + PIM_MINLEN); 2716 } 2717 2718 /* 2719 * Decapsulate the inner IP packet and loopback to forward it 2720 * as a normal multicast packet. Also, make a copy of the 2721 * outer_iphdr + pimhdr + reghdr + encap_iphdr 2722 * to pass to the daemon later, so it can take the appropriate 2723 * actions (e.g., send back PIM_REGISTER_STOP). 2724 * XXX: here m->m_data points to the outer IP header. 2725 */ 2726 mcp = m_copym(m, 0, iphlen + PIM_REG_MINLEN, M_NOWAIT); 2727 if (mcp == NULL) { 2728 CTR1(KTR_IPMF, "%s: m_copym() failed", __func__); 2729 m_freem(m); 2730 return (IPPROTO_DONE); 2731 } 2732 2733 /* Keep statistics */ 2734 /* XXX: registers_bytes include only the encap. mcast pkt */ 2735 PIMSTAT_INC(pims_rcv_registers_msgs); 2736 PIMSTAT_ADD(pims_rcv_registers_bytes, ntohs(encap_ip->ip_len)); 2737 2738 /* 2739 * forward the inner ip packet; point m_data at the inner ip. 2740 */ 2741 m_adj(m, iphlen + PIM_MINLEN); 2742 2743 CTR4(KTR_IPMF, 2744 "%s: forward decap'd REGISTER: src %lx dst %lx vif %d", 2745 __func__, 2746 (u_long)ntohl(encap_ip->ip_src.s_addr), 2747 (u_long)ntohl(encap_ip->ip_dst.s_addr), 2748 (int)V_reg_vif_num); 2749 2750 /* NB: vifp was collected above; can it change on us? */ 2751 if_simloop(vifp, m, dst.sin_family, 0); 2752 2753 /* prepare the register head to send to the mrouting daemon */ 2754 m = mcp; 2755 } 2756 2757 pim_input_to_daemon: 2758 /* 2759 * Pass the PIM message up to the daemon; if it is a Register message, 2760 * pass the 'head' only up to the daemon. This includes the 2761 * outer IP header, PIM header, PIM-Register header and the 2762 * inner IP header. 2763 * XXX: the outer IP header pkt size of a Register is not adjust to 2764 * reflect the fact that the inner multicast data is truncated. 2765 */ 2766 return (rip_input(&m, &off, proto)); 2767 } 2768 2769 static int 2770 sysctl_mfctable(SYSCTL_HANDLER_ARGS) 2771 { 2772 struct mfc *rt; 2773 int error, i; 2774 2775 if (req->newptr) 2776 return (EPERM); 2777 if (V_mfchashtbl == NULL) /* XXX unlocked */ 2778 return (0); 2779 error = sysctl_wire_old_buffer(req, 0); 2780 if (error) 2781 return (error); 2782 2783 MFC_LOCK(); 2784 for (i = 0; i < mfchashsize; i++) { 2785 LIST_FOREACH(rt, &V_mfchashtbl[i], mfc_hash) { 2786 error = SYSCTL_OUT(req, rt, sizeof(struct mfc)); 2787 if (error) 2788 goto out_locked; 2789 } 2790 } 2791 out_locked: 2792 MFC_UNLOCK(); 2793 return (error); 2794 } 2795 2796 static SYSCTL_NODE(_net_inet_ip, OID_AUTO, mfctable, CTLFLAG_RD, 2797 sysctl_mfctable, "IPv4 Multicast Forwarding Table " 2798 "(struct *mfc[mfchashsize], netinet/ip_mroute.h)"); 2799 2800 static void 2801 vnet_mroute_init(const void *unused __unused) 2802 { 2803 2804 V_nexpire = malloc(mfchashsize, M_MRTABLE, M_WAITOK|M_ZERO); 2805 bzero(V_bw_meter_timers, sizeof(V_bw_meter_timers)); 2806 callout_init(&V_expire_upcalls_ch, 1); 2807 callout_init(&V_bw_upcalls_ch, 1); 2808 callout_init(&V_bw_meter_ch, 1); 2809 } 2810 2811 VNET_SYSINIT(vnet_mroute_init, SI_SUB_PROTO_MC, SI_ORDER_ANY, vnet_mroute_init, 2812 NULL); 2813 2814 static void 2815 vnet_mroute_uninit(const void *unused __unused) 2816 { 2817 2818 free(V_nexpire, M_MRTABLE); 2819 V_nexpire = NULL; 2820 } 2821 2822 VNET_SYSUNINIT(vnet_mroute_uninit, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE, 2823 vnet_mroute_uninit, NULL); 2824 2825 static int 2826 ip_mroute_modevent(module_t mod, int type, void *unused) 2827 { 2828 2829 switch (type) { 2830 case MOD_LOAD: 2831 MROUTER_LOCK_INIT(); 2832 2833 if_detach_event_tag = EVENTHANDLER_REGISTER(ifnet_departure_event, 2834 if_detached_event, NULL, EVENTHANDLER_PRI_ANY); 2835 if (if_detach_event_tag == NULL) { 2836 printf("ip_mroute: unable to register " 2837 "ifnet_departure_event handler\n"); 2838 MROUTER_LOCK_DESTROY(); 2839 return (EINVAL); 2840 } 2841 2842 MFC_LOCK_INIT(); 2843 VIF_LOCK_INIT(); 2844 2845 mfchashsize = MFCHASHSIZE; 2846 if (TUNABLE_ULONG_FETCH("net.inet.ip.mfchashsize", &mfchashsize) && 2847 !powerof2(mfchashsize)) { 2848 printf("WARNING: %s not a power of 2; using default\n", 2849 "net.inet.ip.mfchashsize"); 2850 mfchashsize = MFCHASHSIZE; 2851 } 2852 2853 pim_squelch_wholepkt = 0; 2854 TUNABLE_ULONG_FETCH("net.inet.pim.squelch_wholepkt", 2855 &pim_squelch_wholepkt); 2856 2857 pim_encap_cookie = ip_encap_attach(&ipv4_encap_cfg, NULL, M_WAITOK); 2858 if (pim_encap_cookie == NULL) { 2859 printf("ip_mroute: unable to attach pim encap\n"); 2860 VIF_LOCK_DESTROY(); 2861 MFC_LOCK_DESTROY(); 2862 MROUTER_LOCK_DESTROY(); 2863 return (EINVAL); 2864 } 2865 2866 ip_mcast_src = X_ip_mcast_src; 2867 ip_mforward = X_ip_mforward; 2868 ip_mrouter_done = X_ip_mrouter_done; 2869 ip_mrouter_get = X_ip_mrouter_get; 2870 ip_mrouter_set = X_ip_mrouter_set; 2871 2872 ip_rsvp_force_done = X_ip_rsvp_force_done; 2873 ip_rsvp_vif = X_ip_rsvp_vif; 2874 2875 legal_vif_num = X_legal_vif_num; 2876 mrt_ioctl = X_mrt_ioctl; 2877 rsvp_input_p = X_rsvp_input; 2878 break; 2879 2880 case MOD_UNLOAD: 2881 /* 2882 * Typically module unload happens after the user-level 2883 * process has shutdown the kernel services (the check 2884 * below insures someone can't just yank the module out 2885 * from under a running process). But if the module is 2886 * just loaded and then unloaded w/o starting up a user 2887 * process we still need to cleanup. 2888 */ 2889 MROUTER_LOCK(); 2890 if (ip_mrouter_cnt != 0) { 2891 MROUTER_UNLOCK(); 2892 return (EINVAL); 2893 } 2894 ip_mrouter_unloading = 1; 2895 MROUTER_UNLOCK(); 2896 2897 EVENTHANDLER_DEREGISTER(ifnet_departure_event, if_detach_event_tag); 2898 2899 if (pim_encap_cookie) { 2900 ip_encap_detach(pim_encap_cookie); 2901 pim_encap_cookie = NULL; 2902 } 2903 2904 ip_mcast_src = NULL; 2905 ip_mforward = NULL; 2906 ip_mrouter_done = NULL; 2907 ip_mrouter_get = NULL; 2908 ip_mrouter_set = NULL; 2909 2910 ip_rsvp_force_done = NULL; 2911 ip_rsvp_vif = NULL; 2912 2913 legal_vif_num = NULL; 2914 mrt_ioctl = NULL; 2915 rsvp_input_p = NULL; 2916 2917 VIF_LOCK_DESTROY(); 2918 MFC_LOCK_DESTROY(); 2919 MROUTER_LOCK_DESTROY(); 2920 break; 2921 2922 default: 2923 return EOPNOTSUPP; 2924 } 2925 return 0; 2926 } 2927 2928 static moduledata_t ip_mroutemod = { 2929 "ip_mroute", 2930 ip_mroute_modevent, 2931 0 2932 }; 2933 2934 DECLARE_MODULE(ip_mroute, ip_mroutemod, SI_SUB_PROTO_MC, SI_ORDER_MIDDLE); 2935