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