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