1 /* 2 * IP multicast forwarding procedures 3 * 4 * Written by David Waitzman, BBN Labs, August 1988. 5 * Modified by Steve Deering, Stanford, February 1989. 6 * Modified by Mark J. Steiglitz, Stanford, May, 1991 7 * Modified by Van Jacobson, LBL, January 1993 8 * Modified by Ajit Thyagarajan, PARC, August 1993 9 * Modified by Bill Fenner, PARC, April 1995 10 * 11 * MROUTING Revision: 3.5 12 * $Id: ip_mroute.c,v 1.41 1997/09/21 22:01:38 gibbs Exp $ 13 */ 14 15 #include "opt_mrouting.h" 16 17 #include <sys/param.h> 18 #include <sys/systm.h> 19 #include <sys/mbuf.h> 20 #include <sys/socket.h> 21 #include <sys/socketvar.h> 22 #include <sys/protosw.h> 23 #include <sys/time.h> 24 #include <sys/kernel.h> 25 #include <sys/sockio.h> 26 #include <sys/syslog.h> 27 #include <net/if.h> 28 #include <net/route.h> 29 #include <netinet/in.h> 30 #include <netinet/in_systm.h> 31 #include <netinet/ip.h> 32 #include <netinet/ip_var.h> 33 #include <netinet/in_var.h> 34 #include <netinet/igmp.h> 35 #include <netinet/ip_mroute.h> 36 #include <netinet/udp.h> 37 38 #ifndef NTOHL 39 #if BYTE_ORDER != BIG_ENDIAN 40 #define NTOHL(d) ((d) = ntohl((d))) 41 #define NTOHS(d) ((d) = ntohs((u_short)(d))) 42 #define HTONL(d) ((d) = htonl((d))) 43 #define HTONS(d) ((d) = htons((u_short)(d))) 44 #else 45 #define NTOHL(d) 46 #define NTOHS(d) 47 #define HTONL(d) 48 #define HTONS(d) 49 #endif 50 #endif 51 52 #ifndef MROUTING 53 extern u_long _ip_mcast_src __P((int vifi)); 54 extern int _ip_mforward __P((struct ip *ip, struct ifnet *ifp, 55 struct mbuf *m, struct ip_moptions *imo)); 56 extern int _ip_mrouter_done __P((void)); 57 extern int _ip_mrouter_get __P((int cmd, struct socket *so, 58 struct mbuf **m)); 59 extern int _ip_mrouter_set __P((int cmd, struct socket *so, 60 struct mbuf *m)); 61 extern int _mrt_ioctl __P((int req, caddr_t data, struct proc *p)); 62 63 /* 64 * Dummy routines and globals used when multicast routing is not compiled in. 65 */ 66 67 struct socket *ip_mrouter = NULL; 68 static u_int ip_mrtproto = 0; 69 static struct mrtstat mrtstat; 70 u_int rsvpdebug = 0; 71 72 int 73 _ip_mrouter_set(cmd, so, m) 74 int cmd; 75 struct socket *so; 76 struct mbuf *m; 77 { 78 return(EOPNOTSUPP); 79 } 80 81 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = _ip_mrouter_set; 82 83 84 int 85 _ip_mrouter_get(cmd, so, m) 86 int cmd; 87 struct socket *so; 88 struct mbuf **m; 89 { 90 return(EOPNOTSUPP); 91 } 92 93 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = _ip_mrouter_get; 94 95 int 96 _ip_mrouter_done() 97 { 98 return(0); 99 } 100 101 int (*ip_mrouter_done)(void) = _ip_mrouter_done; 102 103 int 104 _ip_mforward(ip, ifp, m, imo) 105 struct ip *ip; 106 struct ifnet *ifp; 107 struct mbuf *m; 108 struct ip_moptions *imo; 109 { 110 return(0); 111 } 112 113 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 114 struct ip_moptions *) = _ip_mforward; 115 116 int 117 _mrt_ioctl(int req, caddr_t data, struct proc *p) 118 { 119 return EOPNOTSUPP; 120 } 121 122 int (*mrt_ioctl)(int, caddr_t, struct proc *) = _mrt_ioctl; 123 124 void 125 rsvp_input(m, iphlen) /* XXX must fixup manually */ 126 struct mbuf *m; 127 int iphlen; 128 { 129 /* Can still get packets with rsvp_on = 0 if there is a local member 130 * of the group to which the RSVP packet is addressed. But in this 131 * case we want to throw the packet away. 132 */ 133 if (!rsvp_on) { 134 m_freem(m); 135 return; 136 } 137 138 if (ip_rsvpd != NULL) { 139 if (rsvpdebug) 140 printf("rsvp_input: Sending packet up old-style socket\n"); 141 rip_input(m, iphlen); 142 return; 143 } 144 /* Drop the packet */ 145 m_freem(m); 146 } 147 148 void ipip_input(struct mbuf *m, int iphlen) { /* XXX must fixup manually */ 149 rip_input(m, iphlen); 150 } 151 152 int (*legal_vif_num)(int) = 0; 153 154 /* 155 * This should never be called, since IP_MULTICAST_VIF should fail, but 156 * just in case it does get called, the code a little lower in ip_output 157 * will assign the packet a local address. 158 */ 159 u_long 160 _ip_mcast_src(int vifi) { return INADDR_ANY; } 161 u_long (*ip_mcast_src)(int) = _ip_mcast_src; 162 163 int 164 ip_rsvp_vif_init(so, m) 165 struct socket *so; 166 struct mbuf *m; 167 { 168 return(EINVAL); 169 } 170 171 int 172 ip_rsvp_vif_done(so, m) 173 struct socket *so; 174 struct mbuf *m; 175 { 176 return(EINVAL); 177 } 178 179 void 180 ip_rsvp_force_done(so) 181 struct socket *so; 182 { 183 return; 184 } 185 186 #else /* MROUTING */ 187 188 #define M_HASCL(m) ((m)->m_flags & M_EXT) 189 190 #define INSIZ sizeof(struct in_addr) 191 #define same(a1, a2) \ 192 (bcmp((caddr_t)(a1), (caddr_t)(a2), INSIZ) == 0) 193 194 #define MT_MRTABLE MT_RTABLE /* since nothing else uses it */ 195 196 /* 197 * Globals. All but ip_mrouter and ip_mrtproto could be static, 198 * except for netstat or debugging purposes. 199 */ 200 #ifndef MROUTE_LKM 201 struct socket *ip_mrouter = NULL; 202 struct mrtstat mrtstat; 203 204 int ip_mrtproto = IGMP_DVMRP; /* for netstat only */ 205 #else /* MROUTE_LKM */ 206 extern void X_ipip_input __P((struct mbuf *m, int iphlen)); 207 extern struct mrtstat mrtstat; 208 static int ip_mrtproto; 209 #endif 210 211 #define NO_RTE_FOUND 0x1 212 #define RTE_FOUND 0x2 213 214 static struct mbuf *mfctable[MFCTBLSIZ]; 215 static u_char nexpire[MFCTBLSIZ]; 216 static struct vif viftable[MAXVIFS]; 217 static u_int mrtdebug = 0; /* debug level */ 218 #define DEBUG_MFC 0x02 219 #define DEBUG_FORWARD 0x04 220 #define DEBUG_EXPIRE 0x08 221 #define DEBUG_XMIT 0x10 222 static u_int tbfdebug = 0; /* tbf debug level */ 223 static u_int rsvpdebug = 0; /* rsvp debug level */ 224 225 static struct callout_handle expire_upcalls_ch; 226 227 #define EXPIRE_TIMEOUT (hz / 4) /* 4x / second */ 228 #define UPCALL_EXPIRE 6 /* number of timeouts */ 229 230 /* 231 * Define the token bucket filter structures 232 * tbftable -> each vif has one of these for storing info 233 */ 234 235 static struct tbf tbftable[MAXVIFS]; 236 #define TBF_REPROCESS (hz / 100) /* 100x / second */ 237 238 /* 239 * 'Interfaces' associated with decapsulator (so we can tell 240 * packets that went through it from ones that get reflected 241 * by a broken gateway). These interfaces are never linked into 242 * the system ifnet list & no routes point to them. I.e., packets 243 * can't be sent this way. They only exist as a placeholder for 244 * multicast source verification. 245 */ 246 static struct ifnet multicast_decap_if[MAXVIFS]; 247 248 #define ENCAP_TTL 64 249 #define ENCAP_PROTO IPPROTO_IPIP /* 4 */ 250 251 /* prototype IP hdr for encapsulated packets */ 252 static struct ip multicast_encap_iphdr = { 253 #if BYTE_ORDER == LITTLE_ENDIAN 254 sizeof(struct ip) >> 2, IPVERSION, 255 #else 256 IPVERSION, sizeof(struct ip) >> 2, 257 #endif 258 0, /* tos */ 259 sizeof(struct ip), /* total length */ 260 0, /* id */ 261 0, /* frag offset */ 262 ENCAP_TTL, ENCAP_PROTO, 263 0, /* checksum */ 264 }; 265 266 /* 267 * Private variables. 268 */ 269 static vifi_t numvifs = 0; 270 static int have_encap_tunnel = 0; 271 272 /* 273 * one-back cache used by ipip_input to locate a tunnel's vif 274 * given a datagram's src ip address. 275 */ 276 static u_long last_encap_src; 277 static struct vif *last_encap_vif; 278 279 static u_long X_ip_mcast_src __P((int vifi)); 280 static int X_ip_mforward __P((struct ip *ip, struct ifnet *ifp, struct mbuf *m, struct ip_moptions *imo)); 281 static int X_ip_mrouter_done __P((void)); 282 static int X_ip_mrouter_get __P((int cmd, struct socket *so, struct mbuf **m)); 283 static int X_ip_mrouter_set __P((int cmd, struct socket *so, struct mbuf *m)); 284 static int X_legal_vif_num __P((int vif)); 285 static int X_mrt_ioctl __P((int cmd, caddr_t data)); 286 287 static int get_sg_cnt(struct sioc_sg_req *); 288 static int get_vif_cnt(struct sioc_vif_req *); 289 static int ip_mrouter_init(struct socket *, struct mbuf *); 290 static int add_vif(struct vifctl *); 291 static int del_vif(vifi_t *); 292 static int add_mfc(struct mfcctl *); 293 static int del_mfc(struct mfcctl *); 294 static int socket_send(struct socket *, struct mbuf *, struct sockaddr_in *); 295 static int get_version(struct mbuf *); 296 static int get_assert(struct mbuf *); 297 static int set_assert(int *); 298 static void expire_upcalls(void *); 299 static int ip_mdq(struct mbuf *, struct ifnet *, struct mfc *, 300 vifi_t); 301 static void phyint_send(struct ip *, struct vif *, struct mbuf *); 302 static void encap_send(struct ip *, struct vif *, struct mbuf *); 303 static void tbf_control(struct vif *, struct mbuf *, struct ip *, u_long); 304 static void tbf_queue(struct vif *, struct mbuf *); 305 static void tbf_process_q(struct vif *); 306 static void tbf_reprocess_q(void *); 307 static int tbf_dq_sel(struct vif *, struct ip *); 308 static void tbf_send_packet(struct vif *, struct mbuf *); 309 static void tbf_update_tokens(struct vif *); 310 static int priority(struct vif *, struct ip *); 311 void multiencap_decap(struct mbuf *); 312 313 /* 314 * whether or not special PIM assert processing is enabled. 315 */ 316 static int pim_assert; 317 /* 318 * Rate limit for assert notification messages, in usec 319 */ 320 #define ASSERT_MSG_TIME 3000000 321 322 /* 323 * Hash function for a source, group entry 324 */ 325 #define MFCHASH(a, g) MFCHASHMOD(((a) >> 20) ^ ((a) >> 10) ^ (a) ^ \ 326 ((g) >> 20) ^ ((g) >> 10) ^ (g)) 327 328 /* 329 * Find a route for a given origin IP address and Multicast group address 330 * Type of service parameter to be added in the future!!! 331 */ 332 333 #define MFCFIND(o, g, rt) { \ 334 register struct mbuf *_mb_rt = mfctable[MFCHASH(o,g)]; \ 335 register struct mfc *_rt = NULL; \ 336 rt = NULL; \ 337 ++mrtstat.mrts_mfc_lookups; \ 338 while (_mb_rt) { \ 339 _rt = mtod(_mb_rt, struct mfc *); \ 340 if ((_rt->mfc_origin.s_addr == o) && \ 341 (_rt->mfc_mcastgrp.s_addr == g) && \ 342 (_mb_rt->m_act == NULL)) { \ 343 rt = _rt; \ 344 break; \ 345 } \ 346 _mb_rt = _mb_rt->m_next; \ 347 } \ 348 if (rt == NULL) { \ 349 ++mrtstat.mrts_mfc_misses; \ 350 } \ 351 } 352 353 354 /* 355 * Macros to compute elapsed time efficiently 356 * Borrowed from Van Jacobson's scheduling code 357 */ 358 #define TV_DELTA(a, b, delta) { \ 359 register int xxs; \ 360 \ 361 delta = (a).tv_usec - (b).tv_usec; \ 362 if ((xxs = (a).tv_sec - (b).tv_sec)) { \ 363 switch (xxs) { \ 364 case 2: \ 365 delta += 1000000; \ 366 /* fall through */ \ 367 case 1: \ 368 delta += 1000000; \ 369 break; \ 370 default: \ 371 delta += (1000000 * xxs); \ 372 } \ 373 } \ 374 } 375 376 #define TV_LT(a, b) (((a).tv_usec < (b).tv_usec && \ 377 (a).tv_sec <= (b).tv_sec) || (a).tv_sec < (b).tv_sec) 378 379 #ifdef UPCALL_TIMING 380 u_long upcall_data[51]; 381 static void collate(struct timeval *); 382 #endif /* UPCALL_TIMING */ 383 384 385 /* 386 * Handle MRT setsockopt commands to modify the multicast routing tables. 387 */ 388 static int 389 X_ip_mrouter_set(cmd, so, m) 390 int cmd; 391 struct socket *so; 392 struct mbuf *m; 393 { 394 if (cmd != MRT_INIT && so != ip_mrouter) return EACCES; 395 396 switch (cmd) { 397 case MRT_INIT: return ip_mrouter_init(so, m); 398 case MRT_DONE: return ip_mrouter_done(); 399 case MRT_ADD_VIF: return add_vif (mtod(m, struct vifctl *)); 400 case MRT_DEL_VIF: return del_vif (mtod(m, vifi_t *)); 401 case MRT_ADD_MFC: return add_mfc (mtod(m, struct mfcctl *)); 402 case MRT_DEL_MFC: return del_mfc (mtod(m, struct mfcctl *)); 403 case MRT_ASSERT: return set_assert(mtod(m, int *)); 404 default: return EOPNOTSUPP; 405 } 406 } 407 408 #ifndef MROUTE_LKM 409 int (*ip_mrouter_set)(int, struct socket *, struct mbuf *) = X_ip_mrouter_set; 410 #endif 411 412 /* 413 * Handle MRT getsockopt commands 414 */ 415 static int 416 X_ip_mrouter_get(cmd, so, m) 417 int cmd; 418 struct socket *so; 419 struct mbuf **m; 420 { 421 struct mbuf *mb; 422 423 if (so != ip_mrouter) return EACCES; 424 425 *m = mb = m_get(M_WAIT, MT_SOOPTS); 426 427 switch (cmd) { 428 case MRT_VERSION: return get_version(mb); 429 case MRT_ASSERT: return get_assert(mb); 430 default: return EOPNOTSUPP; 431 } 432 } 433 434 #ifndef MROUTE_LKM 435 int (*ip_mrouter_get)(int, struct socket *, struct mbuf **) = X_ip_mrouter_get; 436 #endif 437 438 /* 439 * Handle ioctl commands to obtain information from the cache 440 */ 441 static int 442 X_mrt_ioctl(cmd, data) 443 int cmd; 444 caddr_t data; 445 { 446 int error = 0; 447 448 switch (cmd) { 449 case (SIOCGETVIFCNT): 450 return (get_vif_cnt((struct sioc_vif_req *)data)); 451 break; 452 case (SIOCGETSGCNT): 453 return (get_sg_cnt((struct sioc_sg_req *)data)); 454 break; 455 default: 456 return (EINVAL); 457 break; 458 } 459 return error; 460 } 461 462 #ifndef MROUTE_LKM 463 int (*mrt_ioctl)(int, caddr_t) = X_mrt_ioctl; 464 #endif 465 466 /* 467 * returns the packet, byte, rpf-failure count for the source group provided 468 */ 469 static int 470 get_sg_cnt(req) 471 register struct sioc_sg_req *req; 472 { 473 register struct mfc *rt; 474 int s; 475 476 s = splnet(); 477 MFCFIND(req->src.s_addr, req->grp.s_addr, rt); 478 splx(s); 479 if (rt != NULL) { 480 req->pktcnt = rt->mfc_pkt_cnt; 481 req->bytecnt = rt->mfc_byte_cnt; 482 req->wrong_if = rt->mfc_wrong_if; 483 } else 484 req->pktcnt = req->bytecnt = req->wrong_if = 0xffffffff; 485 486 return 0; 487 } 488 489 /* 490 * returns the input and output packet and byte counts on the vif provided 491 */ 492 static int 493 get_vif_cnt(req) 494 register struct sioc_vif_req *req; 495 { 496 register vifi_t vifi = req->vifi; 497 498 if (vifi >= numvifs) return EINVAL; 499 500 req->icount = viftable[vifi].v_pkt_in; 501 req->ocount = viftable[vifi].v_pkt_out; 502 req->ibytes = viftable[vifi].v_bytes_in; 503 req->obytes = viftable[vifi].v_bytes_out; 504 505 return 0; 506 } 507 508 /* 509 * Enable multicast routing 510 */ 511 static int 512 ip_mrouter_init(so, m) 513 struct socket *so; 514 struct mbuf *m; 515 { 516 int *v; 517 518 if (mrtdebug) 519 log(LOG_DEBUG,"ip_mrouter_init: so_type = %d, pr_protocol = %d\n", 520 so->so_type, so->so_proto->pr_protocol); 521 522 if (so->so_type != SOCK_RAW || 523 so->so_proto->pr_protocol != IPPROTO_IGMP) return EOPNOTSUPP; 524 525 if (!m || (m->m_len != sizeof(int *))) 526 return ENOPROTOOPT; 527 528 v = mtod(m, int *); 529 if (*v != 1) 530 return ENOPROTOOPT; 531 532 if (ip_mrouter != NULL) return EADDRINUSE; 533 534 ip_mrouter = so; 535 536 bzero((caddr_t)mfctable, sizeof(mfctable)); 537 bzero((caddr_t)nexpire, sizeof(nexpire)); 538 539 pim_assert = 0; 540 541 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 542 543 if (mrtdebug) 544 log(LOG_DEBUG, "ip_mrouter_init\n"); 545 546 return 0; 547 } 548 549 /* 550 * Disable multicast routing 551 */ 552 static int 553 X_ip_mrouter_done() 554 { 555 vifi_t vifi; 556 int i; 557 struct ifnet *ifp; 558 struct ifreq ifr; 559 struct mbuf *mb_rt; 560 struct mbuf *m; 561 struct rtdetq *rte; 562 int s; 563 564 s = splnet(); 565 566 /* 567 * For each phyint in use, disable promiscuous reception of all IP 568 * multicasts. 569 */ 570 for (vifi = 0; vifi < numvifs; vifi++) { 571 if (viftable[vifi].v_lcl_addr.s_addr != 0 && 572 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 573 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 574 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr 575 = INADDR_ANY; 576 ifp = viftable[vifi].v_ifp; 577 if_allmulti(ifp, 0); 578 } 579 } 580 bzero((caddr_t)tbftable, sizeof(tbftable)); 581 bzero((caddr_t)viftable, sizeof(viftable)); 582 numvifs = 0; 583 pim_assert = 0; 584 585 untimeout(expire_upcalls, (caddr_t)NULL, expire_upcalls_ch); 586 587 /* 588 * Free all multicast forwarding cache entries. 589 */ 590 for (i = 0; i < MFCTBLSIZ; i++) { 591 mb_rt = mfctable[i]; 592 while (mb_rt) { 593 if (mb_rt->m_act != NULL) { 594 while (mb_rt->m_act) { 595 m = mb_rt->m_act; 596 mb_rt->m_act = m->m_act; 597 rte = mtod(m, struct rtdetq *); 598 m_freem(rte->m); 599 m_free(m); 600 } 601 } 602 mb_rt = m_free(mb_rt); 603 } 604 } 605 606 bzero((caddr_t)mfctable, sizeof(mfctable)); 607 608 /* 609 * Reset de-encapsulation cache 610 */ 611 last_encap_src = 0; 612 last_encap_vif = NULL; 613 have_encap_tunnel = 0; 614 615 ip_mrouter = NULL; 616 617 splx(s); 618 619 if (mrtdebug) 620 log(LOG_DEBUG, "ip_mrouter_done\n"); 621 622 return 0; 623 } 624 625 #ifndef MROUTE_LKM 626 int (*ip_mrouter_done)(void) = X_ip_mrouter_done; 627 #endif 628 629 static int 630 get_version(mb) 631 struct mbuf *mb; 632 { 633 int *v; 634 635 v = mtod(mb, int *); 636 637 *v = 0x0305; /* XXX !!!! */ 638 mb->m_len = sizeof(int); 639 640 return 0; 641 } 642 643 /* 644 * Set PIM assert processing global 645 */ 646 static int 647 set_assert(i) 648 int *i; 649 { 650 if ((*i != 1) && (*i != 0)) 651 return EINVAL; 652 653 pim_assert = *i; 654 655 return 0; 656 } 657 658 /* 659 * Get PIM assert processing global 660 */ 661 static int 662 get_assert(m) 663 struct mbuf *m; 664 { 665 int *i; 666 667 i = mtod(m, int *); 668 669 *i = pim_assert; 670 671 return 0; 672 } 673 674 /* 675 * Add a vif to the vif table 676 */ 677 static int 678 add_vif(vifcp) 679 register struct vifctl *vifcp; 680 { 681 register struct vif *vifp = viftable + vifcp->vifc_vifi; 682 static struct sockaddr_in sin = {sizeof sin, AF_INET}; 683 struct ifaddr *ifa; 684 struct ifnet *ifp; 685 struct ifreq ifr; 686 int error, s; 687 struct tbf *v_tbf = tbftable + vifcp->vifc_vifi; 688 689 if (vifcp->vifc_vifi >= MAXVIFS) return EINVAL; 690 if (vifp->v_lcl_addr.s_addr != 0) return EADDRINUSE; 691 692 /* Find the interface with an address in AF_INET family */ 693 sin.sin_addr = vifcp->vifc_lcl_addr; 694 ifa = ifa_ifwithaddr((struct sockaddr *)&sin); 695 if (ifa == 0) return EADDRNOTAVAIL; 696 ifp = ifa->ifa_ifp; 697 698 if (vifcp->vifc_flags & VIFF_TUNNEL) { 699 if ((vifcp->vifc_flags & VIFF_SRCRT) == 0) { 700 /* 701 * An encapsulating tunnel is wanted. Tell ipip_input() to 702 * start paying attention to encapsulated packets. 703 */ 704 if (have_encap_tunnel == 0) { 705 have_encap_tunnel = 1; 706 for (s = 0; s < MAXVIFS; ++s) { 707 multicast_decap_if[s].if_name = "mdecap"; 708 multicast_decap_if[s].if_unit = s; 709 } 710 } 711 /* 712 * Set interface to fake encapsulator interface 713 */ 714 ifp = &multicast_decap_if[vifcp->vifc_vifi]; 715 /* 716 * Prepare cached route entry 717 */ 718 bzero(&vifp->v_route, sizeof(vifp->v_route)); 719 } else { 720 log(LOG_ERR, "source routed tunnels not supported\n"); 721 return EOPNOTSUPP; 722 } 723 } else { 724 /* Make sure the interface supports multicast */ 725 if ((ifp->if_flags & IFF_MULTICAST) == 0) 726 return EOPNOTSUPP; 727 728 /* Enable promiscuous reception of all IP multicasts from the if */ 729 s = splnet(); 730 error = if_allmulti(ifp, 1); 731 splx(s); 732 if (error) 733 return error; 734 } 735 736 s = splnet(); 737 /* define parameters for the tbf structure */ 738 vifp->v_tbf = v_tbf; 739 GET_TIME(vifp->v_tbf->tbf_last_pkt_t); 740 vifp->v_tbf->tbf_n_tok = 0; 741 vifp->v_tbf->tbf_q_len = 0; 742 vifp->v_tbf->tbf_max_q_len = MAXQSIZE; 743 vifp->v_tbf->tbf_q = vifp->v_tbf->tbf_t = NULL; 744 745 vifp->v_flags = vifcp->vifc_flags; 746 vifp->v_threshold = vifcp->vifc_threshold; 747 vifp->v_lcl_addr = vifcp->vifc_lcl_addr; 748 vifp->v_rmt_addr = vifcp->vifc_rmt_addr; 749 vifp->v_ifp = ifp; 750 /* scaling up here allows division by 1024 in critical code */ 751 vifp->v_rate_limit= vifcp->vifc_rate_limit * 1024 / 1000; 752 vifp->v_rsvp_on = 0; 753 vifp->v_rsvpd = NULL; 754 /* initialize per vif pkt counters */ 755 vifp->v_pkt_in = 0; 756 vifp->v_pkt_out = 0; 757 vifp->v_bytes_in = 0; 758 vifp->v_bytes_out = 0; 759 splx(s); 760 761 /* Adjust numvifs up if the vifi is higher than numvifs */ 762 if (numvifs <= vifcp->vifc_vifi) numvifs = vifcp->vifc_vifi + 1; 763 764 if (mrtdebug) 765 log(LOG_DEBUG, "add_vif #%d, lcladdr %x, %s %x, thresh %x, rate %d\n", 766 vifcp->vifc_vifi, 767 ntohl(vifcp->vifc_lcl_addr.s_addr), 768 (vifcp->vifc_flags & VIFF_TUNNEL) ? "rmtaddr" : "mask", 769 ntohl(vifcp->vifc_rmt_addr.s_addr), 770 vifcp->vifc_threshold, 771 vifcp->vifc_rate_limit); 772 773 return 0; 774 } 775 776 /* 777 * Delete a vif from the vif table 778 */ 779 static int 780 del_vif(vifip) 781 vifi_t *vifip; 782 { 783 register struct vif *vifp = viftable + *vifip; 784 register vifi_t vifi; 785 register struct mbuf *m; 786 struct ifnet *ifp; 787 struct ifreq ifr; 788 int s; 789 790 if (*vifip >= numvifs) return EINVAL; 791 if (vifp->v_lcl_addr.s_addr == 0) return EADDRNOTAVAIL; 792 793 s = splnet(); 794 795 if (!(vifp->v_flags & VIFF_TUNNEL)) { 796 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_family = AF_INET; 797 ((struct sockaddr_in *)&(ifr.ifr_addr))->sin_addr.s_addr = INADDR_ANY; 798 ifp = vifp->v_ifp; 799 if_allmulti(ifp, 0); 800 } 801 802 if (vifp == last_encap_vif) { 803 last_encap_vif = 0; 804 last_encap_src = 0; 805 } 806 807 /* 808 * Free packets queued at the interface 809 */ 810 while (vifp->v_tbf->tbf_q) { 811 m = vifp->v_tbf->tbf_q; 812 vifp->v_tbf->tbf_q = m->m_act; 813 m_freem(m); 814 } 815 816 bzero((caddr_t)vifp->v_tbf, sizeof(*(vifp->v_tbf))); 817 bzero((caddr_t)vifp, sizeof (*vifp)); 818 819 /* Adjust numvifs down */ 820 for (vifi = numvifs; vifi > 0; vifi--) 821 if (viftable[vifi-1].v_lcl_addr.s_addr != 0) break; 822 numvifs = vifi; 823 824 splx(s); 825 826 if (mrtdebug) 827 log(LOG_DEBUG, "del_vif %d, numvifs %d\n", *vifip, numvifs); 828 829 return 0; 830 } 831 832 /* 833 * Add an mfc entry 834 */ 835 static int 836 add_mfc(mfccp) 837 struct mfcctl *mfccp; 838 { 839 struct mfc *rt; 840 register struct mbuf *mb_rt; 841 u_long hash; 842 struct mbuf *mb_ntry; 843 struct rtdetq *rte; 844 register u_short nstl; 845 int s; 846 int i; 847 848 MFCFIND(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr, rt); 849 850 /* If an entry already exists, just update the fields */ 851 if (rt) { 852 if (mrtdebug & DEBUG_MFC) 853 log(LOG_DEBUG,"add_mfc update o %x g %x p %x\n", 854 ntohl(mfccp->mfcc_origin.s_addr), 855 ntohl(mfccp->mfcc_mcastgrp.s_addr), 856 mfccp->mfcc_parent); 857 858 s = splnet(); 859 rt->mfc_parent = mfccp->mfcc_parent; 860 for (i = 0; i < numvifs; i++) 861 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 862 splx(s); 863 return 0; 864 } 865 866 /* 867 * Find the entry for which the upcall was made and update 868 */ 869 s = splnet(); 870 hash = MFCHASH(mfccp->mfcc_origin.s_addr, mfccp->mfcc_mcastgrp.s_addr); 871 for (mb_rt = mfctable[hash], nstl = 0; mb_rt; mb_rt = mb_rt->m_next) { 872 873 rt = mtod(mb_rt, struct mfc *); 874 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 875 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr) && 876 (mb_rt->m_act != NULL)) { 877 878 if (nstl++) 879 log(LOG_ERR, "add_mfc %s o %x g %x p %x dbx %x\n", 880 "multiple kernel entries", 881 ntohl(mfccp->mfcc_origin.s_addr), 882 ntohl(mfccp->mfcc_mcastgrp.s_addr), 883 mfccp->mfcc_parent, mb_rt->m_act); 884 885 if (mrtdebug & DEBUG_MFC) 886 log(LOG_DEBUG,"add_mfc o %x g %x p %x dbg %x\n", 887 ntohl(mfccp->mfcc_origin.s_addr), 888 ntohl(mfccp->mfcc_mcastgrp.s_addr), 889 mfccp->mfcc_parent, mb_rt->m_act); 890 891 rt->mfc_origin = mfccp->mfcc_origin; 892 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 893 rt->mfc_parent = mfccp->mfcc_parent; 894 for (i = 0; i < numvifs; i++) 895 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 896 /* initialize pkt counters per src-grp */ 897 rt->mfc_pkt_cnt = 0; 898 rt->mfc_byte_cnt = 0; 899 rt->mfc_wrong_if = 0; 900 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 901 902 rt->mfc_expire = 0; /* Don't clean this guy up */ 903 nexpire[hash]--; 904 905 /* free packets Qed at the end of this entry */ 906 while (mb_rt->m_act) { 907 mb_ntry = mb_rt->m_act; 908 rte = mtod(mb_ntry, struct rtdetq *); 909 /* #ifdef RSVP_ISI */ 910 ip_mdq(rte->m, rte->ifp, rt, -1); 911 /* #endif */ 912 mb_rt->m_act = mb_ntry->m_act; 913 m_freem(rte->m); 914 #ifdef UPCALL_TIMING 915 collate(&(rte->t)); 916 #endif /* UPCALL_TIMING */ 917 m_free(mb_ntry); 918 } 919 } 920 } 921 922 /* 923 * It is possible that an entry is being inserted without an upcall 924 */ 925 if (nstl == 0) { 926 if (mrtdebug & DEBUG_MFC) 927 log(LOG_DEBUG,"add_mfc no upcall h %d o %x g %x p %x\n", 928 hash, ntohl(mfccp->mfcc_origin.s_addr), 929 ntohl(mfccp->mfcc_mcastgrp.s_addr), 930 mfccp->mfcc_parent); 931 932 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 933 934 rt = mtod(mb_rt, struct mfc *); 935 if ((rt->mfc_origin.s_addr == mfccp->mfcc_origin.s_addr) && 936 (rt->mfc_mcastgrp.s_addr == mfccp->mfcc_mcastgrp.s_addr)) { 937 938 rt->mfc_origin = mfccp->mfcc_origin; 939 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 940 rt->mfc_parent = mfccp->mfcc_parent; 941 for (i = 0; i < numvifs; i++) 942 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 943 /* initialize pkt counters per src-grp */ 944 rt->mfc_pkt_cnt = 0; 945 rt->mfc_byte_cnt = 0; 946 rt->mfc_wrong_if = 0; 947 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 948 if (rt->mfc_expire) 949 nexpire[hash]--; 950 rt->mfc_expire = 0; 951 } 952 } 953 if (mb_rt == NULL) { 954 /* no upcall, so make a new entry */ 955 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 956 if (mb_rt == NULL) { 957 splx(s); 958 return ENOBUFS; 959 } 960 961 rt = mtod(mb_rt, struct mfc *); 962 963 /* insert new entry at head of hash chain */ 964 rt->mfc_origin = mfccp->mfcc_origin; 965 rt->mfc_mcastgrp = mfccp->mfcc_mcastgrp; 966 rt->mfc_parent = mfccp->mfcc_parent; 967 for (i = 0; i < numvifs; i++) 968 rt->mfc_ttls[i] = mfccp->mfcc_ttls[i]; 969 /* initialize pkt counters per src-grp */ 970 rt->mfc_pkt_cnt = 0; 971 rt->mfc_byte_cnt = 0; 972 rt->mfc_wrong_if = 0; 973 rt->mfc_last_assert.tv_sec = rt->mfc_last_assert.tv_usec = 0; 974 rt->mfc_expire = 0; 975 976 /* link into table */ 977 mb_rt->m_next = mfctable[hash]; 978 mfctable[hash] = mb_rt; 979 mb_rt->m_act = NULL; 980 } 981 } 982 splx(s); 983 return 0; 984 } 985 986 #ifdef UPCALL_TIMING 987 /* 988 * collect delay statistics on the upcalls 989 */ 990 static void collate(t) 991 register struct timeval *t; 992 { 993 register u_long d; 994 register struct timeval tp; 995 register u_long delta; 996 997 GET_TIME(tp); 998 999 if (TV_LT(*t, tp)) 1000 { 1001 TV_DELTA(tp, *t, delta); 1002 1003 d = delta >> 10; 1004 if (d > 50) 1005 d = 50; 1006 1007 ++upcall_data[d]; 1008 } 1009 } 1010 #endif /* UPCALL_TIMING */ 1011 1012 /* 1013 * Delete an mfc entry 1014 */ 1015 static int 1016 del_mfc(mfccp) 1017 struct mfcctl *mfccp; 1018 { 1019 struct in_addr origin; 1020 struct in_addr mcastgrp; 1021 struct mfc *rt; 1022 struct mbuf *mb_rt; 1023 struct mbuf **nptr; 1024 u_long hash; 1025 int s; 1026 1027 origin = mfccp->mfcc_origin; 1028 mcastgrp = mfccp->mfcc_mcastgrp; 1029 hash = MFCHASH(origin.s_addr, mcastgrp.s_addr); 1030 1031 if (mrtdebug & DEBUG_MFC) 1032 log(LOG_DEBUG,"del_mfc orig %x mcastgrp %x\n", 1033 ntohl(origin.s_addr), ntohl(mcastgrp.s_addr)); 1034 1035 s = splnet(); 1036 1037 nptr = &mfctable[hash]; 1038 while ((mb_rt = *nptr) != NULL) { 1039 rt = mtod(mb_rt, struct mfc *); 1040 if (origin.s_addr == rt->mfc_origin.s_addr && 1041 mcastgrp.s_addr == rt->mfc_mcastgrp.s_addr && 1042 mb_rt->m_act == NULL) 1043 break; 1044 1045 nptr = &mb_rt->m_next; 1046 } 1047 if (mb_rt == NULL) { 1048 splx(s); 1049 return EADDRNOTAVAIL; 1050 } 1051 1052 MFREE(mb_rt, *nptr); 1053 1054 splx(s); 1055 1056 return 0; 1057 } 1058 1059 /* 1060 * Send a message to mrouted on the multicast routing socket 1061 */ 1062 static int 1063 socket_send(s, mm, src) 1064 struct socket *s; 1065 struct mbuf *mm; 1066 struct sockaddr_in *src; 1067 { 1068 if (s) { 1069 if (sbappendaddr(&s->so_rcv, 1070 (struct sockaddr *)src, 1071 mm, (struct mbuf *)0) != 0) { 1072 sorwakeup(s); 1073 return 0; 1074 } 1075 } 1076 m_freem(mm); 1077 return -1; 1078 } 1079 1080 /* 1081 * IP multicast forwarding function. This function assumes that the packet 1082 * pointed to by "ip" has arrived on (or is about to be sent to) the interface 1083 * pointed to by "ifp", and the packet is to be relayed to other networks 1084 * that have members of the packet's destination IP multicast group. 1085 * 1086 * The packet is returned unscathed to the caller, unless it is 1087 * erroneous, in which case a non-zero return value tells the caller to 1088 * discard it. 1089 */ 1090 1091 #define IP_HDR_LEN 20 /* # bytes of fixed IP header (excluding options) */ 1092 #define TUNNEL_LEN 12 /* # bytes of IP option for tunnel encapsulation */ 1093 1094 static int 1095 X_ip_mforward(ip, ifp, m, imo) 1096 register struct ip *ip; 1097 struct ifnet *ifp; 1098 struct mbuf *m; 1099 struct ip_moptions *imo; 1100 { 1101 register struct mfc *rt; 1102 register u_char *ipoptions; 1103 static struct sockaddr_in k_igmpsrc = { sizeof k_igmpsrc, AF_INET }; 1104 static int srctun = 0; 1105 register struct mbuf *mm; 1106 int s; 1107 vifi_t vifi; 1108 struct vif *vifp; 1109 1110 if (mrtdebug & DEBUG_FORWARD) 1111 log(LOG_DEBUG, "ip_mforward: src %x, dst %x, ifp %x\n", 1112 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), ifp); 1113 1114 if (ip->ip_hl < (IP_HDR_LEN + TUNNEL_LEN) >> 2 || 1115 (ipoptions = (u_char *)(ip + 1))[1] != IPOPT_LSRR ) { 1116 /* 1117 * Packet arrived via a physical interface or 1118 * an encapsulated tunnel. 1119 */ 1120 } else { 1121 /* 1122 * Packet arrived through a source-route tunnel. 1123 * Source-route tunnels are no longer supported. 1124 */ 1125 if ((srctun++ % 1000) == 0) 1126 log(LOG_ERR, "ip_mforward: received source-routed packet from %x\n", 1127 ntohl(ip->ip_src.s_addr)); 1128 1129 return 1; 1130 } 1131 1132 if ((imo) && ((vifi = imo->imo_multicast_vif) < numvifs)) { 1133 if (ip->ip_ttl < 255) 1134 ip->ip_ttl++; /* compensate for -1 in *_send routines */ 1135 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1136 vifp = viftable + vifi; 1137 printf("Sending IPPROTO_RSVP from %lx to %lx on vif %d (%s%s%d)\n", 1138 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr), vifi, 1139 (vifp->v_flags & VIFF_TUNNEL) ? "tunnel on " : "", 1140 vifp->v_ifp->if_name, vifp->v_ifp->if_unit); 1141 } 1142 return (ip_mdq(m, ifp, NULL, vifi)); 1143 } 1144 if (rsvpdebug && ip->ip_p == IPPROTO_RSVP) { 1145 printf("Warning: IPPROTO_RSVP from %lx to %lx without vif option\n", 1146 ntohl(ip->ip_src.s_addr), ntohl(ip->ip_dst.s_addr)); 1147 if(!imo) 1148 printf("In fact, no options were specified at all\n"); 1149 } 1150 1151 /* 1152 * Don't forward a packet with time-to-live of zero or one, 1153 * or a packet destined to a local-only group. 1154 */ 1155 if (ip->ip_ttl <= 1 || 1156 ntohl(ip->ip_dst.s_addr) <= INADDR_MAX_LOCAL_GROUP) 1157 return 0; 1158 1159 /* 1160 * Determine forwarding vifs from the forwarding cache table 1161 */ 1162 s = splnet(); 1163 MFCFIND(ip->ip_src.s_addr, ip->ip_dst.s_addr, rt); 1164 1165 /* Entry exists, so forward if necessary */ 1166 if (rt != NULL) { 1167 splx(s); 1168 return (ip_mdq(m, ifp, rt, -1)); 1169 } else { 1170 /* 1171 * If we don't have a route for packet's origin, 1172 * Make a copy of the packet & 1173 * send message to routing daemon 1174 */ 1175 1176 register struct mbuf *mb_rt; 1177 register struct mbuf *mb_ntry; 1178 register struct mbuf *mb0; 1179 register struct rtdetq *rte; 1180 register struct mbuf *rte_m; 1181 register u_long hash; 1182 register int npkts; 1183 int hlen = ip->ip_hl << 2; 1184 #ifdef UPCALL_TIMING 1185 struct timeval tp; 1186 1187 GET_TIME(tp); 1188 #endif 1189 1190 mrtstat.mrts_no_route++; 1191 if (mrtdebug & (DEBUG_FORWARD | DEBUG_MFC)) 1192 log(LOG_DEBUG, "ip_mforward: no rte s %x g %x\n", 1193 ntohl(ip->ip_src.s_addr), 1194 ntohl(ip->ip_dst.s_addr)); 1195 1196 /* 1197 * Allocate mbufs early so that we don't do extra work if we are 1198 * just going to fail anyway. Make sure to pullup the header so 1199 * that other people can't step on it. 1200 */ 1201 MGET(mb_ntry, M_DONTWAIT, MT_DATA); 1202 if (mb_ntry == NULL) { 1203 splx(s); 1204 return ENOBUFS; 1205 } 1206 mb0 = m_copy(m, 0, M_COPYALL); 1207 if (mb0 && (M_HASCL(mb0) || mb0->m_len < hlen)) 1208 mb0 = m_pullup(mb0, hlen); 1209 if (mb0 == NULL) { 1210 m_free(mb_ntry); 1211 splx(s); 1212 return ENOBUFS; 1213 } 1214 1215 /* is there an upcall waiting for this packet? */ 1216 hash = MFCHASH(ip->ip_src.s_addr, ip->ip_dst.s_addr); 1217 for (mb_rt = mfctable[hash]; mb_rt; mb_rt = mb_rt->m_next) { 1218 rt = mtod(mb_rt, struct mfc *); 1219 if ((ip->ip_src.s_addr == rt->mfc_origin.s_addr) && 1220 (ip->ip_dst.s_addr == rt->mfc_mcastgrp.s_addr) && 1221 (mb_rt->m_act != NULL)) 1222 break; 1223 } 1224 1225 if (mb_rt == NULL) { 1226 int i; 1227 struct igmpmsg *im; 1228 1229 /* no upcall, so make a new entry */ 1230 MGET(mb_rt, M_DONTWAIT, MT_MRTABLE); 1231 if (mb_rt == NULL) { 1232 m_free(mb_ntry); 1233 m_freem(mb0); 1234 splx(s); 1235 return ENOBUFS; 1236 } 1237 /* Make a copy of the header to send to the user level process */ 1238 mm = m_copy(mb0, 0, hlen); 1239 if (mm == NULL) { 1240 m_free(mb_ntry); 1241 m_freem(mb0); 1242 m_free(mb_rt); 1243 splx(s); 1244 return ENOBUFS; 1245 } 1246 1247 /* 1248 * Send message to routing daemon to install 1249 * a route into the kernel table 1250 */ 1251 k_igmpsrc.sin_addr = ip->ip_src; 1252 1253 im = mtod(mm, struct igmpmsg *); 1254 im->im_msgtype = IGMPMSG_NOCACHE; 1255 im->im_mbz = 0; 1256 1257 mrtstat.mrts_upcalls++; 1258 1259 if (socket_send(ip_mrouter, mm, &k_igmpsrc) < 0) { 1260 log(LOG_WARNING, "ip_mforward: ip_mrouter socket queue full\n"); 1261 ++mrtstat.mrts_upq_sockfull; 1262 m_free(mb_ntry); 1263 m_freem(mb0); 1264 m_free(mb_rt); 1265 splx(s); 1266 return ENOBUFS; 1267 } 1268 1269 rt = mtod(mb_rt, struct mfc *); 1270 1271 /* insert new entry at head of hash chain */ 1272 rt->mfc_origin.s_addr = ip->ip_src.s_addr; 1273 rt->mfc_mcastgrp.s_addr = ip->ip_dst.s_addr; 1274 rt->mfc_expire = UPCALL_EXPIRE; 1275 nexpire[hash]++; 1276 for (i = 0; i < numvifs; i++) 1277 rt->mfc_ttls[i] = 0; 1278 rt->mfc_parent = -1; 1279 1280 /* link into table */ 1281 mb_rt->m_next = mfctable[hash]; 1282 mfctable[hash] = mb_rt; 1283 mb_rt->m_act = NULL; 1284 1285 rte_m = mb_rt; 1286 } else { 1287 /* determine if q has overflowed */ 1288 for (rte_m = mb_rt, npkts = 0; rte_m->m_act; rte_m = rte_m->m_act) 1289 npkts++; 1290 1291 if (npkts > MAX_UPQ) { 1292 mrtstat.mrts_upq_ovflw++; 1293 m_free(mb_ntry); 1294 m_freem(mb0); 1295 splx(s); 1296 return 0; 1297 } 1298 } 1299 1300 mb_ntry->m_act = NULL; 1301 rte = mtod(mb_ntry, struct rtdetq *); 1302 1303 rte->m = mb0; 1304 rte->ifp = ifp; 1305 #ifdef UPCALL_TIMING 1306 rte->t = tp; 1307 #endif 1308 1309 /* Add this entry to the end of the queue */ 1310 rte_m->m_act = mb_ntry; 1311 1312 splx(s); 1313 1314 return 0; 1315 } 1316 } 1317 1318 #ifndef MROUTE_LKM 1319 int (*ip_mforward)(struct ip *, struct ifnet *, struct mbuf *, 1320 struct ip_moptions *) = X_ip_mforward; 1321 #endif 1322 1323 /* 1324 * Clean up the cache entry if upcall is not serviced 1325 */ 1326 static void 1327 expire_upcalls(void *unused) 1328 { 1329 struct mbuf *mb_rt, *m, **nptr; 1330 struct rtdetq *rte; 1331 struct mfc *mfc; 1332 int i; 1333 int s; 1334 1335 s = splnet(); 1336 for (i = 0; i < MFCTBLSIZ; i++) { 1337 if (nexpire[i] == 0) 1338 continue; 1339 nptr = &mfctable[i]; 1340 for (mb_rt = *nptr; mb_rt != NULL; mb_rt = *nptr) { 1341 mfc = mtod(mb_rt, struct mfc *); 1342 1343 /* 1344 * Skip real cache entries 1345 * Make sure it wasn't marked to not expire (shouldn't happen) 1346 * If it expires now 1347 */ 1348 if (mb_rt->m_act != NULL && 1349 mfc->mfc_expire != 0 && 1350 --mfc->mfc_expire == 0) { 1351 if (mrtdebug & DEBUG_EXPIRE) 1352 log(LOG_DEBUG, "expire_upcalls: expiring (%x %x)\n", 1353 ntohl(mfc->mfc_origin.s_addr), 1354 ntohl(mfc->mfc_mcastgrp.s_addr)); 1355 /* 1356 * drop all the packets 1357 * free the mbuf with the pkt, if, timing info 1358 */ 1359 while (mb_rt->m_act) { 1360 m = mb_rt->m_act; 1361 mb_rt->m_act = m->m_act; 1362 1363 rte = mtod(m, struct rtdetq *); 1364 m_freem(rte->m); 1365 m_free(m); 1366 } 1367 ++mrtstat.mrts_cache_cleanups; 1368 nexpire[i]--; 1369 1370 MFREE(mb_rt, *nptr); 1371 } else { 1372 nptr = &mb_rt->m_next; 1373 } 1374 } 1375 } 1376 splx(s); 1377 expire_upcalls_ch = timeout(expire_upcalls, (caddr_t)NULL, EXPIRE_TIMEOUT); 1378 } 1379 1380 /* 1381 * Packet forwarding routine once entry in the cache is made 1382 */ 1383 static int 1384 ip_mdq(m, ifp, rt, xmt_vif) 1385 register struct mbuf *m; 1386 register struct ifnet *ifp; 1387 register struct mfc *rt; 1388 register vifi_t xmt_vif; 1389 { 1390 register struct ip *ip = mtod(m, struct ip *); 1391 register vifi_t vifi; 1392 register struct vif *vifp; 1393 register int plen = ntohs(ip->ip_len); 1394 1395 /* 1396 * Macro to send packet on vif. Since RSVP packets don't get counted on 1397 * input, they shouldn't get counted on output, so statistics keeping is 1398 * seperate. 1399 */ 1400 #define MC_SEND(ip,vifp,m) { \ 1401 if ((vifp)->v_flags & VIFF_TUNNEL) \ 1402 encap_send((ip), (vifp), (m)); \ 1403 else \ 1404 phyint_send((ip), (vifp), (m)); \ 1405 } 1406 1407 /* 1408 * If xmt_vif is not -1, send on only the requested vif. 1409 * 1410 * (since vifi_t is u_short, -1 becomes MAXUSHORT, which > numvifs.) 1411 */ 1412 if (xmt_vif < numvifs) { 1413 MC_SEND(ip, viftable + xmt_vif, m); 1414 return 1; 1415 } 1416 1417 /* 1418 * Don't forward if it didn't arrive from the parent vif for its origin. 1419 */ 1420 vifi = rt->mfc_parent; 1421 if ((vifi >= numvifs) || (viftable[vifi].v_ifp != ifp)) { 1422 /* came in the wrong interface */ 1423 if (mrtdebug & DEBUG_FORWARD) 1424 log(LOG_DEBUG, "wrong if: ifp %x vifi %d vififp %x\n", 1425 ifp, vifi, viftable[vifi].v_ifp); 1426 ++mrtstat.mrts_wrong_if; 1427 ++rt->mfc_wrong_if; 1428 /* 1429 * If we are doing PIM assert processing, and we are forwarding 1430 * packets on this interface, and it is a broadcast medium 1431 * interface (and not a tunnel), send a message to the routing daemon. 1432 */ 1433 if (pim_assert && rt->mfc_ttls[vifi] && 1434 (ifp->if_flags & IFF_BROADCAST) && 1435 !(viftable[vifi].v_flags & VIFF_TUNNEL)) { 1436 struct sockaddr_in k_igmpsrc; 1437 struct mbuf *mm; 1438 struct igmpmsg *im; 1439 int hlen = ip->ip_hl << 2; 1440 struct timeval now; 1441 register u_long delta; 1442 1443 GET_TIME(now); 1444 1445 TV_DELTA(rt->mfc_last_assert, now, delta); 1446 1447 if (delta > ASSERT_MSG_TIME) { 1448 mm = m_copy(m, 0, hlen); 1449 if (mm && (M_HASCL(mm) || mm->m_len < hlen)) 1450 mm = m_pullup(mm, hlen); 1451 if (mm == NULL) { 1452 return ENOBUFS; 1453 } 1454 1455 rt->mfc_last_assert = now; 1456 1457 im = mtod(mm, struct igmpmsg *); 1458 im->im_msgtype = IGMPMSG_WRONGVIF; 1459 im->im_mbz = 0; 1460 im->im_vif = vifi; 1461 1462 k_igmpsrc.sin_addr = im->im_src; 1463 1464 socket_send(ip_mrouter, mm, &k_igmpsrc); 1465 } 1466 } 1467 return 0; 1468 } 1469 1470 /* If I sourced this packet, it counts as output, else it was input. */ 1471 if (ip->ip_src.s_addr == viftable[vifi].v_lcl_addr.s_addr) { 1472 viftable[vifi].v_pkt_out++; 1473 viftable[vifi].v_bytes_out += plen; 1474 } else { 1475 viftable[vifi].v_pkt_in++; 1476 viftable[vifi].v_bytes_in += plen; 1477 } 1478 rt->mfc_pkt_cnt++; 1479 rt->mfc_byte_cnt += plen; 1480 1481 /* 1482 * For each vif, decide if a copy of the packet should be forwarded. 1483 * Forward if: 1484 * - the ttl exceeds the vif's threshold 1485 * - there are group members downstream on interface 1486 */ 1487 for (vifp = viftable, vifi = 0; vifi < numvifs; vifp++, vifi++) 1488 if ((rt->mfc_ttls[vifi] > 0) && 1489 (ip->ip_ttl > rt->mfc_ttls[vifi])) { 1490 vifp->v_pkt_out++; 1491 vifp->v_bytes_out += plen; 1492 MC_SEND(ip, vifp, m); 1493 } 1494 1495 return 0; 1496 } 1497 1498 /* 1499 * check if a vif number is legal/ok. This is used by ip_output, to export 1500 * numvifs there, 1501 */ 1502 static int 1503 X_legal_vif_num(vif) 1504 int vif; 1505 { 1506 if (vif >= 0 && vif < numvifs) 1507 return(1); 1508 else 1509 return(0); 1510 } 1511 1512 #ifndef MROUTE_LKM 1513 int (*legal_vif_num)(int) = X_legal_vif_num; 1514 #endif 1515 1516 /* 1517 * Return the local address used by this vif 1518 */ 1519 static u_long 1520 X_ip_mcast_src(vifi) 1521 int vifi; 1522 { 1523 if (vifi >= 0 && vifi < numvifs) 1524 return viftable[vifi].v_lcl_addr.s_addr; 1525 else 1526 return INADDR_ANY; 1527 } 1528 1529 #ifndef MROUTE_LKM 1530 u_long (*ip_mcast_src)(int) = X_ip_mcast_src; 1531 #endif 1532 1533 static void 1534 phyint_send(ip, vifp, m) 1535 struct ip *ip; 1536 struct vif *vifp; 1537 struct mbuf *m; 1538 { 1539 register struct mbuf *mb_copy; 1540 register int hlen = ip->ip_hl << 2; 1541 1542 /* 1543 * Make a new reference to the packet; make sure that 1544 * the IP header is actually copied, not just referenced, 1545 * so that ip_output() only scribbles on the copy. 1546 */ 1547 mb_copy = m_copy(m, 0, M_COPYALL); 1548 if (mb_copy && (M_HASCL(mb_copy) || mb_copy->m_len < hlen)) 1549 mb_copy = m_pullup(mb_copy, hlen); 1550 if (mb_copy == NULL) 1551 return; 1552 1553 if (vifp->v_rate_limit <= 0) 1554 tbf_send_packet(vifp, mb_copy); 1555 else 1556 tbf_control(vifp, mb_copy, mtod(mb_copy, struct ip *), ip->ip_len); 1557 } 1558 1559 static void 1560 encap_send(ip, vifp, m) 1561 register struct ip *ip; 1562 register struct vif *vifp; 1563 register struct mbuf *m; 1564 { 1565 register struct mbuf *mb_copy; 1566 register struct ip *ip_copy; 1567 register int i, len = ip->ip_len; 1568 1569 /* 1570 * copy the old packet & pullup it's IP header into the 1571 * new mbuf so we can modify it. Try to fill the new 1572 * mbuf since if we don't the ethernet driver will. 1573 */ 1574 MGETHDR(mb_copy, M_DONTWAIT, MT_HEADER); 1575 if (mb_copy == NULL) 1576 return; 1577 mb_copy->m_data += max_linkhdr; 1578 mb_copy->m_len = sizeof(multicast_encap_iphdr); 1579 1580 if ((mb_copy->m_next = m_copy(m, 0, M_COPYALL)) == NULL) { 1581 m_freem(mb_copy); 1582 return; 1583 } 1584 i = MHLEN - M_LEADINGSPACE(mb_copy); 1585 if (i > len) 1586 i = len; 1587 mb_copy = m_pullup(mb_copy, i); 1588 if (mb_copy == NULL) 1589 return; 1590 mb_copy->m_pkthdr.len = len + sizeof(multicast_encap_iphdr); 1591 1592 /* 1593 * fill in the encapsulating IP header. 1594 */ 1595 ip_copy = mtod(mb_copy, struct ip *); 1596 *ip_copy = multicast_encap_iphdr; 1597 ip_copy->ip_id = htons(ip_id++); 1598 ip_copy->ip_len += len; 1599 ip_copy->ip_src = vifp->v_lcl_addr; 1600 ip_copy->ip_dst = vifp->v_rmt_addr; 1601 1602 /* 1603 * turn the encapsulated IP header back into a valid one. 1604 */ 1605 ip = (struct ip *)((caddr_t)ip_copy + sizeof(multicast_encap_iphdr)); 1606 --ip->ip_ttl; 1607 HTONS(ip->ip_len); 1608 HTONS(ip->ip_off); 1609 ip->ip_sum = 0; 1610 mb_copy->m_data += sizeof(multicast_encap_iphdr); 1611 ip->ip_sum = in_cksum(mb_copy, ip->ip_hl << 2); 1612 mb_copy->m_data -= sizeof(multicast_encap_iphdr); 1613 1614 if (vifp->v_rate_limit <= 0) 1615 tbf_send_packet(vifp, mb_copy); 1616 else 1617 tbf_control(vifp, mb_copy, ip, ip_copy->ip_len); 1618 } 1619 1620 /* 1621 * De-encapsulate a packet and feed it back through ip input (this 1622 * routine is called whenever IP gets a packet with proto type 1623 * ENCAP_PROTO and a local destination address). 1624 */ 1625 void 1626 #ifdef MROUTE_LKM 1627 X_ipip_input(m, iphlen) 1628 #else 1629 ipip_input(m, iphlen) 1630 #endif 1631 register struct mbuf *m; 1632 int iphlen; 1633 { 1634 struct ifnet *ifp = m->m_pkthdr.rcvif; 1635 register struct ip *ip = mtod(m, struct ip *); 1636 register int hlen = ip->ip_hl << 2; 1637 register int s; 1638 register struct ifqueue *ifq; 1639 register struct vif *vifp; 1640 1641 if (!have_encap_tunnel) { 1642 rip_input(m, iphlen); 1643 return; 1644 } 1645 /* 1646 * dump the packet if it's not to a multicast destination or if 1647 * we don't have an encapsulating tunnel with the source. 1648 * Note: This code assumes that the remote site IP address 1649 * uniquely identifies the tunnel (i.e., that this site has 1650 * at most one tunnel with the remote site). 1651 */ 1652 if (! IN_MULTICAST(ntohl(((struct ip *)((char *)ip + hlen))->ip_dst.s_addr))) { 1653 ++mrtstat.mrts_bad_tunnel; 1654 m_freem(m); 1655 return; 1656 } 1657 if (ip->ip_src.s_addr != last_encap_src) { 1658 register struct vif *vife; 1659 1660 vifp = viftable; 1661 vife = vifp + numvifs; 1662 last_encap_src = ip->ip_src.s_addr; 1663 last_encap_vif = 0; 1664 for ( ; vifp < vife; ++vifp) 1665 if (vifp->v_rmt_addr.s_addr == ip->ip_src.s_addr) { 1666 if ((vifp->v_flags & (VIFF_TUNNEL|VIFF_SRCRT)) 1667 == VIFF_TUNNEL) 1668 last_encap_vif = vifp; 1669 break; 1670 } 1671 } 1672 if ((vifp = last_encap_vif) == 0) { 1673 last_encap_src = 0; 1674 mrtstat.mrts_cant_tunnel++; /*XXX*/ 1675 m_freem(m); 1676 if (mrtdebug) 1677 log(LOG_DEBUG, "ip_mforward: no tunnel with %x\n", 1678 ntohl(ip->ip_src.s_addr)); 1679 return; 1680 } 1681 ifp = vifp->v_ifp; 1682 1683 if (hlen > IP_HDR_LEN) 1684 ip_stripoptions(m, (struct mbuf *) 0); 1685 m->m_data += IP_HDR_LEN; 1686 m->m_len -= IP_HDR_LEN; 1687 m->m_pkthdr.len -= IP_HDR_LEN; 1688 m->m_pkthdr.rcvif = ifp; 1689 1690 ifq = &ipintrq; 1691 s = splimp(); 1692 if (IF_QFULL(ifq)) { 1693 IF_DROP(ifq); 1694 m_freem(m); 1695 } else { 1696 IF_ENQUEUE(ifq, m); 1697 /* 1698 * normally we would need a "schednetisr(NETISR_IP)" 1699 * here but we were called by ip_input and it is going 1700 * to loop back & try to dequeue the packet we just 1701 * queued as soon as we return so we avoid the 1702 * unnecessary software interrrupt. 1703 */ 1704 } 1705 splx(s); 1706 } 1707 1708 /* 1709 * Token bucket filter module 1710 */ 1711 1712 static void 1713 tbf_control(vifp, m, ip, p_len) 1714 register struct vif *vifp; 1715 register struct mbuf *m; 1716 register struct ip *ip; 1717 register u_long p_len; 1718 { 1719 register struct tbf *t = vifp->v_tbf; 1720 1721 if (p_len > MAX_BKT_SIZE) { 1722 /* drop if packet is too large */ 1723 mrtstat.mrts_pkt2large++; 1724 m_freem(m); 1725 return; 1726 } 1727 1728 tbf_update_tokens(vifp); 1729 1730 /* if there are enough tokens, 1731 * and the queue is empty, 1732 * send this packet out 1733 */ 1734 1735 if (t->tbf_q_len == 0) { 1736 /* queue empty, send packet if enough tokens */ 1737 if (p_len <= t->tbf_n_tok) { 1738 t->tbf_n_tok -= p_len; 1739 tbf_send_packet(vifp, m); 1740 } else { 1741 /* queue packet and timeout till later */ 1742 tbf_queue(vifp, m); 1743 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1744 } 1745 } else if (t->tbf_q_len < t->tbf_max_q_len) { 1746 /* finite queue length, so queue pkts and process queue */ 1747 tbf_queue(vifp, m); 1748 tbf_process_q(vifp); 1749 } else { 1750 /* queue length too much, try to dq and queue and process */ 1751 if (!tbf_dq_sel(vifp, ip)) { 1752 mrtstat.mrts_q_overflow++; 1753 m_freem(m); 1754 return; 1755 } else { 1756 tbf_queue(vifp, m); 1757 tbf_process_q(vifp); 1758 } 1759 } 1760 return; 1761 } 1762 1763 /* 1764 * adds a packet to the queue at the interface 1765 */ 1766 static void 1767 tbf_queue(vifp, m) 1768 register struct vif *vifp; 1769 register struct mbuf *m; 1770 { 1771 register int s = splnet(); 1772 register struct tbf *t = vifp->v_tbf; 1773 1774 if (t->tbf_t == NULL) { 1775 /* Queue was empty */ 1776 t->tbf_q = m; 1777 } else { 1778 /* Insert at tail */ 1779 t->tbf_t->m_act = m; 1780 } 1781 1782 /* Set new tail pointer */ 1783 t->tbf_t = m; 1784 1785 #ifdef DIAGNOSTIC 1786 /* Make sure we didn't get fed a bogus mbuf */ 1787 if (m->m_act) 1788 panic("tbf_queue: m_act"); 1789 #endif 1790 m->m_act = NULL; 1791 1792 t->tbf_q_len++; 1793 1794 splx(s); 1795 } 1796 1797 1798 /* 1799 * processes the queue at the interface 1800 */ 1801 static void 1802 tbf_process_q(vifp) 1803 register struct vif *vifp; 1804 { 1805 register struct mbuf *m; 1806 register int len; 1807 register int s = splnet(); 1808 register struct tbf *t = vifp->v_tbf; 1809 1810 /* loop through the queue at the interface and send as many packets 1811 * as possible 1812 */ 1813 while (t->tbf_q_len > 0) { 1814 m = t->tbf_q; 1815 1816 len = mtod(m, struct ip *)->ip_len; 1817 1818 /* determine if the packet can be sent */ 1819 if (len <= t->tbf_n_tok) { 1820 /* if so, 1821 * reduce no of tokens, dequeue the packet, 1822 * send the packet. 1823 */ 1824 t->tbf_n_tok -= len; 1825 1826 t->tbf_q = m->m_act; 1827 if (--t->tbf_q_len == 0) 1828 t->tbf_t = NULL; 1829 1830 m->m_act = NULL; 1831 tbf_send_packet(vifp, m); 1832 1833 } else break; 1834 } 1835 splx(s); 1836 } 1837 1838 static void 1839 tbf_reprocess_q(xvifp) 1840 void *xvifp; 1841 { 1842 register struct vif *vifp = xvifp; 1843 if (ip_mrouter == NULL) 1844 return; 1845 1846 tbf_update_tokens(vifp); 1847 1848 tbf_process_q(vifp); 1849 1850 if (vifp->v_tbf->tbf_q_len) 1851 timeout(tbf_reprocess_q, (caddr_t)vifp, TBF_REPROCESS); 1852 } 1853 1854 /* function that will selectively discard a member of the queue 1855 * based on the precedence value and the priority 1856 */ 1857 static int 1858 tbf_dq_sel(vifp, ip) 1859 register struct vif *vifp; 1860 register struct ip *ip; 1861 { 1862 register int s = splnet(); 1863 register u_int p; 1864 register struct mbuf *m, *last; 1865 register struct mbuf **np; 1866 register struct tbf *t = vifp->v_tbf; 1867 1868 p = priority(vifp, ip); 1869 1870 np = &t->tbf_q; 1871 last = NULL; 1872 while ((m = *np) != NULL) { 1873 if (p > priority(vifp, mtod(m, struct ip *))) { 1874 *np = m->m_act; 1875 /* If we're removing the last packet, fix the tail pointer */ 1876 if (m == t->tbf_t) 1877 t->tbf_t = last; 1878 m_freem(m); 1879 /* it's impossible for the queue to be empty, but 1880 * we check anyway. */ 1881 if (--t->tbf_q_len == 0) 1882 t->tbf_t = NULL; 1883 splx(s); 1884 mrtstat.mrts_drop_sel++; 1885 return(1); 1886 } 1887 np = &m->m_act; 1888 last = m; 1889 } 1890 splx(s); 1891 return(0); 1892 } 1893 1894 static void 1895 tbf_send_packet(vifp, m) 1896 register struct vif *vifp; 1897 register struct mbuf *m; 1898 { 1899 struct ip_moptions imo; 1900 int error; 1901 static struct route ro; 1902 int s = splnet(); 1903 1904 if (vifp->v_flags & VIFF_TUNNEL) { 1905 /* If tunnel options */ 1906 ip_output(m, (struct mbuf *)0, &vifp->v_route, 1907 IP_FORWARDING, (struct ip_moptions *)0); 1908 } else { 1909 imo.imo_multicast_ifp = vifp->v_ifp; 1910 imo.imo_multicast_ttl = mtod(m, struct ip *)->ip_ttl - 1; 1911 imo.imo_multicast_loop = 1; 1912 imo.imo_multicast_vif = -1; 1913 1914 /* 1915 * Re-entrancy should not be a problem here, because 1916 * the packets that we send out and are looped back at us 1917 * should get rejected because they appear to come from 1918 * the loopback interface, thus preventing looping. 1919 */ 1920 error = ip_output(m, (struct mbuf *)0, &ro, 1921 IP_FORWARDING, &imo); 1922 1923 if (mrtdebug & DEBUG_XMIT) 1924 log(LOG_DEBUG, "phyint_send on vif %d err %d\n", 1925 vifp - viftable, error); 1926 } 1927 splx(s); 1928 } 1929 1930 /* determine the current time and then 1931 * the elapsed time (between the last time and time now) 1932 * in milliseconds & update the no. of tokens in the bucket 1933 */ 1934 static void 1935 tbf_update_tokens(vifp) 1936 register struct vif *vifp; 1937 { 1938 struct timeval tp; 1939 register u_long tm; 1940 register int s = splnet(); 1941 register struct tbf *t = vifp->v_tbf; 1942 1943 GET_TIME(tp); 1944 1945 TV_DELTA(tp, t->tbf_last_pkt_t, tm); 1946 1947 /* 1948 * This formula is actually 1949 * "time in seconds" * "bytes/second". 1950 * 1951 * (tm / 1000000) * (v_rate_limit * 1000 * (1000/1024) / 8) 1952 * 1953 * The (1000/1024) was introduced in add_vif to optimize 1954 * this divide into a shift. 1955 */ 1956 t->tbf_n_tok += tm * vifp->v_rate_limit / 1024 / 8; 1957 t->tbf_last_pkt_t = tp; 1958 1959 if (t->tbf_n_tok > MAX_BKT_SIZE) 1960 t->tbf_n_tok = MAX_BKT_SIZE; 1961 1962 splx(s); 1963 } 1964 1965 static int 1966 priority(vifp, ip) 1967 register struct vif *vifp; 1968 register struct ip *ip; 1969 { 1970 register int prio; 1971 1972 /* temporary hack; may add general packet classifier some day */ 1973 1974 /* 1975 * The UDP port space is divided up into four priority ranges: 1976 * [0, 16384) : unclassified - lowest priority 1977 * [16384, 32768) : audio - highest priority 1978 * [32768, 49152) : whiteboard - medium priority 1979 * [49152, 65536) : video - low priority 1980 */ 1981 if (ip->ip_p == IPPROTO_UDP) { 1982 struct udphdr *udp = (struct udphdr *)(((char *)ip) + (ip->ip_hl << 2)); 1983 switch (ntohs(udp->uh_dport) & 0xc000) { 1984 case 0x4000: 1985 prio = 70; 1986 break; 1987 case 0x8000: 1988 prio = 60; 1989 break; 1990 case 0xc000: 1991 prio = 55; 1992 break; 1993 default: 1994 prio = 50; 1995 break; 1996 } 1997 if (tbfdebug > 1) 1998 log(LOG_DEBUG, "port %x prio%d\n", ntohs(udp->uh_dport), prio); 1999 } else { 2000 prio = 50; 2001 } 2002 return prio; 2003 } 2004 2005 /* 2006 * End of token bucket filter modifications 2007 */ 2008 2009 int 2010 ip_rsvp_vif_init(so, m) 2011 struct socket *so; 2012 struct mbuf *m; 2013 { 2014 int i; 2015 register int s; 2016 2017 if (rsvpdebug) 2018 printf("ip_rsvp_vif_init: so_type = %d, pr_protocol = %d\n", 2019 so->so_type, so->so_proto->pr_protocol); 2020 2021 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2022 return EOPNOTSUPP; 2023 2024 /* Check mbuf. */ 2025 if (m == NULL || m->m_len != sizeof(int)) { 2026 return EINVAL; 2027 } 2028 i = *(mtod(m, int *)); 2029 2030 if (rsvpdebug) 2031 printf("ip_rsvp_vif_init: vif = %d rsvp_on = %d\n",i,rsvp_on); 2032 2033 s = splnet(); 2034 2035 /* Check vif. */ 2036 if (!legal_vif_num(i)) { 2037 splx(s); 2038 return EADDRNOTAVAIL; 2039 } 2040 2041 /* Check if socket is available. */ 2042 if (viftable[i].v_rsvpd != NULL) { 2043 splx(s); 2044 return EADDRINUSE; 2045 } 2046 2047 viftable[i].v_rsvpd = so; 2048 /* This may seem silly, but we need to be sure we don't over-increment 2049 * the RSVP counter, in case something slips up. 2050 */ 2051 if (!viftable[i].v_rsvp_on) { 2052 viftable[i].v_rsvp_on = 1; 2053 rsvp_on++; 2054 } 2055 2056 splx(s); 2057 return 0; 2058 } 2059 2060 int 2061 ip_rsvp_vif_done(so, m) 2062 struct socket *so; 2063 struct mbuf *m; 2064 { 2065 int i; 2066 register int s; 2067 2068 if (rsvpdebug) 2069 printf("ip_rsvp_vif_done: so_type = %d, pr_protocol = %d\n", 2070 so->so_type, so->so_proto->pr_protocol); 2071 2072 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2073 return EOPNOTSUPP; 2074 2075 /* Check mbuf. */ 2076 if (m == NULL || m->m_len != sizeof(int)) { 2077 return EINVAL; 2078 } 2079 i = *(mtod(m, int *)); 2080 2081 s = splnet(); 2082 2083 /* Check vif. */ 2084 if (!legal_vif_num(i)) { 2085 splx(s); 2086 return EADDRNOTAVAIL; 2087 } 2088 2089 if (rsvpdebug) 2090 printf("ip_rsvp_vif_done: v_rsvpd = %p so = %p\n", 2091 viftable[i].v_rsvpd, so); 2092 2093 viftable[i].v_rsvpd = NULL; 2094 /* This may seem silly, but we need to be sure we don't over-decrement 2095 * the RSVP counter, in case something slips up. 2096 */ 2097 if (viftable[i].v_rsvp_on) { 2098 viftable[i].v_rsvp_on = 0; 2099 rsvp_on--; 2100 } 2101 2102 splx(s); 2103 return 0; 2104 } 2105 2106 void 2107 ip_rsvp_force_done(so) 2108 struct socket *so; 2109 { 2110 int vifi; 2111 register int s; 2112 2113 /* Don't bother if it is not the right type of socket. */ 2114 if (so->so_type != SOCK_RAW || so->so_proto->pr_protocol != IPPROTO_RSVP) 2115 return; 2116 2117 s = splnet(); 2118 2119 /* The socket may be attached to more than one vif...this 2120 * is perfectly legal. 2121 */ 2122 for (vifi = 0; vifi < numvifs; vifi++) { 2123 if (viftable[vifi].v_rsvpd == so) { 2124 viftable[vifi].v_rsvpd = NULL; 2125 /* This may seem silly, but we need to be sure we don't 2126 * over-decrement the RSVP counter, in case something slips up. 2127 */ 2128 if (viftable[vifi].v_rsvp_on) { 2129 viftable[vifi].v_rsvp_on = 0; 2130 rsvp_on--; 2131 } 2132 } 2133 } 2134 2135 splx(s); 2136 return; 2137 } 2138 2139 void 2140 rsvp_input(m, iphlen) 2141 struct mbuf *m; 2142 int iphlen; 2143 { 2144 int vifi; 2145 register struct ip *ip = mtod(m, struct ip *); 2146 static struct sockaddr_in rsvp_src = { sizeof rsvp_src, AF_INET }; 2147 register int s; 2148 struct ifnet *ifp; 2149 2150 if (rsvpdebug) 2151 printf("rsvp_input: rsvp_on %d\n",rsvp_on); 2152 2153 /* Can still get packets with rsvp_on = 0 if there is a local member 2154 * of the group to which the RSVP packet is addressed. But in this 2155 * case we want to throw the packet away. 2156 */ 2157 if (!rsvp_on) { 2158 m_freem(m); 2159 return; 2160 } 2161 2162 /* If the old-style non-vif-associated socket is set, then use 2163 * it and ignore the new ones. 2164 */ 2165 if (ip_rsvpd != NULL) { 2166 if (rsvpdebug) 2167 printf("rsvp_input: Sending packet up old-style socket\n"); 2168 rip_input(m, iphlen); 2169 return; 2170 } 2171 2172 s = splnet(); 2173 2174 if (rsvpdebug) 2175 printf("rsvp_input: check vifs\n"); 2176 2177 #ifdef DIAGNOSTIC 2178 if (!(m->m_flags & M_PKTHDR)) 2179 panic("rsvp_input no hdr"); 2180 #endif 2181 2182 ifp = m->m_pkthdr.rcvif; 2183 /* Find which vif the packet arrived on. */ 2184 for (vifi = 0; vifi < numvifs; vifi++) { 2185 if (viftable[vifi].v_ifp == ifp) 2186 break; 2187 } 2188 2189 if (vifi == numvifs) { 2190 /* Can't find vif packet arrived on. Drop packet. */ 2191 if (rsvpdebug) 2192 printf("rsvp_input: Can't find vif for packet...dropping it.\n"); 2193 m_freem(m); 2194 splx(s); 2195 return; 2196 } 2197 2198 if (rsvpdebug) 2199 printf("rsvp_input: check socket\n"); 2200 2201 if (viftable[vifi].v_rsvpd == NULL) { 2202 /* drop packet, since there is no specific socket for this 2203 * interface */ 2204 if (rsvpdebug) 2205 printf("rsvp_input: No socket defined for vif %d\n",vifi); 2206 m_freem(m); 2207 splx(s); 2208 return; 2209 } 2210 rsvp_src.sin_addr = ip->ip_src; 2211 2212 if (rsvpdebug && m) 2213 printf("rsvp_input: m->m_len = %d, sbspace() = %ld\n", 2214 m->m_len,sbspace(&(viftable[vifi].v_rsvpd->so_rcv))); 2215 2216 if (socket_send(viftable[vifi].v_rsvpd, m, &rsvp_src) < 0) 2217 if (rsvpdebug) 2218 printf("rsvp_input: Failed to append to socket\n"); 2219 else 2220 if (rsvpdebug) 2221 printf("rsvp_input: send packet up\n"); 2222 2223 splx(s); 2224 } 2225 2226 #ifdef MROUTE_LKM 2227 #include <sys/conf.h> 2228 #include <sys/exec.h> 2229 #include <sys/sysent.h> 2230 #include <sys/lkm.h> 2231 2232 MOD_MISC("ip_mroute_mod") 2233 2234 static int 2235 ip_mroute_mod_handle(struct lkm_table *lkmtp, int cmd) 2236 { 2237 int i; 2238 struct lkm_misc *args = lkmtp->private.lkm_misc; 2239 int err = 0; 2240 2241 switch(cmd) { 2242 static int (*old_ip_mrouter_cmd)(); 2243 static int (*old_ip_mrouter_done)(); 2244 static int (*old_ip_mforward)(); 2245 static int (*old_mrt_ioctl)(); 2246 static void (*old_proto4_input)(); 2247 static int (*old_legal_vif_num)(); 2248 extern struct protosw inetsw[]; 2249 2250 case LKM_E_LOAD: 2251 if(lkmexists(lkmtp) || ip_mrtproto) 2252 return(EEXIST); 2253 old_ip_mrouter_cmd = ip_mrouter_cmd; 2254 ip_mrouter_cmd = X_ip_mrouter_cmd; 2255 old_ip_mrouter_done = ip_mrouter_done; 2256 ip_mrouter_done = X_ip_mrouter_done; 2257 old_ip_mforward = ip_mforward; 2258 ip_mforward = X_ip_mforward; 2259 old_mrt_ioctl = mrt_ioctl; 2260 mrt_ioctl = X_mrt_ioctl; 2261 old_proto4_input = inetsw[ip_protox[ENCAP_PROTO]].pr_input; 2262 inetsw[ip_protox[ENCAP_PROTO]].pr_input = X_ipip_input; 2263 old_legal_vif_num = legal_vif_num; 2264 legal_vif_num = X_legal_vif_num; 2265 ip_mrtproto = IGMP_DVMRP; 2266 2267 printf("\nIP multicast routing loaded\n"); 2268 break; 2269 2270 case LKM_E_UNLOAD: 2271 if (ip_mrouter) 2272 return EINVAL; 2273 2274 ip_mrouter_cmd = old_ip_mrouter_cmd; 2275 ip_mrouter_done = old_ip_mrouter_done; 2276 ip_mforward = old_ip_mforward; 2277 mrt_ioctl = old_mrt_ioctl; 2278 inetsw[ip_protox[ENCAP_PROTO]].pr_input = old_proto4_input; 2279 legal_vif_num = old_legal_vif_num; 2280 ip_mrtproto = 0; 2281 break; 2282 2283 default: 2284 err = EINVAL; 2285 break; 2286 } 2287 2288 return(err); 2289 } 2290 2291 int 2292 ip_mroute_mod(struct lkm_table *lkmtp, int cmd, int ver) { 2293 DISPATCH(lkmtp, cmd, ver, ip_mroute_mod_handle, ip_mroute_mod_handle, 2294 nosys); 2295 } 2296 2297 #endif /* MROUTE_LKM */ 2298 #endif /* MROUTING */ 2299