1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_bootp.h" 36 #include "opt_ipfw.h" 37 #include "opt_ipstealth.h" 38 #include "opt_ipsec.h" 39 #include "opt_route.h" 40 #include "opt_rss.h" 41 42 #include <sys/param.h> 43 #include <sys/systm.h> 44 #include <sys/hhook.h> 45 #include <sys/mbuf.h> 46 #include <sys/malloc.h> 47 #include <sys/domain.h> 48 #include <sys/protosw.h> 49 #include <sys/socket.h> 50 #include <sys/time.h> 51 #include <sys/kernel.h> 52 #include <sys/lock.h> 53 #include <sys/rmlock.h> 54 #include <sys/rwlock.h> 55 #include <sys/sdt.h> 56 #include <sys/syslog.h> 57 #include <sys/sysctl.h> 58 59 #include <net/pfil.h> 60 #include <net/if.h> 61 #include <net/if_types.h> 62 #include <net/if_var.h> 63 #include <net/if_dl.h> 64 #include <net/route.h> 65 #include <net/netisr.h> 66 #include <net/rss_config.h> 67 #include <net/vnet.h> 68 69 #include <netinet/in.h> 70 #include <netinet/in_kdtrace.h> 71 #include <netinet/in_systm.h> 72 #include <netinet/in_var.h> 73 #include <netinet/ip.h> 74 #include <netinet/in_pcb.h> 75 #include <netinet/ip_var.h> 76 #include <netinet/ip_fw.h> 77 #include <netinet/ip_icmp.h> 78 #include <netinet/ip_options.h> 79 #include <machine/in_cksum.h> 80 #include <netinet/ip_carp.h> 81 #ifdef IPSEC 82 #include <netinet/ip_ipsec.h> 83 #include <netipsec/ipsec.h> 84 #include <netipsec/key.h> 85 #endif /* IPSEC */ 86 #include <netinet/in_rss.h> 87 88 #include <sys/socketvar.h> 89 90 #include <security/mac/mac_framework.h> 91 92 #ifdef CTASSERT 93 CTASSERT(sizeof(struct ip) == 20); 94 #endif 95 96 /* IP reassembly functions are defined in ip_reass.c. */ 97 extern void ipreass_init(void); 98 extern void ipreass_drain(void); 99 extern void ipreass_slowtimo(void); 100 #ifdef VIMAGE 101 extern void ipreass_destroy(void); 102 #endif 103 104 struct rmlock in_ifaddr_lock; 105 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 106 107 VNET_DEFINE(int, rsvp_on); 108 109 VNET_DEFINE(int, ipforwarding); 110 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, 111 &VNET_NAME(ipforwarding), 0, 112 "Enable IP forwarding between interfaces"); 113 114 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 115 #define V_ipsendredirects VNET(ipsendredirects) 116 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, 117 &VNET_NAME(ipsendredirects), 0, 118 "Enable sending IP redirects"); 119 120 /* 121 * XXX - Setting ip_checkinterface mostly implements the receive side of 122 * the Strong ES model described in RFC 1122, but since the routing table 123 * and transmit implementation do not implement the Strong ES model, 124 * setting this to 1 results in an odd hybrid. 125 * 126 * XXX - ip_checkinterface currently must be disabled if you use ipnat 127 * to translate the destination address to another local interface. 128 * 129 * XXX - ip_checkinterface must be disabled if you add IP aliases 130 * to the loopback interface instead of the interface where the 131 * packets for those addresses are received. 132 */ 133 static VNET_DEFINE(int, ip_checkinterface); 134 #define V_ip_checkinterface VNET(ip_checkinterface) 135 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW, 136 &VNET_NAME(ip_checkinterface), 0, 137 "Verify packet arrives on correct interface"); 138 139 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 140 141 static struct netisr_handler ip_nh = { 142 .nh_name = "ip", 143 .nh_handler = ip_input, 144 .nh_proto = NETISR_IP, 145 #ifdef RSS 146 .nh_m2cpuid = rss_soft_m2cpuid_v4, 147 .nh_policy = NETISR_POLICY_CPU, 148 .nh_dispatch = NETISR_DISPATCH_HYBRID, 149 #else 150 .nh_policy = NETISR_POLICY_FLOW, 151 #endif 152 }; 153 154 #ifdef RSS 155 /* 156 * Directly dispatched frames are currently assumed 157 * to have a flowid already calculated. 158 * 159 * It should likely have something that assert it 160 * actually has valid flow details. 161 */ 162 static struct netisr_handler ip_direct_nh = { 163 .nh_name = "ip_direct", 164 .nh_handler = ip_direct_input, 165 .nh_proto = NETISR_IP_DIRECT, 166 .nh_m2cpuid = rss_soft_m2cpuid_v4, 167 .nh_policy = NETISR_POLICY_CPU, 168 .nh_dispatch = NETISR_DISPATCH_HYBRID, 169 }; 170 #endif 171 172 extern struct domain inetdomain; 173 extern struct protosw inetsw[]; 174 u_char ip_protox[IPPROTO_MAX]; 175 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 176 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 177 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 178 179 #ifdef IPCTL_DEFMTU 180 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 181 &ip_mtu, 0, "Default MTU"); 182 #endif 183 184 #ifdef IPSTEALTH 185 VNET_DEFINE(int, ipstealth); 186 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, 187 &VNET_NAME(ipstealth), 0, 188 "IP stealth mode, no TTL decrementation on forwarding"); 189 #endif 190 191 /* 192 * IP statistics are stored in the "array" of counter(9)s. 193 */ 194 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 195 VNET_PCPUSTAT_SYSINIT(ipstat); 196 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 197 "IP statistics (struct ipstat, netinet/ip_var.h)"); 198 199 #ifdef VIMAGE 200 VNET_PCPUSTAT_SYSUNINIT(ipstat); 201 #endif /* VIMAGE */ 202 203 /* 204 * Kernel module interface for updating ipstat. The argument is an index 205 * into ipstat treated as an array. 206 */ 207 void 208 kmod_ipstat_inc(int statnum) 209 { 210 211 counter_u64_add(VNET(ipstat)[statnum], 1); 212 } 213 214 void 215 kmod_ipstat_dec(int statnum) 216 { 217 218 counter_u64_add(VNET(ipstat)[statnum], -1); 219 } 220 221 static int 222 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 223 { 224 int error, qlimit; 225 226 netisr_getqlimit(&ip_nh, &qlimit); 227 error = sysctl_handle_int(oidp, &qlimit, 0, req); 228 if (error || !req->newptr) 229 return (error); 230 if (qlimit < 1) 231 return (EINVAL); 232 return (netisr_setqlimit(&ip_nh, qlimit)); 233 } 234 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 235 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 236 "Maximum size of the IP input queue"); 237 238 static int 239 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 240 { 241 u_int64_t qdrops_long; 242 int error, qdrops; 243 244 netisr_getqdrops(&ip_nh, &qdrops_long); 245 qdrops = qdrops_long; 246 error = sysctl_handle_int(oidp, &qdrops, 0, req); 247 if (error || !req->newptr) 248 return (error); 249 if (qdrops != 0) 250 return (EINVAL); 251 netisr_clearqdrops(&ip_nh); 252 return (0); 253 } 254 255 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 256 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 257 "Number of packets dropped from the IP input queue"); 258 259 #ifdef RSS 260 static int 261 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 262 { 263 int error, qlimit; 264 265 netisr_getqlimit(&ip_direct_nh, &qlimit); 266 error = sysctl_handle_int(oidp, &qlimit, 0, req); 267 if (error || !req->newptr) 268 return (error); 269 if (qlimit < 1) 270 return (EINVAL); 271 return (netisr_setqlimit(&ip_direct_nh, qlimit)); 272 } 273 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen, 274 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I", 275 "Maximum size of the IP direct input queue"); 276 277 static int 278 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) 279 { 280 u_int64_t qdrops_long; 281 int error, qdrops; 282 283 netisr_getqdrops(&ip_direct_nh, &qdrops_long); 284 qdrops = qdrops_long; 285 error = sysctl_handle_int(oidp, &qdrops, 0, req); 286 if (error || !req->newptr) 287 return (error); 288 if (qdrops != 0) 289 return (EINVAL); 290 netisr_clearqdrops(&ip_direct_nh); 291 return (0); 292 } 293 294 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops, 295 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", 296 "Number of packets dropped from the IP direct input queue"); 297 #endif /* RSS */ 298 299 /* 300 * IP initialization: fill in IP protocol switch table. 301 * All protocols not implemented in kernel go to raw IP protocol handler. 302 */ 303 void 304 ip_init(void) 305 { 306 struct protosw *pr; 307 int i; 308 309 TAILQ_INIT(&V_in_ifaddrhead); 310 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 311 312 /* Initialize IP reassembly queue. */ 313 ipreass_init(); 314 315 /* Initialize packet filter hooks. */ 316 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 317 V_inet_pfil_hook.ph_af = AF_INET; 318 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 319 printf("%s: WARNING: unable to register pfil hook, " 320 "error %d\n", __func__, i); 321 322 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET, 323 &V_ipsec_hhh_in[HHOOK_IPSEC_INET], 324 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 325 printf("%s: WARNING: unable to register input helper hook\n", 326 __func__); 327 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET, 328 &V_ipsec_hhh_out[HHOOK_IPSEC_INET], 329 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 330 printf("%s: WARNING: unable to register output helper hook\n", 331 __func__); 332 333 /* Skip initialization of globals for non-default instances. */ 334 if (!IS_DEFAULT_VNET(curvnet)) 335 return; 336 337 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 338 if (pr == NULL) 339 panic("ip_init: PF_INET not found"); 340 341 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 342 for (i = 0; i < IPPROTO_MAX; i++) 343 ip_protox[i] = pr - inetsw; 344 /* 345 * Cycle through IP protocols and put them into the appropriate place 346 * in ip_protox[]. 347 */ 348 for (pr = inetdomain.dom_protosw; 349 pr < inetdomain.dom_protoswNPROTOSW; pr++) 350 if (pr->pr_domain->dom_family == PF_INET && 351 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 352 /* Be careful to only index valid IP protocols. */ 353 if (pr->pr_protocol < IPPROTO_MAX) 354 ip_protox[pr->pr_protocol] = pr - inetsw; 355 } 356 357 netisr_register(&ip_nh); 358 #ifdef RSS 359 netisr_register(&ip_direct_nh); 360 #endif 361 } 362 363 #ifdef VIMAGE 364 void 365 ip_destroy(void) 366 { 367 int error; 368 369 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0) 370 printf("%s: WARNING: unable to unregister pfil hook, " 371 "error %d\n", __func__, error); 372 373 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]); 374 if (error != 0) { 375 printf("%s: WARNING: unable to deregister input helper hook " 376 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: " 377 "error %d returned\n", __func__, error); 378 } 379 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]); 380 if (error != 0) { 381 printf("%s: WARNING: unable to deregister output helper hook " 382 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: " 383 "error %d returned\n", __func__, error); 384 } 385 /* Cleanup in_ifaddr hash table; should be empty. */ 386 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 387 388 /* Destroy IP reassembly queue. */ 389 ipreass_destroy(); 390 } 391 #endif 392 393 #ifdef RSS 394 /* 395 * IP direct input routine. 396 * 397 * This is called when reinjecting completed fragments where 398 * all of the previous checking and book-keeping has been done. 399 */ 400 void 401 ip_direct_input(struct mbuf *m) 402 { 403 struct ip *ip; 404 int hlen; 405 406 ip = mtod(m, struct ip *); 407 hlen = ip->ip_hl << 2; 408 409 IPSTAT_INC(ips_delivered); 410 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 411 return; 412 } 413 #endif 414 415 /* 416 * Ip input routine. Checksum and byte swap header. If fragmented 417 * try to reassemble. Process options. Pass to next level. 418 */ 419 void 420 ip_input(struct mbuf *m) 421 { 422 struct ip *ip = NULL; 423 struct in_ifaddr *ia = NULL; 424 struct ifaddr *ifa; 425 struct ifnet *ifp; 426 int checkif, hlen = 0; 427 uint16_t sum, ip_len; 428 int dchg = 0; /* dest changed after fw */ 429 struct in_addr odst; /* original dst address */ 430 431 M_ASSERTPKTHDR(m); 432 433 if (m->m_flags & M_FASTFWD_OURS) { 434 m->m_flags &= ~M_FASTFWD_OURS; 435 /* Set up some basics that will be used later. */ 436 ip = mtod(m, struct ip *); 437 hlen = ip->ip_hl << 2; 438 ip_len = ntohs(ip->ip_len); 439 goto ours; 440 } 441 442 IPSTAT_INC(ips_total); 443 444 if (m->m_pkthdr.len < sizeof(struct ip)) 445 goto tooshort; 446 447 if (m->m_len < sizeof (struct ip) && 448 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 449 IPSTAT_INC(ips_toosmall); 450 return; 451 } 452 ip = mtod(m, struct ip *); 453 454 if (ip->ip_v != IPVERSION) { 455 IPSTAT_INC(ips_badvers); 456 goto bad; 457 } 458 459 hlen = ip->ip_hl << 2; 460 if (hlen < sizeof(struct ip)) { /* minimum header length */ 461 IPSTAT_INC(ips_badhlen); 462 goto bad; 463 } 464 if (hlen > m->m_len) { 465 if ((m = m_pullup(m, hlen)) == NULL) { 466 IPSTAT_INC(ips_badhlen); 467 return; 468 } 469 ip = mtod(m, struct ip *); 470 } 471 472 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 473 474 /* 127/8 must not appear on wire - RFC1122 */ 475 ifp = m->m_pkthdr.rcvif; 476 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 477 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 478 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 479 IPSTAT_INC(ips_badaddr); 480 goto bad; 481 } 482 } 483 484 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 485 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 486 } else { 487 if (hlen == sizeof(struct ip)) { 488 sum = in_cksum_hdr(ip); 489 } else { 490 sum = in_cksum(m, hlen); 491 } 492 } 493 if (sum) { 494 IPSTAT_INC(ips_badsum); 495 goto bad; 496 } 497 498 #ifdef ALTQ 499 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 500 /* packet is dropped by traffic conditioner */ 501 return; 502 #endif 503 504 ip_len = ntohs(ip->ip_len); 505 if (ip_len < hlen) { 506 IPSTAT_INC(ips_badlen); 507 goto bad; 508 } 509 510 /* 511 * Check that the amount of data in the buffers 512 * is as at least much as the IP header would have us expect. 513 * Trim mbufs if longer than we expect. 514 * Drop packet if shorter than we expect. 515 */ 516 if (m->m_pkthdr.len < ip_len) { 517 tooshort: 518 IPSTAT_INC(ips_tooshort); 519 goto bad; 520 } 521 if (m->m_pkthdr.len > ip_len) { 522 if (m->m_len == m->m_pkthdr.len) { 523 m->m_len = ip_len; 524 m->m_pkthdr.len = ip_len; 525 } else 526 m_adj(m, ip_len - m->m_pkthdr.len); 527 } 528 529 /* Try to forward the packet, but if we fail continue */ 530 #ifdef IPSEC 531 /* For now we do not handle IPSEC in tryforward. */ 532 if (!key_havesp(IPSEC_DIR_INBOUND) && !key_havesp(IPSEC_DIR_OUTBOUND) && 533 (V_ipforwarding == 1)) 534 if (ip_tryforward(m) == NULL) 535 return; 536 /* 537 * Bypass packet filtering for packets previously handled by IPsec. 538 */ 539 if (ip_ipsec_filtertunnel(m)) 540 goto passin; 541 #else 542 if (V_ipforwarding == 1) 543 if (ip_tryforward(m) == NULL) 544 return; 545 #endif /* IPSEC */ 546 547 /* 548 * Run through list of hooks for input packets. 549 * 550 * NB: Beware of the destination address changing (e.g. 551 * by NAT rewriting). When this happens, tell 552 * ip_forward to do the right thing. 553 */ 554 555 /* Jump over all PFIL processing if hooks are not active. */ 556 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 557 goto passin; 558 559 odst = ip->ip_dst; 560 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 561 return; 562 if (m == NULL) /* consumed by filter */ 563 return; 564 565 ip = mtod(m, struct ip *); 566 dchg = (odst.s_addr != ip->ip_dst.s_addr); 567 ifp = m->m_pkthdr.rcvif; 568 569 if (m->m_flags & M_FASTFWD_OURS) { 570 m->m_flags &= ~M_FASTFWD_OURS; 571 goto ours; 572 } 573 if (m->m_flags & M_IP_NEXTHOP) { 574 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 575 /* 576 * Directly ship the packet on. This allows 577 * forwarding packets originally destined to us 578 * to some other directly connected host. 579 */ 580 ip_forward(m, 1); 581 return; 582 } 583 } 584 passin: 585 586 /* 587 * Process options and, if not destined for us, 588 * ship it on. ip_dooptions returns 1 when an 589 * error was detected (causing an icmp message 590 * to be sent and the original packet to be freed). 591 */ 592 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 593 return; 594 595 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 596 * matter if it is destined to another node, or whether it is 597 * a multicast one, RSVP wants it! and prevents it from being forwarded 598 * anywhere else. Also checks if the rsvp daemon is running before 599 * grabbing the packet. 600 */ 601 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 602 goto ours; 603 604 /* 605 * Check our list of addresses, to see if the packet is for us. 606 * If we don't have any addresses, assume any unicast packet 607 * we receive might be for us (and let the upper layers deal 608 * with it). 609 */ 610 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 611 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 612 goto ours; 613 614 /* 615 * Enable a consistency check between the destination address 616 * and the arrival interface for a unicast packet (the RFC 1122 617 * strong ES model) if IP forwarding is disabled and the packet 618 * is not locally generated and the packet is not subject to 619 * 'ipfw fwd'. 620 * 621 * XXX - Checking also should be disabled if the destination 622 * address is ipnat'ed to a different interface. 623 * 624 * XXX - Checking is incompatible with IP aliases added 625 * to the loopback interface instead of the interface where 626 * the packets are received. 627 * 628 * XXX - This is the case for carp vhost IPs as well so we 629 * insert a workaround. If the packet got here, we already 630 * checked with carp_iamatch() and carp_forus(). 631 */ 632 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 633 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 634 ifp->if_carp == NULL && (dchg == 0); 635 636 /* 637 * Check for exact addresses in the hash bucket. 638 */ 639 /* IN_IFADDR_RLOCK(); */ 640 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 641 /* 642 * If the address matches, verify that the packet 643 * arrived via the correct interface if checking is 644 * enabled. 645 */ 646 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 647 (!checkif || ia->ia_ifp == ifp)) { 648 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 649 counter_u64_add(ia->ia_ifa.ifa_ibytes, 650 m->m_pkthdr.len); 651 /* IN_IFADDR_RUNLOCK(); */ 652 goto ours; 653 } 654 } 655 /* IN_IFADDR_RUNLOCK(); */ 656 657 /* 658 * Check for broadcast addresses. 659 * 660 * Only accept broadcast packets that arrive via the matching 661 * interface. Reception of forwarded directed broadcasts would 662 * be handled via ip_forward() and ether_output() with the loopback 663 * into the stack for SIMPLEX interfaces handled by ether_output(). 664 */ 665 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 666 IF_ADDR_RLOCK(ifp); 667 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 668 if (ifa->ifa_addr->sa_family != AF_INET) 669 continue; 670 ia = ifatoia(ifa); 671 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 672 ip->ip_dst.s_addr) { 673 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 674 counter_u64_add(ia->ia_ifa.ifa_ibytes, 675 m->m_pkthdr.len); 676 IF_ADDR_RUNLOCK(ifp); 677 goto ours; 678 } 679 #ifdef BOOTP_COMPAT 680 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 681 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 682 counter_u64_add(ia->ia_ifa.ifa_ibytes, 683 m->m_pkthdr.len); 684 IF_ADDR_RUNLOCK(ifp); 685 goto ours; 686 } 687 #endif 688 } 689 IF_ADDR_RUNLOCK(ifp); 690 ia = NULL; 691 } 692 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 693 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 694 IPSTAT_INC(ips_cantforward); 695 m_freem(m); 696 return; 697 } 698 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 699 if (V_ip_mrouter) { 700 /* 701 * If we are acting as a multicast router, all 702 * incoming multicast packets are passed to the 703 * kernel-level multicast forwarding function. 704 * The packet is returned (relatively) intact; if 705 * ip_mforward() returns a non-zero value, the packet 706 * must be discarded, else it may be accepted below. 707 */ 708 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 709 IPSTAT_INC(ips_cantforward); 710 m_freem(m); 711 return; 712 } 713 714 /* 715 * The process-level routing daemon needs to receive 716 * all multicast IGMP packets, whether or not this 717 * host belongs to their destination groups. 718 */ 719 if (ip->ip_p == IPPROTO_IGMP) 720 goto ours; 721 IPSTAT_INC(ips_forward); 722 } 723 /* 724 * Assume the packet is for us, to avoid prematurely taking 725 * a lock on the in_multi hash. Protocols must perform 726 * their own filtering and update statistics accordingly. 727 */ 728 goto ours; 729 } 730 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 731 goto ours; 732 if (ip->ip_dst.s_addr == INADDR_ANY) 733 goto ours; 734 735 /* 736 * Not for us; forward if possible and desirable. 737 */ 738 if (V_ipforwarding == 0) { 739 IPSTAT_INC(ips_cantforward); 740 m_freem(m); 741 } else { 742 ip_forward(m, dchg); 743 } 744 return; 745 746 ours: 747 #ifdef IPSTEALTH 748 /* 749 * IPSTEALTH: Process non-routing options only 750 * if the packet is destined for us. 751 */ 752 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) 753 return; 754 #endif /* IPSTEALTH */ 755 756 /* 757 * Attempt reassembly; if it succeeds, proceed. 758 * ip_reass() will return a different mbuf. 759 */ 760 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 761 /* XXXGL: shouldn't we save & set m_flags? */ 762 m = ip_reass(m); 763 if (m == NULL) 764 return; 765 ip = mtod(m, struct ip *); 766 /* Get the header length of the reassembled packet */ 767 hlen = ip->ip_hl << 2; 768 } 769 770 #ifdef IPSEC 771 /* 772 * enforce IPsec policy checking if we are seeing last header. 773 * note that we do not visit this with protocols with pcb layer 774 * code - like udp/tcp/raw ip. 775 */ 776 if (ip_ipsec_input(m, ip->ip_p) != 0) 777 goto bad; 778 #endif /* IPSEC */ 779 780 /* 781 * Switch out to protocol's input routine. 782 */ 783 IPSTAT_INC(ips_delivered); 784 785 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 786 return; 787 bad: 788 m_freem(m); 789 } 790 791 /* 792 * IP timer processing; 793 * if a timer expires on a reassembly 794 * queue, discard it. 795 */ 796 void 797 ip_slowtimo(void) 798 { 799 VNET_ITERATOR_DECL(vnet_iter); 800 801 VNET_LIST_RLOCK_NOSLEEP(); 802 VNET_FOREACH(vnet_iter) { 803 CURVNET_SET(vnet_iter); 804 ipreass_slowtimo(); 805 CURVNET_RESTORE(); 806 } 807 VNET_LIST_RUNLOCK_NOSLEEP(); 808 } 809 810 void 811 ip_drain(void) 812 { 813 VNET_ITERATOR_DECL(vnet_iter); 814 815 VNET_LIST_RLOCK_NOSLEEP(); 816 VNET_FOREACH(vnet_iter) { 817 CURVNET_SET(vnet_iter); 818 ipreass_drain(); 819 CURVNET_RESTORE(); 820 } 821 VNET_LIST_RUNLOCK_NOSLEEP(); 822 } 823 824 /* 825 * The protocol to be inserted into ip_protox[] must be already registered 826 * in inetsw[], either statically or through pf_proto_register(). 827 */ 828 int 829 ipproto_register(short ipproto) 830 { 831 struct protosw *pr; 832 833 /* Sanity checks. */ 834 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 835 return (EPROTONOSUPPORT); 836 837 /* 838 * The protocol slot must not be occupied by another protocol 839 * already. An index pointing to IPPROTO_RAW is unused. 840 */ 841 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 842 if (pr == NULL) 843 return (EPFNOSUPPORT); 844 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 845 return (EEXIST); 846 847 /* Find the protocol position in inetsw[] and set the index. */ 848 for (pr = inetdomain.dom_protosw; 849 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 850 if (pr->pr_domain->dom_family == PF_INET && 851 pr->pr_protocol && pr->pr_protocol == ipproto) { 852 ip_protox[pr->pr_protocol] = pr - inetsw; 853 return (0); 854 } 855 } 856 return (EPROTONOSUPPORT); 857 } 858 859 int 860 ipproto_unregister(short ipproto) 861 { 862 struct protosw *pr; 863 864 /* Sanity checks. */ 865 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 866 return (EPROTONOSUPPORT); 867 868 /* Check if the protocol was indeed registered. */ 869 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 870 if (pr == NULL) 871 return (EPFNOSUPPORT); 872 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 873 return (ENOENT); 874 875 /* Reset the protocol slot to IPPROTO_RAW. */ 876 ip_protox[ipproto] = pr - inetsw; 877 return (0); 878 } 879 880 u_char inetctlerrmap[PRC_NCMDS] = { 881 0, 0, 0, 0, 882 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 883 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 884 EMSGSIZE, EHOSTUNREACH, 0, 0, 885 0, 0, EHOSTUNREACH, 0, 886 ENOPROTOOPT, ECONNREFUSED 887 }; 888 889 /* 890 * Forward a packet. If some error occurs return the sender 891 * an icmp packet. Note we can't always generate a meaningful 892 * icmp message because icmp doesn't have a large enough repertoire 893 * of codes and types. 894 * 895 * If not forwarding, just drop the packet. This could be confusing 896 * if ipforwarding was zero but some routing protocol was advancing 897 * us as a gateway to somewhere. However, we must let the routing 898 * protocol deal with that. 899 * 900 * The srcrt parameter indicates whether the packet is being forwarded 901 * via a source route. 902 */ 903 void 904 ip_forward(struct mbuf *m, int srcrt) 905 { 906 struct ip *ip = mtod(m, struct ip *); 907 struct in_ifaddr *ia; 908 struct mbuf *mcopy; 909 struct sockaddr_in *sin; 910 struct in_addr dest; 911 struct route ro; 912 int error, type = 0, code = 0, mtu = 0; 913 914 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 915 IPSTAT_INC(ips_cantforward); 916 m_freem(m); 917 return; 918 } 919 #ifdef IPSEC 920 if (ip_ipsec_fwd(m) != 0) { 921 IPSTAT_INC(ips_cantforward); 922 m_freem(m); 923 return; 924 } 925 #endif /* IPSEC */ 926 #ifdef IPSTEALTH 927 if (!V_ipstealth) { 928 #endif 929 if (ip->ip_ttl <= IPTTLDEC) { 930 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 931 0, 0); 932 return; 933 } 934 #ifdef IPSTEALTH 935 } 936 #endif 937 938 bzero(&ro, sizeof(ro)); 939 sin = (struct sockaddr_in *)&ro.ro_dst; 940 sin->sin_family = AF_INET; 941 sin->sin_len = sizeof(*sin); 942 sin->sin_addr = ip->ip_dst; 943 #ifdef RADIX_MPATH 944 rtalloc_mpath_fib(&ro, 945 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 946 M_GETFIB(m)); 947 #else 948 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 949 #endif 950 if (ro.ro_rt != NULL) { 951 ia = ifatoia(ro.ro_rt->rt_ifa); 952 ifa_ref(&ia->ia_ifa); 953 } else 954 ia = NULL; 955 #ifndef IPSEC 956 /* 957 * 'ia' may be NULL if there is no route for this destination. 958 * In case of IPsec, Don't discard it just yet, but pass it to 959 * ip_output in case of outgoing IPsec policy. 960 */ 961 if (!srcrt && ia == NULL) { 962 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 963 RO_RTFREE(&ro); 964 return; 965 } 966 #endif 967 968 /* 969 * Save the IP header and at most 8 bytes of the payload, 970 * in case we need to generate an ICMP message to the src. 971 * 972 * XXX this can be optimized a lot by saving the data in a local 973 * buffer on the stack (72 bytes at most), and only allocating the 974 * mbuf if really necessary. The vast majority of the packets 975 * are forwarded without having to send an ICMP back (either 976 * because unnecessary, or because rate limited), so we are 977 * really we are wasting a lot of work here. 978 * 979 * We don't use m_copy() because it might return a reference 980 * to a shared cluster. Both this function and ip_output() 981 * assume exclusive access to the IP header in `m', so any 982 * data in a cluster may change before we reach icmp_error(). 983 */ 984 mcopy = m_gethdr(M_NOWAIT, m->m_type); 985 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { 986 /* 987 * It's probably ok if the pkthdr dup fails (because 988 * the deep copy of the tag chain failed), but for now 989 * be conservative and just discard the copy since 990 * code below may some day want the tags. 991 */ 992 m_free(mcopy); 993 mcopy = NULL; 994 } 995 if (mcopy != NULL) { 996 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 997 mcopy->m_pkthdr.len = mcopy->m_len; 998 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 999 } 1000 1001 #ifdef IPSTEALTH 1002 if (!V_ipstealth) { 1003 #endif 1004 ip->ip_ttl -= IPTTLDEC; 1005 #ifdef IPSTEALTH 1006 } 1007 #endif 1008 1009 /* 1010 * If forwarding packet using same interface that it came in on, 1011 * perhaps should send a redirect to sender to shortcut a hop. 1012 * Only send redirect if source is sending directly to us, 1013 * and if packet was not source routed (or has any options). 1014 * Also, don't send redirect if forwarding using a default route 1015 * or a route modified by a redirect. 1016 */ 1017 dest.s_addr = 0; 1018 if (!srcrt && V_ipsendredirects && 1019 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1020 struct rtentry *rt; 1021 1022 rt = ro.ro_rt; 1023 1024 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1025 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1026 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1027 u_long src = ntohl(ip->ip_src.s_addr); 1028 1029 if (RTA(rt) && 1030 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1031 if (rt->rt_flags & RTF_GATEWAY) 1032 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1033 else 1034 dest.s_addr = ip->ip_dst.s_addr; 1035 /* Router requirements says to only send host redirects */ 1036 type = ICMP_REDIRECT; 1037 code = ICMP_REDIRECT_HOST; 1038 } 1039 } 1040 } 1041 1042 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1043 1044 if (error == EMSGSIZE && ro.ro_rt) 1045 mtu = ro.ro_rt->rt_mtu; 1046 RO_RTFREE(&ro); 1047 1048 if (error) 1049 IPSTAT_INC(ips_cantforward); 1050 else { 1051 IPSTAT_INC(ips_forward); 1052 if (type) 1053 IPSTAT_INC(ips_redirectsent); 1054 else { 1055 if (mcopy) 1056 m_freem(mcopy); 1057 if (ia != NULL) 1058 ifa_free(&ia->ia_ifa); 1059 return; 1060 } 1061 } 1062 if (mcopy == NULL) { 1063 if (ia != NULL) 1064 ifa_free(&ia->ia_ifa); 1065 return; 1066 } 1067 1068 switch (error) { 1069 1070 case 0: /* forwarded, but need redirect */ 1071 /* type, code set above */ 1072 break; 1073 1074 case ENETUNREACH: 1075 case EHOSTUNREACH: 1076 case ENETDOWN: 1077 case EHOSTDOWN: 1078 default: 1079 type = ICMP_UNREACH; 1080 code = ICMP_UNREACH_HOST; 1081 break; 1082 1083 case EMSGSIZE: 1084 type = ICMP_UNREACH; 1085 code = ICMP_UNREACH_NEEDFRAG; 1086 1087 #ifdef IPSEC 1088 /* 1089 * If IPsec is configured for this path, 1090 * override any possibly mtu value set by ip_output. 1091 */ 1092 mtu = ip_ipsec_mtu(mcopy, mtu); 1093 #endif /* IPSEC */ 1094 /* 1095 * If the MTU was set before make sure we are below the 1096 * interface MTU. 1097 * If the MTU wasn't set before use the interface mtu or 1098 * fall back to the next smaller mtu step compared to the 1099 * current packet size. 1100 */ 1101 if (mtu != 0) { 1102 if (ia != NULL) 1103 mtu = min(mtu, ia->ia_ifp->if_mtu); 1104 } else { 1105 if (ia != NULL) 1106 mtu = ia->ia_ifp->if_mtu; 1107 else 1108 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1109 } 1110 IPSTAT_INC(ips_cantfrag); 1111 break; 1112 1113 case ENOBUFS: 1114 case EACCES: /* ipfw denied packet */ 1115 m_freem(mcopy); 1116 if (ia != NULL) 1117 ifa_free(&ia->ia_ifa); 1118 return; 1119 } 1120 if (ia != NULL) 1121 ifa_free(&ia->ia_ifa); 1122 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1123 } 1124 1125 void 1126 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1127 struct mbuf *m) 1128 { 1129 1130 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1131 struct bintime bt; 1132 1133 bintime(&bt); 1134 if (inp->inp_socket->so_options & SO_BINTIME) { 1135 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), 1136 SCM_BINTIME, SOL_SOCKET); 1137 if (*mp) 1138 mp = &(*mp)->m_next; 1139 } 1140 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1141 struct timeval tv; 1142 1143 bintime2timeval(&bt, &tv); 1144 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), 1145 SCM_TIMESTAMP, SOL_SOCKET); 1146 if (*mp) 1147 mp = &(*mp)->m_next; 1148 } 1149 } 1150 if (inp->inp_flags & INP_RECVDSTADDR) { 1151 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, 1152 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1153 if (*mp) 1154 mp = &(*mp)->m_next; 1155 } 1156 if (inp->inp_flags & INP_RECVTTL) { 1157 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, 1158 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1159 if (*mp) 1160 mp = &(*mp)->m_next; 1161 } 1162 #ifdef notyet 1163 /* XXX 1164 * Moving these out of udp_input() made them even more broken 1165 * than they already were. 1166 */ 1167 /* options were tossed already */ 1168 if (inp->inp_flags & INP_RECVOPTS) { 1169 *mp = sbcreatecontrol((caddr_t)opts_deleted_above, 1170 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1171 if (*mp) 1172 mp = &(*mp)->m_next; 1173 } 1174 /* ip_srcroute doesn't do what we want here, need to fix */ 1175 if (inp->inp_flags & INP_RECVRETOPTS) { 1176 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), 1177 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1178 if (*mp) 1179 mp = &(*mp)->m_next; 1180 } 1181 #endif 1182 if (inp->inp_flags & INP_RECVIF) { 1183 struct ifnet *ifp; 1184 struct sdlbuf { 1185 struct sockaddr_dl sdl; 1186 u_char pad[32]; 1187 } sdlbuf; 1188 struct sockaddr_dl *sdp; 1189 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1190 1191 if ((ifp = m->m_pkthdr.rcvif) && 1192 ifp->if_index && ifp->if_index <= V_if_index) { 1193 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1194 /* 1195 * Change our mind and don't try copy. 1196 */ 1197 if (sdp->sdl_family != AF_LINK || 1198 sdp->sdl_len > sizeof(sdlbuf)) { 1199 goto makedummy; 1200 } 1201 bcopy(sdp, sdl2, sdp->sdl_len); 1202 } else { 1203 makedummy: 1204 sdl2->sdl_len = 1205 offsetof(struct sockaddr_dl, sdl_data[0]); 1206 sdl2->sdl_family = AF_LINK; 1207 sdl2->sdl_index = 0; 1208 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1209 } 1210 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, 1211 IP_RECVIF, IPPROTO_IP); 1212 if (*mp) 1213 mp = &(*mp)->m_next; 1214 } 1215 if (inp->inp_flags & INP_RECVTOS) { 1216 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, 1217 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1218 if (*mp) 1219 mp = &(*mp)->m_next; 1220 } 1221 1222 if (inp->inp_flags2 & INP_RECVFLOWID) { 1223 uint32_t flowid, flow_type; 1224 1225 flowid = m->m_pkthdr.flowid; 1226 flow_type = M_HASHTYPE_GET(m); 1227 1228 /* 1229 * XXX should handle the failure of one or the 1230 * other - don't populate both? 1231 */ 1232 *mp = sbcreatecontrol((caddr_t) &flowid, 1233 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP); 1234 if (*mp) 1235 mp = &(*mp)->m_next; 1236 *mp = sbcreatecontrol((caddr_t) &flow_type, 1237 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP); 1238 if (*mp) 1239 mp = &(*mp)->m_next; 1240 } 1241 1242 #ifdef RSS 1243 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1244 uint32_t flowid, flow_type; 1245 uint32_t rss_bucketid; 1246 1247 flowid = m->m_pkthdr.flowid; 1248 flow_type = M_HASHTYPE_GET(m); 1249 1250 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1251 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1252 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP); 1253 if (*mp) 1254 mp = &(*mp)->m_next; 1255 } 1256 } 1257 #endif 1258 } 1259 1260 /* 1261 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1262 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1263 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1264 * compiled. 1265 */ 1266 static VNET_DEFINE(int, ip_rsvp_on); 1267 VNET_DEFINE(struct socket *, ip_rsvpd); 1268 1269 #define V_ip_rsvp_on VNET(ip_rsvp_on) 1270 1271 int 1272 ip_rsvp_init(struct socket *so) 1273 { 1274 1275 if (so->so_type != SOCK_RAW || 1276 so->so_proto->pr_protocol != IPPROTO_RSVP) 1277 return EOPNOTSUPP; 1278 1279 if (V_ip_rsvpd != NULL) 1280 return EADDRINUSE; 1281 1282 V_ip_rsvpd = so; 1283 /* 1284 * This may seem silly, but we need to be sure we don't over-increment 1285 * the RSVP counter, in case something slips up. 1286 */ 1287 if (!V_ip_rsvp_on) { 1288 V_ip_rsvp_on = 1; 1289 V_rsvp_on++; 1290 } 1291 1292 return 0; 1293 } 1294 1295 int 1296 ip_rsvp_done(void) 1297 { 1298 1299 V_ip_rsvpd = NULL; 1300 /* 1301 * This may seem silly, but we need to be sure we don't over-decrement 1302 * the RSVP counter, in case something slips up. 1303 */ 1304 if (V_ip_rsvp_on) { 1305 V_ip_rsvp_on = 0; 1306 V_rsvp_on--; 1307 } 1308 return 0; 1309 } 1310 1311 int 1312 rsvp_input(struct mbuf **mp, int *offp, int proto) 1313 { 1314 struct mbuf *m; 1315 1316 m = *mp; 1317 *mp = NULL; 1318 1319 if (rsvp_input_p) { /* call the real one if loaded */ 1320 *mp = m; 1321 rsvp_input_p(mp, offp, proto); 1322 return (IPPROTO_DONE); 1323 } 1324 1325 /* Can still get packets with rsvp_on = 0 if there is a local member 1326 * of the group to which the RSVP packet is addressed. But in this 1327 * case we want to throw the packet away. 1328 */ 1329 1330 if (!V_rsvp_on) { 1331 m_freem(m); 1332 return (IPPROTO_DONE); 1333 } 1334 1335 if (V_ip_rsvpd != NULL) { 1336 *mp = m; 1337 rip_input(mp, offp, proto); 1338 return (IPPROTO_DONE); 1339 } 1340 /* Drop the packet */ 1341 m_freem(m); 1342 return (IPPROTO_DONE); 1343 } 1344