1 /*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1988, 1993 5 * The Regents of the University of California. All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 1. Redistributions of source code must retain the above copyright 11 * notice, this list of conditions and the following disclaimer. 12 * 2. Redistributions in binary form must reproduce the above copyright 13 * notice, this list of conditions and the following disclaimer in the 14 * documentation and/or other materials provided with the distribution. 15 * 3. Neither the name of the University nor the names of its contributors 16 * may be used to endorse or promote products derived from this software 17 * without specific prior written permission. 18 * 19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 29 * SUCH DAMAGE. 30 * 31 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 32 */ 33 34 #include <sys/cdefs.h> 35 __FBSDID("$FreeBSD$"); 36 37 #include "opt_bootp.h" 38 #include "opt_ipstealth.h" 39 #include "opt_ipsec.h" 40 #include "opt_route.h" 41 #include "opt_rss.h" 42 #include "opt_sctp.h" 43 44 #include <sys/param.h> 45 #include <sys/systm.h> 46 #include <sys/hhook.h> 47 #include <sys/mbuf.h> 48 #include <sys/malloc.h> 49 #include <sys/domain.h> 50 #include <sys/protosw.h> 51 #include <sys/socket.h> 52 #include <sys/time.h> 53 #include <sys/kernel.h> 54 #include <sys/lock.h> 55 #include <sys/rmlock.h> 56 #include <sys/rwlock.h> 57 #include <sys/sdt.h> 58 #include <sys/syslog.h> 59 #include <sys/sysctl.h> 60 61 #include <net/if.h> 62 #include <net/if_types.h> 63 #include <net/if_var.h> 64 #include <net/if_dl.h> 65 #include <net/pfil.h> 66 #include <net/route.h> 67 #include <net/route/nhop.h> 68 #include <net/netisr.h> 69 #include <net/rss_config.h> 70 #include <net/vnet.h> 71 72 #include <netinet/in.h> 73 #include <netinet/in_kdtrace.h> 74 #include <netinet/in_systm.h> 75 #include <netinet/in_var.h> 76 #include <netinet/ip.h> 77 #include <netinet/in_fib.h> 78 #include <netinet/in_pcb.h> 79 #include <netinet/ip_var.h> 80 #include <netinet/ip_encap.h> 81 #include <netinet/ip_fw.h> 82 #include <netinet/ip_icmp.h> 83 #include <netinet/igmp_var.h> 84 #include <netinet/ip_options.h> 85 #include <machine/in_cksum.h> 86 #include <netinet/ip_carp.h> 87 #include <netinet/in_rss.h> 88 #ifdef SCTP 89 #include <netinet/sctp_var.h> 90 #endif 91 92 #include <netipsec/ipsec_support.h> 93 94 #include <sys/socketvar.h> 95 96 #include <security/mac/mac_framework.h> 97 98 #ifdef CTASSERT 99 CTASSERT(sizeof(struct ip) == 20); 100 #endif 101 102 /* IP reassembly functions are defined in ip_reass.c. */ 103 extern void ipreass_init(void); 104 extern void ipreass_vnet_init(void); 105 #ifdef VIMAGE 106 extern void ipreass_destroy(void); 107 #endif 108 109 VNET_DEFINE(int, rsvp_on); 110 111 VNET_DEFINE(int, ipforwarding); 112 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, 113 &VNET_NAME(ipforwarding), 0, 114 "Enable IP forwarding between interfaces"); 115 116 /* 117 * Respond with an ICMP host redirect when we forward a packet out of 118 * the same interface on which it was received. See RFC 792. 119 */ 120 VNET_DEFINE(int, ipsendredirects) = 1; 121 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, 122 &VNET_NAME(ipsendredirects), 0, 123 "Enable sending IP redirects"); 124 125 VNET_DEFINE_STATIC(bool, ip_strong_es) = false; 126 #define V_ip_strong_es VNET(ip_strong_es) 127 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, rfc1122_strong_es, 128 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_strong_es), false, 129 "Packet's IP destination address must match address on arrival interface"); 130 131 VNET_DEFINE_STATIC(bool, ip_sav) = true; 132 #define V_ip_sav VNET(ip_sav) 133 SYSCTL_BOOL(_net_inet_ip, OID_AUTO, source_address_validation, 134 CTLFLAG_VNET | CTLFLAG_RW, &VNET_NAME(ip_sav), true, 135 "Drop incoming packets with source address that is a local address"); 136 137 VNET_DEFINE(pfil_head_t, inet_pfil_head); /* Packet filter hooks */ 138 139 static struct netisr_handler ip_nh = { 140 .nh_name = "ip", 141 .nh_handler = ip_input, 142 .nh_proto = NETISR_IP, 143 #ifdef RSS 144 .nh_m2cpuid = rss_soft_m2cpuid_v4, 145 .nh_policy = NETISR_POLICY_CPU, 146 .nh_dispatch = NETISR_DISPATCH_HYBRID, 147 #else 148 .nh_policy = NETISR_POLICY_FLOW, 149 #endif 150 }; 151 152 #ifdef RSS 153 /* 154 * Directly dispatched frames are currently assumed 155 * to have a flowid already calculated. 156 * 157 * It should likely have something that assert it 158 * actually has valid flow details. 159 */ 160 static struct netisr_handler ip_direct_nh = { 161 .nh_name = "ip_direct", 162 .nh_handler = ip_direct_input, 163 .nh_proto = NETISR_IP_DIRECT, 164 .nh_m2cpuid = rss_soft_m2cpuid_v4, 165 .nh_policy = NETISR_POLICY_CPU, 166 .nh_dispatch = NETISR_DISPATCH_HYBRID, 167 }; 168 #endif 169 170 ipproto_input_t *ip_protox[IPPROTO_MAX] = { 171 [0 ... IPPROTO_MAX - 1] = rip_input }; 172 ipproto_ctlinput_t *ip_ctlprotox[IPPROTO_MAX] = { 173 [0 ... IPPROTO_MAX - 1] = rip_ctlinput }; 174 175 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 /* Make sure it is safe to use hashinit(9) on CK_LIST. */ 180 CTASSERT(sizeof(struct in_ifaddrhashhead) == sizeof(LIST_HEAD(, in_addr))); 181 182 #ifdef IPCTL_DEFMTU 183 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 184 &ip_mtu, 0, "Default MTU"); 185 #endif 186 187 #ifdef IPSTEALTH 188 VNET_DEFINE(int, ipstealth); 189 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, 190 &VNET_NAME(ipstealth), 0, 191 "IP stealth mode, no TTL decrementation on forwarding"); 192 #endif 193 194 /* 195 * IP statistics are stored in the "array" of counter(9)s. 196 */ 197 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 198 VNET_PCPUSTAT_SYSINIT(ipstat); 199 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 200 "IP statistics (struct ipstat, netinet/ip_var.h)"); 201 202 #ifdef VIMAGE 203 VNET_PCPUSTAT_SYSUNINIT(ipstat); 204 #endif /* VIMAGE */ 205 206 /* 207 * Kernel module interface for updating ipstat. The argument is an index 208 * into ipstat treated as an array. 209 */ 210 void 211 kmod_ipstat_inc(int statnum) 212 { 213 214 counter_u64_add(VNET(ipstat)[statnum], 1); 215 } 216 217 void 218 kmod_ipstat_dec(int statnum) 219 { 220 221 counter_u64_add(VNET(ipstat)[statnum], -1); 222 } 223 224 static int 225 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 226 { 227 int error, qlimit; 228 229 netisr_getqlimit(&ip_nh, &qlimit); 230 error = sysctl_handle_int(oidp, &qlimit, 0, req); 231 if (error || !req->newptr) 232 return (error); 233 if (qlimit < 1) 234 return (EINVAL); 235 return (netisr_setqlimit(&ip_nh, qlimit)); 236 } 237 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 238 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 0, 0, 239 sysctl_netinet_intr_queue_maxlen, "I", 240 "Maximum size of the IP input queue"); 241 242 static int 243 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 244 { 245 u_int64_t qdrops_long; 246 int error, qdrops; 247 248 netisr_getqdrops(&ip_nh, &qdrops_long); 249 qdrops = qdrops_long; 250 error = sysctl_handle_int(oidp, &qdrops, 0, req); 251 if (error || !req->newptr) 252 return (error); 253 if (qdrops != 0) 254 return (EINVAL); 255 netisr_clearqdrops(&ip_nh); 256 return (0); 257 } 258 259 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 260 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 261 0, 0, sysctl_netinet_intr_queue_drops, "I", 262 "Number of packets dropped from the IP input queue"); 263 264 #ifdef RSS 265 static int 266 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 267 { 268 int error, qlimit; 269 270 netisr_getqlimit(&ip_direct_nh, &qlimit); 271 error = sysctl_handle_int(oidp, &qlimit, 0, req); 272 if (error || !req->newptr) 273 return (error); 274 if (qlimit < 1) 275 return (EINVAL); 276 return (netisr_setqlimit(&ip_direct_nh, qlimit)); 277 } 278 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 279 CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_MPSAFE, 280 0, 0, sysctl_netinet_intr_direct_queue_maxlen, 281 "I", "Maximum size of the IP direct input queue"); 282 283 static int 284 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) 285 { 286 u_int64_t qdrops_long; 287 int error, qdrops; 288 289 netisr_getqdrops(&ip_direct_nh, &qdrops_long); 290 qdrops = qdrops_long; 291 error = sysctl_handle_int(oidp, &qdrops, 0, req); 292 if (error || !req->newptr) 293 return (error); 294 if (qdrops != 0) 295 return (EINVAL); 296 netisr_clearqdrops(&ip_direct_nh); 297 return (0); 298 } 299 300 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops, 301 CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_MPSAFE, 0, 0, 302 sysctl_netinet_intr_direct_queue_drops, "I", 303 "Number of packets dropped from the IP direct input queue"); 304 #endif /* RSS */ 305 306 /* 307 * IP initialization: fill in IP protocol switch table. 308 * All protocols not implemented in kernel go to raw IP protocol handler. 309 */ 310 static void 311 ip_vnet_init(void *arg __unused) 312 { 313 struct pfil_head_args args; 314 315 CK_STAILQ_INIT(&V_in_ifaddrhead); 316 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 317 318 /* Initialize IP reassembly queue. */ 319 ipreass_vnet_init(); 320 321 /* Initialize packet filter hooks. */ 322 args.pa_version = PFIL_VERSION; 323 args.pa_flags = PFIL_IN | PFIL_OUT; 324 args.pa_type = PFIL_TYPE_IP4; 325 args.pa_headname = PFIL_INET_NAME; 326 V_inet_pfil_head = pfil_head_register(&args); 327 328 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET, 329 &V_ipsec_hhh_in[HHOOK_IPSEC_INET], 330 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 331 printf("%s: WARNING: unable to register input helper hook\n", 332 __func__); 333 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET, 334 &V_ipsec_hhh_out[HHOOK_IPSEC_INET], 335 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 336 printf("%s: WARNING: unable to register output helper hook\n", 337 __func__); 338 339 #ifdef VIMAGE 340 netisr_register_vnet(&ip_nh); 341 #ifdef RSS 342 netisr_register_vnet(&ip_direct_nh); 343 #endif 344 #endif 345 } 346 VNET_SYSINIT(ip_vnet_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_FOURTH, 347 ip_vnet_init, NULL); 348 349 static void 350 ip_init(const void *unused __unused) 351 { 352 353 ipreass_init(); 354 355 /* 356 * Register statically compiled protocols, that are unlikely to 357 * ever become dynamic. 358 */ 359 IPPROTO_REGISTER(IPPROTO_ICMP, icmp_input, NULL); 360 IPPROTO_REGISTER(IPPROTO_IGMP, igmp_input, NULL); 361 IPPROTO_REGISTER(IPPROTO_RSVP, rsvp_input, NULL); 362 IPPROTO_REGISTER(IPPROTO_IPV4, encap4_input, NULL); 363 IPPROTO_REGISTER(IPPROTO_MOBILE, encap4_input, NULL); 364 IPPROTO_REGISTER(IPPROTO_ETHERIP, encap4_input, NULL); 365 IPPROTO_REGISTER(IPPROTO_GRE, encap4_input, NULL); 366 IPPROTO_REGISTER(IPPROTO_IPV6, encap4_input, NULL); 367 IPPROTO_REGISTER(IPPROTO_PIM, encap4_input, NULL); 368 #ifdef SCTP /* XXX: has a loadable & static version */ 369 IPPROTO_REGISTER(IPPROTO_SCTP, sctp_input, sctp_ctlinput); 370 #endif 371 372 netisr_register(&ip_nh); 373 #ifdef RSS 374 netisr_register(&ip_direct_nh); 375 #endif 376 } 377 SYSINIT(ip_init, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_init, NULL); 378 379 #ifdef VIMAGE 380 static void 381 ip_destroy(void *unused __unused) 382 { 383 int error; 384 385 #ifdef RSS 386 netisr_unregister_vnet(&ip_direct_nh); 387 #endif 388 netisr_unregister_vnet(&ip_nh); 389 390 pfil_head_unregister(V_inet_pfil_head); 391 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]); 392 if (error != 0) { 393 printf("%s: WARNING: unable to deregister input helper hook " 394 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: " 395 "error %d returned\n", __func__, error); 396 } 397 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]); 398 if (error != 0) { 399 printf("%s: WARNING: unable to deregister output helper hook " 400 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: " 401 "error %d returned\n", __func__, error); 402 } 403 404 /* Remove the IPv4 addresses from all interfaces. */ 405 in_ifscrub_all(); 406 407 /* Make sure the IPv4 routes are gone as well. */ 408 rib_flush_routes_family(AF_INET); 409 410 /* Destroy IP reassembly queue. */ 411 ipreass_destroy(); 412 413 /* Cleanup in_ifaddr hash table; should be empty. */ 414 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 415 } 416 417 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL); 418 #endif 419 420 #ifdef RSS 421 /* 422 * IP direct input routine. 423 * 424 * This is called when reinjecting completed fragments where 425 * all of the previous checking and book-keeping has been done. 426 */ 427 void 428 ip_direct_input(struct mbuf *m) 429 { 430 struct ip *ip; 431 int hlen; 432 433 ip = mtod(m, struct ip *); 434 hlen = ip->ip_hl << 2; 435 436 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 437 if (IPSEC_ENABLED(ipv4)) { 438 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0) 439 return; 440 } 441 #endif /* IPSEC */ 442 IPSTAT_INC(ips_delivered); 443 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p); 444 } 445 #endif 446 447 /* 448 * Ip input routine. Checksum and byte swap header. If fragmented 449 * try to reassemble. Process options. Pass to next level. 450 */ 451 void 452 ip_input(struct mbuf *m) 453 { 454 struct ip *ip = NULL; 455 struct in_ifaddr *ia = NULL; 456 struct ifaddr *ifa; 457 struct ifnet *ifp; 458 int hlen = 0; 459 uint16_t sum, ip_len; 460 int dchg = 0; /* dest changed after fw */ 461 struct in_addr odst; /* original dst address */ 462 bool strong_es; 463 464 M_ASSERTPKTHDR(m); 465 NET_EPOCH_ASSERT(); 466 467 if (m->m_flags & M_FASTFWD_OURS) { 468 m->m_flags &= ~M_FASTFWD_OURS; 469 /* Set up some basics that will be used later. */ 470 ip = mtod(m, struct ip *); 471 hlen = ip->ip_hl << 2; 472 ip_len = ntohs(ip->ip_len); 473 goto ours; 474 } 475 476 IPSTAT_INC(ips_total); 477 478 if (__predict_false(m->m_pkthdr.len < sizeof(struct ip))) 479 goto tooshort; 480 481 if (m->m_len < sizeof(struct ip)) { 482 m = m_pullup(m, sizeof(struct ip)); 483 if (__predict_false(m == NULL)) { 484 IPSTAT_INC(ips_toosmall); 485 return; 486 } 487 } 488 ip = mtod(m, struct ip *); 489 490 if (__predict_false(ip->ip_v != IPVERSION)) { 491 IPSTAT_INC(ips_badvers); 492 goto bad; 493 } 494 495 hlen = ip->ip_hl << 2; 496 if (__predict_false(hlen < sizeof(struct ip))) { /* minimum header length */ 497 IPSTAT_INC(ips_badhlen); 498 goto bad; 499 } 500 if (hlen > m->m_len) { 501 m = m_pullup(m, hlen); 502 if (__predict_false(m == NULL)) { 503 IPSTAT_INC(ips_badhlen); 504 return; 505 } 506 ip = mtod(m, struct ip *); 507 } 508 509 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 510 511 /* IN_LOOPBACK must not appear on the wire - RFC1122 */ 512 ifp = m->m_pkthdr.rcvif; 513 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) || 514 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) { 515 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 516 IPSTAT_INC(ips_badaddr); 517 goto bad; 518 } 519 } 520 521 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 522 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 523 } else { 524 if (hlen == sizeof(struct ip)) { 525 sum = in_cksum_hdr(ip); 526 } else { 527 sum = in_cksum(m, hlen); 528 } 529 } 530 if (__predict_false(sum)) { 531 IPSTAT_INC(ips_badsum); 532 goto bad; 533 } 534 535 ip_len = ntohs(ip->ip_len); 536 if (__predict_false(ip_len < hlen)) { 537 IPSTAT_INC(ips_badlen); 538 goto bad; 539 } 540 541 /* 542 * Check that the amount of data in the buffers 543 * is as at least much as the IP header would have us expect. 544 * Trim mbufs if longer than we expect. 545 * Drop packet if shorter than we expect. 546 */ 547 if (__predict_false(m->m_pkthdr.len < ip_len)) { 548 tooshort: 549 IPSTAT_INC(ips_tooshort); 550 goto bad; 551 } 552 if (m->m_pkthdr.len > ip_len) { 553 if (m->m_len == m->m_pkthdr.len) { 554 m->m_len = ip_len; 555 m->m_pkthdr.len = ip_len; 556 } else 557 m_adj(m, ip_len - m->m_pkthdr.len); 558 } 559 560 /* 561 * Try to forward the packet, but if we fail continue. 562 * ip_tryforward() may generate redirects these days. 563 * XXX the logic below falling through to normal processing 564 * if redirects are required should be revisited as well. 565 * ip_tryforward() does inbound and outbound packet firewall 566 * processing. If firewall has decided that destination becomes 567 * our local address, it sets M_FASTFWD_OURS flag. In this 568 * case skip another inbound firewall processing and update 569 * ip pointer. 570 */ 571 if (V_ipforwarding != 0 572 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 573 && (!IPSEC_ENABLED(ipv4) || 574 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0) 575 #endif 576 ) { 577 /* 578 * ip_dooptions() was run so we can ignore the source route (or 579 * any IP options case) case for redirects in ip_tryforward(). 580 */ 581 if ((m = ip_tryforward(m)) == NULL) 582 return; 583 if (m->m_flags & M_FASTFWD_OURS) { 584 m->m_flags &= ~M_FASTFWD_OURS; 585 ip = mtod(m, struct ip *); 586 goto ours; 587 } 588 } 589 590 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 591 /* 592 * Bypass packet filtering for packets previously handled by IPsec. 593 */ 594 if (IPSEC_ENABLED(ipv4) && 595 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0) 596 goto passin; 597 #endif 598 599 /* 600 * Run through list of hooks for input packets. 601 * 602 * NB: Beware of the destination address changing (e.g. 603 * by NAT rewriting). When this happens, tell 604 * ip_forward to do the right thing. 605 */ 606 607 /* Jump over all PFIL processing if hooks are not active. */ 608 if (!PFIL_HOOKED_IN(V_inet_pfil_head)) 609 goto passin; 610 611 odst = ip->ip_dst; 612 if (pfil_mbuf_in(V_inet_pfil_head, &m, ifp, NULL) != 613 PFIL_PASS) 614 return; 615 if (m == NULL) /* consumed by filter */ 616 return; 617 618 ip = mtod(m, struct ip *); 619 dchg = (odst.s_addr != ip->ip_dst.s_addr); 620 621 if (m->m_flags & M_FASTFWD_OURS) { 622 m->m_flags &= ~M_FASTFWD_OURS; 623 goto ours; 624 } 625 if (m->m_flags & M_IP_NEXTHOP) { 626 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 627 /* 628 * Directly ship the packet on. This allows 629 * forwarding packets originally destined to us 630 * to some other directly connected host. 631 */ 632 ip_forward(m, 1); 633 return; 634 } 635 } 636 passin: 637 638 /* 639 * Process options and, if not destined for us, 640 * ship it on. ip_dooptions returns 1 when an 641 * error was detected (causing an icmp message 642 * to be sent and the original packet to be freed). 643 */ 644 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 645 return; 646 647 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 648 * matter if it is destined to another node, or whether it is 649 * a multicast one, RSVP wants it! and prevents it from being forwarded 650 * anywhere else. Also checks if the rsvp daemon is running before 651 * grabbing the packet. 652 */ 653 if (ip->ip_p == IPPROTO_RSVP && V_rsvp_on) 654 goto ours; 655 656 /* 657 * Check our list of addresses, to see if the packet is for us. 658 * If we don't have any addresses, assume any unicast packet 659 * we receive might be for us (and let the upper layers deal 660 * with it). 661 */ 662 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) && 663 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 664 goto ours; 665 666 /* 667 * Enable a consistency check between the destination address 668 * and the arrival interface for a unicast packet (the RFC 1122 669 * strong ES model) with a list of additional predicates: 670 * - if IP forwarding is disabled 671 * - the packet is not locally generated 672 * - the packet is not subject to 'ipfw fwd' 673 * - Interface is not running CARP. If the packet got here, we already 674 * checked it with carp_iamatch() and carp_forus(). 675 */ 676 strong_es = V_ip_strong_es && (V_ipforwarding == 0) && 677 ((ifp->if_flags & IFF_LOOPBACK) == 0) && 678 ifp->if_carp == NULL && (dchg == 0); 679 680 /* 681 * Check for exact addresses in the hash bucket. 682 */ 683 CK_LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 684 if (IA_SIN(ia)->sin_addr.s_addr != ip->ip_dst.s_addr) 685 continue; 686 687 /* 688 * net.inet.ip.rfc1122_strong_es: the address matches, verify 689 * that the packet arrived via the correct interface. 690 */ 691 if (__predict_false(strong_es && ia->ia_ifp != ifp)) { 692 IPSTAT_INC(ips_badaddr); 693 goto bad; 694 } 695 696 /* 697 * net.inet.ip.source_address_validation: drop incoming 698 * packets that pretend to be ours. 699 */ 700 if (V_ip_sav && !(ifp->if_flags & IFF_LOOPBACK) && 701 __predict_false(in_localip_fib(ip->ip_src, ifp->if_fib))) { 702 IPSTAT_INC(ips_badaddr); 703 goto bad; 704 } 705 706 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 707 counter_u64_add(ia->ia_ifa.ifa_ibytes, m->m_pkthdr.len); 708 goto ours; 709 } 710 711 /* 712 * Check for broadcast addresses. 713 * 714 * Only accept broadcast packets that arrive via the matching 715 * interface. Reception of forwarded directed broadcasts would 716 * be handled via ip_forward() and ether_output() with the loopback 717 * into the stack for SIMPLEX interfaces handled by ether_output(). 718 */ 719 if (ifp->if_flags & IFF_BROADCAST) { 720 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 721 if (ifa->ifa_addr->sa_family != AF_INET) 722 continue; 723 ia = ifatoia(ifa); 724 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 725 ip->ip_dst.s_addr) { 726 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 727 counter_u64_add(ia->ia_ifa.ifa_ibytes, 728 m->m_pkthdr.len); 729 goto ours; 730 } 731 #ifdef BOOTP_COMPAT 732 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 733 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 734 counter_u64_add(ia->ia_ifa.ifa_ibytes, 735 m->m_pkthdr.len); 736 goto ours; 737 } 738 #endif 739 } 740 ia = NULL; 741 } 742 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 743 /* 744 * RFC 3927 2.7: Do not forward multicast packets from 745 * IN_LINKLOCAL. 746 */ 747 if (V_ip_mrouter && !IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) { 748 /* 749 * If we are acting as a multicast router, all 750 * incoming multicast packets are passed to the 751 * kernel-level multicast forwarding function. 752 * The packet is returned (relatively) intact; if 753 * ip_mforward() returns a non-zero value, the packet 754 * must be discarded, else it may be accepted below. 755 */ 756 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 757 IPSTAT_INC(ips_cantforward); 758 m_freem(m); 759 return; 760 } 761 762 /* 763 * The process-level routing daemon needs to receive 764 * all multicast IGMP packets, whether or not this 765 * host belongs to their destination groups. 766 */ 767 if (ip->ip_p == IPPROTO_IGMP) { 768 goto ours; 769 } 770 IPSTAT_INC(ips_forward); 771 } 772 /* 773 * Assume the packet is for us, to avoid prematurely taking 774 * a lock on the in_multi hash. Protocols must perform 775 * their own filtering and update statistics accordingly. 776 */ 777 goto ours; 778 } 779 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 780 goto ours; 781 if (ip->ip_dst.s_addr == INADDR_ANY) 782 goto ours; 783 /* RFC 3927 2.7: Do not forward packets to or from IN_LINKLOCAL. */ 784 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr)) || 785 IN_LINKLOCAL(ntohl(ip->ip_src.s_addr))) { 786 IPSTAT_INC(ips_cantforward); 787 m_freem(m); 788 return; 789 } 790 791 /* 792 * Not for us; forward if possible and desirable. 793 */ 794 if (V_ipforwarding == 0) { 795 IPSTAT_INC(ips_cantforward); 796 m_freem(m); 797 } else { 798 ip_forward(m, dchg); 799 } 800 return; 801 802 ours: 803 #ifdef IPSTEALTH 804 /* 805 * IPSTEALTH: Process non-routing options only 806 * if the packet is destined for us. 807 */ 808 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) 809 return; 810 #endif /* IPSTEALTH */ 811 812 /* 813 * Attempt reassembly; if it succeeds, proceed. 814 * ip_reass() will return a different mbuf. 815 */ 816 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 817 /* XXXGL: shouldn't we save & set m_flags? */ 818 m = ip_reass(m); 819 if (m == NULL) 820 return; 821 ip = mtod(m, struct ip *); 822 /* Get the header length of the reassembled packet */ 823 hlen = ip->ip_hl << 2; 824 } 825 826 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 827 if (IPSEC_ENABLED(ipv4)) { 828 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0) 829 return; 830 } 831 #endif /* IPSEC */ 832 833 /* 834 * Switch out to protocol's input routine. 835 */ 836 IPSTAT_INC(ips_delivered); 837 838 ip_protox[ip->ip_p](&m, &hlen, ip->ip_p); 839 return; 840 bad: 841 m_freem(m); 842 } 843 844 int 845 ipproto_register(uint8_t proto, ipproto_input_t input, ipproto_ctlinput_t ctl) 846 { 847 848 MPASS(proto > 0); 849 850 /* 851 * The protocol slot must not be occupied by another protocol 852 * already. An index pointing to rip_input() is unused. 853 */ 854 if (ip_protox[proto] == rip_input) { 855 ip_protox[proto] = input; 856 ip_ctlprotox[proto] = ctl; 857 return (0); 858 } else 859 return (EEXIST); 860 } 861 862 int 863 ipproto_unregister(uint8_t proto) 864 { 865 866 MPASS(proto > 0); 867 868 if (ip_protox[proto] != rip_input) { 869 ip_protox[proto] = rip_input; 870 ip_ctlprotox[proto] = rip_ctlinput; 871 return (0); 872 } else 873 return (ENOENT); 874 } 875 876 u_char inetctlerrmap[PRC_NCMDS] = { 877 0, 0, 0, 0, 878 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 879 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 880 EMSGSIZE, EHOSTUNREACH, 0, 0, 881 0, 0, EHOSTUNREACH, 0, 882 ENOPROTOOPT, ECONNREFUSED 883 }; 884 885 /* 886 * Forward a packet. If some error occurs return the sender 887 * an icmp packet. Note we can't always generate a meaningful 888 * icmp message because icmp doesn't have a large enough repertoire 889 * of codes and types. 890 * 891 * If not forwarding, just drop the packet. This could be confusing 892 * if ipforwarding was zero but some routing protocol was advancing 893 * us as a gateway to somewhere. However, we must let the routing 894 * protocol deal with that. 895 * 896 * The srcrt parameter indicates whether the packet is being forwarded 897 * via a source route. 898 */ 899 void 900 ip_forward(struct mbuf *m, int srcrt) 901 { 902 struct ip *ip = mtod(m, struct ip *); 903 struct in_ifaddr *ia; 904 struct mbuf *mcopy; 905 struct sockaddr_in *sin; 906 struct in_addr dest; 907 struct route ro; 908 uint32_t flowid; 909 int error, type = 0, code = 0, mtu = 0; 910 911 NET_EPOCH_ASSERT(); 912 913 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 914 IPSTAT_INC(ips_cantforward); 915 m_freem(m); 916 return; 917 } 918 if ( 919 #ifdef IPSTEALTH 920 V_ipstealth == 0 && 921 #endif 922 ip->ip_ttl <= IPTTLDEC) { 923 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); 924 return; 925 } 926 927 bzero(&ro, sizeof(ro)); 928 sin = (struct sockaddr_in *)&ro.ro_dst; 929 sin->sin_family = AF_INET; 930 sin->sin_len = sizeof(*sin); 931 sin->sin_addr = ip->ip_dst; 932 flowid = m->m_pkthdr.flowid; 933 ro.ro_nh = fib4_lookup(M_GETFIB(m), ip->ip_dst, 0, NHR_REF, flowid); 934 if (ro.ro_nh != NULL) { 935 ia = ifatoia(ro.ro_nh->nh_ifa); 936 } else 937 ia = NULL; 938 /* 939 * Save the IP header and at most 8 bytes of the payload, 940 * in case we need to generate an ICMP message to the src. 941 * 942 * XXX this can be optimized a lot by saving the data in a local 943 * buffer on the stack (72 bytes at most), and only allocating the 944 * mbuf if really necessary. The vast majority of the packets 945 * are forwarded without having to send an ICMP back (either 946 * because unnecessary, or because rate limited), so we are 947 * really we are wasting a lot of work here. 948 * 949 * We don't use m_copym() because it might return a reference 950 * to a shared cluster. Both this function and ip_output() 951 * assume exclusive access to the IP header in `m', so any 952 * data in a cluster may change before we reach icmp_error(). 953 */ 954 mcopy = m_gethdr(M_NOWAIT, m->m_type); 955 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { 956 /* 957 * It's probably ok if the pkthdr dup fails (because 958 * the deep copy of the tag chain failed), but for now 959 * be conservative and just discard the copy since 960 * code below may some day want the tags. 961 */ 962 m_free(mcopy); 963 mcopy = NULL; 964 } 965 if (mcopy != NULL) { 966 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 967 mcopy->m_pkthdr.len = mcopy->m_len; 968 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 969 } 970 #ifdef IPSTEALTH 971 if (V_ipstealth == 0) 972 #endif 973 ip->ip_ttl -= IPTTLDEC; 974 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 975 if (IPSEC_ENABLED(ipv4)) { 976 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) { 977 /* mbuf consumed by IPsec */ 978 RO_NHFREE(&ro); 979 m_freem(mcopy); 980 if (error != EINPROGRESS) 981 IPSTAT_INC(ips_cantforward); 982 return; 983 } 984 /* No IPsec processing required */ 985 } 986 #endif /* IPSEC */ 987 /* 988 * If forwarding packet using same interface that it came in on, 989 * perhaps should send a redirect to sender to shortcut a hop. 990 * Only send redirect if source is sending directly to us, 991 * and if packet was not source routed (or has any options). 992 * Also, don't send redirect if forwarding using a default route 993 * or a route modified by a redirect. 994 */ 995 dest.s_addr = 0; 996 if (!srcrt && V_ipsendredirects && 997 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 998 struct nhop_object *nh; 999 1000 nh = ro.ro_nh; 1001 1002 if (nh != NULL && ((nh->nh_flags & (NHF_REDIRECT|NHF_DEFAULT)) == 0)) { 1003 struct in_ifaddr *nh_ia = (struct in_ifaddr *)(nh->nh_ifa); 1004 u_long src = ntohl(ip->ip_src.s_addr); 1005 1006 if (nh_ia != NULL && 1007 (src & nh_ia->ia_subnetmask) == nh_ia->ia_subnet) { 1008 /* Router requirements says to only send host redirects */ 1009 type = ICMP_REDIRECT; 1010 code = ICMP_REDIRECT_HOST; 1011 if (nh->nh_flags & NHF_GATEWAY) { 1012 if (nh->gw_sa.sa_family == AF_INET) 1013 dest.s_addr = nh->gw4_sa.sin_addr.s_addr; 1014 else /* Do not redirect in case gw is AF_INET6 */ 1015 type = 0; 1016 } else 1017 dest.s_addr = ip->ip_dst.s_addr; 1018 } 1019 } 1020 } 1021 1022 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1023 1024 if (error == EMSGSIZE && ro.ro_nh) 1025 mtu = ro.ro_nh->nh_mtu; 1026 RO_NHFREE(&ro); 1027 1028 if (error) 1029 IPSTAT_INC(ips_cantforward); 1030 else { 1031 IPSTAT_INC(ips_forward); 1032 if (type) 1033 IPSTAT_INC(ips_redirectsent); 1034 else { 1035 if (mcopy) 1036 m_freem(mcopy); 1037 return; 1038 } 1039 } 1040 if (mcopy == NULL) 1041 return; 1042 1043 switch (error) { 1044 case 0: /* forwarded, but need redirect */ 1045 /* type, code set above */ 1046 break; 1047 1048 case ENETUNREACH: 1049 case EHOSTUNREACH: 1050 case ENETDOWN: 1051 case EHOSTDOWN: 1052 default: 1053 type = ICMP_UNREACH; 1054 code = ICMP_UNREACH_HOST; 1055 break; 1056 1057 case EMSGSIZE: 1058 type = ICMP_UNREACH; 1059 code = ICMP_UNREACH_NEEDFRAG; 1060 /* 1061 * If the MTU was set before make sure we are below the 1062 * interface MTU. 1063 * If the MTU wasn't set before use the interface mtu or 1064 * fall back to the next smaller mtu step compared to the 1065 * current packet size. 1066 */ 1067 if (mtu != 0) { 1068 if (ia != NULL) 1069 mtu = min(mtu, ia->ia_ifp->if_mtu); 1070 } else { 1071 if (ia != NULL) 1072 mtu = ia->ia_ifp->if_mtu; 1073 else 1074 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1075 } 1076 IPSTAT_INC(ips_cantfrag); 1077 break; 1078 1079 case ENOBUFS: 1080 case EACCES: /* ipfw denied packet */ 1081 m_freem(mcopy); 1082 return; 1083 } 1084 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1085 } 1086 1087 #define CHECK_SO_CT(sp, ct) \ 1088 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0) 1089 1090 void 1091 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1092 struct mbuf *m) 1093 { 1094 bool stamped; 1095 1096 stamped = false; 1097 if ((inp->inp_socket->so_options & SO_BINTIME) || 1098 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) { 1099 struct bintime boottimebin, bt; 1100 struct timespec ts1; 1101 1102 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1103 M_TSTMP)) { 1104 mbuf_tstmp2timespec(m, &ts1); 1105 timespec2bintime(&ts1, &bt); 1106 getboottimebin(&boottimebin); 1107 bintime_add(&bt, &boottimebin); 1108 } else { 1109 bintime(&bt); 1110 } 1111 *mp = sbcreatecontrol(&bt, sizeof(bt), SCM_BINTIME, 1112 SOL_SOCKET, M_NOWAIT); 1113 if (*mp != NULL) { 1114 mp = &(*mp)->m_next; 1115 stamped = true; 1116 } 1117 } 1118 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) { 1119 struct bintime boottimebin, bt1; 1120 struct timespec ts1; 1121 struct timeval tv; 1122 1123 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1124 M_TSTMP)) { 1125 mbuf_tstmp2timespec(m, &ts1); 1126 timespec2bintime(&ts1, &bt1); 1127 getboottimebin(&boottimebin); 1128 bintime_add(&bt1, &boottimebin); 1129 bintime2timeval(&bt1, &tv); 1130 } else { 1131 microtime(&tv); 1132 } 1133 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), SCM_TIMESTAMP, 1134 SOL_SOCKET, M_NOWAIT); 1135 if (*mp != NULL) { 1136 mp = &(*mp)->m_next; 1137 stamped = true; 1138 } 1139 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) { 1140 struct bintime boottimebin; 1141 struct timespec ts, ts1; 1142 1143 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1144 M_TSTMP)) { 1145 mbuf_tstmp2timespec(m, &ts); 1146 getboottimebin(&boottimebin); 1147 bintime2timespec(&boottimebin, &ts1); 1148 timespecadd(&ts, &ts1, &ts); 1149 } else { 1150 nanotime(&ts); 1151 } 1152 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_REALTIME, 1153 SOL_SOCKET, M_NOWAIT); 1154 if (*mp != NULL) { 1155 mp = &(*mp)->m_next; 1156 stamped = true; 1157 } 1158 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) { 1159 struct timespec ts; 1160 1161 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1162 M_TSTMP)) 1163 mbuf_tstmp2timespec(m, &ts); 1164 else 1165 nanouptime(&ts); 1166 *mp = sbcreatecontrol(&ts, sizeof(ts), SCM_MONOTONIC, 1167 SOL_SOCKET, M_NOWAIT); 1168 if (*mp != NULL) { 1169 mp = &(*mp)->m_next; 1170 stamped = true; 1171 } 1172 } 1173 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1174 M_TSTMP)) { 1175 struct sock_timestamp_info sti; 1176 1177 bzero(&sti, sizeof(sti)); 1178 sti.st_info_flags = ST_INFO_HW; 1179 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1180 sti.st_info_flags |= ST_INFO_HW_HPREC; 1181 *mp = sbcreatecontrol(&sti, sizeof(sti), SCM_TIME_INFO, 1182 SOL_SOCKET, M_NOWAIT); 1183 if (*mp != NULL) 1184 mp = &(*mp)->m_next; 1185 } 1186 if (inp->inp_flags & INP_RECVDSTADDR) { 1187 *mp = sbcreatecontrol(&ip->ip_dst, sizeof(struct in_addr), 1188 IP_RECVDSTADDR, IPPROTO_IP, M_NOWAIT); 1189 if (*mp) 1190 mp = &(*mp)->m_next; 1191 } 1192 if (inp->inp_flags & INP_RECVTTL) { 1193 *mp = sbcreatecontrol(&ip->ip_ttl, sizeof(u_char), IP_RECVTTL, 1194 IPPROTO_IP, M_NOWAIT); 1195 if (*mp) 1196 mp = &(*mp)->m_next; 1197 } 1198 #ifdef notyet 1199 /* XXX 1200 * Moving these out of udp_input() made them even more broken 1201 * than they already were. 1202 */ 1203 /* options were tossed already */ 1204 if (inp->inp_flags & INP_RECVOPTS) { 1205 *mp = sbcreatecontrol(opts_deleted_above, 1206 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP, M_NOWAIT); 1207 if (*mp) 1208 mp = &(*mp)->m_next; 1209 } 1210 /* ip_srcroute doesn't do what we want here, need to fix */ 1211 if (inp->inp_flags & INP_RECVRETOPTS) { 1212 *mp = sbcreatecontrol(ip_srcroute(m), sizeof(struct in_addr), 1213 IP_RECVRETOPTS, IPPROTO_IP, M_NOWAIT); 1214 if (*mp) 1215 mp = &(*mp)->m_next; 1216 } 1217 #endif 1218 if (inp->inp_flags & INP_RECVIF) { 1219 struct ifnet *ifp; 1220 struct sdlbuf { 1221 struct sockaddr_dl sdl; 1222 u_char pad[32]; 1223 } sdlbuf; 1224 struct sockaddr_dl *sdp; 1225 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1226 1227 if ((ifp = m->m_pkthdr.rcvif)) { 1228 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1229 /* 1230 * Change our mind and don't try copy. 1231 */ 1232 if (sdp->sdl_family != AF_LINK || 1233 sdp->sdl_len > sizeof(sdlbuf)) { 1234 goto makedummy; 1235 } 1236 bcopy(sdp, sdl2, sdp->sdl_len); 1237 } else { 1238 makedummy: 1239 sdl2->sdl_len = 1240 offsetof(struct sockaddr_dl, sdl_data[0]); 1241 sdl2->sdl_family = AF_LINK; 1242 sdl2->sdl_index = 0; 1243 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1244 } 1245 *mp = sbcreatecontrol(sdl2, sdl2->sdl_len, IP_RECVIF, 1246 IPPROTO_IP, M_NOWAIT); 1247 if (*mp) 1248 mp = &(*mp)->m_next; 1249 } 1250 if (inp->inp_flags & INP_RECVTOS) { 1251 *mp = sbcreatecontrol(&ip->ip_tos, sizeof(u_char), IP_RECVTOS, 1252 IPPROTO_IP, M_NOWAIT); 1253 if (*mp) 1254 mp = &(*mp)->m_next; 1255 } 1256 1257 if (inp->inp_flags2 & INP_RECVFLOWID) { 1258 uint32_t flowid, flow_type; 1259 1260 flowid = m->m_pkthdr.flowid; 1261 flow_type = M_HASHTYPE_GET(m); 1262 1263 /* 1264 * XXX should handle the failure of one or the 1265 * other - don't populate both? 1266 */ 1267 *mp = sbcreatecontrol(&flowid, sizeof(uint32_t), IP_FLOWID, 1268 IPPROTO_IP, M_NOWAIT); 1269 if (*mp) 1270 mp = &(*mp)->m_next; 1271 *mp = sbcreatecontrol(&flow_type, sizeof(uint32_t), 1272 IP_FLOWTYPE, IPPROTO_IP, M_NOWAIT); 1273 if (*mp) 1274 mp = &(*mp)->m_next; 1275 } 1276 1277 #ifdef RSS 1278 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1279 uint32_t flowid, flow_type; 1280 uint32_t rss_bucketid; 1281 1282 flowid = m->m_pkthdr.flowid; 1283 flow_type = M_HASHTYPE_GET(m); 1284 1285 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1286 *mp = sbcreatecontrol(&rss_bucketid, sizeof(uint32_t), 1287 IP_RSSBUCKETID, IPPROTO_IP, M_NOWAIT); 1288 if (*mp) 1289 mp = &(*mp)->m_next; 1290 } 1291 } 1292 #endif 1293 } 1294 1295 /* 1296 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1297 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1298 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1299 * compiled. 1300 */ 1301 VNET_DEFINE_STATIC(int, ip_rsvp_on); 1302 VNET_DEFINE(struct socket *, ip_rsvpd); 1303 1304 #define V_ip_rsvp_on VNET(ip_rsvp_on) 1305 1306 int 1307 ip_rsvp_init(struct socket *so) 1308 { 1309 1310 if (V_ip_rsvpd != NULL) 1311 return EADDRINUSE; 1312 1313 V_ip_rsvpd = so; 1314 /* 1315 * This may seem silly, but we need to be sure we don't over-increment 1316 * the RSVP counter, in case something slips up. 1317 */ 1318 if (!V_ip_rsvp_on) { 1319 V_ip_rsvp_on = 1; 1320 V_rsvp_on++; 1321 } 1322 1323 return 0; 1324 } 1325 1326 int 1327 ip_rsvp_done(void) 1328 { 1329 1330 V_ip_rsvpd = NULL; 1331 /* 1332 * This may seem silly, but we need to be sure we don't over-decrement 1333 * the RSVP counter, in case something slips up. 1334 */ 1335 if (V_ip_rsvp_on) { 1336 V_ip_rsvp_on = 0; 1337 V_rsvp_on--; 1338 } 1339 return 0; 1340 } 1341 1342 int 1343 rsvp_input(struct mbuf **mp, int *offp, int proto) 1344 { 1345 struct mbuf *m; 1346 1347 m = *mp; 1348 *mp = NULL; 1349 1350 if (rsvp_input_p) { /* call the real one if loaded */ 1351 *mp = m; 1352 rsvp_input_p(mp, offp, proto); 1353 return (IPPROTO_DONE); 1354 } 1355 1356 /* Can still get packets with rsvp_on = 0 if there is a local member 1357 * of the group to which the RSVP packet is addressed. But in this 1358 * case we want to throw the packet away. 1359 */ 1360 1361 if (!V_rsvp_on) { 1362 m_freem(m); 1363 return (IPPROTO_DONE); 1364 } 1365 1366 if (V_ip_rsvpd != NULL) { 1367 *mp = m; 1368 rip_input(mp, offp, proto); 1369 return (IPPROTO_DONE); 1370 } 1371 /* Drop the packet */ 1372 m_freem(m); 1373 return (IPPROTO_DONE); 1374 } 1375