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 43 #include <sys/param.h> 44 #include <sys/systm.h> 45 #include <sys/hhook.h> 46 #include <sys/mbuf.h> 47 #include <sys/malloc.h> 48 #include <sys/domain.h> 49 #include <sys/protosw.h> 50 #include <sys/socket.h> 51 #include <sys/time.h> 52 #include <sys/kernel.h> 53 #include <sys/lock.h> 54 #include <sys/rmlock.h> 55 #include <sys/rwlock.h> 56 #include <sys/sdt.h> 57 #include <sys/syslog.h> 58 #include <sys/sysctl.h> 59 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/pfil.h> 65 #include <net/route.h> 66 #include <net/netisr.h> 67 #include <net/rss_config.h> 68 #include <net/vnet.h> 69 70 #include <netinet/in.h> 71 #include <netinet/in_kdtrace.h> 72 #include <netinet/in_systm.h> 73 #include <netinet/in_var.h> 74 #include <netinet/ip.h> 75 #include <netinet/in_pcb.h> 76 #include <netinet/ip_var.h> 77 #include <netinet/ip_fw.h> 78 #include <netinet/ip_icmp.h> 79 #include <netinet/ip_options.h> 80 #include <machine/in_cksum.h> 81 #include <netinet/ip_carp.h> 82 #include <netinet/in_rss.h> 83 84 #include <netipsec/ipsec_support.h> 85 86 #include <sys/socketvar.h> 87 88 #include <security/mac/mac_framework.h> 89 90 #ifdef CTASSERT 91 CTASSERT(sizeof(struct ip) == 20); 92 #endif 93 94 /* IP reassembly functions are defined in ip_reass.c. */ 95 extern void ipreass_init(void); 96 extern void ipreass_drain(void); 97 extern void ipreass_slowtimo(void); 98 #ifdef VIMAGE 99 extern void ipreass_destroy(void); 100 #endif 101 102 struct rmlock in_ifaddr_lock; 103 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 104 105 VNET_DEFINE(int, rsvp_on); 106 107 VNET_DEFINE(int, ipforwarding); 108 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, 109 &VNET_NAME(ipforwarding), 0, 110 "Enable IP forwarding between interfaces"); 111 112 VNET_DEFINE_STATIC(int, ipsendredirects) = 1; /* XXX */ 113 #define V_ipsendredirects VNET(ipsendredirects) 114 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, 115 &VNET_NAME(ipsendredirects), 0, 116 "Enable sending IP redirects"); 117 118 /* 119 * XXX - Setting ip_checkinterface mostly implements the receive side of 120 * the Strong ES model described in RFC 1122, but since the routing table 121 * and transmit implementation do not implement the Strong ES model, 122 * setting this to 1 results in an odd hybrid. 123 * 124 * XXX - ip_checkinterface currently must be disabled if you use ipnat 125 * to translate the destination address to another local interface. 126 * 127 * XXX - ip_checkinterface must be disabled if you add IP aliases 128 * to the loopback interface instead of the interface where the 129 * packets for those addresses are received. 130 */ 131 VNET_DEFINE_STATIC(int, ip_checkinterface); 132 #define V_ip_checkinterface VNET(ip_checkinterface) 133 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW, 134 &VNET_NAME(ip_checkinterface), 0, 135 "Verify packet arrives on correct interface"); 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 extern struct domain inetdomain; 171 extern struct protosw inetsw[]; 172 u_char ip_protox[IPPROTO_MAX]; 173 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 174 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 175 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 176 177 #ifdef IPCTL_DEFMTU 178 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 179 &ip_mtu, 0, "Default MTU"); 180 #endif 181 182 #ifdef IPSTEALTH 183 VNET_DEFINE(int, ipstealth); 184 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, 185 &VNET_NAME(ipstealth), 0, 186 "IP stealth mode, no TTL decrementation on forwarding"); 187 #endif 188 189 /* 190 * IP statistics are stored in the "array" of counter(9)s. 191 */ 192 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 193 VNET_PCPUSTAT_SYSINIT(ipstat); 194 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 195 "IP statistics (struct ipstat, netinet/ip_var.h)"); 196 197 #ifdef VIMAGE 198 VNET_PCPUSTAT_SYSUNINIT(ipstat); 199 #endif /* VIMAGE */ 200 201 /* 202 * Kernel module interface for updating ipstat. The argument is an index 203 * into ipstat treated as an array. 204 */ 205 void 206 kmod_ipstat_inc(int statnum) 207 { 208 209 counter_u64_add(VNET(ipstat)[statnum], 1); 210 } 211 212 void 213 kmod_ipstat_dec(int statnum) 214 { 215 216 counter_u64_add(VNET(ipstat)[statnum], -1); 217 } 218 219 static int 220 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 221 { 222 int error, qlimit; 223 224 netisr_getqlimit(&ip_nh, &qlimit); 225 error = sysctl_handle_int(oidp, &qlimit, 0, req); 226 if (error || !req->newptr) 227 return (error); 228 if (qlimit < 1) 229 return (EINVAL); 230 return (netisr_setqlimit(&ip_nh, qlimit)); 231 } 232 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 233 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 234 "Maximum size of the IP input queue"); 235 236 static int 237 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 238 { 239 u_int64_t qdrops_long; 240 int error, qdrops; 241 242 netisr_getqdrops(&ip_nh, &qdrops_long); 243 qdrops = qdrops_long; 244 error = sysctl_handle_int(oidp, &qdrops, 0, req); 245 if (error || !req->newptr) 246 return (error); 247 if (qdrops != 0) 248 return (EINVAL); 249 netisr_clearqdrops(&ip_nh); 250 return (0); 251 } 252 253 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 254 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 255 "Number of packets dropped from the IP input queue"); 256 257 #ifdef RSS 258 static int 259 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 260 { 261 int error, qlimit; 262 263 netisr_getqlimit(&ip_direct_nh, &qlimit); 264 error = sysctl_handle_int(oidp, &qlimit, 0, req); 265 if (error || !req->newptr) 266 return (error); 267 if (qlimit < 1) 268 return (EINVAL); 269 return (netisr_setqlimit(&ip_direct_nh, qlimit)); 270 } 271 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 272 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, 273 "I", "Maximum size of the IP direct input queue"); 274 275 static int 276 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) 277 { 278 u_int64_t qdrops_long; 279 int error, qdrops; 280 281 netisr_getqdrops(&ip_direct_nh, &qdrops_long); 282 qdrops = qdrops_long; 283 error = sysctl_handle_int(oidp, &qdrops, 0, req); 284 if (error || !req->newptr) 285 return (error); 286 if (qdrops != 0) 287 return (EINVAL); 288 netisr_clearqdrops(&ip_direct_nh); 289 return (0); 290 } 291 292 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops, 293 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", 294 "Number of packets dropped from the IP direct input queue"); 295 #endif /* RSS */ 296 297 /* 298 * IP initialization: fill in IP protocol switch table. 299 * All protocols not implemented in kernel go to raw IP protocol handler. 300 */ 301 void 302 ip_init(void) 303 { 304 struct pfil_head_args args; 305 struct protosw *pr; 306 int i; 307 308 CK_STAILQ_INIT(&V_in_ifaddrhead); 309 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 310 311 /* Initialize IP reassembly queue. */ 312 ipreass_init(); 313 314 /* Initialize packet filter hooks. */ 315 args.pa_version = PFIL_VERSION; 316 args.pa_flags = PFIL_IN | PFIL_OUT; 317 args.pa_type = PFIL_TYPE_IP4; 318 args.pa_headname = PFIL_INET_NAME; 319 V_inet_pfil_head = pfil_head_register(&args); 320 321 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET, 322 &V_ipsec_hhh_in[HHOOK_IPSEC_INET], 323 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 324 printf("%s: WARNING: unable to register input helper hook\n", 325 __func__); 326 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET, 327 &V_ipsec_hhh_out[HHOOK_IPSEC_INET], 328 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 329 printf("%s: WARNING: unable to register output helper hook\n", 330 __func__); 331 332 /* Skip initialization of globals for non-default instances. */ 333 #ifdef VIMAGE 334 if (!IS_DEFAULT_VNET(curvnet)) { 335 netisr_register_vnet(&ip_nh); 336 #ifdef RSS 337 netisr_register_vnet(&ip_direct_nh); 338 #endif 339 return; 340 } 341 #endif 342 343 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 344 if (pr == NULL) 345 panic("ip_init: PF_INET not found"); 346 347 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 348 for (i = 0; i < IPPROTO_MAX; i++) 349 ip_protox[i] = pr - inetsw; 350 /* 351 * Cycle through IP protocols and put them into the appropriate place 352 * in ip_protox[]. 353 */ 354 for (pr = inetdomain.dom_protosw; 355 pr < inetdomain.dom_protoswNPROTOSW; pr++) 356 if (pr->pr_domain->dom_family == PF_INET && 357 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 358 /* Be careful to only index valid IP protocols. */ 359 if (pr->pr_protocol < IPPROTO_MAX) 360 ip_protox[pr->pr_protocol] = pr - inetsw; 361 } 362 363 netisr_register(&ip_nh); 364 #ifdef RSS 365 netisr_register(&ip_direct_nh); 366 #endif 367 } 368 369 #ifdef VIMAGE 370 static void 371 ip_destroy(void *unused __unused) 372 { 373 struct ifnet *ifp; 374 int error; 375 376 #ifdef RSS 377 netisr_unregister_vnet(&ip_direct_nh); 378 #endif 379 netisr_unregister_vnet(&ip_nh); 380 381 pfil_head_unregister(V_inet_pfil_head); 382 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]); 383 if (error != 0) { 384 printf("%s: WARNING: unable to deregister input helper hook " 385 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: " 386 "error %d returned\n", __func__, error); 387 } 388 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]); 389 if (error != 0) { 390 printf("%s: WARNING: unable to deregister output helper hook " 391 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: " 392 "error %d returned\n", __func__, error); 393 } 394 395 /* Remove the IPv4 addresses from all interfaces. */ 396 in_ifscrub_all(); 397 398 /* Make sure the IPv4 routes are gone as well. */ 399 IFNET_RLOCK(); 400 CK_STAILQ_FOREACH(ifp, &V_ifnet, if_link) 401 rt_flushifroutes_af(ifp, AF_INET); 402 IFNET_RUNLOCK(); 403 404 /* Destroy IP reassembly queue. */ 405 ipreass_destroy(); 406 407 /* Cleanup in_ifaddr hash table; should be empty. */ 408 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 409 } 410 411 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL); 412 #endif 413 414 #ifdef RSS 415 /* 416 * IP direct input routine. 417 * 418 * This is called when reinjecting completed fragments where 419 * all of the previous checking and book-keeping has been done. 420 */ 421 void 422 ip_direct_input(struct mbuf *m) 423 { 424 struct ip *ip; 425 int hlen; 426 427 ip = mtod(m, struct ip *); 428 hlen = ip->ip_hl << 2; 429 430 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 431 if (IPSEC_ENABLED(ipv4)) { 432 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0) 433 return; 434 } 435 #endif /* IPSEC */ 436 IPSTAT_INC(ips_delivered); 437 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 438 return; 439 } 440 #endif 441 442 /* 443 * Ip input routine. Checksum and byte swap header. If fragmented 444 * try to reassemble. Process options. Pass to next level. 445 */ 446 void 447 ip_input(struct mbuf *m) 448 { 449 struct rm_priotracker in_ifa_tracker; 450 struct ip *ip = NULL; 451 struct in_ifaddr *ia = NULL; 452 struct ifaddr *ifa; 453 struct ifnet *ifp; 454 int checkif, hlen = 0; 455 uint16_t sum, ip_len; 456 int dchg = 0; /* dest changed after fw */ 457 struct in_addr odst; /* original dst address */ 458 459 M_ASSERTPKTHDR(m); 460 NET_EPOCH_ASSERT(); 461 462 if (m->m_flags & M_FASTFWD_OURS) { 463 m->m_flags &= ~M_FASTFWD_OURS; 464 /* Set up some basics that will be used later. */ 465 ip = mtod(m, struct ip *); 466 hlen = ip->ip_hl << 2; 467 ip_len = ntohs(ip->ip_len); 468 goto ours; 469 } 470 471 IPSTAT_INC(ips_total); 472 473 if (m->m_pkthdr.len < sizeof(struct ip)) 474 goto tooshort; 475 476 if (m->m_len < sizeof (struct ip) && 477 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 478 IPSTAT_INC(ips_toosmall); 479 return; 480 } 481 ip = mtod(m, struct ip *); 482 483 if (ip->ip_v != IPVERSION) { 484 IPSTAT_INC(ips_badvers); 485 goto bad; 486 } 487 488 hlen = ip->ip_hl << 2; 489 if (hlen < sizeof(struct ip)) { /* minimum header length */ 490 IPSTAT_INC(ips_badhlen); 491 goto bad; 492 } 493 if (hlen > m->m_len) { 494 if ((m = m_pullup(m, hlen)) == NULL) { 495 IPSTAT_INC(ips_badhlen); 496 return; 497 } 498 ip = mtod(m, struct ip *); 499 } 500 501 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 502 503 /* IN_LOOPBACK must not appear on the wire - RFC1122 */ 504 ifp = m->m_pkthdr.rcvif; 505 if (IN_LOOPBACK(ntohl(ip->ip_dst.s_addr)) || 506 IN_LOOPBACK(ntohl(ip->ip_src.s_addr))) { 507 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 508 IPSTAT_INC(ips_badaddr); 509 goto bad; 510 } 511 } 512 513 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 514 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 515 } else { 516 if (hlen == sizeof(struct ip)) { 517 sum = in_cksum_hdr(ip); 518 } else { 519 sum = in_cksum(m, hlen); 520 } 521 } 522 if (sum) { 523 IPSTAT_INC(ips_badsum); 524 goto bad; 525 } 526 527 #ifdef ALTQ 528 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 529 /* packet is dropped by traffic conditioner */ 530 return; 531 #endif 532 533 ip_len = ntohs(ip->ip_len); 534 if (ip_len < hlen) { 535 IPSTAT_INC(ips_badlen); 536 goto bad; 537 } 538 539 /* 540 * Check that the amount of data in the buffers 541 * is as at least much as the IP header would have us expect. 542 * Trim mbufs if longer than we expect. 543 * Drop packet if shorter than we expect. 544 */ 545 if (m->m_pkthdr.len < ip_len) { 546 tooshort: 547 IPSTAT_INC(ips_tooshort); 548 goto bad; 549 } 550 if (m->m_pkthdr.len > ip_len) { 551 if (m->m_len == m->m_pkthdr.len) { 552 m->m_len = ip_len; 553 m->m_pkthdr.len = ip_len; 554 } else 555 m_adj(m, ip_len - m->m_pkthdr.len); 556 } 557 558 /* 559 * Try to forward the packet, but if we fail continue. 560 * ip_tryforward() does not generate redirects, so fall 561 * through to normal processing if redirects are required. 562 * ip_tryforward() does inbound and outbound packet firewall 563 * processing. If firewall has decided that destination becomes 564 * our local address, it sets M_FASTFWD_OURS flag. In this 565 * case skip another inbound firewall processing and update 566 * ip pointer. 567 */ 568 if (V_ipforwarding != 0 && V_ipsendredirects == 0 569 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 570 && (!IPSEC_ENABLED(ipv4) || 571 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0) 572 #endif 573 ) { 574 if ((m = ip_tryforward(m)) == NULL) 575 return; 576 if (m->m_flags & M_FASTFWD_OURS) { 577 m->m_flags &= ~M_FASTFWD_OURS; 578 ip = mtod(m, struct ip *); 579 goto ours; 580 } 581 } 582 583 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 584 /* 585 * Bypass packet filtering for packets previously handled by IPsec. 586 */ 587 if (IPSEC_ENABLED(ipv4) && 588 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0) 589 goto passin; 590 #endif 591 592 /* 593 * Run through list of hooks for input packets. 594 * 595 * NB: Beware of the destination address changing (e.g. 596 * by NAT rewriting). When this happens, tell 597 * ip_forward to do the right thing. 598 */ 599 600 /* Jump over all PFIL processing if hooks are not active. */ 601 if (!PFIL_HOOKED_IN(V_inet_pfil_head)) 602 goto passin; 603 604 odst = ip->ip_dst; 605 if (pfil_run_hooks(V_inet_pfil_head, &m, ifp, PFIL_IN, NULL) != 606 PFIL_PASS) 607 return; 608 if (m == NULL) /* consumed by filter */ 609 return; 610 611 ip = mtod(m, struct ip *); 612 dchg = (odst.s_addr != ip->ip_dst.s_addr); 613 ifp = m->m_pkthdr.rcvif; 614 615 if (m->m_flags & M_FASTFWD_OURS) { 616 m->m_flags &= ~M_FASTFWD_OURS; 617 goto ours; 618 } 619 if (m->m_flags & M_IP_NEXTHOP) { 620 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 621 /* 622 * Directly ship the packet on. This allows 623 * forwarding packets originally destined to us 624 * to some other directly connected host. 625 */ 626 ip_forward(m, 1); 627 return; 628 } 629 } 630 passin: 631 632 /* 633 * Process options and, if not destined for us, 634 * ship it on. ip_dooptions returns 1 when an 635 * error was detected (causing an icmp message 636 * to be sent and the original packet to be freed). 637 */ 638 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 639 return; 640 641 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 642 * matter if it is destined to another node, or whether it is 643 * a multicast one, RSVP wants it! and prevents it from being forwarded 644 * anywhere else. Also checks if the rsvp daemon is running before 645 * grabbing the packet. 646 */ 647 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 648 goto ours; 649 650 /* 651 * Check our list of addresses, to see if the packet is for us. 652 * If we don't have any addresses, assume any unicast packet 653 * we receive might be for us (and let the upper layers deal 654 * with it). 655 */ 656 if (CK_STAILQ_EMPTY(&V_in_ifaddrhead) && 657 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 658 goto ours; 659 660 /* 661 * Enable a consistency check between the destination address 662 * and the arrival interface for a unicast packet (the RFC 1122 663 * strong ES model) if IP forwarding is disabled and the packet 664 * is not locally generated and the packet is not subject to 665 * 'ipfw fwd'. 666 * 667 * XXX - Checking also should be disabled if the destination 668 * address is ipnat'ed to a different interface. 669 * 670 * XXX - Checking is incompatible with IP aliases added 671 * to the loopback interface instead of the interface where 672 * the packets are received. 673 * 674 * XXX - This is the case for carp vhost IPs as well so we 675 * insert a workaround. If the packet got here, we already 676 * checked with carp_iamatch() and carp_forus(). 677 */ 678 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 679 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 680 ifp->if_carp == NULL && (dchg == 0); 681 682 /* 683 * Check for exact addresses in the hash bucket. 684 */ 685 IN_IFADDR_RLOCK(&in_ifa_tracker); 686 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 687 /* 688 * If the address matches, verify that the packet 689 * arrived via the correct interface if checking is 690 * enabled. 691 */ 692 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 693 (!checkif || ia->ia_ifp == ifp)) { 694 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 695 counter_u64_add(ia->ia_ifa.ifa_ibytes, 696 m->m_pkthdr.len); 697 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 698 goto ours; 699 } 700 } 701 IN_IFADDR_RUNLOCK(&in_ifa_tracker); 702 703 /* 704 * Check for broadcast addresses. 705 * 706 * Only accept broadcast packets that arrive via the matching 707 * interface. Reception of forwarded directed broadcasts would 708 * be handled via ip_forward() and ether_output() with the loopback 709 * into the stack for SIMPLEX interfaces handled by ether_output(). 710 */ 711 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 712 CK_STAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 713 if (ifa->ifa_addr->sa_family != AF_INET) 714 continue; 715 ia = ifatoia(ifa); 716 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 717 ip->ip_dst.s_addr) { 718 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 719 counter_u64_add(ia->ia_ifa.ifa_ibytes, 720 m->m_pkthdr.len); 721 goto ours; 722 } 723 #ifdef BOOTP_COMPAT 724 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 725 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 726 counter_u64_add(ia->ia_ifa.ifa_ibytes, 727 m->m_pkthdr.len); 728 goto ours; 729 } 730 #endif 731 } 732 ia = NULL; 733 } 734 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 735 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 736 IPSTAT_INC(ips_cantforward); 737 m_freem(m); 738 return; 739 } 740 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 741 if (V_ip_mrouter) { 742 /* 743 * If we are acting as a multicast router, all 744 * incoming multicast packets are passed to the 745 * kernel-level multicast forwarding function. 746 * The packet is returned (relatively) intact; if 747 * ip_mforward() returns a non-zero value, the packet 748 * must be discarded, else it may be accepted below. 749 */ 750 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 751 IPSTAT_INC(ips_cantforward); 752 m_freem(m); 753 return; 754 } 755 756 /* 757 * The process-level routing daemon needs to receive 758 * all multicast IGMP packets, whether or not this 759 * host belongs to their destination groups. 760 */ 761 if (ip->ip_p == IPPROTO_IGMP) 762 goto ours; 763 IPSTAT_INC(ips_forward); 764 } 765 /* 766 * Assume the packet is for us, to avoid prematurely taking 767 * a lock on the in_multi hash. Protocols must perform 768 * their own filtering and update statistics accordingly. 769 */ 770 goto ours; 771 } 772 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 773 goto ours; 774 if (ip->ip_dst.s_addr == INADDR_ANY) 775 goto ours; 776 777 /* 778 * Not for us; forward if possible and desirable. 779 */ 780 if (V_ipforwarding == 0) { 781 IPSTAT_INC(ips_cantforward); 782 m_freem(m); 783 } else { 784 ip_forward(m, dchg); 785 } 786 return; 787 788 ours: 789 #ifdef IPSTEALTH 790 /* 791 * IPSTEALTH: Process non-routing options only 792 * if the packet is destined for us. 793 */ 794 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) 795 return; 796 #endif /* IPSTEALTH */ 797 798 /* 799 * Attempt reassembly; if it succeeds, proceed. 800 * ip_reass() will return a different mbuf. 801 */ 802 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 803 /* XXXGL: shouldn't we save & set m_flags? */ 804 m = ip_reass(m); 805 if (m == NULL) 806 return; 807 ip = mtod(m, struct ip *); 808 /* Get the header length of the reassembled packet */ 809 hlen = ip->ip_hl << 2; 810 } 811 812 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 813 if (IPSEC_ENABLED(ipv4)) { 814 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0) 815 return; 816 } 817 #endif /* IPSEC */ 818 819 /* 820 * Switch out to protocol's input routine. 821 */ 822 IPSTAT_INC(ips_delivered); 823 824 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 825 return; 826 bad: 827 m_freem(m); 828 } 829 830 /* 831 * IP timer processing; 832 * if a timer expires on a reassembly 833 * queue, discard it. 834 */ 835 void 836 ip_slowtimo(void) 837 { 838 VNET_ITERATOR_DECL(vnet_iter); 839 840 VNET_LIST_RLOCK_NOSLEEP(); 841 VNET_FOREACH(vnet_iter) { 842 CURVNET_SET(vnet_iter); 843 ipreass_slowtimo(); 844 CURVNET_RESTORE(); 845 } 846 VNET_LIST_RUNLOCK_NOSLEEP(); 847 } 848 849 void 850 ip_drain(void) 851 { 852 VNET_ITERATOR_DECL(vnet_iter); 853 854 VNET_LIST_RLOCK_NOSLEEP(); 855 VNET_FOREACH(vnet_iter) { 856 CURVNET_SET(vnet_iter); 857 ipreass_drain(); 858 CURVNET_RESTORE(); 859 } 860 VNET_LIST_RUNLOCK_NOSLEEP(); 861 } 862 863 /* 864 * The protocol to be inserted into ip_protox[] must be already registered 865 * in inetsw[], either statically or through pf_proto_register(). 866 */ 867 int 868 ipproto_register(short ipproto) 869 { 870 struct protosw *pr; 871 872 /* Sanity checks. */ 873 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 874 return (EPROTONOSUPPORT); 875 876 /* 877 * The protocol slot must not be occupied by another protocol 878 * already. An index pointing to IPPROTO_RAW is unused. 879 */ 880 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 881 if (pr == NULL) 882 return (EPFNOSUPPORT); 883 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 884 return (EEXIST); 885 886 /* Find the protocol position in inetsw[] and set the index. */ 887 for (pr = inetdomain.dom_protosw; 888 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 889 if (pr->pr_domain->dom_family == PF_INET && 890 pr->pr_protocol && pr->pr_protocol == ipproto) { 891 ip_protox[pr->pr_protocol] = pr - inetsw; 892 return (0); 893 } 894 } 895 return (EPROTONOSUPPORT); 896 } 897 898 int 899 ipproto_unregister(short ipproto) 900 { 901 struct protosw *pr; 902 903 /* Sanity checks. */ 904 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 905 return (EPROTONOSUPPORT); 906 907 /* Check if the protocol was indeed registered. */ 908 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 909 if (pr == NULL) 910 return (EPFNOSUPPORT); 911 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 912 return (ENOENT); 913 914 /* Reset the protocol slot to IPPROTO_RAW. */ 915 ip_protox[ipproto] = pr - inetsw; 916 return (0); 917 } 918 919 u_char inetctlerrmap[PRC_NCMDS] = { 920 0, 0, 0, 0, 921 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 922 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 923 EMSGSIZE, EHOSTUNREACH, 0, 0, 924 0, 0, EHOSTUNREACH, 0, 925 ENOPROTOOPT, ECONNREFUSED 926 }; 927 928 /* 929 * Forward a packet. If some error occurs return the sender 930 * an icmp packet. Note we can't always generate a meaningful 931 * icmp message because icmp doesn't have a large enough repertoire 932 * of codes and types. 933 * 934 * If not forwarding, just drop the packet. This could be confusing 935 * if ipforwarding was zero but some routing protocol was advancing 936 * us as a gateway to somewhere. However, we must let the routing 937 * protocol deal with that. 938 * 939 * The srcrt parameter indicates whether the packet is being forwarded 940 * via a source route. 941 */ 942 void 943 ip_forward(struct mbuf *m, int srcrt) 944 { 945 struct ip *ip = mtod(m, struct ip *); 946 struct in_ifaddr *ia; 947 struct mbuf *mcopy; 948 struct sockaddr_in *sin; 949 struct in_addr dest; 950 struct route ro; 951 int error, type = 0, code = 0, mtu = 0; 952 953 NET_EPOCH_ASSERT(); 954 955 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 956 IPSTAT_INC(ips_cantforward); 957 m_freem(m); 958 return; 959 } 960 if ( 961 #ifdef IPSTEALTH 962 V_ipstealth == 0 && 963 #endif 964 ip->ip_ttl <= IPTTLDEC) { 965 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); 966 return; 967 } 968 969 bzero(&ro, sizeof(ro)); 970 sin = (struct sockaddr_in *)&ro.ro_dst; 971 sin->sin_family = AF_INET; 972 sin->sin_len = sizeof(*sin); 973 sin->sin_addr = ip->ip_dst; 974 #ifdef RADIX_MPATH 975 rtalloc_mpath_fib(&ro, 976 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 977 M_GETFIB(m)); 978 #else 979 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 980 #endif 981 if (ro.ro_rt != NULL) { 982 ia = ifatoia(ro.ro_rt->rt_ifa); 983 } else 984 ia = NULL; 985 /* 986 * Save the IP header and at most 8 bytes of the payload, 987 * in case we need to generate an ICMP message to the src. 988 * 989 * XXX this can be optimized a lot by saving the data in a local 990 * buffer on the stack (72 bytes at most), and only allocating the 991 * mbuf if really necessary. The vast majority of the packets 992 * are forwarded without having to send an ICMP back (either 993 * because unnecessary, or because rate limited), so we are 994 * really we are wasting a lot of work here. 995 * 996 * We don't use m_copym() because it might return a reference 997 * to a shared cluster. Both this function and ip_output() 998 * assume exclusive access to the IP header in `m', so any 999 * data in a cluster may change before we reach icmp_error(). 1000 */ 1001 mcopy = m_gethdr(M_NOWAIT, m->m_type); 1002 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { 1003 /* 1004 * It's probably ok if the pkthdr dup fails (because 1005 * the deep copy of the tag chain failed), but for now 1006 * be conservative and just discard the copy since 1007 * code below may some day want the tags. 1008 */ 1009 m_free(mcopy); 1010 mcopy = NULL; 1011 } 1012 if (mcopy != NULL) { 1013 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 1014 mcopy->m_pkthdr.len = mcopy->m_len; 1015 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1016 } 1017 #ifdef IPSTEALTH 1018 if (V_ipstealth == 0) 1019 #endif 1020 ip->ip_ttl -= IPTTLDEC; 1021 #if defined(IPSEC) || defined(IPSEC_SUPPORT) 1022 if (IPSEC_ENABLED(ipv4)) { 1023 if ((error = IPSEC_FORWARD(ipv4, m)) != 0) { 1024 /* mbuf consumed by IPsec */ 1025 m_freem(mcopy); 1026 if (error != EINPROGRESS) 1027 IPSTAT_INC(ips_cantforward); 1028 return; 1029 } 1030 /* No IPsec processing required */ 1031 } 1032 #endif /* IPSEC */ 1033 /* 1034 * If forwarding packet using same interface that it came in on, 1035 * perhaps should send a redirect to sender to shortcut a hop. 1036 * Only send redirect if source is sending directly to us, 1037 * and if packet was not source routed (or has any options). 1038 * Also, don't send redirect if forwarding using a default route 1039 * or a route modified by a redirect. 1040 */ 1041 dest.s_addr = 0; 1042 if (!srcrt && V_ipsendredirects && 1043 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1044 struct rtentry *rt; 1045 1046 rt = ro.ro_rt; 1047 1048 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1049 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1050 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1051 u_long src = ntohl(ip->ip_src.s_addr); 1052 1053 if (RTA(rt) && 1054 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1055 if (rt->rt_flags & RTF_GATEWAY) 1056 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1057 else 1058 dest.s_addr = ip->ip_dst.s_addr; 1059 /* Router requirements says to only send host redirects */ 1060 type = ICMP_REDIRECT; 1061 code = ICMP_REDIRECT_HOST; 1062 } 1063 } 1064 } 1065 1066 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1067 1068 if (error == EMSGSIZE && ro.ro_rt) 1069 mtu = ro.ro_rt->rt_mtu; 1070 RO_RTFREE(&ro); 1071 1072 if (error) 1073 IPSTAT_INC(ips_cantforward); 1074 else { 1075 IPSTAT_INC(ips_forward); 1076 if (type) 1077 IPSTAT_INC(ips_redirectsent); 1078 else { 1079 if (mcopy) 1080 m_freem(mcopy); 1081 return; 1082 } 1083 } 1084 if (mcopy == NULL) 1085 return; 1086 1087 1088 switch (error) { 1089 1090 case 0: /* forwarded, but need redirect */ 1091 /* type, code set above */ 1092 break; 1093 1094 case ENETUNREACH: 1095 case EHOSTUNREACH: 1096 case ENETDOWN: 1097 case EHOSTDOWN: 1098 default: 1099 type = ICMP_UNREACH; 1100 code = ICMP_UNREACH_HOST; 1101 break; 1102 1103 case EMSGSIZE: 1104 type = ICMP_UNREACH; 1105 code = ICMP_UNREACH_NEEDFRAG; 1106 /* 1107 * If the MTU was set before make sure we are below the 1108 * interface MTU. 1109 * If the MTU wasn't set before use the interface mtu or 1110 * fall back to the next smaller mtu step compared to the 1111 * current packet size. 1112 */ 1113 if (mtu != 0) { 1114 if (ia != NULL) 1115 mtu = min(mtu, ia->ia_ifp->if_mtu); 1116 } else { 1117 if (ia != NULL) 1118 mtu = ia->ia_ifp->if_mtu; 1119 else 1120 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1121 } 1122 IPSTAT_INC(ips_cantfrag); 1123 break; 1124 1125 case ENOBUFS: 1126 case EACCES: /* ipfw denied packet */ 1127 m_freem(mcopy); 1128 return; 1129 } 1130 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1131 } 1132 1133 #define CHECK_SO_CT(sp, ct) \ 1134 (((sp->so_options & SO_TIMESTAMP) && (sp->so_ts_clock == ct)) ? 1 : 0) 1135 1136 void 1137 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1138 struct mbuf *m) 1139 { 1140 bool stamped; 1141 1142 stamped = false; 1143 if ((inp->inp_socket->so_options & SO_BINTIME) || 1144 CHECK_SO_CT(inp->inp_socket, SO_TS_BINTIME)) { 1145 struct bintime boottimebin, bt; 1146 struct timespec ts1; 1147 1148 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1149 M_TSTMP)) { 1150 mbuf_tstmp2timespec(m, &ts1); 1151 timespec2bintime(&ts1, &bt); 1152 getboottimebin(&boottimebin); 1153 bintime_add(&bt, &boottimebin); 1154 } else { 1155 bintime(&bt); 1156 } 1157 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), 1158 SCM_BINTIME, SOL_SOCKET); 1159 if (*mp != NULL) { 1160 mp = &(*mp)->m_next; 1161 stamped = true; 1162 } 1163 } 1164 if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME_MICRO)) { 1165 struct bintime boottimebin, bt1; 1166 struct timespec ts1;; 1167 struct timeval tv; 1168 1169 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1170 M_TSTMP)) { 1171 mbuf_tstmp2timespec(m, &ts1); 1172 timespec2bintime(&ts1, &bt1); 1173 getboottimebin(&boottimebin); 1174 bintime_add(&bt1, &boottimebin); 1175 bintime2timeval(&bt1, &tv); 1176 } else { 1177 microtime(&tv); 1178 } 1179 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), 1180 SCM_TIMESTAMP, SOL_SOCKET); 1181 if (*mp != NULL) { 1182 mp = &(*mp)->m_next; 1183 stamped = true; 1184 } 1185 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_REALTIME)) { 1186 struct bintime boottimebin; 1187 struct timespec ts, ts1; 1188 1189 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1190 M_TSTMP)) { 1191 mbuf_tstmp2timespec(m, &ts); 1192 getboottimebin(&boottimebin); 1193 bintime2timespec(&boottimebin, &ts1); 1194 timespecadd(&ts, &ts1, &ts); 1195 } else { 1196 nanotime(&ts); 1197 } 1198 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts), 1199 SCM_REALTIME, SOL_SOCKET); 1200 if (*mp != NULL) { 1201 mp = &(*mp)->m_next; 1202 stamped = true; 1203 } 1204 } else if (CHECK_SO_CT(inp->inp_socket, SO_TS_MONOTONIC)) { 1205 struct timespec ts; 1206 1207 if ((m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1208 M_TSTMP)) 1209 mbuf_tstmp2timespec(m, &ts); 1210 else 1211 nanouptime(&ts); 1212 *mp = sbcreatecontrol((caddr_t)&ts, sizeof(ts), 1213 SCM_MONOTONIC, SOL_SOCKET); 1214 if (*mp != NULL) { 1215 mp = &(*mp)->m_next; 1216 stamped = true; 1217 } 1218 } 1219 if (stamped && (m->m_flags & (M_PKTHDR | M_TSTMP)) == (M_PKTHDR | 1220 M_TSTMP)) { 1221 struct sock_timestamp_info sti; 1222 1223 bzero(&sti, sizeof(sti)); 1224 sti.st_info_flags = ST_INFO_HW; 1225 if ((m->m_flags & M_TSTMP_HPREC) != 0) 1226 sti.st_info_flags |= ST_INFO_HW_HPREC; 1227 *mp = sbcreatecontrol((caddr_t)&sti, sizeof(sti), SCM_TIME_INFO, 1228 SOL_SOCKET); 1229 if (*mp != NULL) 1230 mp = &(*mp)->m_next; 1231 } 1232 if (inp->inp_flags & INP_RECVDSTADDR) { 1233 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, 1234 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1235 if (*mp) 1236 mp = &(*mp)->m_next; 1237 } 1238 if (inp->inp_flags & INP_RECVTTL) { 1239 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, 1240 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1241 if (*mp) 1242 mp = &(*mp)->m_next; 1243 } 1244 #ifdef notyet 1245 /* XXX 1246 * Moving these out of udp_input() made them even more broken 1247 * than they already were. 1248 */ 1249 /* options were tossed already */ 1250 if (inp->inp_flags & INP_RECVOPTS) { 1251 *mp = sbcreatecontrol((caddr_t)opts_deleted_above, 1252 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1253 if (*mp) 1254 mp = &(*mp)->m_next; 1255 } 1256 /* ip_srcroute doesn't do what we want here, need to fix */ 1257 if (inp->inp_flags & INP_RECVRETOPTS) { 1258 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), 1259 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1260 if (*mp) 1261 mp = &(*mp)->m_next; 1262 } 1263 #endif 1264 if (inp->inp_flags & INP_RECVIF) { 1265 struct ifnet *ifp; 1266 struct sdlbuf { 1267 struct sockaddr_dl sdl; 1268 u_char pad[32]; 1269 } sdlbuf; 1270 struct sockaddr_dl *sdp; 1271 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1272 1273 if ((ifp = m->m_pkthdr.rcvif) && 1274 ifp->if_index && ifp->if_index <= V_if_index) { 1275 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1276 /* 1277 * Change our mind and don't try copy. 1278 */ 1279 if (sdp->sdl_family != AF_LINK || 1280 sdp->sdl_len > sizeof(sdlbuf)) { 1281 goto makedummy; 1282 } 1283 bcopy(sdp, sdl2, sdp->sdl_len); 1284 } else { 1285 makedummy: 1286 sdl2->sdl_len = 1287 offsetof(struct sockaddr_dl, sdl_data[0]); 1288 sdl2->sdl_family = AF_LINK; 1289 sdl2->sdl_index = 0; 1290 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1291 } 1292 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, 1293 IP_RECVIF, IPPROTO_IP); 1294 if (*mp) 1295 mp = &(*mp)->m_next; 1296 } 1297 if (inp->inp_flags & INP_RECVTOS) { 1298 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, 1299 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1300 if (*mp) 1301 mp = &(*mp)->m_next; 1302 } 1303 1304 if (inp->inp_flags2 & INP_RECVFLOWID) { 1305 uint32_t flowid, flow_type; 1306 1307 flowid = m->m_pkthdr.flowid; 1308 flow_type = M_HASHTYPE_GET(m); 1309 1310 /* 1311 * XXX should handle the failure of one or the 1312 * other - don't populate both? 1313 */ 1314 *mp = sbcreatecontrol((caddr_t) &flowid, 1315 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP); 1316 if (*mp) 1317 mp = &(*mp)->m_next; 1318 *mp = sbcreatecontrol((caddr_t) &flow_type, 1319 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP); 1320 if (*mp) 1321 mp = &(*mp)->m_next; 1322 } 1323 1324 #ifdef RSS 1325 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1326 uint32_t flowid, flow_type; 1327 uint32_t rss_bucketid; 1328 1329 flowid = m->m_pkthdr.flowid; 1330 flow_type = M_HASHTYPE_GET(m); 1331 1332 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1333 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1334 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP); 1335 if (*mp) 1336 mp = &(*mp)->m_next; 1337 } 1338 } 1339 #endif 1340 } 1341 1342 /* 1343 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1344 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1345 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1346 * compiled. 1347 */ 1348 VNET_DEFINE_STATIC(int, ip_rsvp_on); 1349 VNET_DEFINE(struct socket *, ip_rsvpd); 1350 1351 #define V_ip_rsvp_on VNET(ip_rsvp_on) 1352 1353 int 1354 ip_rsvp_init(struct socket *so) 1355 { 1356 1357 if (so->so_type != SOCK_RAW || 1358 so->so_proto->pr_protocol != IPPROTO_RSVP) 1359 return EOPNOTSUPP; 1360 1361 if (V_ip_rsvpd != NULL) 1362 return EADDRINUSE; 1363 1364 V_ip_rsvpd = so; 1365 /* 1366 * This may seem silly, but we need to be sure we don't over-increment 1367 * the RSVP counter, in case something slips up. 1368 */ 1369 if (!V_ip_rsvp_on) { 1370 V_ip_rsvp_on = 1; 1371 V_rsvp_on++; 1372 } 1373 1374 return 0; 1375 } 1376 1377 int 1378 ip_rsvp_done(void) 1379 { 1380 1381 V_ip_rsvpd = NULL; 1382 /* 1383 * This may seem silly, but we need to be sure we don't over-decrement 1384 * the RSVP counter, in case something slips up. 1385 */ 1386 if (V_ip_rsvp_on) { 1387 V_ip_rsvp_on = 0; 1388 V_rsvp_on--; 1389 } 1390 return 0; 1391 } 1392 1393 int 1394 rsvp_input(struct mbuf **mp, int *offp, int proto) 1395 { 1396 struct mbuf *m; 1397 1398 m = *mp; 1399 *mp = NULL; 1400 1401 if (rsvp_input_p) { /* call the real one if loaded */ 1402 *mp = m; 1403 rsvp_input_p(mp, offp, proto); 1404 return (IPPROTO_DONE); 1405 } 1406 1407 /* Can still get packets with rsvp_on = 0 if there is a local member 1408 * of the group to which the RSVP packet is addressed. But in this 1409 * case we want to throw the packet away. 1410 */ 1411 1412 if (!V_rsvp_on) { 1413 m_freem(m); 1414 return (IPPROTO_DONE); 1415 } 1416 1417 if (V_ip_rsvpd != NULL) { 1418 *mp = m; 1419 rip_input(mp, offp, proto); 1420 return (IPPROTO_DONE); 1421 } 1422 /* Drop the packet */ 1423 m_freem(m); 1424 return (IPPROTO_DONE); 1425 } 1426