ip_input.c revision 338343
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: stable/11/sys/netinet/ip_input.c 338343 2018-08-28 07:24:09Z ae $"); 34 35#include "opt_bootp.h" 36#include "opt_ipstealth.h" 37#include "opt_ipsec.h" 38#include "opt_route.h" 39#include "opt_rss.h" 40 41#include <sys/param.h> 42#include <sys/systm.h> 43#include <sys/hhook.h> 44#include <sys/mbuf.h> 45#include <sys/malloc.h> 46#include <sys/domain.h> 47#include <sys/protosw.h> 48#include <sys/socket.h> 49#include <sys/time.h> 50#include <sys/kernel.h> 51#include <sys/lock.h> 52#include <sys/rmlock.h> 53#include <sys/rwlock.h> 54#include <sys/sdt.h> 55#include <sys/syslog.h> 56#include <sys/sysctl.h> 57 58#include <net/pfil.h> 59#include <net/if.h> 60#include <net/if_types.h> 61#include <net/if_var.h> 62#include <net/if_dl.h> 63#include <net/route.h> 64#include <net/netisr.h> 65#include <net/rss_config.h> 66#include <net/vnet.h> 67 68#include <netinet/in.h> 69#include <netinet/in_kdtrace.h> 70#include <netinet/in_systm.h> 71#include <netinet/in_var.h> 72#include <netinet/ip.h> 73#include <netinet/in_pcb.h> 74#include <netinet/ip_var.h> 75#include <netinet/ip_fw.h> 76#include <netinet/ip_icmp.h> 77#include <netinet/ip_options.h> 78#include <machine/in_cksum.h> 79#include <netinet/ip_carp.h> 80#include <netinet/in_rss.h> 81 82#include <netipsec/ipsec_support.h> 83 84#include <sys/socketvar.h> 85 86#include <security/mac/mac_framework.h> 87 88#ifdef CTASSERT 89CTASSERT(sizeof(struct ip) == 20); 90#endif 91 92/* IP reassembly functions are defined in ip_reass.c. */ 93extern void ipreass_init(void); 94extern void ipreass_drain(void); 95extern void ipreass_slowtimo(void); 96#ifdef VIMAGE 97extern void ipreass_destroy(void); 98#endif 99 100struct rmlock in_ifaddr_lock; 101RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 102 103VNET_DEFINE(int, rsvp_on); 104 105VNET_DEFINE(int, ipforwarding); 106SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, 107 &VNET_NAME(ipforwarding), 0, 108 "Enable IP forwarding between interfaces"); 109 110static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 111#define V_ipsendredirects VNET(ipsendredirects) 112SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, 113 &VNET_NAME(ipsendredirects), 0, 114 "Enable sending IP redirects"); 115 116/* 117 * XXX - Setting ip_checkinterface mostly implements the receive side of 118 * the Strong ES model described in RFC 1122, but since the routing table 119 * and transmit implementation do not implement the Strong ES model, 120 * setting this to 1 results in an odd hybrid. 121 * 122 * XXX - ip_checkinterface currently must be disabled if you use ipnat 123 * to translate the destination address to another local interface. 124 * 125 * XXX - ip_checkinterface must be disabled if you add IP aliases 126 * to the loopback interface instead of the interface where the 127 * packets for those addresses are received. 128 */ 129static VNET_DEFINE(int, ip_checkinterface); 130#define V_ip_checkinterface VNET(ip_checkinterface) 131SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW, 132 &VNET_NAME(ip_checkinterface), 0, 133 "Verify packet arrives on correct interface"); 134 135VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 136 137static struct netisr_handler ip_nh = { 138 .nh_name = "ip", 139 .nh_handler = ip_input, 140 .nh_proto = NETISR_IP, 141#ifdef RSS 142 .nh_m2cpuid = rss_soft_m2cpuid_v4, 143 .nh_policy = NETISR_POLICY_CPU, 144 .nh_dispatch = NETISR_DISPATCH_HYBRID, 145#else 146 .nh_policy = NETISR_POLICY_FLOW, 147#endif 148}; 149 150#ifdef RSS 151/* 152 * Directly dispatched frames are currently assumed 153 * to have a flowid already calculated. 154 * 155 * It should likely have something that assert it 156 * actually has valid flow details. 157 */ 158static struct netisr_handler ip_direct_nh = { 159 .nh_name = "ip_direct", 160 .nh_handler = ip_direct_input, 161 .nh_proto = NETISR_IP_DIRECT, 162 .nh_m2cpuid = rss_soft_m2cpuid_v4, 163 .nh_policy = NETISR_POLICY_CPU, 164 .nh_dispatch = NETISR_DISPATCH_HYBRID, 165}; 166#endif 167 168extern struct domain inetdomain; 169extern struct protosw inetsw[]; 170u_char ip_protox[IPPROTO_MAX]; 171VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 172VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 173VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 174 175#ifdef IPCTL_DEFMTU 176SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 177 &ip_mtu, 0, "Default MTU"); 178#endif 179 180#ifdef IPSTEALTH 181VNET_DEFINE(int, ipstealth); 182SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, 183 &VNET_NAME(ipstealth), 0, 184 "IP stealth mode, no TTL decrementation on forwarding"); 185#endif 186 187/* 188 * IP statistics are stored in the "array" of counter(9)s. 189 */ 190VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 191VNET_PCPUSTAT_SYSINIT(ipstat); 192SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 193 "IP statistics (struct ipstat, netinet/ip_var.h)"); 194 195#ifdef VIMAGE 196VNET_PCPUSTAT_SYSUNINIT(ipstat); 197#endif /* VIMAGE */ 198 199/* 200 * Kernel module interface for updating ipstat. The argument is an index 201 * into ipstat treated as an array. 202 */ 203void 204kmod_ipstat_inc(int statnum) 205{ 206 207 counter_u64_add(VNET(ipstat)[statnum], 1); 208} 209 210void 211kmod_ipstat_dec(int statnum) 212{ 213 214 counter_u64_add(VNET(ipstat)[statnum], -1); 215} 216 217static int 218sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 219{ 220 int error, qlimit; 221 222 netisr_getqlimit(&ip_nh, &qlimit); 223 error = sysctl_handle_int(oidp, &qlimit, 0, req); 224 if (error || !req->newptr) 225 return (error); 226 if (qlimit < 1) 227 return (EINVAL); 228 return (netisr_setqlimit(&ip_nh, qlimit)); 229} 230SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 231 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 232 "Maximum size of the IP input queue"); 233 234static int 235sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 236{ 237 u_int64_t qdrops_long; 238 int error, qdrops; 239 240 netisr_getqdrops(&ip_nh, &qdrops_long); 241 qdrops = qdrops_long; 242 error = sysctl_handle_int(oidp, &qdrops, 0, req); 243 if (error || !req->newptr) 244 return (error); 245 if (qdrops != 0) 246 return (EINVAL); 247 netisr_clearqdrops(&ip_nh); 248 return (0); 249} 250 251SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 252 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 253 "Number of packets dropped from the IP input queue"); 254 255#ifdef RSS 256static int 257sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 258{ 259 int error, qlimit; 260 261 netisr_getqlimit(&ip_direct_nh, &qlimit); 262 error = sysctl_handle_int(oidp, &qlimit, 0, req); 263 if (error || !req->newptr) 264 return (error); 265 if (qlimit < 1) 266 return (EINVAL); 267 return (netisr_setqlimit(&ip_direct_nh, qlimit)); 268} 269SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQMAXLEN, intr_direct_queue_maxlen, 270 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, 271 "I", "Maximum size of the IP direct input queue"); 272 273static int 274sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) 275{ 276 u_int64_t qdrops_long; 277 int error, qdrops; 278 279 netisr_getqdrops(&ip_direct_nh, &qdrops_long); 280 qdrops = qdrops_long; 281 error = sysctl_handle_int(oidp, &qdrops, 0, req); 282 if (error || !req->newptr) 283 return (error); 284 if (qdrops != 0) 285 return (EINVAL); 286 netisr_clearqdrops(&ip_direct_nh); 287 return (0); 288} 289 290SYSCTL_PROC(_net_inet_ip, IPCTL_INTRDQDROPS, intr_direct_queue_drops, 291 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", 292 "Number of packets dropped from the IP direct input queue"); 293#endif /* RSS */ 294 295/* 296 * IP initialization: fill in IP protocol switch table. 297 * All protocols not implemented in kernel go to raw IP protocol handler. 298 */ 299void 300ip_init(void) 301{ 302 struct protosw *pr; 303 int i; 304 305 TAILQ_INIT(&V_in_ifaddrhead); 306 V_in_ifaddrhashtbl = hashinit(INADDR_NHASH, M_IFADDR, &V_in_ifaddrhmask); 307 308 /* Initialize IP reassembly queue. */ 309 ipreass_init(); 310 311 /* Initialize packet filter hooks. */ 312 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 313 V_inet_pfil_hook.ph_af = AF_INET; 314 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 315 printf("%s: WARNING: unable to register pfil hook, " 316 "error %d\n", __func__, i); 317 318 if (hhook_head_register(HHOOK_TYPE_IPSEC_IN, AF_INET, 319 &V_ipsec_hhh_in[HHOOK_IPSEC_INET], 320 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 321 printf("%s: WARNING: unable to register input helper hook\n", 322 __func__); 323 if (hhook_head_register(HHOOK_TYPE_IPSEC_OUT, AF_INET, 324 &V_ipsec_hhh_out[HHOOK_IPSEC_INET], 325 HHOOK_WAITOK | HHOOK_HEADISINVNET) != 0) 326 printf("%s: WARNING: unable to register output helper hook\n", 327 __func__); 328 329 /* Skip initialization of globals for non-default instances. */ 330#ifdef VIMAGE 331 if (!IS_DEFAULT_VNET(curvnet)) { 332 netisr_register_vnet(&ip_nh); 333#ifdef RSS 334 netisr_register_vnet(&ip_direct_nh); 335#endif 336 return; 337 } 338#endif 339 340 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 341 if (pr == NULL) 342 panic("ip_init: PF_INET not found"); 343 344 /* Initialize the entire ip_protox[] array to IPPROTO_RAW. */ 345 for (i = 0; i < IPPROTO_MAX; i++) 346 ip_protox[i] = pr - inetsw; 347 /* 348 * Cycle through IP protocols and put them into the appropriate place 349 * in ip_protox[]. 350 */ 351 for (pr = inetdomain.dom_protosw; 352 pr < inetdomain.dom_protoswNPROTOSW; pr++) 353 if (pr->pr_domain->dom_family == PF_INET && 354 pr->pr_protocol && pr->pr_protocol != IPPROTO_RAW) { 355 /* Be careful to only index valid IP protocols. */ 356 if (pr->pr_protocol < IPPROTO_MAX) 357 ip_protox[pr->pr_protocol] = pr - inetsw; 358 } 359 360 netisr_register(&ip_nh); 361#ifdef RSS 362 netisr_register(&ip_direct_nh); 363#endif 364} 365 366#ifdef VIMAGE 367static void 368ip_destroy(void *unused __unused) 369{ 370 struct ifnet *ifp; 371 int error; 372 373#ifdef RSS 374 netisr_unregister_vnet(&ip_direct_nh); 375#endif 376 netisr_unregister_vnet(&ip_nh); 377 378 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0) 379 printf("%s: WARNING: unable to unregister pfil hook, " 380 "error %d\n", __func__, error); 381 382 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 TAILQ_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 411VNET_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 */ 421void 422ip_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 */ 446void 447ip_input(struct mbuf *m) 448{ 449 struct ip *ip = NULL; 450 struct in_ifaddr *ia = NULL; 451 struct ifaddr *ifa; 452 struct ifnet *ifp; 453 int checkif, hlen = 0; 454 uint16_t sum, ip_len; 455 int dchg = 0; /* dest changed after fw */ 456 struct in_addr odst; /* original dst address */ 457 458 M_ASSERTPKTHDR(m); 459 460 if (m->m_flags & M_FASTFWD_OURS) { 461 m->m_flags &= ~M_FASTFWD_OURS; 462 /* Set up some basics that will be used later. */ 463 ip = mtod(m, struct ip *); 464 hlen = ip->ip_hl << 2; 465 ip_len = ntohs(ip->ip_len); 466 goto ours; 467 } 468 469 IPSTAT_INC(ips_total); 470 471 if (m->m_pkthdr.len < sizeof(struct ip)) 472 goto tooshort; 473 474 if (m->m_len < sizeof (struct ip) && 475 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 476 IPSTAT_INC(ips_toosmall); 477 return; 478 } 479 ip = mtod(m, struct ip *); 480 481 if (ip->ip_v != IPVERSION) { 482 IPSTAT_INC(ips_badvers); 483 goto bad; 484 } 485 486 hlen = ip->ip_hl << 2; 487 if (hlen < sizeof(struct ip)) { /* minimum header length */ 488 IPSTAT_INC(ips_badhlen); 489 goto bad; 490 } 491 if (hlen > m->m_len) { 492 if ((m = m_pullup(m, hlen)) == NULL) { 493 IPSTAT_INC(ips_badhlen); 494 return; 495 } 496 ip = mtod(m, struct ip *); 497 } 498 499 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 500 501 /* 127/8 must not appear on wire - RFC1122 */ 502 ifp = m->m_pkthdr.rcvif; 503 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 504 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 505 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 506 IPSTAT_INC(ips_badaddr); 507 goto bad; 508 } 509 } 510 511 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 512 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 513 } else { 514 if (hlen == sizeof(struct ip)) { 515 sum = in_cksum_hdr(ip); 516 } else { 517 sum = in_cksum(m, hlen); 518 } 519 } 520 if (sum) { 521 IPSTAT_INC(ips_badsum); 522 goto bad; 523 } 524 525#ifdef ALTQ 526 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 527 /* packet is dropped by traffic conditioner */ 528 return; 529#endif 530 531 ip_len = ntohs(ip->ip_len); 532 if (ip_len < hlen) { 533 IPSTAT_INC(ips_badlen); 534 goto bad; 535 } 536 537 /* 538 * Check that the amount of data in the buffers 539 * is as at least much as the IP header would have us expect. 540 * Trim mbufs if longer than we expect. 541 * Drop packet if shorter than we expect. 542 */ 543 if (m->m_pkthdr.len < ip_len) { 544tooshort: 545 IPSTAT_INC(ips_tooshort); 546 goto bad; 547 } 548 if (m->m_pkthdr.len > ip_len) { 549 if (m->m_len == m->m_pkthdr.len) { 550 m->m_len = ip_len; 551 m->m_pkthdr.len = ip_len; 552 } else 553 m_adj(m, ip_len - m->m_pkthdr.len); 554 } 555 556 /* 557 * Try to forward the packet, but if we fail continue. 558 * ip_tryforward() does not generate redirects, so fall 559 * through to normal processing if redirects are required. 560 * ip_tryforward() does inbound and outbound packet firewall 561 * processing. If firewall has decided that destination becomes 562 * our local address, it sets M_FASTFWD_OURS flag. In this 563 * case skip another inbound firewall processing and update 564 * ip pointer. 565 */ 566 if (V_ipforwarding != 0 && V_ipsendredirects == 0 567#if defined(IPSEC) || defined(IPSEC_SUPPORT) 568 && (!IPSEC_ENABLED(ipv4) || 569 IPSEC_CAPS(ipv4, m, IPSEC_CAP_OPERABLE) == 0) 570#endif 571 ) { 572 if ((m = ip_tryforward(m)) == NULL) 573 return; 574 if (m->m_flags & M_FASTFWD_OURS) { 575 m->m_flags &= ~M_FASTFWD_OURS; 576 ip = mtod(m, struct ip *); 577 goto ours; 578 } 579 } 580 581#if defined(IPSEC) || defined(IPSEC_SUPPORT) 582 /* 583 * Bypass packet filtering for packets previously handled by IPsec. 584 */ 585 if (IPSEC_ENABLED(ipv4) && 586 IPSEC_CAPS(ipv4, m, IPSEC_CAP_BYPASS_FILTER) != 0) 587 goto passin; 588#endif 589 590 /* 591 * Run through list of hooks for input packets. 592 * 593 * NB: Beware of the destination address changing (e.g. 594 * by NAT rewriting). When this happens, tell 595 * ip_forward to do the right thing. 596 */ 597 598 /* Jump over all PFIL processing if hooks are not active. */ 599 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 600 goto passin; 601 602 odst = ip->ip_dst; 603 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, 0, NULL) != 0) 604 return; 605 if (m == NULL) /* consumed by filter */ 606 return; 607 608 ip = mtod(m, struct ip *); 609 dchg = (odst.s_addr != ip->ip_dst.s_addr); 610 ifp = m->m_pkthdr.rcvif; 611 612 if (m->m_flags & M_FASTFWD_OURS) { 613 m->m_flags &= ~M_FASTFWD_OURS; 614 goto ours; 615 } 616 if (m->m_flags & M_IP_NEXTHOP) { 617 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 618 /* 619 * Directly ship the packet on. This allows 620 * forwarding packets originally destined to us 621 * to some other directly connected host. 622 */ 623 ip_forward(m, 1); 624 return; 625 } 626 } 627passin: 628 629 /* 630 * Process options and, if not destined for us, 631 * ship it on. ip_dooptions returns 1 when an 632 * error was detected (causing an icmp message 633 * to be sent and the original packet to be freed). 634 */ 635 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 636 return; 637 638 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 639 * matter if it is destined to another node, or whether it is 640 * a multicast one, RSVP wants it! and prevents it from being forwarded 641 * anywhere else. Also checks if the rsvp daemon is running before 642 * grabbing the packet. 643 */ 644 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 645 goto ours; 646 647 /* 648 * Check our list of addresses, to see if the packet is for us. 649 * If we don't have any addresses, assume any unicast packet 650 * we receive might be for us (and let the upper layers deal 651 * with it). 652 */ 653 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 654 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 655 goto ours; 656 657 /* 658 * Enable a consistency check between the destination address 659 * and the arrival interface for a unicast packet (the RFC 1122 660 * strong ES model) if IP forwarding is disabled and the packet 661 * is not locally generated and the packet is not subject to 662 * 'ipfw fwd'. 663 * 664 * XXX - Checking also should be disabled if the destination 665 * address is ipnat'ed to a different interface. 666 * 667 * XXX - Checking is incompatible with IP aliases added 668 * to the loopback interface instead of the interface where 669 * the packets are received. 670 * 671 * XXX - This is the case for carp vhost IPs as well so we 672 * insert a workaround. If the packet got here, we already 673 * checked with carp_iamatch() and carp_forus(). 674 */ 675 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 676 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 677 ifp->if_carp == NULL && (dchg == 0); 678 679 /* 680 * Check for exact addresses in the hash bucket. 681 */ 682 /* IN_IFADDR_RLOCK(); */ 683 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 684 /* 685 * If the address matches, verify that the packet 686 * arrived via the correct interface if checking is 687 * enabled. 688 */ 689 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 690 (!checkif || ia->ia_ifp == ifp)) { 691 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 692 counter_u64_add(ia->ia_ifa.ifa_ibytes, 693 m->m_pkthdr.len); 694 /* IN_IFADDR_RUNLOCK(); */ 695 goto ours; 696 } 697 } 698 /* IN_IFADDR_RUNLOCK(); */ 699 700 /* 701 * Check for broadcast addresses. 702 * 703 * Only accept broadcast packets that arrive via the matching 704 * interface. Reception of forwarded directed broadcasts would 705 * be handled via ip_forward() and ether_output() with the loopback 706 * into the stack for SIMPLEX interfaces handled by ether_output(). 707 */ 708 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 709 IF_ADDR_RLOCK(ifp); 710 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 711 if (ifa->ifa_addr->sa_family != AF_INET) 712 continue; 713 ia = ifatoia(ifa); 714 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 715 ip->ip_dst.s_addr) { 716 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 717 counter_u64_add(ia->ia_ifa.ifa_ibytes, 718 m->m_pkthdr.len); 719 IF_ADDR_RUNLOCK(ifp); 720 goto ours; 721 } 722#ifdef BOOTP_COMPAT 723 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 724 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 725 counter_u64_add(ia->ia_ifa.ifa_ibytes, 726 m->m_pkthdr.len); 727 IF_ADDR_RUNLOCK(ifp); 728 goto ours; 729 } 730#endif 731 } 732 IF_ADDR_RUNLOCK(ifp); 733 ia = NULL; 734 } 735 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 736 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 737 IPSTAT_INC(ips_cantforward); 738 m_freem(m); 739 return; 740 } 741 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 742 if (V_ip_mrouter) { 743 /* 744 * If we are acting as a multicast router, all 745 * incoming multicast packets are passed to the 746 * kernel-level multicast forwarding function. 747 * The packet is returned (relatively) intact; if 748 * ip_mforward() returns a non-zero value, the packet 749 * must be discarded, else it may be accepted below. 750 */ 751 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 752 IPSTAT_INC(ips_cantforward); 753 m_freem(m); 754 return; 755 } 756 757 /* 758 * The process-level routing daemon needs to receive 759 * all multicast IGMP packets, whether or not this 760 * host belongs to their destination groups. 761 */ 762 if (ip->ip_p == IPPROTO_IGMP) 763 goto ours; 764 IPSTAT_INC(ips_forward); 765 } 766 /* 767 * Assume the packet is for us, to avoid prematurely taking 768 * a lock on the in_multi hash. Protocols must perform 769 * their own filtering and update statistics accordingly. 770 */ 771 goto ours; 772 } 773 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 774 goto ours; 775 if (ip->ip_dst.s_addr == INADDR_ANY) 776 goto ours; 777 778 /* 779 * Not for us; forward if possible and desirable. 780 */ 781 if (V_ipforwarding == 0) { 782 IPSTAT_INC(ips_cantforward); 783 m_freem(m); 784 } else { 785 ip_forward(m, dchg); 786 } 787 return; 788 789ours: 790#ifdef IPSTEALTH 791 /* 792 * IPSTEALTH: Process non-routing options only 793 * if the packet is destined for us. 794 */ 795 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) 796 return; 797#endif /* IPSTEALTH */ 798 799 /* 800 * Attempt reassembly; if it succeeds, proceed. 801 * ip_reass() will return a different mbuf. 802 */ 803 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 804 /* XXXGL: shouldn't we save & set m_flags? */ 805 m = ip_reass(m); 806 if (m == NULL) 807 return; 808 ip = mtod(m, struct ip *); 809 /* Get the header length of the reassembled packet */ 810 hlen = ip->ip_hl << 2; 811 } 812 813#if defined(IPSEC) || defined(IPSEC_SUPPORT) 814 if (IPSEC_ENABLED(ipv4)) { 815 if (IPSEC_INPUT(ipv4, m, hlen, ip->ip_p) != 0) 816 return; 817 } 818#endif /* IPSEC */ 819 820 /* 821 * Switch out to protocol's input routine. 822 */ 823 IPSTAT_INC(ips_delivered); 824 825 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 826 return; 827bad: 828 m_freem(m); 829} 830 831/* 832 * IP timer processing; 833 * if a timer expires on a reassembly 834 * queue, discard it. 835 */ 836void 837ip_slowtimo(void) 838{ 839 VNET_ITERATOR_DECL(vnet_iter); 840 841 VNET_LIST_RLOCK_NOSLEEP(); 842 VNET_FOREACH(vnet_iter) { 843 CURVNET_SET(vnet_iter); 844 ipreass_slowtimo(); 845 CURVNET_RESTORE(); 846 } 847 VNET_LIST_RUNLOCK_NOSLEEP(); 848} 849 850void 851ip_drain(void) 852{ 853 VNET_ITERATOR_DECL(vnet_iter); 854 855 VNET_LIST_RLOCK_NOSLEEP(); 856 VNET_FOREACH(vnet_iter) { 857 CURVNET_SET(vnet_iter); 858 ipreass_drain(); 859 CURVNET_RESTORE(); 860 } 861 VNET_LIST_RUNLOCK_NOSLEEP(); 862} 863 864/* 865 * The protocol to be inserted into ip_protox[] must be already registered 866 * in inetsw[], either statically or through pf_proto_register(). 867 */ 868int 869ipproto_register(short ipproto) 870{ 871 struct protosw *pr; 872 873 /* Sanity checks. */ 874 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 875 return (EPROTONOSUPPORT); 876 877 /* 878 * The protocol slot must not be occupied by another protocol 879 * already. An index pointing to IPPROTO_RAW is unused. 880 */ 881 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 882 if (pr == NULL) 883 return (EPFNOSUPPORT); 884 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 885 return (EEXIST); 886 887 /* Find the protocol position in inetsw[] and set the index. */ 888 for (pr = inetdomain.dom_protosw; 889 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 890 if (pr->pr_domain->dom_family == PF_INET && 891 pr->pr_protocol && pr->pr_protocol == ipproto) { 892 ip_protox[pr->pr_protocol] = pr - inetsw; 893 return (0); 894 } 895 } 896 return (EPROTONOSUPPORT); 897} 898 899int 900ipproto_unregister(short ipproto) 901{ 902 struct protosw *pr; 903 904 /* Sanity checks. */ 905 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 906 return (EPROTONOSUPPORT); 907 908 /* Check if the protocol was indeed registered. */ 909 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 910 if (pr == NULL) 911 return (EPFNOSUPPORT); 912 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 913 return (ENOENT); 914 915 /* Reset the protocol slot to IPPROTO_RAW. */ 916 ip_protox[ipproto] = pr - inetsw; 917 return (0); 918} 919 920u_char inetctlerrmap[PRC_NCMDS] = { 921 0, 0, 0, 0, 922 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 923 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 924 EMSGSIZE, EHOSTUNREACH, 0, 0, 925 0, 0, EHOSTUNREACH, 0, 926 ENOPROTOOPT, ECONNREFUSED 927}; 928 929/* 930 * Forward a packet. If some error occurs return the sender 931 * an icmp packet. Note we can't always generate a meaningful 932 * icmp message because icmp doesn't have a large enough repertoire 933 * of codes and types. 934 * 935 * If not forwarding, just drop the packet. This could be confusing 936 * if ipforwarding was zero but some routing protocol was advancing 937 * us as a gateway to somewhere. However, we must let the routing 938 * protocol deal with that. 939 * 940 * The srcrt parameter indicates whether the packet is being forwarded 941 * via a source route. 942 */ 943void 944ip_forward(struct mbuf *m, int srcrt) 945{ 946 struct ip *ip = mtod(m, struct ip *); 947 struct in_ifaddr *ia; 948 struct mbuf *mcopy; 949 struct sockaddr_in *sin; 950 struct in_addr dest; 951 struct route ro; 952 int error, type = 0, code = 0, mtu = 0; 953 954 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 955 IPSTAT_INC(ips_cantforward); 956 m_freem(m); 957 return; 958 } 959 if ( 960#ifdef IPSTEALTH 961 V_ipstealth == 0 && 962#endif 963 ip->ip_ttl <= IPTTLDEC) { 964 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 0, 0); 965 return; 966 } 967 968 bzero(&ro, sizeof(ro)); 969 sin = (struct sockaddr_in *)&ro.ro_dst; 970 sin->sin_family = AF_INET; 971 sin->sin_len = sizeof(*sin); 972 sin->sin_addr = ip->ip_dst; 973#ifdef RADIX_MPATH 974 rtalloc_mpath_fib(&ro, 975 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 976 M_GETFIB(m)); 977#else 978 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 979#endif 980 if (ro.ro_rt != NULL) { 981 ia = ifatoia(ro.ro_rt->rt_ifa); 982 ifa_ref(&ia->ia_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_copy() 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 if (ia != NULL) 1082 ifa_free(&ia->ia_ifa); 1083 return; 1084 } 1085 } 1086 if (mcopy == NULL) { 1087 if (ia != NULL) 1088 ifa_free(&ia->ia_ifa); 1089 return; 1090 } 1091 1092 switch (error) { 1093 1094 case 0: /* forwarded, but need redirect */ 1095 /* type, code set above */ 1096 break; 1097 1098 case ENETUNREACH: 1099 case EHOSTUNREACH: 1100 case ENETDOWN: 1101 case EHOSTDOWN: 1102 default: 1103 type = ICMP_UNREACH; 1104 code = ICMP_UNREACH_HOST; 1105 break; 1106 1107 case EMSGSIZE: 1108 type = ICMP_UNREACH; 1109 code = ICMP_UNREACH_NEEDFRAG; 1110 /* 1111 * If the MTU was set before make sure we are below the 1112 * interface MTU. 1113 * If the MTU wasn't set before use the interface mtu or 1114 * fall back to the next smaller mtu step compared to the 1115 * current packet size. 1116 */ 1117 if (mtu != 0) { 1118 if (ia != NULL) 1119 mtu = min(mtu, ia->ia_ifp->if_mtu); 1120 } else { 1121 if (ia != NULL) 1122 mtu = ia->ia_ifp->if_mtu; 1123 else 1124 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1125 } 1126 IPSTAT_INC(ips_cantfrag); 1127 break; 1128 1129 case ENOBUFS: 1130 case EACCES: /* ipfw denied packet */ 1131 m_freem(mcopy); 1132 if (ia != NULL) 1133 ifa_free(&ia->ia_ifa); 1134 return; 1135 } 1136 if (ia != NULL) 1137 ifa_free(&ia->ia_ifa); 1138 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1139} 1140 1141void 1142ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1143 struct mbuf *m) 1144{ 1145 1146 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1147 struct bintime bt; 1148 1149 bintime(&bt); 1150 if (inp->inp_socket->so_options & SO_BINTIME) { 1151 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), 1152 SCM_BINTIME, SOL_SOCKET); 1153 if (*mp) 1154 mp = &(*mp)->m_next; 1155 } 1156 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1157 struct timeval tv; 1158 1159 bintime2timeval(&bt, &tv); 1160 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), 1161 SCM_TIMESTAMP, SOL_SOCKET); 1162 if (*mp) 1163 mp = &(*mp)->m_next; 1164 } 1165 } 1166 if (inp->inp_flags & INP_RECVDSTADDR) { 1167 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, 1168 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1169 if (*mp) 1170 mp = &(*mp)->m_next; 1171 } 1172 if (inp->inp_flags & INP_RECVTTL) { 1173 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, 1174 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1175 if (*mp) 1176 mp = &(*mp)->m_next; 1177 } 1178#ifdef notyet 1179 /* XXX 1180 * Moving these out of udp_input() made them even more broken 1181 * than they already were. 1182 */ 1183 /* options were tossed already */ 1184 if (inp->inp_flags & INP_RECVOPTS) { 1185 *mp = sbcreatecontrol((caddr_t)opts_deleted_above, 1186 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1187 if (*mp) 1188 mp = &(*mp)->m_next; 1189 } 1190 /* ip_srcroute doesn't do what we want here, need to fix */ 1191 if (inp->inp_flags & INP_RECVRETOPTS) { 1192 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), 1193 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1194 if (*mp) 1195 mp = &(*mp)->m_next; 1196 } 1197#endif 1198 if (inp->inp_flags & INP_RECVIF) { 1199 struct ifnet *ifp; 1200 struct sdlbuf { 1201 struct sockaddr_dl sdl; 1202 u_char pad[32]; 1203 } sdlbuf; 1204 struct sockaddr_dl *sdp; 1205 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1206 1207 if ((ifp = m->m_pkthdr.rcvif) && 1208 ifp->if_index && ifp->if_index <= V_if_index) { 1209 sdp = (struct sockaddr_dl *)ifp->if_addr->ifa_addr; 1210 /* 1211 * Change our mind and don't try copy. 1212 */ 1213 if (sdp->sdl_family != AF_LINK || 1214 sdp->sdl_len > sizeof(sdlbuf)) { 1215 goto makedummy; 1216 } 1217 bcopy(sdp, sdl2, sdp->sdl_len); 1218 } else { 1219makedummy: 1220 sdl2->sdl_len = 1221 offsetof(struct sockaddr_dl, sdl_data[0]); 1222 sdl2->sdl_family = AF_LINK; 1223 sdl2->sdl_index = 0; 1224 sdl2->sdl_nlen = sdl2->sdl_alen = sdl2->sdl_slen = 0; 1225 } 1226 *mp = sbcreatecontrol((caddr_t)sdl2, sdl2->sdl_len, 1227 IP_RECVIF, IPPROTO_IP); 1228 if (*mp) 1229 mp = &(*mp)->m_next; 1230 } 1231 if (inp->inp_flags & INP_RECVTOS) { 1232 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, 1233 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 1234 if (*mp) 1235 mp = &(*mp)->m_next; 1236 } 1237 1238 if (inp->inp_flags2 & INP_RECVFLOWID) { 1239 uint32_t flowid, flow_type; 1240 1241 flowid = m->m_pkthdr.flowid; 1242 flow_type = M_HASHTYPE_GET(m); 1243 1244 /* 1245 * XXX should handle the failure of one or the 1246 * other - don't populate both? 1247 */ 1248 *mp = sbcreatecontrol((caddr_t) &flowid, 1249 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP); 1250 if (*mp) 1251 mp = &(*mp)->m_next; 1252 *mp = sbcreatecontrol((caddr_t) &flow_type, 1253 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP); 1254 if (*mp) 1255 mp = &(*mp)->m_next; 1256 } 1257 1258#ifdef RSS 1259 if (inp->inp_flags2 & INP_RECVRSSBUCKETID) { 1260 uint32_t flowid, flow_type; 1261 uint32_t rss_bucketid; 1262 1263 flowid = m->m_pkthdr.flowid; 1264 flow_type = M_HASHTYPE_GET(m); 1265 1266 if (rss_hash2bucket(flowid, flow_type, &rss_bucketid) == 0) { 1267 *mp = sbcreatecontrol((caddr_t) &rss_bucketid, 1268 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP); 1269 if (*mp) 1270 mp = &(*mp)->m_next; 1271 } 1272 } 1273#endif 1274} 1275 1276/* 1277 * XXXRW: Multicast routing code in ip_mroute.c is generally MPSAFE, but the 1278 * ip_rsvp and ip_rsvp_on variables need to be interlocked with rsvp_on 1279 * locking. This code remains in ip_input.c as ip_mroute.c is optionally 1280 * compiled. 1281 */ 1282static VNET_DEFINE(int, ip_rsvp_on); 1283VNET_DEFINE(struct socket *, ip_rsvpd); 1284 1285#define V_ip_rsvp_on VNET(ip_rsvp_on) 1286 1287int 1288ip_rsvp_init(struct socket *so) 1289{ 1290 1291 if (so->so_type != SOCK_RAW || 1292 so->so_proto->pr_protocol != IPPROTO_RSVP) 1293 return EOPNOTSUPP; 1294 1295 if (V_ip_rsvpd != NULL) 1296 return EADDRINUSE; 1297 1298 V_ip_rsvpd = so; 1299 /* 1300 * This may seem silly, but we need to be sure we don't over-increment 1301 * the RSVP counter, in case something slips up. 1302 */ 1303 if (!V_ip_rsvp_on) { 1304 V_ip_rsvp_on = 1; 1305 V_rsvp_on++; 1306 } 1307 1308 return 0; 1309} 1310 1311int 1312ip_rsvp_done(void) 1313{ 1314 1315 V_ip_rsvpd = NULL; 1316 /* 1317 * This may seem silly, but we need to be sure we don't over-decrement 1318 * the RSVP counter, in case something slips up. 1319 */ 1320 if (V_ip_rsvp_on) { 1321 V_ip_rsvp_on = 0; 1322 V_rsvp_on--; 1323 } 1324 return 0; 1325} 1326 1327int 1328rsvp_input(struct mbuf **mp, int *offp, int proto) 1329{ 1330 struct mbuf *m; 1331 1332 m = *mp; 1333 *mp = NULL; 1334 1335 if (rsvp_input_p) { /* call the real one if loaded */ 1336 *mp = m; 1337 rsvp_input_p(mp, offp, proto); 1338 return (IPPROTO_DONE); 1339 } 1340 1341 /* Can still get packets with rsvp_on = 0 if there is a local member 1342 * of the group to which the RSVP packet is addressed. But in this 1343 * case we want to throw the packet away. 1344 */ 1345 1346 if (!V_rsvp_on) { 1347 m_freem(m); 1348 return (IPPROTO_DONE); 1349 } 1350 1351 if (V_ip_rsvpd != NULL) { 1352 *mp = m; 1353 rip_input(mp, offp, proto); 1354 return (IPPROTO_DONE); 1355 } 1356 /* Drop the packet */ 1357 m_freem(m); 1358 return (IPPROTO_DONE); 1359} 1360