1 /*- 2 * Copyright (c) 1982, 1986, 1988, 1993 3 * The Regents of the University of California. All rights reserved. 4 * 5 * Redistribution and use in source and binary forms, with or without 6 * modification, are permitted provided that the following conditions 7 * are met: 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 2. Redistributions in binary form must reproduce the above copyright 11 * notice, this list of conditions and the following disclaimer in the 12 * documentation and/or other materials provided with the distribution. 13 * 4. Neither the name of the University nor the names of its contributors 14 * may be used to endorse or promote products derived from this software 15 * without specific prior written permission. 16 * 17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 27 * SUCH DAMAGE. 28 * 29 * @(#)ip_input.c 8.2 (Berkeley) 1/4/94 30 */ 31 32 #include <sys/cdefs.h> 33 __FBSDID("$FreeBSD$"); 34 35 #include "opt_bootp.h" 36 #include "opt_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 #ifdef IPSEC 81 #include <netinet/ip_ipsec.h> 82 #include <netipsec/ipsec.h> 83 #include <netipsec/key.h> 84 #endif /* IPSEC */ 85 #include <netinet/in_rss.h> 86 87 #include <sys/socketvar.h> 88 89 #include <security/mac/mac_framework.h> 90 91 #ifdef CTASSERT 92 CTASSERT(sizeof(struct ip) == 20); 93 #endif 94 95 /* IP reassembly functions are defined in ip_reass.c. */ 96 extern void ipreass_init(void); 97 extern void ipreass_drain(void); 98 extern void ipreass_slowtimo(void); 99 #ifdef VIMAGE 100 extern void ipreass_destroy(void); 101 #endif 102 103 struct rmlock in_ifaddr_lock; 104 RM_SYSINIT(in_ifaddr_lock, &in_ifaddr_lock, "in_ifaddr_lock"); 105 106 VNET_DEFINE(int, rsvp_on); 107 108 VNET_DEFINE(int, ipforwarding); 109 SYSCTL_INT(_net_inet_ip, IPCTL_FORWARDING, forwarding, CTLFLAG_VNET | CTLFLAG_RW, 110 &VNET_NAME(ipforwarding), 0, 111 "Enable IP forwarding between interfaces"); 112 113 static VNET_DEFINE(int, ipsendredirects) = 1; /* XXX */ 114 #define V_ipsendredirects VNET(ipsendredirects) 115 SYSCTL_INT(_net_inet_ip, IPCTL_SENDREDIRECTS, redirect, CTLFLAG_VNET | CTLFLAG_RW, 116 &VNET_NAME(ipsendredirects), 0, 117 "Enable sending IP redirects"); 118 119 /* 120 * XXX - Setting ip_checkinterface mostly implements the receive side of 121 * the Strong ES model described in RFC 1122, but since the routing table 122 * and transmit implementation do not implement the Strong ES model, 123 * setting this to 1 results in an odd hybrid. 124 * 125 * XXX - ip_checkinterface currently must be disabled if you use ipnat 126 * to translate the destination address to another local interface. 127 * 128 * XXX - ip_checkinterface must be disabled if you add IP aliases 129 * to the loopback interface instead of the interface where the 130 * packets for those addresses are received. 131 */ 132 static VNET_DEFINE(int, ip_checkinterface); 133 #define V_ip_checkinterface VNET(ip_checkinterface) 134 SYSCTL_INT(_net_inet_ip, OID_AUTO, check_interface, CTLFLAG_VNET | CTLFLAG_RW, 135 &VNET_NAME(ip_checkinterface), 0, 136 "Verify packet arrives on correct interface"); 137 138 VNET_DEFINE(struct pfil_head, inet_pfil_hook); /* Packet filter hooks */ 139 140 static struct netisr_handler ip_nh = { 141 .nh_name = "ip", 142 .nh_handler = ip_input, 143 .nh_proto = NETISR_IP, 144 #ifdef RSS 145 .nh_m2cpuid = rss_soft_m2cpuid_v4, 146 .nh_policy = NETISR_POLICY_CPU, 147 .nh_dispatch = NETISR_DISPATCH_HYBRID, 148 #else 149 .nh_policy = NETISR_POLICY_FLOW, 150 #endif 151 }; 152 153 #ifdef RSS 154 /* 155 * Directly dispatched frames are currently assumed 156 * to have a flowid already calculated. 157 * 158 * It should likely have something that assert it 159 * actually has valid flow details. 160 */ 161 static struct netisr_handler ip_direct_nh = { 162 .nh_name = "ip_direct", 163 .nh_handler = ip_direct_input, 164 .nh_proto = NETISR_IP_DIRECT, 165 .nh_m2cpuid = rss_soft_m2cpuid_v4, 166 .nh_policy = NETISR_POLICY_CPU, 167 .nh_dispatch = NETISR_DISPATCH_HYBRID, 168 }; 169 #endif 170 171 extern struct domain inetdomain; 172 extern struct protosw inetsw[]; 173 u_char ip_protox[IPPROTO_MAX]; 174 VNET_DEFINE(struct in_ifaddrhead, in_ifaddrhead); /* first inet address */ 175 VNET_DEFINE(struct in_ifaddrhashhead *, in_ifaddrhashtbl); /* inet addr hash table */ 176 VNET_DEFINE(u_long, in_ifaddrhmask); /* mask for hash table */ 177 178 #ifdef IPCTL_DEFMTU 179 SYSCTL_INT(_net_inet_ip, IPCTL_DEFMTU, mtu, CTLFLAG_RW, 180 &ip_mtu, 0, "Default MTU"); 181 #endif 182 183 #ifdef IPSTEALTH 184 VNET_DEFINE(int, ipstealth); 185 SYSCTL_INT(_net_inet_ip, OID_AUTO, stealth, CTLFLAG_VNET | CTLFLAG_RW, 186 &VNET_NAME(ipstealth), 0, 187 "IP stealth mode, no TTL decrementation on forwarding"); 188 #endif 189 190 /* 191 * IP statistics are stored in the "array" of counter(9)s. 192 */ 193 VNET_PCPUSTAT_DEFINE(struct ipstat, ipstat); 194 VNET_PCPUSTAT_SYSINIT(ipstat); 195 SYSCTL_VNET_PCPUSTAT(_net_inet_ip, IPCTL_STATS, stats, struct ipstat, ipstat, 196 "IP statistics (struct ipstat, netinet/ip_var.h)"); 197 198 #ifdef VIMAGE 199 VNET_PCPUSTAT_SYSUNINIT(ipstat); 200 #endif /* VIMAGE */ 201 202 /* 203 * Kernel module interface for updating ipstat. The argument is an index 204 * into ipstat treated as an array. 205 */ 206 void 207 kmod_ipstat_inc(int statnum) 208 { 209 210 counter_u64_add(VNET(ipstat)[statnum], 1); 211 } 212 213 void 214 kmod_ipstat_dec(int statnum) 215 { 216 217 counter_u64_add(VNET(ipstat)[statnum], -1); 218 } 219 220 static int 221 sysctl_netinet_intr_queue_maxlen(SYSCTL_HANDLER_ARGS) 222 { 223 int error, qlimit; 224 225 netisr_getqlimit(&ip_nh, &qlimit); 226 error = sysctl_handle_int(oidp, &qlimit, 0, req); 227 if (error || !req->newptr) 228 return (error); 229 if (qlimit < 1) 230 return (EINVAL); 231 return (netisr_setqlimit(&ip_nh, qlimit)); 232 } 233 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_queue_maxlen, 234 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_queue_maxlen, "I", 235 "Maximum size of the IP input queue"); 236 237 static int 238 sysctl_netinet_intr_queue_drops(SYSCTL_HANDLER_ARGS) 239 { 240 u_int64_t qdrops_long; 241 int error, qdrops; 242 243 netisr_getqdrops(&ip_nh, &qdrops_long); 244 qdrops = qdrops_long; 245 error = sysctl_handle_int(oidp, &qdrops, 0, req); 246 if (error || !req->newptr) 247 return (error); 248 if (qdrops != 0) 249 return (EINVAL); 250 netisr_clearqdrops(&ip_nh); 251 return (0); 252 } 253 254 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_queue_drops, 255 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_queue_drops, "I", 256 "Number of packets dropped from the IP input queue"); 257 258 #ifdef RSS 259 static int 260 sysctl_netinet_intr_direct_queue_maxlen(SYSCTL_HANDLER_ARGS) 261 { 262 int error, qlimit; 263 264 netisr_getqlimit(&ip_direct_nh, &qlimit); 265 error = sysctl_handle_int(oidp, &qlimit, 0, req); 266 if (error || !req->newptr) 267 return (error); 268 if (qlimit < 1) 269 return (EINVAL); 270 return (netisr_setqlimit(&ip_direct_nh, qlimit)); 271 } 272 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQMAXLEN, intr_direct_queue_maxlen, 273 CTLTYPE_INT|CTLFLAG_RW, 0, 0, sysctl_netinet_intr_direct_queue_maxlen, "I", 274 "Maximum size of the IP direct input queue"); 275 276 static int 277 sysctl_netinet_intr_direct_queue_drops(SYSCTL_HANDLER_ARGS) 278 { 279 u_int64_t qdrops_long; 280 int error, qdrops; 281 282 netisr_getqdrops(&ip_direct_nh, &qdrops_long); 283 qdrops = qdrops_long; 284 error = sysctl_handle_int(oidp, &qdrops, 0, req); 285 if (error || !req->newptr) 286 return (error); 287 if (qdrops != 0) 288 return (EINVAL); 289 netisr_clearqdrops(&ip_direct_nh); 290 return (0); 291 } 292 293 SYSCTL_PROC(_net_inet_ip, IPCTL_INTRQDROPS, intr_direct_queue_drops, 294 CTLTYPE_INT|CTLFLAG_RD, 0, 0, sysctl_netinet_intr_direct_queue_drops, "I", 295 "Number of packets dropped from the IP direct input queue"); 296 #endif /* RSS */ 297 298 /* 299 * IP initialization: fill in IP protocol switch table. 300 * All protocols not implemented in kernel go to raw IP protocol handler. 301 */ 302 void 303 ip_init(void) 304 { 305 struct protosw *pr; 306 int i; 307 308 TAILQ_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 V_inet_pfil_hook.ph_type = PFIL_TYPE_AF; 316 V_inet_pfil_hook.ph_af = AF_INET; 317 if ((i = pfil_head_register(&V_inet_pfil_hook)) != 0) 318 printf("%s: WARNING: unable to register pfil hook, " 319 "error %d\n", __func__, i); 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 if ((error = pfil_head_unregister(&V_inet_pfil_hook)) != 0) 382 printf("%s: WARNING: unable to unregister pfil hook, " 383 "error %d\n", __func__, error); 384 385 error = hhook_head_deregister(V_ipsec_hhh_in[HHOOK_IPSEC_INET]); 386 if (error != 0) { 387 printf("%s: WARNING: unable to deregister input helper hook " 388 "type HHOOK_TYPE_IPSEC_IN, id HHOOK_IPSEC_INET: " 389 "error %d returned\n", __func__, error); 390 } 391 error = hhook_head_deregister(V_ipsec_hhh_out[HHOOK_IPSEC_INET]); 392 if (error != 0) { 393 printf("%s: WARNING: unable to deregister output helper hook " 394 "type HHOOK_TYPE_IPSEC_OUT, id HHOOK_IPSEC_INET: " 395 "error %d returned\n", __func__, error); 396 } 397 398 /* Remove the IPv4 addresses from all interfaces. */ 399 in_ifscrub_all(); 400 401 /* Make sure the IPv4 routes are gone as well. */ 402 IFNET_RLOCK(); 403 TAILQ_FOREACH(ifp, &V_ifnet, if_link) 404 rt_flushifroutes_af(ifp, AF_INET); 405 IFNET_RUNLOCK(); 406 407 /* Destroy IP reassembly queue. */ 408 ipreass_destroy(); 409 410 /* Cleanup in_ifaddr hash table; should be empty. */ 411 hashdestroy(V_in_ifaddrhashtbl, M_IFADDR, V_in_ifaddrhmask); 412 } 413 414 VNET_SYSUNINIT(ip, SI_SUB_PROTO_DOMAIN, SI_ORDER_THIRD, ip_destroy, NULL); 415 #endif 416 417 #ifdef RSS 418 /* 419 * IP direct input routine. 420 * 421 * This is called when reinjecting completed fragments where 422 * all of the previous checking and book-keeping has been done. 423 */ 424 void 425 ip_direct_input(struct mbuf *m) 426 { 427 struct ip *ip; 428 int hlen; 429 430 ip = mtod(m, struct ip *); 431 hlen = ip->ip_hl << 2; 432 433 IPSTAT_INC(ips_delivered); 434 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 435 return; 436 } 437 #endif 438 439 /* 440 * Ip input routine. Checksum and byte swap header. If fragmented 441 * try to reassemble. Process options. Pass to next level. 442 */ 443 void 444 ip_input(struct mbuf *m) 445 { 446 struct ip *ip = NULL; 447 struct in_ifaddr *ia = NULL; 448 struct ifaddr *ifa; 449 struct ifnet *ifp; 450 int checkif, hlen = 0; 451 uint16_t sum, ip_len; 452 int dchg = 0; /* dest changed after fw */ 453 struct in_addr odst; /* original dst address */ 454 455 M_ASSERTPKTHDR(m); 456 457 if (m->m_flags & M_FASTFWD_OURS) { 458 m->m_flags &= ~M_FASTFWD_OURS; 459 /* Set up some basics that will be used later. */ 460 ip = mtod(m, struct ip *); 461 hlen = ip->ip_hl << 2; 462 ip_len = ntohs(ip->ip_len); 463 goto ours; 464 } 465 466 IPSTAT_INC(ips_total); 467 468 if (m->m_pkthdr.len < sizeof(struct ip)) 469 goto tooshort; 470 471 if (m->m_len < sizeof (struct ip) && 472 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 473 IPSTAT_INC(ips_toosmall); 474 return; 475 } 476 ip = mtod(m, struct ip *); 477 478 if (ip->ip_v != IPVERSION) { 479 IPSTAT_INC(ips_badvers); 480 goto bad; 481 } 482 483 hlen = ip->ip_hl << 2; 484 if (hlen < sizeof(struct ip)) { /* minimum header length */ 485 IPSTAT_INC(ips_badhlen); 486 goto bad; 487 } 488 if (hlen > m->m_len) { 489 if ((m = m_pullup(m, hlen)) == NULL) { 490 IPSTAT_INC(ips_badhlen); 491 return; 492 } 493 ip = mtod(m, struct ip *); 494 } 495 496 IP_PROBE(receive, NULL, NULL, ip, m->m_pkthdr.rcvif, ip, NULL); 497 498 /* 127/8 must not appear on wire - RFC1122 */ 499 ifp = m->m_pkthdr.rcvif; 500 if ((ntohl(ip->ip_dst.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET || 501 (ntohl(ip->ip_src.s_addr) >> IN_CLASSA_NSHIFT) == IN_LOOPBACKNET) { 502 if ((ifp->if_flags & IFF_LOOPBACK) == 0) { 503 IPSTAT_INC(ips_badaddr); 504 goto bad; 505 } 506 } 507 508 if (m->m_pkthdr.csum_flags & CSUM_IP_CHECKED) { 509 sum = !(m->m_pkthdr.csum_flags & CSUM_IP_VALID); 510 } else { 511 if (hlen == sizeof(struct ip)) { 512 sum = in_cksum_hdr(ip); 513 } else { 514 sum = in_cksum(m, hlen); 515 } 516 } 517 if (sum) { 518 IPSTAT_INC(ips_badsum); 519 goto bad; 520 } 521 522 #ifdef ALTQ 523 if (altq_input != NULL && (*altq_input)(m, AF_INET) == 0) 524 /* packet is dropped by traffic conditioner */ 525 return; 526 #endif 527 528 ip_len = ntohs(ip->ip_len); 529 if (ip_len < hlen) { 530 IPSTAT_INC(ips_badlen); 531 goto bad; 532 } 533 534 /* 535 * Check that the amount of data in the buffers 536 * is as at least much as the IP header would have us expect. 537 * Trim mbufs if longer than we expect. 538 * Drop packet if shorter than we expect. 539 */ 540 if (m->m_pkthdr.len < ip_len) { 541 tooshort: 542 IPSTAT_INC(ips_tooshort); 543 goto bad; 544 } 545 if (m->m_pkthdr.len > ip_len) { 546 if (m->m_len == m->m_pkthdr.len) { 547 m->m_len = ip_len; 548 m->m_pkthdr.len = ip_len; 549 } else 550 m_adj(m, ip_len - m->m_pkthdr.len); 551 } 552 553 /* 554 * Try to forward the packet, but if we fail continue. 555 * ip_tryforward() does inbound and outbound packet firewall 556 * processing. If firewall has decided that destination becomes 557 * our local address, it sets M_FASTFWD_OURS flag. In this 558 * case skip another inbound firewall processing and update 559 * ip pointer. 560 */ 561 if (V_ipforwarding != 0 562 #ifdef IPSEC 563 && !key_havesp(IPSEC_DIR_INBOUND) 564 && !key_havesp(IPSEC_DIR_OUTBOUND) 565 #endif 566 ) { 567 if ((m = ip_tryforward(m)) == NULL) 568 return; 569 if (m->m_flags & M_FASTFWD_OURS) { 570 m->m_flags &= ~M_FASTFWD_OURS; 571 ip = mtod(m, struct ip *); 572 goto ours; 573 } 574 } 575 #ifdef IPSEC 576 /* 577 * Bypass packet filtering for packets previously handled by IPsec. 578 */ 579 if (ip_ipsec_filtertunnel(m)) 580 goto passin; 581 #endif 582 /* 583 * Run through list of hooks for input packets. 584 * 585 * NB: Beware of the destination address changing (e.g. 586 * by NAT rewriting). When this happens, tell 587 * ip_forward to do the right thing. 588 */ 589 590 /* Jump over all PFIL processing if hooks are not active. */ 591 if (!PFIL_HOOKED(&V_inet_pfil_hook)) 592 goto passin; 593 594 odst = ip->ip_dst; 595 if (pfil_run_hooks(&V_inet_pfil_hook, &m, ifp, PFIL_IN, NULL) != 0) 596 return; 597 if (m == NULL) /* consumed by filter */ 598 return; 599 600 ip = mtod(m, struct ip *); 601 dchg = (odst.s_addr != ip->ip_dst.s_addr); 602 ifp = m->m_pkthdr.rcvif; 603 604 if (m->m_flags & M_FASTFWD_OURS) { 605 m->m_flags &= ~M_FASTFWD_OURS; 606 goto ours; 607 } 608 if (m->m_flags & M_IP_NEXTHOP) { 609 if (m_tag_find(m, PACKET_TAG_IPFORWARD, NULL) != NULL) { 610 /* 611 * Directly ship the packet on. This allows 612 * forwarding packets originally destined to us 613 * to some other directly connected host. 614 */ 615 ip_forward(m, 1); 616 return; 617 } 618 } 619 passin: 620 621 /* 622 * Process options and, if not destined for us, 623 * ship it on. ip_dooptions returns 1 when an 624 * error was detected (causing an icmp message 625 * to be sent and the original packet to be freed). 626 */ 627 if (hlen > sizeof (struct ip) && ip_dooptions(m, 0)) 628 return; 629 630 /* greedy RSVP, snatches any PATH packet of the RSVP protocol and no 631 * matter if it is destined to another node, or whether it is 632 * a multicast one, RSVP wants it! and prevents it from being forwarded 633 * anywhere else. Also checks if the rsvp daemon is running before 634 * grabbing the packet. 635 */ 636 if (V_rsvp_on && ip->ip_p==IPPROTO_RSVP) 637 goto ours; 638 639 /* 640 * Check our list of addresses, to see if the packet is for us. 641 * If we don't have any addresses, assume any unicast packet 642 * we receive might be for us (and let the upper layers deal 643 * with it). 644 */ 645 if (TAILQ_EMPTY(&V_in_ifaddrhead) && 646 (m->m_flags & (M_MCAST|M_BCAST)) == 0) 647 goto ours; 648 649 /* 650 * Enable a consistency check between the destination address 651 * and the arrival interface for a unicast packet (the RFC 1122 652 * strong ES model) if IP forwarding is disabled and the packet 653 * is not locally generated and the packet is not subject to 654 * 'ipfw fwd'. 655 * 656 * XXX - Checking also should be disabled if the destination 657 * address is ipnat'ed to a different interface. 658 * 659 * XXX - Checking is incompatible with IP aliases added 660 * to the loopback interface instead of the interface where 661 * the packets are received. 662 * 663 * XXX - This is the case for carp vhost IPs as well so we 664 * insert a workaround. If the packet got here, we already 665 * checked with carp_iamatch() and carp_forus(). 666 */ 667 checkif = V_ip_checkinterface && (V_ipforwarding == 0) && 668 ifp != NULL && ((ifp->if_flags & IFF_LOOPBACK) == 0) && 669 ifp->if_carp == NULL && (dchg == 0); 670 671 /* 672 * Check for exact addresses in the hash bucket. 673 */ 674 /* IN_IFADDR_RLOCK(); */ 675 LIST_FOREACH(ia, INADDR_HASH(ip->ip_dst.s_addr), ia_hash) { 676 /* 677 * If the address matches, verify that the packet 678 * arrived via the correct interface if checking is 679 * enabled. 680 */ 681 if (IA_SIN(ia)->sin_addr.s_addr == ip->ip_dst.s_addr && 682 (!checkif || ia->ia_ifp == ifp)) { 683 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 684 counter_u64_add(ia->ia_ifa.ifa_ibytes, 685 m->m_pkthdr.len); 686 /* IN_IFADDR_RUNLOCK(); */ 687 goto ours; 688 } 689 } 690 /* IN_IFADDR_RUNLOCK(); */ 691 692 /* 693 * Check for broadcast addresses. 694 * 695 * Only accept broadcast packets that arrive via the matching 696 * interface. Reception of forwarded directed broadcasts would 697 * be handled via ip_forward() and ether_output() with the loopback 698 * into the stack for SIMPLEX interfaces handled by ether_output(). 699 */ 700 if (ifp != NULL && ifp->if_flags & IFF_BROADCAST) { 701 IF_ADDR_RLOCK(ifp); 702 TAILQ_FOREACH(ifa, &ifp->if_addrhead, ifa_link) { 703 if (ifa->ifa_addr->sa_family != AF_INET) 704 continue; 705 ia = ifatoia(ifa); 706 if (satosin(&ia->ia_broadaddr)->sin_addr.s_addr == 707 ip->ip_dst.s_addr) { 708 counter_u64_add(ia->ia_ifa.ifa_ipackets, 1); 709 counter_u64_add(ia->ia_ifa.ifa_ibytes, 710 m->m_pkthdr.len); 711 IF_ADDR_RUNLOCK(ifp); 712 goto ours; 713 } 714 #ifdef BOOTP_COMPAT 715 if (IA_SIN(ia)->sin_addr.s_addr == INADDR_ANY) { 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 #endif 723 } 724 IF_ADDR_RUNLOCK(ifp); 725 ia = NULL; 726 } 727 /* RFC 3927 2.7: Do not forward datagrams for 169.254.0.0/16. */ 728 if (IN_LINKLOCAL(ntohl(ip->ip_dst.s_addr))) { 729 IPSTAT_INC(ips_cantforward); 730 m_freem(m); 731 return; 732 } 733 if (IN_MULTICAST(ntohl(ip->ip_dst.s_addr))) { 734 if (V_ip_mrouter) { 735 /* 736 * If we are acting as a multicast router, all 737 * incoming multicast packets are passed to the 738 * kernel-level multicast forwarding function. 739 * The packet is returned (relatively) intact; if 740 * ip_mforward() returns a non-zero value, the packet 741 * must be discarded, else it may be accepted below. 742 */ 743 if (ip_mforward && ip_mforward(ip, ifp, m, 0) != 0) { 744 IPSTAT_INC(ips_cantforward); 745 m_freem(m); 746 return; 747 } 748 749 /* 750 * The process-level routing daemon needs to receive 751 * all multicast IGMP packets, whether or not this 752 * host belongs to their destination groups. 753 */ 754 if (ip->ip_p == IPPROTO_IGMP) 755 goto ours; 756 IPSTAT_INC(ips_forward); 757 } 758 /* 759 * Assume the packet is for us, to avoid prematurely taking 760 * a lock on the in_multi hash. Protocols must perform 761 * their own filtering and update statistics accordingly. 762 */ 763 goto ours; 764 } 765 if (ip->ip_dst.s_addr == (u_long)INADDR_BROADCAST) 766 goto ours; 767 if (ip->ip_dst.s_addr == INADDR_ANY) 768 goto ours; 769 770 /* 771 * Not for us; forward if possible and desirable. 772 */ 773 if (V_ipforwarding == 0) { 774 IPSTAT_INC(ips_cantforward); 775 m_freem(m); 776 } else { 777 ip_forward(m, dchg); 778 } 779 return; 780 781 ours: 782 #ifdef IPSTEALTH 783 /* 784 * IPSTEALTH: Process non-routing options only 785 * if the packet is destined for us. 786 */ 787 if (V_ipstealth && hlen > sizeof (struct ip) && ip_dooptions(m, 1)) 788 return; 789 #endif /* IPSTEALTH */ 790 791 /* 792 * Attempt reassembly; if it succeeds, proceed. 793 * ip_reass() will return a different mbuf. 794 */ 795 if (ip->ip_off & htons(IP_MF | IP_OFFMASK)) { 796 /* XXXGL: shouldn't we save & set m_flags? */ 797 m = ip_reass(m); 798 if (m == NULL) 799 return; 800 ip = mtod(m, struct ip *); 801 /* Get the header length of the reassembled packet */ 802 hlen = ip->ip_hl << 2; 803 } 804 805 #ifdef IPSEC 806 /* 807 * enforce IPsec policy checking if we are seeing last header. 808 * note that we do not visit this with protocols with pcb layer 809 * code - like udp/tcp/raw ip. 810 */ 811 if (ip_ipsec_input(m, ip->ip_p) != 0) 812 goto bad; 813 #endif /* IPSEC */ 814 815 /* 816 * Switch out to protocol's input routine. 817 */ 818 IPSTAT_INC(ips_delivered); 819 820 (*inetsw[ip_protox[ip->ip_p]].pr_input)(&m, &hlen, ip->ip_p); 821 return; 822 bad: 823 m_freem(m); 824 } 825 826 /* 827 * IP timer processing; 828 * if a timer expires on a reassembly 829 * queue, discard it. 830 */ 831 void 832 ip_slowtimo(void) 833 { 834 VNET_ITERATOR_DECL(vnet_iter); 835 836 VNET_LIST_RLOCK_NOSLEEP(); 837 VNET_FOREACH(vnet_iter) { 838 CURVNET_SET(vnet_iter); 839 ipreass_slowtimo(); 840 CURVNET_RESTORE(); 841 } 842 VNET_LIST_RUNLOCK_NOSLEEP(); 843 } 844 845 void 846 ip_drain(void) 847 { 848 VNET_ITERATOR_DECL(vnet_iter); 849 850 VNET_LIST_RLOCK_NOSLEEP(); 851 VNET_FOREACH(vnet_iter) { 852 CURVNET_SET(vnet_iter); 853 ipreass_drain(); 854 CURVNET_RESTORE(); 855 } 856 VNET_LIST_RUNLOCK_NOSLEEP(); 857 } 858 859 /* 860 * The protocol to be inserted into ip_protox[] must be already registered 861 * in inetsw[], either statically or through pf_proto_register(). 862 */ 863 int 864 ipproto_register(short ipproto) 865 { 866 struct protosw *pr; 867 868 /* Sanity checks. */ 869 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 870 return (EPROTONOSUPPORT); 871 872 /* 873 * The protocol slot must not be occupied by another protocol 874 * already. An index pointing to IPPROTO_RAW is unused. 875 */ 876 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 877 if (pr == NULL) 878 return (EPFNOSUPPORT); 879 if (ip_protox[ipproto] != pr - inetsw) /* IPPROTO_RAW */ 880 return (EEXIST); 881 882 /* Find the protocol position in inetsw[] and set the index. */ 883 for (pr = inetdomain.dom_protosw; 884 pr < inetdomain.dom_protoswNPROTOSW; pr++) { 885 if (pr->pr_domain->dom_family == PF_INET && 886 pr->pr_protocol && pr->pr_protocol == ipproto) { 887 ip_protox[pr->pr_protocol] = pr - inetsw; 888 return (0); 889 } 890 } 891 return (EPROTONOSUPPORT); 892 } 893 894 int 895 ipproto_unregister(short ipproto) 896 { 897 struct protosw *pr; 898 899 /* Sanity checks. */ 900 if (ipproto <= 0 || ipproto >= IPPROTO_MAX) 901 return (EPROTONOSUPPORT); 902 903 /* Check if the protocol was indeed registered. */ 904 pr = pffindproto(PF_INET, IPPROTO_RAW, SOCK_RAW); 905 if (pr == NULL) 906 return (EPFNOSUPPORT); 907 if (ip_protox[ipproto] == pr - inetsw) /* IPPROTO_RAW */ 908 return (ENOENT); 909 910 /* Reset the protocol slot to IPPROTO_RAW. */ 911 ip_protox[ipproto] = pr - inetsw; 912 return (0); 913 } 914 915 u_char inetctlerrmap[PRC_NCMDS] = { 916 0, 0, 0, 0, 917 0, EMSGSIZE, EHOSTDOWN, EHOSTUNREACH, 918 EHOSTUNREACH, EHOSTUNREACH, ECONNREFUSED, ECONNREFUSED, 919 EMSGSIZE, EHOSTUNREACH, 0, 0, 920 0, 0, EHOSTUNREACH, 0, 921 ENOPROTOOPT, ECONNREFUSED 922 }; 923 924 /* 925 * Forward a packet. If some error occurs return the sender 926 * an icmp packet. Note we can't always generate a meaningful 927 * icmp message because icmp doesn't have a large enough repertoire 928 * of codes and types. 929 * 930 * If not forwarding, just drop the packet. This could be confusing 931 * if ipforwarding was zero but some routing protocol was advancing 932 * us as a gateway to somewhere. However, we must let the routing 933 * protocol deal with that. 934 * 935 * The srcrt parameter indicates whether the packet is being forwarded 936 * via a source route. 937 */ 938 void 939 ip_forward(struct mbuf *m, int srcrt) 940 { 941 struct ip *ip = mtod(m, struct ip *); 942 struct in_ifaddr *ia; 943 struct mbuf *mcopy; 944 struct sockaddr_in *sin; 945 struct in_addr dest; 946 struct route ro; 947 int error, type = 0, code = 0, mtu = 0; 948 949 if (m->m_flags & (M_BCAST|M_MCAST) || in_canforward(ip->ip_dst) == 0) { 950 IPSTAT_INC(ips_cantforward); 951 m_freem(m); 952 return; 953 } 954 #ifdef IPSEC 955 if (ip_ipsec_fwd(m) != 0) { 956 IPSTAT_INC(ips_cantforward); 957 m_freem(m); 958 return; 959 } 960 #endif /* IPSEC */ 961 #ifdef IPSTEALTH 962 if (!V_ipstealth) { 963 #endif 964 if (ip->ip_ttl <= IPTTLDEC) { 965 icmp_error(m, ICMP_TIMXCEED, ICMP_TIMXCEED_INTRANS, 966 0, 0); 967 return; 968 } 969 #ifdef IPSTEALTH 970 } 971 #endif 972 973 bzero(&ro, sizeof(ro)); 974 sin = (struct sockaddr_in *)&ro.ro_dst; 975 sin->sin_family = AF_INET; 976 sin->sin_len = sizeof(*sin); 977 sin->sin_addr = ip->ip_dst; 978 #ifdef RADIX_MPATH 979 rtalloc_mpath_fib(&ro, 980 ntohl(ip->ip_src.s_addr ^ ip->ip_dst.s_addr), 981 M_GETFIB(m)); 982 #else 983 in_rtalloc_ign(&ro, 0, M_GETFIB(m)); 984 #endif 985 if (ro.ro_rt != NULL) { 986 ia = ifatoia(ro.ro_rt->rt_ifa); 987 ifa_ref(&ia->ia_ifa); 988 } else 989 ia = NULL; 990 #ifndef IPSEC 991 /* 992 * 'ia' may be NULL if there is no route for this destination. 993 * In case of IPsec, Don't discard it just yet, but pass it to 994 * ip_output in case of outgoing IPsec policy. 995 */ 996 if (!srcrt && ia == NULL) { 997 icmp_error(m, ICMP_UNREACH, ICMP_UNREACH_HOST, 0, 0); 998 RO_RTFREE(&ro); 999 return; 1000 } 1001 #endif 1002 1003 /* 1004 * Save the IP header and at most 8 bytes of the payload, 1005 * in case we need to generate an ICMP message to the src. 1006 * 1007 * XXX this can be optimized a lot by saving the data in a local 1008 * buffer on the stack (72 bytes at most), and only allocating the 1009 * mbuf if really necessary. The vast majority of the packets 1010 * are forwarded without having to send an ICMP back (either 1011 * because unnecessary, or because rate limited), so we are 1012 * really we are wasting a lot of work here. 1013 * 1014 * We don't use m_copym() because it might return a reference 1015 * to a shared cluster. Both this function and ip_output() 1016 * assume exclusive access to the IP header in `m', so any 1017 * data in a cluster may change before we reach icmp_error(). 1018 */ 1019 mcopy = m_gethdr(M_NOWAIT, m->m_type); 1020 if (mcopy != NULL && !m_dup_pkthdr(mcopy, m, M_NOWAIT)) { 1021 /* 1022 * It's probably ok if the pkthdr dup fails (because 1023 * the deep copy of the tag chain failed), but for now 1024 * be conservative and just discard the copy since 1025 * code below may some day want the tags. 1026 */ 1027 m_free(mcopy); 1028 mcopy = NULL; 1029 } 1030 if (mcopy != NULL) { 1031 mcopy->m_len = min(ntohs(ip->ip_len), M_TRAILINGSPACE(mcopy)); 1032 mcopy->m_pkthdr.len = mcopy->m_len; 1033 m_copydata(m, 0, mcopy->m_len, mtod(mcopy, caddr_t)); 1034 } 1035 1036 #ifdef IPSTEALTH 1037 if (!V_ipstealth) { 1038 #endif 1039 ip->ip_ttl -= IPTTLDEC; 1040 #ifdef IPSTEALTH 1041 } 1042 #endif 1043 1044 /* 1045 * If forwarding packet using same interface that it came in on, 1046 * perhaps should send a redirect to sender to shortcut a hop. 1047 * Only send redirect if source is sending directly to us, 1048 * and if packet was not source routed (or has any options). 1049 * Also, don't send redirect if forwarding using a default route 1050 * or a route modified by a redirect. 1051 */ 1052 dest.s_addr = 0; 1053 if (!srcrt && V_ipsendredirects && 1054 ia != NULL && ia->ia_ifp == m->m_pkthdr.rcvif) { 1055 struct rtentry *rt; 1056 1057 rt = ro.ro_rt; 1058 1059 if (rt && (rt->rt_flags & (RTF_DYNAMIC|RTF_MODIFIED)) == 0 && 1060 satosin(rt_key(rt))->sin_addr.s_addr != 0) { 1061 #define RTA(rt) ((struct in_ifaddr *)(rt->rt_ifa)) 1062 u_long src = ntohl(ip->ip_src.s_addr); 1063 1064 if (RTA(rt) && 1065 (src & RTA(rt)->ia_subnetmask) == RTA(rt)->ia_subnet) { 1066 if (rt->rt_flags & RTF_GATEWAY) 1067 dest.s_addr = satosin(rt->rt_gateway)->sin_addr.s_addr; 1068 else 1069 dest.s_addr = ip->ip_dst.s_addr; 1070 /* Router requirements says to only send host redirects */ 1071 type = ICMP_REDIRECT; 1072 code = ICMP_REDIRECT_HOST; 1073 } 1074 } 1075 } 1076 1077 error = ip_output(m, NULL, &ro, IP_FORWARDING, NULL, NULL); 1078 1079 if (error == EMSGSIZE && ro.ro_rt) 1080 mtu = ro.ro_rt->rt_mtu; 1081 RO_RTFREE(&ro); 1082 1083 if (error) 1084 IPSTAT_INC(ips_cantforward); 1085 else { 1086 IPSTAT_INC(ips_forward); 1087 if (type) 1088 IPSTAT_INC(ips_redirectsent); 1089 else { 1090 if (mcopy) 1091 m_freem(mcopy); 1092 if (ia != NULL) 1093 ifa_free(&ia->ia_ifa); 1094 return; 1095 } 1096 } 1097 if (mcopy == NULL) { 1098 if (ia != NULL) 1099 ifa_free(&ia->ia_ifa); 1100 return; 1101 } 1102 1103 switch (error) { 1104 1105 case 0: /* forwarded, but need redirect */ 1106 /* type, code set above */ 1107 break; 1108 1109 case ENETUNREACH: 1110 case EHOSTUNREACH: 1111 case ENETDOWN: 1112 case EHOSTDOWN: 1113 default: 1114 type = ICMP_UNREACH; 1115 code = ICMP_UNREACH_HOST; 1116 break; 1117 1118 case EMSGSIZE: 1119 type = ICMP_UNREACH; 1120 code = ICMP_UNREACH_NEEDFRAG; 1121 1122 #ifdef IPSEC 1123 /* 1124 * If IPsec is configured for this path, 1125 * override any possibly mtu value set by ip_output. 1126 */ 1127 mtu = ip_ipsec_mtu(mcopy, mtu); 1128 #endif /* IPSEC */ 1129 /* 1130 * If the MTU was set before make sure we are below the 1131 * interface MTU. 1132 * If the MTU wasn't set before use the interface mtu or 1133 * fall back to the next smaller mtu step compared to the 1134 * current packet size. 1135 */ 1136 if (mtu != 0) { 1137 if (ia != NULL) 1138 mtu = min(mtu, ia->ia_ifp->if_mtu); 1139 } else { 1140 if (ia != NULL) 1141 mtu = ia->ia_ifp->if_mtu; 1142 else 1143 mtu = ip_next_mtu(ntohs(ip->ip_len), 0); 1144 } 1145 IPSTAT_INC(ips_cantfrag); 1146 break; 1147 1148 case ENOBUFS: 1149 case EACCES: /* ipfw denied packet */ 1150 m_freem(mcopy); 1151 if (ia != NULL) 1152 ifa_free(&ia->ia_ifa); 1153 return; 1154 } 1155 if (ia != NULL) 1156 ifa_free(&ia->ia_ifa); 1157 icmp_error(mcopy, type, code, dest.s_addr, mtu); 1158 } 1159 1160 void 1161 ip_savecontrol(struct inpcb *inp, struct mbuf **mp, struct ip *ip, 1162 struct mbuf *m) 1163 { 1164 1165 if (inp->inp_socket->so_options & (SO_BINTIME | SO_TIMESTAMP)) { 1166 struct bintime bt; 1167 1168 bintime(&bt); 1169 if (inp->inp_socket->so_options & SO_BINTIME) { 1170 *mp = sbcreatecontrol((caddr_t)&bt, sizeof(bt), 1171 SCM_BINTIME, SOL_SOCKET); 1172 if (*mp) 1173 mp = &(*mp)->m_next; 1174 } 1175 if (inp->inp_socket->so_options & SO_TIMESTAMP) { 1176 struct timeval tv; 1177 1178 bintime2timeval(&bt, &tv); 1179 *mp = sbcreatecontrol((caddr_t)&tv, sizeof(tv), 1180 SCM_TIMESTAMP, SOL_SOCKET); 1181 if (*mp) 1182 mp = &(*mp)->m_next; 1183 } 1184 } 1185 if (inp->inp_flags & INP_RECVDSTADDR) { 1186 *mp = sbcreatecontrol((caddr_t)&ip->ip_dst, 1187 sizeof(struct in_addr), IP_RECVDSTADDR, IPPROTO_IP); 1188 if (*mp) 1189 mp = &(*mp)->m_next; 1190 } 1191 if (inp->inp_flags & INP_RECVTTL) { 1192 *mp = sbcreatecontrol((caddr_t)&ip->ip_ttl, 1193 sizeof(u_char), IP_RECVTTL, IPPROTO_IP); 1194 if (*mp) 1195 mp = &(*mp)->m_next; 1196 } 1197 #ifdef notyet 1198 /* XXX 1199 * Moving these out of udp_input() made them even more broken 1200 * than they already were. 1201 */ 1202 /* options were tossed already */ 1203 if (inp->inp_flags & INP_RECVOPTS) { 1204 *mp = sbcreatecontrol((caddr_t)opts_deleted_above, 1205 sizeof(struct in_addr), IP_RECVOPTS, IPPROTO_IP); 1206 if (*mp) 1207 mp = &(*mp)->m_next; 1208 } 1209 /* ip_srcroute doesn't do what we want here, need to fix */ 1210 if (inp->inp_flags & INP_RECVRETOPTS) { 1211 *mp = sbcreatecontrol((caddr_t)ip_srcroute(m), 1212 sizeof(struct in_addr), IP_RECVRETOPTS, IPPROTO_IP); 1213 if (*mp) 1214 mp = &(*mp)->m_next; 1215 } 1216 #endif 1217 if (inp->inp_flags & INP_RECVIF) { 1218 struct ifnet *ifp; 1219 struct sdlbuf { 1220 struct sockaddr_dl sdl; 1221 u_char pad[32]; 1222 } sdlbuf; 1223 struct sockaddr_dl *sdp; 1224 struct sockaddr_dl *sdl2 = &sdlbuf.sdl; 1225 1226 if ((ifp = m->m_pkthdr.rcvif) && 1227 ifp->if_index && ifp->if_index <= V_if_index) { 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((caddr_t)sdl2, sdl2->sdl_len, 1246 IP_RECVIF, IPPROTO_IP); 1247 if (*mp) 1248 mp = &(*mp)->m_next; 1249 } 1250 if (inp->inp_flags & INP_RECVTOS) { 1251 *mp = sbcreatecontrol((caddr_t)&ip->ip_tos, 1252 sizeof(u_char), IP_RECVTOS, IPPROTO_IP); 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((caddr_t) &flowid, 1268 sizeof(uint32_t), IP_FLOWID, IPPROTO_IP); 1269 if (*mp) 1270 mp = &(*mp)->m_next; 1271 *mp = sbcreatecontrol((caddr_t) &flow_type, 1272 sizeof(uint32_t), IP_FLOWTYPE, IPPROTO_IP); 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((caddr_t) &rss_bucketid, 1287 sizeof(uint32_t), IP_RSSBUCKETID, IPPROTO_IP); 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 static VNET_DEFINE(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 (so->so_type != SOCK_RAW || 1311 so->so_proto->pr_protocol != IPPROTO_RSVP) 1312 return EOPNOTSUPP; 1313 1314 if (V_ip_rsvpd != NULL) 1315 return EADDRINUSE; 1316 1317 V_ip_rsvpd = so; 1318 /* 1319 * This may seem silly, but we need to be sure we don't over-increment 1320 * the RSVP counter, in case something slips up. 1321 */ 1322 if (!V_ip_rsvp_on) { 1323 V_ip_rsvp_on = 1; 1324 V_rsvp_on++; 1325 } 1326 1327 return 0; 1328 } 1329 1330 int 1331 ip_rsvp_done(void) 1332 { 1333 1334 V_ip_rsvpd = NULL; 1335 /* 1336 * This may seem silly, but we need to be sure we don't over-decrement 1337 * the RSVP counter, in case something slips up. 1338 */ 1339 if (V_ip_rsvp_on) { 1340 V_ip_rsvp_on = 0; 1341 V_rsvp_on--; 1342 } 1343 return 0; 1344 } 1345 1346 int 1347 rsvp_input(struct mbuf **mp, int *offp, int proto) 1348 { 1349 struct mbuf *m; 1350 1351 m = *mp; 1352 *mp = NULL; 1353 1354 if (rsvp_input_p) { /* call the real one if loaded */ 1355 *mp = m; 1356 rsvp_input_p(mp, offp, proto); 1357 return (IPPROTO_DONE); 1358 } 1359 1360 /* Can still get packets with rsvp_on = 0 if there is a local member 1361 * of the group to which the RSVP packet is addressed. But in this 1362 * case we want to throw the packet away. 1363 */ 1364 1365 if (!V_rsvp_on) { 1366 m_freem(m); 1367 return (IPPROTO_DONE); 1368 } 1369 1370 if (V_ip_rsvpd != NULL) { 1371 *mp = m; 1372 rip_input(mp, offp, proto); 1373 return (IPPROTO_DONE); 1374 } 1375 /* Drop the packet */ 1376 m_freem(m); 1377 return (IPPROTO_DONE); 1378 } 1379