1 /*- 2 * Copyright (c) 2002, 2003 Sam Leffler, Errno Consulting 3 * 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 * 14 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND 15 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 16 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 17 * ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE 18 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 19 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 20 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 21 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 22 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 23 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 24 * SUCH DAMAGE. 25 * 26 * $FreeBSD$ 27 */ 28 29 /* 30 * IPsec output processing. 31 */ 32 #include "opt_inet.h" 33 #include "opt_inet6.h" 34 #include "opt_ipsec.h" 35 #include "opt_enc.h" 36 37 #include <sys/param.h> 38 #include <sys/systm.h> 39 #include <sys/mbuf.h> 40 #include <sys/domain.h> 41 #include <sys/protosw.h> 42 #include <sys/socket.h> 43 #include <sys/errno.h> 44 #include <sys/syslog.h> 45 46 #include <net/if.h> 47 #include <net/pfil.h> 48 #include <net/route.h> 49 #include <net/vnet.h> 50 51 #include <netinet/in.h> 52 #include <netinet/in_systm.h> 53 #include <netinet/ip.h> 54 #include <netinet/ip_var.h> 55 #include <netinet/in_var.h> 56 #include <netinet/ip_ecn.h> 57 #ifdef INET6 58 #include <netinet6/ip6_ecn.h> 59 #endif 60 61 #include <netinet/ip6.h> 62 #ifdef INET6 63 #include <netinet6/ip6_var.h> 64 #endif 65 #include <netinet/in_pcb.h> 66 #ifdef INET6 67 #include <netinet/icmp6.h> 68 #endif 69 70 #include <netipsec/ipsec.h> 71 #ifdef INET6 72 #include <netipsec/ipsec6.h> 73 #endif 74 #include <netipsec/ah_var.h> 75 #include <netipsec/esp_var.h> 76 #include <netipsec/ipcomp_var.h> 77 78 #include <netipsec/xform.h> 79 80 #include <netipsec/key.h> 81 #include <netipsec/keydb.h> 82 #include <netipsec/key_debug.h> 83 84 #include <machine/in_cksum.h> 85 86 #ifdef IPSEC_NAT_T 87 #include <netinet/udp.h> 88 #endif 89 90 #ifdef DEV_ENC 91 #include <net/if_enc.h> 92 #endif 93 94 95 int 96 ipsec_process_done(struct mbuf *m, struct ipsecrequest *isr) 97 { 98 struct tdb_ident *tdbi; 99 struct m_tag *mtag; 100 struct secasvar *sav; 101 struct secasindex *saidx; 102 int error; 103 104 IPSEC_ASSERT(m != NULL, ("null mbuf")); 105 IPSEC_ASSERT(isr != NULL, ("null ISR")); 106 sav = isr->sav; 107 IPSEC_ASSERT(sav != NULL, ("null SA")); 108 IPSEC_ASSERT(sav->sah != NULL, ("null SAH")); 109 110 saidx = &sav->sah->saidx; 111 switch (saidx->dst.sa.sa_family) { 112 #ifdef INET 113 case AF_INET: 114 /* Fix the header length, for AH processing. */ 115 mtod(m, struct ip *)->ip_len = htons(m->m_pkthdr.len); 116 break; 117 #endif /* INET */ 118 #ifdef INET6 119 case AF_INET6: 120 /* Fix the header length, for AH processing. */ 121 if (m->m_pkthdr.len < sizeof (struct ip6_hdr)) { 122 error = ENXIO; 123 goto bad; 124 } 125 if (m->m_pkthdr.len - sizeof (struct ip6_hdr) > IPV6_MAXPACKET) { 126 /* No jumbogram support. */ 127 error = ENXIO; /*?*/ 128 goto bad; 129 } 130 mtod(m, struct ip6_hdr *)->ip6_plen = 131 htons(m->m_pkthdr.len - sizeof(struct ip6_hdr)); 132 break; 133 #endif /* INET6 */ 134 default: 135 DPRINTF(("%s: unknown protocol family %u\n", __func__, 136 saidx->dst.sa.sa_family)); 137 error = ENXIO; 138 goto bad; 139 } 140 141 /* 142 * Add a record of what we've done or what needs to be done to the 143 * packet. 144 */ 145 mtag = m_tag_get(PACKET_TAG_IPSEC_OUT_DONE, 146 sizeof(struct tdb_ident), M_NOWAIT); 147 if (mtag == NULL) { 148 DPRINTF(("%s: could not get packet tag\n", __func__)); 149 error = ENOMEM; 150 goto bad; 151 } 152 153 tdbi = (struct tdb_ident *)(mtag + 1); 154 tdbi->dst = saidx->dst; 155 tdbi->proto = saidx->proto; 156 tdbi->spi = sav->spi; 157 m_tag_prepend(m, mtag); 158 159 /* 160 * If there's another (bundled) SA to apply, do so. 161 * Note that this puts a burden on the kernel stack size. 162 * If this is a problem we'll need to introduce a queue 163 * to set the packet on so we can unwind the stack before 164 * doing further processing. 165 */ 166 if (isr->next) { 167 V_ipsec4stat.ips_out_bundlesa++; 168 /* XXX-BZ currently only support same AF bundles. */ 169 switch (saidx->dst.sa.sa_family) { 170 #ifdef INET 171 case AF_INET: 172 return ipsec4_process_packet(m, isr->next, 0, 0); 173 /* NOTREACHED */ 174 #endif 175 #ifdef notyet 176 #ifdef INET6 177 case AF_INET6: 178 /* XXX */ 179 ipsec6_output_trans() 180 ipsec6_output_tunnel() 181 /* NOTREACHED */ 182 #endif /* INET6 */ 183 #endif 184 default: 185 DPRINTF(("%s: unknown protocol family %u\n", __func__, 186 saidx->dst.sa.sa_family)); 187 error = ENXIO; 188 goto bad; 189 } 190 } 191 key_sa_recordxfer(sav, m); /* record data transfer */ 192 193 m_addr_changed(m); 194 195 /* 196 * We're done with IPsec processing, transmit the packet using the 197 * appropriate network protocol (IP or IPv6). SPD lookup will be 198 * performed again there. 199 */ 200 switch (saidx->dst.sa.sa_family) { 201 #ifdef INET 202 struct ip *ip; 203 case AF_INET: 204 ip = mtod(m, struct ip *); 205 ip->ip_len = ntohs(ip->ip_len); 206 ip->ip_off = ntohs(ip->ip_off); 207 208 #ifdef IPSEC_NAT_T 209 /* 210 * If NAT-T is enabled, now that all IPsec processing is done 211 * insert UDP encapsulation header after IP header. 212 */ 213 if (sav->natt_type) { 214 #ifdef _IP_VHL 215 const int hlen = IP_VHL_HL(ip->ip_vhl); 216 #else 217 const int hlen = (ip->ip_hl << 2); 218 #endif 219 int size, off; 220 struct mbuf *mi; 221 struct udphdr *udp; 222 223 size = sizeof(struct udphdr); 224 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) { 225 /* 226 * draft-ietf-ipsec-nat-t-ike-0[01].txt and 227 * draft-ietf-ipsec-udp-encaps-(00/)01.txt, 228 * ignoring possible AH mode 229 * non-IKE marker + non-ESP marker 230 * from draft-ietf-ipsec-udp-encaps-00.txt. 231 */ 232 size += sizeof(u_int64_t); 233 } 234 mi = m_makespace(m, hlen, size, &off); 235 if (mi == NULL) { 236 DPRINTF(("%s: m_makespace for udphdr failed\n", 237 __func__)); 238 error = ENOBUFS; 239 goto bad; 240 } 241 242 udp = (struct udphdr *)(mtod(mi, caddr_t) + off); 243 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) 244 udp->uh_sport = htons(UDP_ENCAP_ESPINUDP_PORT); 245 else 246 udp->uh_sport = 247 KEY_PORTFROMSADDR(&sav->sah->saidx.src); 248 udp->uh_dport = KEY_PORTFROMSADDR(&sav->sah->saidx.dst); 249 udp->uh_sum = 0; 250 udp->uh_ulen = htons(m->m_pkthdr.len - hlen); 251 ip->ip_len = m->m_pkthdr.len; 252 ip->ip_p = IPPROTO_UDP; 253 254 if (sav->natt_type == UDP_ENCAP_ESPINUDP_NON_IKE) 255 *(u_int64_t *)(udp + 1) = 0; 256 } 257 #endif /* IPSEC_NAT_T */ 258 259 return ip_output(m, NULL, NULL, IP_RAWOUTPUT, NULL, NULL); 260 #endif /* INET */ 261 #ifdef INET6 262 case AF_INET6: 263 /* 264 * We don't need massage, IPv6 header fields are always in 265 * net endian. 266 */ 267 return ip6_output(m, NULL, NULL, 0, NULL, NULL, NULL); 268 #endif /* INET6 */ 269 } 270 panic("ipsec_process_done"); 271 bad: 272 m_freem(m); 273 return (error); 274 } 275 276 static struct ipsecrequest * 277 ipsec_nextisr( 278 struct mbuf *m, 279 struct ipsecrequest *isr, 280 int af, 281 struct secasindex *saidx, 282 int *error 283 ) 284 { 285 #define IPSEC_OSTAT(x,y,z) (isr->saidx.proto == IPPROTO_ESP ? (x)++ : \ 286 isr->saidx.proto == IPPROTO_AH ? (y)++ : (z)++) 287 struct secasvar *sav; 288 289 IPSECREQUEST_LOCK_ASSERT(isr); 290 291 IPSEC_ASSERT(af == AF_INET || af == AF_INET6, 292 ("invalid address family %u", af)); 293 again: 294 /* 295 * Craft SA index to search for proper SA. Note that 296 * we only fillin unspecified SA peers for transport 297 * mode; for tunnel mode they must already be filled in. 298 */ 299 *saidx = isr->saidx; 300 if (isr->saidx.mode == IPSEC_MODE_TRANSPORT) { 301 /* Fillin unspecified SA peers only for transport mode */ 302 if (af == AF_INET) { 303 struct sockaddr_in *sin; 304 struct ip *ip = mtod(m, struct ip *); 305 306 if (saidx->src.sa.sa_len == 0) { 307 sin = &saidx->src.sin; 308 sin->sin_len = sizeof(*sin); 309 sin->sin_family = AF_INET; 310 sin->sin_port = IPSEC_PORT_ANY; 311 sin->sin_addr = ip->ip_src; 312 } 313 if (saidx->dst.sa.sa_len == 0) { 314 sin = &saidx->dst.sin; 315 sin->sin_len = sizeof(*sin); 316 sin->sin_family = AF_INET; 317 sin->sin_port = IPSEC_PORT_ANY; 318 sin->sin_addr = ip->ip_dst; 319 } 320 } else { 321 struct sockaddr_in6 *sin6; 322 struct ip6_hdr *ip6 = mtod(m, struct ip6_hdr *); 323 324 if (saidx->src.sin6.sin6_len == 0) { 325 sin6 = (struct sockaddr_in6 *)&saidx->src; 326 sin6->sin6_len = sizeof(*sin6); 327 sin6->sin6_family = AF_INET6; 328 sin6->sin6_port = IPSEC_PORT_ANY; 329 sin6->sin6_addr = ip6->ip6_src; 330 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_src)) { 331 /* fix scope id for comparing SPD */ 332 sin6->sin6_addr.s6_addr16[1] = 0; 333 sin6->sin6_scope_id = 334 ntohs(ip6->ip6_src.s6_addr16[1]); 335 } 336 } 337 if (saidx->dst.sin6.sin6_len == 0) { 338 sin6 = (struct sockaddr_in6 *)&saidx->dst; 339 sin6->sin6_len = sizeof(*sin6); 340 sin6->sin6_family = AF_INET6; 341 sin6->sin6_port = IPSEC_PORT_ANY; 342 sin6->sin6_addr = ip6->ip6_dst; 343 if (IN6_IS_SCOPE_LINKLOCAL(&ip6->ip6_dst)) { 344 /* fix scope id for comparing SPD */ 345 sin6->sin6_addr.s6_addr16[1] = 0; 346 sin6->sin6_scope_id = 347 ntohs(ip6->ip6_dst.s6_addr16[1]); 348 } 349 } 350 } 351 } 352 353 /* 354 * Lookup SA and validate it. 355 */ 356 *error = key_checkrequest(isr, saidx); 357 if (*error != 0) { 358 /* 359 * IPsec processing is required, but no SA found. 360 * I assume that key_acquire() had been called 361 * to get/establish the SA. Here I discard 362 * this packet because it is responsibility for 363 * upper layer to retransmit the packet. 364 */ 365 V_ipsec4stat.ips_out_nosa++; 366 goto bad; 367 } 368 sav = isr->sav; 369 if (sav == NULL) { 370 IPSEC_ASSERT(ipsec_get_reqlevel(isr) == IPSEC_LEVEL_USE, 371 ("no SA found, but required; level %u", 372 ipsec_get_reqlevel(isr))); 373 IPSECREQUEST_UNLOCK(isr); 374 isr = isr->next; 375 /* 376 * If isr is NULL, we found a 'use' policy w/o SA. 377 * Return w/o error and w/o isr so we can drop out 378 * and continue w/o IPsec processing. 379 */ 380 if (isr == NULL) 381 return isr; 382 IPSECREQUEST_LOCK(isr); 383 goto again; 384 } 385 386 /* 387 * Check system global policy controls. 388 */ 389 if ((isr->saidx.proto == IPPROTO_ESP && !V_esp_enable) || 390 (isr->saidx.proto == IPPROTO_AH && !V_ah_enable) || 391 (isr->saidx.proto == IPPROTO_IPCOMP && !V_ipcomp_enable)) { 392 DPRINTF(("%s: IPsec outbound packet dropped due" 393 " to policy (check your sysctls)\n", __func__)); 394 IPSEC_OSTAT(V_espstat.esps_pdrops, V_ahstat.ahs_pdrops, 395 V_ipcompstat.ipcomps_pdrops); 396 *error = EHOSTUNREACH; 397 goto bad; 398 } 399 400 /* 401 * Sanity check the SA contents for the caller 402 * before they invoke the xform output method. 403 */ 404 if (sav->tdb_xform == NULL) { 405 DPRINTF(("%s: no transform for SA\n", __func__)); 406 IPSEC_OSTAT(V_espstat.esps_noxform, V_ahstat.ahs_noxform, 407 V_ipcompstat.ipcomps_noxform); 408 *error = EHOSTUNREACH; 409 goto bad; 410 } 411 return isr; 412 bad: 413 IPSEC_ASSERT(*error != 0, ("error return w/ no error code")); 414 IPSECREQUEST_UNLOCK(isr); 415 return NULL; 416 #undef IPSEC_OSTAT 417 } 418 419 #ifdef INET 420 /* 421 * IPsec output logic for IPv4. 422 */ 423 int 424 ipsec4_process_packet( 425 struct mbuf *m, 426 struct ipsecrequest *isr, 427 int flags, 428 int tunalready) 429 { 430 struct secasindex saidx; 431 struct secasvar *sav; 432 struct ip *ip; 433 int error, i, off; 434 435 IPSEC_ASSERT(m != NULL, ("null mbuf")); 436 IPSEC_ASSERT(isr != NULL, ("null isr")); 437 438 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */ 439 440 isr = ipsec_nextisr(m, isr, AF_INET, &saidx, &error); 441 if (isr == NULL) { 442 if (error != 0) 443 goto bad; 444 return EJUSTRETURN; 445 } 446 447 sav = isr->sav; 448 449 #ifdef DEV_ENC 450 encif->if_opackets++; 451 encif->if_obytes += m->m_pkthdr.len; 452 453 /* pass the mbuf to enc0 for bpf processing */ 454 ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_BEFORE); 455 /* pass the mbuf to enc0 for packet filtering */ 456 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0) 457 goto bad; 458 #endif 459 460 if (!tunalready) { 461 union sockaddr_union *dst = &sav->sah->saidx.dst; 462 int setdf; 463 464 /* 465 * Collect IP_DF state from the outer header. 466 */ 467 if (dst->sa.sa_family == AF_INET) { 468 if (m->m_len < sizeof (struct ip) && 469 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 470 error = ENOBUFS; 471 goto bad; 472 } 473 ip = mtod(m, struct ip *); 474 /* Honor system-wide control of how to handle IP_DF */ 475 switch (V_ip4_ipsec_dfbit) { 476 case 0: /* clear in outer header */ 477 case 1: /* set in outer header */ 478 setdf = V_ip4_ipsec_dfbit; 479 break; 480 default: /* propagate to outer header */ 481 setdf = ntohs(ip->ip_off & IP_DF); 482 break; 483 } 484 } else { 485 ip = NULL; /* keep compiler happy */ 486 setdf = 0; 487 } 488 /* Do the appropriate encapsulation, if necessary */ 489 if (isr->saidx.mode == IPSEC_MODE_TUNNEL || /* Tunnel requ'd */ 490 dst->sa.sa_family != AF_INET || /* PF mismatch */ 491 #if 0 492 (sav->flags & SADB_X_SAFLAGS_TUNNEL) || /* Tunnel requ'd */ 493 sav->tdb_xform->xf_type == XF_IP4 || /* ditto */ 494 #endif 495 (dst->sa.sa_family == AF_INET && /* Proxy */ 496 dst->sin.sin_addr.s_addr != INADDR_ANY && 497 dst->sin.sin_addr.s_addr != ip->ip_dst.s_addr)) { 498 struct mbuf *mp; 499 500 /* Fix IPv4 header checksum and length */ 501 if (m->m_len < sizeof (struct ip) && 502 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 503 error = ENOBUFS; 504 goto bad; 505 } 506 ip = mtod(m, struct ip *); 507 ip->ip_len = htons(m->m_pkthdr.len); 508 ip->ip_sum = 0; 509 #ifdef _IP_VHL 510 if (ip->ip_vhl == IP_VHL_BORING) 511 ip->ip_sum = in_cksum_hdr(ip); 512 else 513 ip->ip_sum = in_cksum(m, 514 _IP_VHL_HL(ip->ip_vhl) << 2); 515 #else 516 ip->ip_sum = in_cksum(m, ip->ip_hl << 2); 517 #endif 518 519 /* Encapsulate the packet */ 520 error = ipip_output(m, isr, &mp, 0, 0); 521 if (mp == NULL && !error) { 522 /* Should never happen. */ 523 DPRINTF(("%s: ipip_output returns no mbuf and " 524 "no error!", __func__)); 525 error = EFAULT; 526 } 527 if (error) { 528 if (mp) { 529 /* XXX: Should never happen! */ 530 m_freem(mp); 531 } 532 m = NULL; /* ipip_output() already freed it */ 533 goto bad; 534 } 535 m = mp, mp = NULL; 536 /* 537 * ipip_output clears IP_DF in the new header. If 538 * we need to propagate IP_DF from the outer header, 539 * then we have to do it here. 540 * 541 * XXX shouldn't assume what ipip_output does. 542 */ 543 if (dst->sa.sa_family == AF_INET && setdf) { 544 if (m->m_len < sizeof (struct ip) && 545 (m = m_pullup(m, sizeof (struct ip))) == NULL) { 546 error = ENOBUFS; 547 goto bad; 548 } 549 ip = mtod(m, struct ip *); 550 ip->ip_off = ntohs(ip->ip_off); 551 ip->ip_off |= IP_DF; 552 ip->ip_off = htons(ip->ip_off); 553 } 554 } 555 } 556 557 #ifdef DEV_ENC 558 /* pass the mbuf to enc0 for bpf processing */ 559 ipsec_bpf(m, sav, AF_INET, ENC_OUT|ENC_AFTER); 560 /* pass the mbuf to enc0 for packet filtering */ 561 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0) 562 goto bad; 563 #endif 564 565 /* 566 * Dispatch to the appropriate IPsec transform logic. The 567 * packet will be returned for transmission after crypto 568 * processing, etc. are completed. For encapsulation we 569 * bypass this call because of the explicit call done above 570 * (necessary to deal with IP_DF handling for IPv4). 571 * 572 * NB: m & sav are ``passed to caller'' who's reponsible for 573 * for reclaiming their resources. 574 */ 575 if (sav->tdb_xform->xf_type != XF_IP4) { 576 ip = mtod(m, struct ip *); 577 i = ip->ip_hl << 2; 578 off = offsetof(struct ip, ip_p); 579 error = (*sav->tdb_xform->xf_output)(m, isr, NULL, i, off); 580 } else { 581 error = ipsec_process_done(m, isr); 582 } 583 IPSECREQUEST_UNLOCK(isr); 584 return error; 585 bad: 586 if (isr) 587 IPSECREQUEST_UNLOCK(isr); 588 if (m) 589 m_freem(m); 590 return error; 591 } 592 #endif 593 594 #ifdef INET6 595 /* 596 * Chop IP6 header from the payload. 597 */ 598 static struct mbuf * 599 ipsec6_splithdr(struct mbuf *m) 600 { 601 struct mbuf *mh; 602 struct ip6_hdr *ip6; 603 int hlen; 604 605 IPSEC_ASSERT(m->m_len >= sizeof (struct ip6_hdr), 606 ("first mbuf too short, len %u", m->m_len)); 607 ip6 = mtod(m, struct ip6_hdr *); 608 hlen = sizeof(struct ip6_hdr); 609 if (m->m_len > hlen) { 610 MGETHDR(mh, M_DONTWAIT, MT_DATA); 611 if (!mh) { 612 m_freem(m); 613 return NULL; 614 } 615 M_MOVE_PKTHDR(mh, m); 616 MH_ALIGN(mh, hlen); 617 m->m_len -= hlen; 618 m->m_data += hlen; 619 mh->m_next = m; 620 m = mh; 621 m->m_len = hlen; 622 bcopy((caddr_t)ip6, mtod(m, caddr_t), hlen); 623 } else if (m->m_len < hlen) { 624 m = m_pullup(m, hlen); 625 if (!m) 626 return NULL; 627 } 628 return m; 629 } 630 631 /* 632 * IPsec output logic for IPv6, transport mode. 633 */ 634 int 635 ipsec6_output_trans( 636 struct ipsec_output_state *state, 637 u_char *nexthdrp, 638 struct mbuf *mprev, 639 struct secpolicy *sp, 640 int flags, 641 int *tun) 642 { 643 struct ipsecrequest *isr; 644 struct secasindex saidx; 645 int error = 0; 646 struct mbuf *m; 647 648 IPSEC_ASSERT(state != NULL, ("null state")); 649 IPSEC_ASSERT(state->m != NULL, ("null m")); 650 IPSEC_ASSERT(nexthdrp != NULL, ("null nexthdrp")); 651 IPSEC_ASSERT(mprev != NULL, ("null mprev")); 652 IPSEC_ASSERT(sp != NULL, ("null sp")); 653 IPSEC_ASSERT(tun != NULL, ("null tun")); 654 655 KEYDEBUG(KEYDEBUG_IPSEC_DATA, 656 printf("%s: applied SP\n", __func__); 657 kdebug_secpolicy(sp)); 658 659 isr = sp->req; 660 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) { 661 /* the rest will be handled by ipsec6_output_tunnel() */ 662 *tun = 1; /* need tunnel-mode processing */ 663 return 0; 664 } 665 666 *tun = 0; 667 m = state->m; 668 669 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */ 670 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error); 671 if (isr == NULL) { 672 if (error != 0) { 673 #ifdef notdef 674 /* XXX should notification be done for all errors ? */ 675 /* 676 * Notify the fact that the packet is discarded 677 * to ourselves. I believe this is better than 678 * just silently discarding. (jinmei@kame.net) 679 * XXX: should we restrict the error to TCP packets? 680 * XXX: should we directly notify sockets via 681 * pfctlinputs? 682 */ 683 icmp6_error(m, ICMP6_DST_UNREACH, 684 ICMP6_DST_UNREACH_ADMIN, 0); 685 m = NULL; /* NB: icmp6_error frees mbuf */ 686 #endif 687 goto bad; 688 } 689 return EJUSTRETURN; 690 } 691 692 error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL, 693 sizeof (struct ip6_hdr), 694 offsetof(struct ip6_hdr, 695 ip6_nxt)); 696 IPSECREQUEST_UNLOCK(isr); 697 return error; 698 bad: 699 if (isr) 700 IPSECREQUEST_UNLOCK(isr); 701 if (m) 702 m_freem(m); 703 state->m = NULL; 704 return error; 705 } 706 707 static int 708 ipsec6_encapsulate(struct mbuf *m, struct secasvar *sav) 709 { 710 struct ip6_hdr *oip6; 711 struct ip6_hdr *ip6; 712 size_t plen; 713 714 /* can't tunnel between different AFs */ 715 if (sav->sah->saidx.src.sa.sa_family != AF_INET6 || 716 sav->sah->saidx.dst.sa.sa_family != AF_INET6) { 717 m_freem(m); 718 return EINVAL; 719 } 720 IPSEC_ASSERT(m->m_len == sizeof (struct ip6_hdr), 721 ("mbuf wrong size; len %u", m->m_len)); 722 723 724 /* 725 * grow the mbuf to accomodate the new IPv6 header. 726 */ 727 plen = m->m_pkthdr.len; 728 if (M_LEADINGSPACE(m->m_next) < sizeof(struct ip6_hdr)) { 729 struct mbuf *n; 730 MGET(n, M_DONTWAIT, MT_DATA); 731 if (!n) { 732 m_freem(m); 733 return ENOBUFS; 734 } 735 n->m_len = sizeof(struct ip6_hdr); 736 n->m_next = m->m_next; 737 m->m_next = n; 738 m->m_pkthdr.len += sizeof(struct ip6_hdr); 739 oip6 = mtod(n, struct ip6_hdr *); 740 } else { 741 m->m_next->m_len += sizeof(struct ip6_hdr); 742 m->m_next->m_data -= sizeof(struct ip6_hdr); 743 m->m_pkthdr.len += sizeof(struct ip6_hdr); 744 oip6 = mtod(m->m_next, struct ip6_hdr *); 745 } 746 ip6 = mtod(m, struct ip6_hdr *); 747 bcopy((caddr_t)ip6, (caddr_t)oip6, sizeof(struct ip6_hdr)); 748 749 /* Fake link-local scope-class addresses */ 750 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_src)) 751 oip6->ip6_src.s6_addr16[1] = 0; 752 if (IN6_IS_SCOPE_LINKLOCAL(&oip6->ip6_dst)) 753 oip6->ip6_dst.s6_addr16[1] = 0; 754 755 /* construct new IPv6 header. see RFC 2401 5.1.2.2 */ 756 /* ECN consideration. */ 757 ip6_ecn_ingress(V_ip6_ipsec_ecn, &ip6->ip6_flow, &oip6->ip6_flow); 758 if (plen < IPV6_MAXPACKET - sizeof(struct ip6_hdr)) 759 ip6->ip6_plen = htons(plen); 760 else { 761 /* ip6->ip6_plen will be updated in ip6_output() */ 762 } 763 ip6->ip6_nxt = IPPROTO_IPV6; 764 ip6->ip6_src = sav->sah->saidx.src.sin6.sin6_addr; 765 ip6->ip6_dst = sav->sah->saidx.dst.sin6.sin6_addr; 766 ip6->ip6_hlim = IPV6_DEFHLIM; 767 768 /* XXX Should ip6_src be updated later ? */ 769 770 return 0; 771 } 772 773 /* 774 * IPsec output logic for IPv6, tunnel mode. 775 */ 776 int 777 ipsec6_output_tunnel(struct ipsec_output_state *state, struct secpolicy *sp, int flags) 778 { 779 struct ip6_hdr *ip6; 780 struct ipsecrequest *isr; 781 struct secasindex saidx; 782 int error; 783 struct sockaddr_in6 *dst6; 784 struct mbuf *m; 785 786 IPSEC_ASSERT(state != NULL, ("null state")); 787 IPSEC_ASSERT(state->m != NULL, ("null m")); 788 IPSEC_ASSERT(sp != NULL, ("null sp")); 789 790 KEYDEBUG(KEYDEBUG_IPSEC_DATA, 791 printf("%s: applied SP\n", __func__); 792 kdebug_secpolicy(sp)); 793 794 m = state->m; 795 /* 796 * transport mode ipsec (before the 1st tunnel mode) is already 797 * processed by ipsec6_output_trans(). 798 */ 799 for (isr = sp->req; isr; isr = isr->next) { 800 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) 801 break; 802 } 803 804 IPSECREQUEST_LOCK(isr); /* insure SA contents don't change */ 805 isr = ipsec_nextisr(m, isr, AF_INET6, &saidx, &error); 806 if (isr == NULL) { 807 if (error != 0) 808 goto bad; 809 return EJUSTRETURN; 810 } 811 812 #ifdef DEV_ENC 813 encif->if_opackets++; 814 encif->if_obytes += m->m_pkthdr.len; 815 816 /* pass the mbuf to enc0 for bpf processing */ 817 ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_BEFORE); 818 /* pass the mbuf to enc0 for packet filtering */ 819 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_BEFORE)) != 0) 820 goto bad; 821 #endif 822 823 /* 824 * There may be the case that SA status will be changed when 825 * we are refering to one. So calling splsoftnet(). 826 */ 827 if (isr->saidx.mode == IPSEC_MODE_TUNNEL) { 828 /* 829 * build IPsec tunnel. 830 */ 831 /* XXX should be processed with other familiy */ 832 if (isr->sav->sah->saidx.src.sa.sa_family != AF_INET6) { 833 ipseclog((LOG_ERR, "%s: family mismatched between " 834 "inner and outer, spi=%u\n", __func__, 835 ntohl(isr->sav->spi))); 836 V_ipsec6stat.ips_out_inval++; 837 error = EAFNOSUPPORT; 838 goto bad; 839 } 840 841 m = ipsec6_splithdr(m); 842 if (!m) { 843 V_ipsec6stat.ips_out_nomem++; 844 error = ENOMEM; 845 goto bad; 846 } 847 error = ipsec6_encapsulate(m, isr->sav); 848 if (error) { 849 m = NULL; 850 goto bad; 851 } 852 ip6 = mtod(m, struct ip6_hdr *); 853 854 state->ro = 855 (struct route *)&isr->sav->sah->route_cache.sin6_route; 856 state->dst = (struct sockaddr *)&state->ro->ro_dst; 857 dst6 = (struct sockaddr_in6 *)state->dst; 858 if (state->ro->ro_rt 859 && ((state->ro->ro_rt->rt_flags & RTF_UP) == 0 860 || !IN6_ARE_ADDR_EQUAL(&dst6->sin6_addr, &ip6->ip6_dst))) { 861 RTFREE(state->ro->ro_rt); 862 state->ro->ro_rt = NULL; 863 } 864 if (state->ro->ro_rt == NULL) { 865 bzero(dst6, sizeof(*dst6)); 866 dst6->sin6_family = AF_INET6; 867 dst6->sin6_len = sizeof(*dst6); 868 dst6->sin6_addr = ip6->ip6_dst; 869 rtalloc_ign_fib(state->ro, 0UL, M_GETFIB(m)); 870 } 871 if (state->ro->ro_rt == NULL) { 872 V_ip6stat.ip6s_noroute++; 873 V_ipsec6stat.ips_out_noroute++; 874 error = EHOSTUNREACH; 875 goto bad; 876 } 877 878 /* adjust state->dst if tunnel endpoint is offlink */ 879 if (state->ro->ro_rt->rt_flags & RTF_GATEWAY) 880 state->dst = (struct sockaddr *)state->ro->ro_rt->rt_gateway; 881 } 882 883 m = ipsec6_splithdr(m); 884 if (!m) { 885 V_ipsec6stat.ips_out_nomem++; 886 error = ENOMEM; 887 goto bad; 888 } 889 ip6 = mtod(m, struct ip6_hdr *); 890 891 #ifdef DEV_ENC 892 /* pass the mbuf to enc0 for bpf processing */ 893 ipsec_bpf(m, isr->sav, AF_INET6, ENC_OUT|ENC_AFTER); 894 /* pass the mbuf to enc0 for packet filtering */ 895 if ((error = ipsec_filter(&m, PFIL_OUT, ENC_OUT|ENC_AFTER)) != 0) 896 goto bad; 897 #endif 898 899 error = (*isr->sav->tdb_xform->xf_output)(m, isr, NULL, 900 sizeof (struct ip6_hdr), 901 offsetof(struct ip6_hdr, ip6_nxt)); 902 IPSECREQUEST_UNLOCK(isr); 903 return error; 904 bad: 905 if (isr) 906 IPSECREQUEST_UNLOCK(isr); 907 if (m) 908 m_freem(m); 909 state->m = NULL; 910 return error; 911 } 912 #endif /*INET6*/ 913