1 /* SCTP kernel implementation 2 * Copyright (c) 1999-2000 Cisco, Inc. 3 * Copyright (c) 1999-2001 Motorola, Inc. 4 * Copyright (c) 2001-2003 International Business Machines, Corp. 5 * Copyright (c) 2001 Intel Corp. 6 * Copyright (c) 2001 Nokia, Inc. 7 * Copyright (c) 2001 La Monte H.P. Yarroll 8 * 9 * This file is part of the SCTP kernel implementation 10 * 11 * These functions handle all input from the IP layer into SCTP. 12 * 13 * This SCTP implementation is free software; 14 * you can redistribute it and/or modify it under the terms of 15 * the GNU General Public License as published by 16 * the Free Software Foundation; either version 2, or (at your option) 17 * any later version. 18 * 19 * This SCTP implementation is distributed in the hope that it 20 * will be useful, but WITHOUT ANY WARRANTY; without even the implied 21 * ************************ 22 * warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. 23 * See the GNU General Public License for more details. 24 * 25 * You should have received a copy of the GNU General Public License 26 * along with GNU CC; see the file COPYING. If not, see 27 * <http://www.gnu.org/licenses/>. 28 * 29 * Please send any bug reports or fixes you make to the 30 * email address(es): 31 * lksctp developers <linux-sctp@vger.kernel.org> 32 * 33 * Written or modified by: 34 * La Monte H.P. Yarroll <piggy@acm.org> 35 * Karl Knutson <karl@athena.chicago.il.us> 36 * Xingang Guo <xingang.guo@intel.com> 37 * Jon Grimm <jgrimm@us.ibm.com> 38 * Hui Huang <hui.huang@nokia.com> 39 * Daisy Chang <daisyc@us.ibm.com> 40 * Sridhar Samudrala <sri@us.ibm.com> 41 * Ardelle Fan <ardelle.fan@intel.com> 42 */ 43 44 #include <linux/types.h> 45 #include <linux/list.h> /* For struct list_head */ 46 #include <linux/socket.h> 47 #include <linux/ip.h> 48 #include <linux/time.h> /* For struct timeval */ 49 #include <linux/slab.h> 50 #include <net/ip.h> 51 #include <net/icmp.h> 52 #include <net/snmp.h> 53 #include <net/sock.h> 54 #include <net/xfrm.h> 55 #include <net/sctp/sctp.h> 56 #include <net/sctp/sm.h> 57 #include <net/sctp/checksum.h> 58 #include <net/net_namespace.h> 59 60 /* Forward declarations for internal helpers. */ 61 static int sctp_rcv_ootb(struct sk_buff *); 62 static struct sctp_association *__sctp_rcv_lookup(struct net *net, 63 struct sk_buff *skb, 64 const union sctp_addr *paddr, 65 const union sctp_addr *laddr, 66 struct sctp_transport **transportp); 67 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(struct net *net, 68 const union sctp_addr *laddr); 69 static struct sctp_association *__sctp_lookup_association( 70 struct net *net, 71 const union sctp_addr *local, 72 const union sctp_addr *peer, 73 struct sctp_transport **pt); 74 75 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb); 76 77 78 /* Calculate the SCTP checksum of an SCTP packet. */ 79 static inline int sctp_rcv_checksum(struct net *net, struct sk_buff *skb) 80 { 81 struct sctphdr *sh = sctp_hdr(skb); 82 __le32 cmp = sh->checksum; 83 __le32 val = sctp_compute_cksum(skb, 0); 84 85 if (val != cmp) { 86 /* CRC failure, dump it. */ 87 __SCTP_INC_STATS(net, SCTP_MIB_CHECKSUMERRORS); 88 return -1; 89 } 90 return 0; 91 } 92 93 /* 94 * This is the routine which IP calls when receiving an SCTP packet. 95 */ 96 int sctp_rcv(struct sk_buff *skb) 97 { 98 struct sock *sk; 99 struct sctp_association *asoc; 100 struct sctp_endpoint *ep = NULL; 101 struct sctp_ep_common *rcvr; 102 struct sctp_transport *transport = NULL; 103 struct sctp_chunk *chunk; 104 union sctp_addr src; 105 union sctp_addr dest; 106 int family; 107 struct sctp_af *af; 108 struct net *net = dev_net(skb->dev); 109 110 if (skb->pkt_type != PACKET_HOST) 111 goto discard_it; 112 113 __SCTP_INC_STATS(net, SCTP_MIB_INSCTPPACKS); 114 115 /* If packet is too small to contain a single chunk, let's not 116 * waste time on it anymore. 117 */ 118 if (skb->len < sizeof(struct sctphdr) + sizeof(struct sctp_chunkhdr) + 119 skb_transport_offset(skb)) 120 goto discard_it; 121 122 /* If the packet is fragmented and we need to do crc checking, 123 * it's better to just linearize it otherwise crc computing 124 * takes longer. 125 */ 126 if ((!(skb_shinfo(skb)->gso_type & SKB_GSO_SCTP) && 127 skb_linearize(skb)) || 128 !pskb_may_pull(skb, sizeof(struct sctphdr))) 129 goto discard_it; 130 131 /* Pull up the IP header. */ 132 __skb_pull(skb, skb_transport_offset(skb)); 133 134 skb->csum_valid = 0; /* Previous value not applicable */ 135 if (skb_csum_unnecessary(skb)) 136 __skb_decr_checksum_unnecessary(skb); 137 else if (!sctp_checksum_disable && 138 !(skb_shinfo(skb)->gso_type & SKB_GSO_SCTP) && 139 sctp_rcv_checksum(net, skb) < 0) 140 goto discard_it; 141 skb->csum_valid = 1; 142 143 __skb_pull(skb, sizeof(struct sctphdr)); 144 145 family = ipver2af(ip_hdr(skb)->version); 146 af = sctp_get_af_specific(family); 147 if (unlikely(!af)) 148 goto discard_it; 149 SCTP_INPUT_CB(skb)->af = af; 150 151 /* Initialize local addresses for lookups. */ 152 af->from_skb(&src, skb, 1); 153 af->from_skb(&dest, skb, 0); 154 155 /* If the packet is to or from a non-unicast address, 156 * silently discard the packet. 157 * 158 * This is not clearly defined in the RFC except in section 159 * 8.4 - OOTB handling. However, based on the book "Stream Control 160 * Transmission Protocol" 2.1, "It is important to note that the 161 * IP address of an SCTP transport address must be a routable 162 * unicast address. In other words, IP multicast addresses and 163 * IP broadcast addresses cannot be used in an SCTP transport 164 * address." 165 */ 166 if (!af->addr_valid(&src, NULL, skb) || 167 !af->addr_valid(&dest, NULL, skb)) 168 goto discard_it; 169 170 asoc = __sctp_rcv_lookup(net, skb, &src, &dest, &transport); 171 172 if (!asoc) 173 ep = __sctp_rcv_lookup_endpoint(net, &dest); 174 175 /* Retrieve the common input handling substructure. */ 176 rcvr = asoc ? &asoc->base : &ep->base; 177 sk = rcvr->sk; 178 179 /* 180 * If a frame arrives on an interface and the receiving socket is 181 * bound to another interface, via SO_BINDTODEVICE, treat it as OOTB 182 */ 183 if (sk->sk_bound_dev_if && (sk->sk_bound_dev_if != af->skb_iif(skb))) { 184 if (transport) { 185 sctp_transport_put(transport); 186 asoc = NULL; 187 transport = NULL; 188 } else { 189 sctp_endpoint_put(ep); 190 ep = NULL; 191 } 192 sk = net->sctp.ctl_sock; 193 ep = sctp_sk(sk)->ep; 194 sctp_endpoint_hold(ep); 195 rcvr = &ep->base; 196 } 197 198 /* 199 * RFC 2960, 8.4 - Handle "Out of the blue" Packets. 200 * An SCTP packet is called an "out of the blue" (OOTB) 201 * packet if it is correctly formed, i.e., passed the 202 * receiver's checksum check, but the receiver is not 203 * able to identify the association to which this 204 * packet belongs. 205 */ 206 if (!asoc) { 207 if (sctp_rcv_ootb(skb)) { 208 __SCTP_INC_STATS(net, SCTP_MIB_OUTOFBLUES); 209 goto discard_release; 210 } 211 } 212 213 if (!xfrm_policy_check(sk, XFRM_POLICY_IN, skb, family)) 214 goto discard_release; 215 nf_reset(skb); 216 217 if (sk_filter(sk, skb)) 218 goto discard_release; 219 220 /* Create an SCTP packet structure. */ 221 chunk = sctp_chunkify(skb, asoc, sk, GFP_ATOMIC); 222 if (!chunk) 223 goto discard_release; 224 SCTP_INPUT_CB(skb)->chunk = chunk; 225 226 /* Remember what endpoint is to handle this packet. */ 227 chunk->rcvr = rcvr; 228 229 /* Remember the SCTP header. */ 230 chunk->sctp_hdr = sctp_hdr(skb); 231 232 /* Set the source and destination addresses of the incoming chunk. */ 233 sctp_init_addrs(chunk, &src, &dest); 234 235 /* Remember where we came from. */ 236 chunk->transport = transport; 237 238 /* Acquire access to the sock lock. Note: We are safe from other 239 * bottom halves on this lock, but a user may be in the lock too, 240 * so check if it is busy. 241 */ 242 bh_lock_sock(sk); 243 244 if (sk != rcvr->sk) { 245 /* Our cached sk is different from the rcvr->sk. This is 246 * because migrate()/accept() may have moved the association 247 * to a new socket and released all the sockets. So now we 248 * are holding a lock on the old socket while the user may 249 * be doing something with the new socket. Switch our veiw 250 * of the current sk. 251 */ 252 bh_unlock_sock(sk); 253 sk = rcvr->sk; 254 bh_lock_sock(sk); 255 } 256 257 if (sock_owned_by_user(sk)) { 258 if (sctp_add_backlog(sk, skb)) { 259 bh_unlock_sock(sk); 260 sctp_chunk_free(chunk); 261 skb = NULL; /* sctp_chunk_free already freed the skb */ 262 goto discard_release; 263 } 264 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_BACKLOG); 265 } else { 266 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_SOFTIRQ); 267 sctp_inq_push(&chunk->rcvr->inqueue, chunk); 268 } 269 270 bh_unlock_sock(sk); 271 272 /* Release the asoc/ep ref we took in the lookup calls. */ 273 if (transport) 274 sctp_transport_put(transport); 275 else 276 sctp_endpoint_put(ep); 277 278 return 0; 279 280 discard_it: 281 __SCTP_INC_STATS(net, SCTP_MIB_IN_PKT_DISCARDS); 282 kfree_skb(skb); 283 return 0; 284 285 discard_release: 286 /* Release the asoc/ep ref we took in the lookup calls. */ 287 if (transport) 288 sctp_transport_put(transport); 289 else 290 sctp_endpoint_put(ep); 291 292 goto discard_it; 293 } 294 295 /* Process the backlog queue of the socket. Every skb on 296 * the backlog holds a ref on an association or endpoint. 297 * We hold this ref throughout the state machine to make 298 * sure that the structure we need is still around. 299 */ 300 int sctp_backlog_rcv(struct sock *sk, struct sk_buff *skb) 301 { 302 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk; 303 struct sctp_inq *inqueue = &chunk->rcvr->inqueue; 304 struct sctp_transport *t = chunk->transport; 305 struct sctp_ep_common *rcvr = NULL; 306 int backloged = 0; 307 308 rcvr = chunk->rcvr; 309 310 /* If the rcvr is dead then the association or endpoint 311 * has been deleted and we can safely drop the chunk 312 * and refs that we are holding. 313 */ 314 if (rcvr->dead) { 315 sctp_chunk_free(chunk); 316 goto done; 317 } 318 319 if (unlikely(rcvr->sk != sk)) { 320 /* In this case, the association moved from one socket to 321 * another. We are currently sitting on the backlog of the 322 * old socket, so we need to move. 323 * However, since we are here in the process context we 324 * need to take make sure that the user doesn't own 325 * the new socket when we process the packet. 326 * If the new socket is user-owned, queue the chunk to the 327 * backlog of the new socket without dropping any refs. 328 * Otherwise, we can safely push the chunk on the inqueue. 329 */ 330 331 sk = rcvr->sk; 332 local_bh_disable(); 333 bh_lock_sock(sk); 334 335 if (sock_owned_by_user(sk)) { 336 if (sk_add_backlog(sk, skb, sk->sk_rcvbuf)) 337 sctp_chunk_free(chunk); 338 else 339 backloged = 1; 340 } else 341 sctp_inq_push(inqueue, chunk); 342 343 bh_unlock_sock(sk); 344 local_bh_enable(); 345 346 /* If the chunk was backloged again, don't drop refs */ 347 if (backloged) 348 return 0; 349 } else { 350 sctp_inq_push(inqueue, chunk); 351 } 352 353 done: 354 /* Release the refs we took in sctp_add_backlog */ 355 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type) 356 sctp_transport_put(t); 357 else if (SCTP_EP_TYPE_SOCKET == rcvr->type) 358 sctp_endpoint_put(sctp_ep(rcvr)); 359 else 360 BUG(); 361 362 return 0; 363 } 364 365 static int sctp_add_backlog(struct sock *sk, struct sk_buff *skb) 366 { 367 struct sctp_chunk *chunk = SCTP_INPUT_CB(skb)->chunk; 368 struct sctp_transport *t = chunk->transport; 369 struct sctp_ep_common *rcvr = chunk->rcvr; 370 int ret; 371 372 ret = sk_add_backlog(sk, skb, sk->sk_rcvbuf); 373 if (!ret) { 374 /* Hold the assoc/ep while hanging on the backlog queue. 375 * This way, we know structures we need will not disappear 376 * from us 377 */ 378 if (SCTP_EP_TYPE_ASSOCIATION == rcvr->type) 379 sctp_transport_hold(t); 380 else if (SCTP_EP_TYPE_SOCKET == rcvr->type) 381 sctp_endpoint_hold(sctp_ep(rcvr)); 382 else 383 BUG(); 384 } 385 return ret; 386 387 } 388 389 /* Handle icmp frag needed error. */ 390 void sctp_icmp_frag_needed(struct sock *sk, struct sctp_association *asoc, 391 struct sctp_transport *t, __u32 pmtu) 392 { 393 if (!t || (t->pathmtu <= pmtu)) 394 return; 395 396 if (sock_owned_by_user(sk)) { 397 asoc->pmtu_pending = 1; 398 t->pmtu_pending = 1; 399 return; 400 } 401 402 if (t->param_flags & SPP_PMTUD_ENABLE) { 403 /* Update transports view of the MTU */ 404 sctp_transport_update_pmtu(t, pmtu); 405 406 /* Update association pmtu. */ 407 sctp_assoc_sync_pmtu(asoc); 408 } 409 410 /* Retransmit with the new pmtu setting. 411 * Normally, if PMTU discovery is disabled, an ICMP Fragmentation 412 * Needed will never be sent, but if a message was sent before 413 * PMTU discovery was disabled that was larger than the PMTU, it 414 * would not be fragmented, so it must be re-transmitted fragmented. 415 */ 416 sctp_retransmit(&asoc->outqueue, t, SCTP_RTXR_PMTUD); 417 } 418 419 void sctp_icmp_redirect(struct sock *sk, struct sctp_transport *t, 420 struct sk_buff *skb) 421 { 422 struct dst_entry *dst; 423 424 if (!t) 425 return; 426 dst = sctp_transport_dst_check(t); 427 if (dst) 428 dst->ops->redirect(dst, sk, skb); 429 } 430 431 /* 432 * SCTP Implementer's Guide, 2.37 ICMP handling procedures 433 * 434 * ICMP8) If the ICMP code is a "Unrecognized next header type encountered" 435 * or a "Protocol Unreachable" treat this message as an abort 436 * with the T bit set. 437 * 438 * This function sends an event to the state machine, which will abort the 439 * association. 440 * 441 */ 442 void sctp_icmp_proto_unreachable(struct sock *sk, 443 struct sctp_association *asoc, 444 struct sctp_transport *t) 445 { 446 if (sock_owned_by_user(sk)) { 447 if (timer_pending(&t->proto_unreach_timer)) 448 return; 449 else { 450 if (!mod_timer(&t->proto_unreach_timer, 451 jiffies + (HZ/20))) 452 sctp_association_hold(asoc); 453 } 454 } else { 455 struct net *net = sock_net(sk); 456 457 pr_debug("%s: unrecognized next header type " 458 "encountered!\n", __func__); 459 460 if (del_timer(&t->proto_unreach_timer)) 461 sctp_association_put(asoc); 462 463 sctp_do_sm(net, SCTP_EVENT_T_OTHER, 464 SCTP_ST_OTHER(SCTP_EVENT_ICMP_PROTO_UNREACH), 465 asoc->state, asoc->ep, asoc, t, 466 GFP_ATOMIC); 467 } 468 } 469 470 /* Common lookup code for icmp/icmpv6 error handler. */ 471 struct sock *sctp_err_lookup(struct net *net, int family, struct sk_buff *skb, 472 struct sctphdr *sctphdr, 473 struct sctp_association **app, 474 struct sctp_transport **tpp) 475 { 476 union sctp_addr saddr; 477 union sctp_addr daddr; 478 struct sctp_af *af; 479 struct sock *sk = NULL; 480 struct sctp_association *asoc; 481 struct sctp_transport *transport = NULL; 482 struct sctp_init_chunk *chunkhdr; 483 __u32 vtag = ntohl(sctphdr->vtag); 484 int len = skb->len - ((void *)sctphdr - (void *)skb->data); 485 486 *app = NULL; *tpp = NULL; 487 488 af = sctp_get_af_specific(family); 489 if (unlikely(!af)) { 490 return NULL; 491 } 492 493 /* Initialize local addresses for lookups. */ 494 af->from_skb(&saddr, skb, 1); 495 af->from_skb(&daddr, skb, 0); 496 497 /* Look for an association that matches the incoming ICMP error 498 * packet. 499 */ 500 asoc = __sctp_lookup_association(net, &saddr, &daddr, &transport); 501 if (!asoc) 502 return NULL; 503 504 sk = asoc->base.sk; 505 506 /* RFC 4960, Appendix C. ICMP Handling 507 * 508 * ICMP6) An implementation MUST validate that the Verification Tag 509 * contained in the ICMP message matches the Verification Tag of 510 * the peer. If the Verification Tag is not 0 and does NOT 511 * match, discard the ICMP message. If it is 0 and the ICMP 512 * message contains enough bytes to verify that the chunk type is 513 * an INIT chunk and that the Initiate Tag matches the tag of the 514 * peer, continue with ICMP7. If the ICMP message is too short 515 * or the chunk type or the Initiate Tag does not match, silently 516 * discard the packet. 517 */ 518 if (vtag == 0) { 519 chunkhdr = (void *)sctphdr + sizeof(struct sctphdr); 520 if (len < sizeof(struct sctphdr) + sizeof(sctp_chunkhdr_t) 521 + sizeof(__be32) || 522 chunkhdr->chunk_hdr.type != SCTP_CID_INIT || 523 ntohl(chunkhdr->init_hdr.init_tag) != asoc->c.my_vtag) { 524 goto out; 525 } 526 } else if (vtag != asoc->c.peer_vtag) { 527 goto out; 528 } 529 530 bh_lock_sock(sk); 531 532 /* If too many ICMPs get dropped on busy 533 * servers this needs to be solved differently. 534 */ 535 if (sock_owned_by_user(sk)) 536 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS); 537 538 *app = asoc; 539 *tpp = transport; 540 return sk; 541 542 out: 543 sctp_transport_put(transport); 544 return NULL; 545 } 546 547 /* Common cleanup code for icmp/icmpv6 error handler. */ 548 void sctp_err_finish(struct sock *sk, struct sctp_transport *t) 549 { 550 bh_unlock_sock(sk); 551 sctp_transport_put(t); 552 } 553 554 /* 555 * This routine is called by the ICMP module when it gets some 556 * sort of error condition. If err < 0 then the socket should 557 * be closed and the error returned to the user. If err > 0 558 * it's just the icmp type << 8 | icmp code. After adjustment 559 * header points to the first 8 bytes of the sctp header. We need 560 * to find the appropriate port. 561 * 562 * The locking strategy used here is very "optimistic". When 563 * someone else accesses the socket the ICMP is just dropped 564 * and for some paths there is no check at all. 565 * A more general error queue to queue errors for later handling 566 * is probably better. 567 * 568 */ 569 void sctp_v4_err(struct sk_buff *skb, __u32 info) 570 { 571 const struct iphdr *iph = (const struct iphdr *)skb->data; 572 const int ihlen = iph->ihl * 4; 573 const int type = icmp_hdr(skb)->type; 574 const int code = icmp_hdr(skb)->code; 575 struct sock *sk; 576 struct sctp_association *asoc = NULL; 577 struct sctp_transport *transport; 578 struct inet_sock *inet; 579 __u16 saveip, savesctp; 580 int err; 581 struct net *net = dev_net(skb->dev); 582 583 /* Fix up skb to look at the embedded net header. */ 584 saveip = skb->network_header; 585 savesctp = skb->transport_header; 586 skb_reset_network_header(skb); 587 skb_set_transport_header(skb, ihlen); 588 sk = sctp_err_lookup(net, AF_INET, skb, sctp_hdr(skb), &asoc, &transport); 589 /* Put back, the original values. */ 590 skb->network_header = saveip; 591 skb->transport_header = savesctp; 592 if (!sk) { 593 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS); 594 return; 595 } 596 /* Warning: The sock lock is held. Remember to call 597 * sctp_err_finish! 598 */ 599 600 switch (type) { 601 case ICMP_PARAMETERPROB: 602 err = EPROTO; 603 break; 604 case ICMP_DEST_UNREACH: 605 if (code > NR_ICMP_UNREACH) 606 goto out_unlock; 607 608 /* PMTU discovery (RFC1191) */ 609 if (ICMP_FRAG_NEEDED == code) { 610 sctp_icmp_frag_needed(sk, asoc, transport, 611 SCTP_TRUNC4(info)); 612 goto out_unlock; 613 } else { 614 if (ICMP_PROT_UNREACH == code) { 615 sctp_icmp_proto_unreachable(sk, asoc, 616 transport); 617 goto out_unlock; 618 } 619 } 620 err = icmp_err_convert[code].errno; 621 break; 622 case ICMP_TIME_EXCEEDED: 623 /* Ignore any time exceeded errors due to fragment reassembly 624 * timeouts. 625 */ 626 if (ICMP_EXC_FRAGTIME == code) 627 goto out_unlock; 628 629 err = EHOSTUNREACH; 630 break; 631 case ICMP_REDIRECT: 632 sctp_icmp_redirect(sk, transport, skb); 633 /* Fall through to out_unlock. */ 634 default: 635 goto out_unlock; 636 } 637 638 inet = inet_sk(sk); 639 if (!sock_owned_by_user(sk) && inet->recverr) { 640 sk->sk_err = err; 641 sk->sk_error_report(sk); 642 } else { /* Only an error on timeout */ 643 sk->sk_err_soft = err; 644 } 645 646 out_unlock: 647 sctp_err_finish(sk, transport); 648 } 649 650 /* 651 * RFC 2960, 8.4 - Handle "Out of the blue" Packets. 652 * 653 * This function scans all the chunks in the OOTB packet to determine if 654 * the packet should be discarded right away. If a response might be needed 655 * for this packet, or, if further processing is possible, the packet will 656 * be queued to a proper inqueue for the next phase of handling. 657 * 658 * Output: 659 * Return 0 - If further processing is needed. 660 * Return 1 - If the packet can be discarded right away. 661 */ 662 static int sctp_rcv_ootb(struct sk_buff *skb) 663 { 664 sctp_chunkhdr_t *ch, _ch; 665 int ch_end, offset = 0; 666 667 /* Scan through all the chunks in the packet. */ 668 do { 669 /* Make sure we have at least the header there */ 670 if (offset + sizeof(sctp_chunkhdr_t) > skb->len) 671 break; 672 673 ch = skb_header_pointer(skb, offset, sizeof(*ch), &_ch); 674 675 /* Break out if chunk length is less then minimal. */ 676 if (ntohs(ch->length) < sizeof(sctp_chunkhdr_t)) 677 break; 678 679 ch_end = offset + SCTP_PAD4(ntohs(ch->length)); 680 if (ch_end > skb->len) 681 break; 682 683 /* RFC 8.4, 2) If the OOTB packet contains an ABORT chunk, the 684 * receiver MUST silently discard the OOTB packet and take no 685 * further action. 686 */ 687 if (SCTP_CID_ABORT == ch->type) 688 goto discard; 689 690 /* RFC 8.4, 6) If the packet contains a SHUTDOWN COMPLETE 691 * chunk, the receiver should silently discard the packet 692 * and take no further action. 693 */ 694 if (SCTP_CID_SHUTDOWN_COMPLETE == ch->type) 695 goto discard; 696 697 /* RFC 4460, 2.11.2 698 * This will discard packets with INIT chunk bundled as 699 * subsequent chunks in the packet. When INIT is first, 700 * the normal INIT processing will discard the chunk. 701 */ 702 if (SCTP_CID_INIT == ch->type && (void *)ch != skb->data) 703 goto discard; 704 705 offset = ch_end; 706 } while (ch_end < skb->len); 707 708 return 0; 709 710 discard: 711 return 1; 712 } 713 714 /* Insert endpoint into the hash table. */ 715 static void __sctp_hash_endpoint(struct sctp_endpoint *ep) 716 { 717 struct net *net = sock_net(ep->base.sk); 718 struct sctp_ep_common *epb; 719 struct sctp_hashbucket *head; 720 721 epb = &ep->base; 722 723 epb->hashent = sctp_ep_hashfn(net, epb->bind_addr.port); 724 head = &sctp_ep_hashtable[epb->hashent]; 725 726 write_lock(&head->lock); 727 hlist_add_head(&epb->node, &head->chain); 728 write_unlock(&head->lock); 729 } 730 731 /* Add an endpoint to the hash. Local BH-safe. */ 732 void sctp_hash_endpoint(struct sctp_endpoint *ep) 733 { 734 local_bh_disable(); 735 __sctp_hash_endpoint(ep); 736 local_bh_enable(); 737 } 738 739 /* Remove endpoint from the hash table. */ 740 static void __sctp_unhash_endpoint(struct sctp_endpoint *ep) 741 { 742 struct net *net = sock_net(ep->base.sk); 743 struct sctp_hashbucket *head; 744 struct sctp_ep_common *epb; 745 746 epb = &ep->base; 747 748 epb->hashent = sctp_ep_hashfn(net, epb->bind_addr.port); 749 750 head = &sctp_ep_hashtable[epb->hashent]; 751 752 write_lock(&head->lock); 753 hlist_del_init(&epb->node); 754 write_unlock(&head->lock); 755 } 756 757 /* Remove endpoint from the hash. Local BH-safe. */ 758 void sctp_unhash_endpoint(struct sctp_endpoint *ep) 759 { 760 local_bh_disable(); 761 __sctp_unhash_endpoint(ep); 762 local_bh_enable(); 763 } 764 765 /* Look up an endpoint. */ 766 static struct sctp_endpoint *__sctp_rcv_lookup_endpoint(struct net *net, 767 const union sctp_addr *laddr) 768 { 769 struct sctp_hashbucket *head; 770 struct sctp_ep_common *epb; 771 struct sctp_endpoint *ep; 772 int hash; 773 774 hash = sctp_ep_hashfn(net, ntohs(laddr->v4.sin_port)); 775 head = &sctp_ep_hashtable[hash]; 776 read_lock(&head->lock); 777 sctp_for_each_hentry(epb, &head->chain) { 778 ep = sctp_ep(epb); 779 if (sctp_endpoint_is_match(ep, net, laddr)) 780 goto hit; 781 } 782 783 ep = sctp_sk(net->sctp.ctl_sock)->ep; 784 785 hit: 786 sctp_endpoint_hold(ep); 787 read_unlock(&head->lock); 788 return ep; 789 } 790 791 /* rhashtable for transport */ 792 struct sctp_hash_cmp_arg { 793 const union sctp_addr *paddr; 794 const struct net *net; 795 u16 lport; 796 }; 797 798 static inline int sctp_hash_cmp(struct rhashtable_compare_arg *arg, 799 const void *ptr) 800 { 801 struct sctp_transport *t = (struct sctp_transport *)ptr; 802 const struct sctp_hash_cmp_arg *x = arg->key; 803 int err = 1; 804 805 if (!sctp_cmp_addr_exact(&t->ipaddr, x->paddr)) 806 return err; 807 if (!sctp_transport_hold(t)) 808 return err; 809 810 if (!net_eq(sock_net(t->asoc->base.sk), x->net)) 811 goto out; 812 if (x->lport != htons(t->asoc->base.bind_addr.port)) 813 goto out; 814 815 err = 0; 816 out: 817 sctp_transport_put(t); 818 return err; 819 } 820 821 static inline u32 sctp_hash_obj(const void *data, u32 len, u32 seed) 822 { 823 const struct sctp_transport *t = data; 824 const union sctp_addr *paddr = &t->ipaddr; 825 const struct net *net = sock_net(t->asoc->base.sk); 826 u16 lport = htons(t->asoc->base.bind_addr.port); 827 u32 addr; 828 829 if (paddr->sa.sa_family == AF_INET6) 830 addr = jhash(&paddr->v6.sin6_addr, 16, seed); 831 else 832 addr = paddr->v4.sin_addr.s_addr; 833 834 return jhash_3words(addr, ((__u32)paddr->v4.sin_port) << 16 | 835 (__force __u32)lport, net_hash_mix(net), seed); 836 } 837 838 static inline u32 sctp_hash_key(const void *data, u32 len, u32 seed) 839 { 840 const struct sctp_hash_cmp_arg *x = data; 841 const union sctp_addr *paddr = x->paddr; 842 const struct net *net = x->net; 843 u16 lport = x->lport; 844 u32 addr; 845 846 if (paddr->sa.sa_family == AF_INET6) 847 addr = jhash(&paddr->v6.sin6_addr, 16, seed); 848 else 849 addr = paddr->v4.sin_addr.s_addr; 850 851 return jhash_3words(addr, ((__u32)paddr->v4.sin_port) << 16 | 852 (__force __u32)lport, net_hash_mix(net), seed); 853 } 854 855 static const struct rhashtable_params sctp_hash_params = { 856 .head_offset = offsetof(struct sctp_transport, node), 857 .hashfn = sctp_hash_key, 858 .obj_hashfn = sctp_hash_obj, 859 .obj_cmpfn = sctp_hash_cmp, 860 .automatic_shrinking = true, 861 }; 862 863 int sctp_transport_hashtable_init(void) 864 { 865 return rhltable_init(&sctp_transport_hashtable, &sctp_hash_params); 866 } 867 868 void sctp_transport_hashtable_destroy(void) 869 { 870 rhltable_destroy(&sctp_transport_hashtable); 871 } 872 873 int sctp_hash_transport(struct sctp_transport *t) 874 { 875 struct sctp_transport *transport; 876 struct rhlist_head *tmp, *list; 877 struct sctp_hash_cmp_arg arg; 878 int err; 879 880 if (t->asoc->temp) 881 return 0; 882 883 arg.net = sock_net(t->asoc->base.sk); 884 arg.paddr = &t->ipaddr; 885 arg.lport = htons(t->asoc->base.bind_addr.port); 886 887 rcu_read_lock(); 888 list = rhltable_lookup(&sctp_transport_hashtable, &arg, 889 sctp_hash_params); 890 891 rhl_for_each_entry_rcu(transport, tmp, list, node) 892 if (transport->asoc->ep == t->asoc->ep) { 893 rcu_read_unlock(); 894 err = -EEXIST; 895 goto out; 896 } 897 rcu_read_unlock(); 898 899 err = rhltable_insert_key(&sctp_transport_hashtable, &arg, 900 &t->node, sctp_hash_params); 901 902 out: 903 if (err) 904 pr_err_once("insert transport fail, errno %d\n", err); 905 906 return err; 907 } 908 909 void sctp_unhash_transport(struct sctp_transport *t) 910 { 911 if (t->asoc->temp) 912 return; 913 914 rhltable_remove(&sctp_transport_hashtable, &t->node, 915 sctp_hash_params); 916 } 917 918 /* return a transport with holding it */ 919 struct sctp_transport *sctp_addrs_lookup_transport( 920 struct net *net, 921 const union sctp_addr *laddr, 922 const union sctp_addr *paddr) 923 { 924 struct rhlist_head *tmp, *list; 925 struct sctp_transport *t; 926 struct sctp_hash_cmp_arg arg = { 927 .paddr = paddr, 928 .net = net, 929 .lport = laddr->v4.sin_port, 930 }; 931 932 list = rhltable_lookup(&sctp_transport_hashtable, &arg, 933 sctp_hash_params); 934 935 rhl_for_each_entry_rcu(t, tmp, list, node) { 936 if (!sctp_transport_hold(t)) 937 continue; 938 939 if (sctp_bind_addr_match(&t->asoc->base.bind_addr, 940 laddr, sctp_sk(t->asoc->base.sk))) 941 return t; 942 sctp_transport_put(t); 943 } 944 945 return NULL; 946 } 947 948 /* return a transport without holding it, as it's only used under sock lock */ 949 struct sctp_transport *sctp_epaddr_lookup_transport( 950 const struct sctp_endpoint *ep, 951 const union sctp_addr *paddr) 952 { 953 struct net *net = sock_net(ep->base.sk); 954 struct rhlist_head *tmp, *list; 955 struct sctp_transport *t; 956 struct sctp_hash_cmp_arg arg = { 957 .paddr = paddr, 958 .net = net, 959 .lport = htons(ep->base.bind_addr.port), 960 }; 961 962 list = rhltable_lookup(&sctp_transport_hashtable, &arg, 963 sctp_hash_params); 964 965 rhl_for_each_entry_rcu(t, tmp, list, node) 966 if (ep == t->asoc->ep) 967 return t; 968 969 return NULL; 970 } 971 972 /* Look up an association. */ 973 static struct sctp_association *__sctp_lookup_association( 974 struct net *net, 975 const union sctp_addr *local, 976 const union sctp_addr *peer, 977 struct sctp_transport **pt) 978 { 979 struct sctp_transport *t; 980 struct sctp_association *asoc = NULL; 981 982 t = sctp_addrs_lookup_transport(net, local, peer); 983 if (!t) 984 goto out; 985 986 asoc = t->asoc; 987 *pt = t; 988 989 out: 990 return asoc; 991 } 992 993 /* Look up an association. protected by RCU read lock */ 994 static 995 struct sctp_association *sctp_lookup_association(struct net *net, 996 const union sctp_addr *laddr, 997 const union sctp_addr *paddr, 998 struct sctp_transport **transportp) 999 { 1000 struct sctp_association *asoc; 1001 1002 rcu_read_lock(); 1003 asoc = __sctp_lookup_association(net, laddr, paddr, transportp); 1004 rcu_read_unlock(); 1005 1006 return asoc; 1007 } 1008 1009 /* Is there an association matching the given local and peer addresses? */ 1010 int sctp_has_association(struct net *net, 1011 const union sctp_addr *laddr, 1012 const union sctp_addr *paddr) 1013 { 1014 struct sctp_association *asoc; 1015 struct sctp_transport *transport; 1016 1017 if ((asoc = sctp_lookup_association(net, laddr, paddr, &transport))) { 1018 sctp_transport_put(transport); 1019 return 1; 1020 } 1021 1022 return 0; 1023 } 1024 1025 /* 1026 * SCTP Implementors Guide, 2.18 Handling of address 1027 * parameters within the INIT or INIT-ACK. 1028 * 1029 * D) When searching for a matching TCB upon reception of an INIT 1030 * or INIT-ACK chunk the receiver SHOULD use not only the 1031 * source address of the packet (containing the INIT or 1032 * INIT-ACK) but the receiver SHOULD also use all valid 1033 * address parameters contained within the chunk. 1034 * 1035 * 2.18.3 Solution description 1036 * 1037 * This new text clearly specifies to an implementor the need 1038 * to look within the INIT or INIT-ACK. Any implementation that 1039 * does not do this, may not be able to establish associations 1040 * in certain circumstances. 1041 * 1042 */ 1043 static struct sctp_association *__sctp_rcv_init_lookup(struct net *net, 1044 struct sk_buff *skb, 1045 const union sctp_addr *laddr, struct sctp_transport **transportp) 1046 { 1047 struct sctp_association *asoc; 1048 union sctp_addr addr; 1049 union sctp_addr *paddr = &addr; 1050 struct sctphdr *sh = sctp_hdr(skb); 1051 union sctp_params params; 1052 sctp_init_chunk_t *init; 1053 struct sctp_af *af; 1054 1055 /* 1056 * This code will NOT touch anything inside the chunk--it is 1057 * strictly READ-ONLY. 1058 * 1059 * RFC 2960 3 SCTP packet Format 1060 * 1061 * Multiple chunks can be bundled into one SCTP packet up to 1062 * the MTU size, except for the INIT, INIT ACK, and SHUTDOWN 1063 * COMPLETE chunks. These chunks MUST NOT be bundled with any 1064 * other chunk in a packet. See Section 6.10 for more details 1065 * on chunk bundling. 1066 */ 1067 1068 /* Find the start of the TLVs and the end of the chunk. This is 1069 * the region we search for address parameters. 1070 */ 1071 init = (sctp_init_chunk_t *)skb->data; 1072 1073 /* Walk the parameters looking for embedded addresses. */ 1074 sctp_walk_params(params, init, init_hdr.params) { 1075 1076 /* Note: Ignoring hostname addresses. */ 1077 af = sctp_get_af_specific(param_type2af(params.p->type)); 1078 if (!af) 1079 continue; 1080 1081 af->from_addr_param(paddr, params.addr, sh->source, 0); 1082 1083 asoc = __sctp_lookup_association(net, laddr, paddr, transportp); 1084 if (asoc) 1085 return asoc; 1086 } 1087 1088 return NULL; 1089 } 1090 1091 /* ADD-IP, Section 5.2 1092 * When an endpoint receives an ASCONF Chunk from the remote peer 1093 * special procedures may be needed to identify the association the 1094 * ASCONF Chunk is associated with. To properly find the association 1095 * the following procedures SHOULD be followed: 1096 * 1097 * D2) If the association is not found, use the address found in the 1098 * Address Parameter TLV combined with the port number found in the 1099 * SCTP common header. If found proceed to rule D4. 1100 * 1101 * D2-ext) If more than one ASCONF Chunks are packed together, use the 1102 * address found in the ASCONF Address Parameter TLV of each of the 1103 * subsequent ASCONF Chunks. If found, proceed to rule D4. 1104 */ 1105 static struct sctp_association *__sctp_rcv_asconf_lookup( 1106 struct net *net, 1107 sctp_chunkhdr_t *ch, 1108 const union sctp_addr *laddr, 1109 __be16 peer_port, 1110 struct sctp_transport **transportp) 1111 { 1112 sctp_addip_chunk_t *asconf = (struct sctp_addip_chunk *)ch; 1113 struct sctp_af *af; 1114 union sctp_addr_param *param; 1115 union sctp_addr paddr; 1116 1117 /* Skip over the ADDIP header and find the Address parameter */ 1118 param = (union sctp_addr_param *)(asconf + 1); 1119 1120 af = sctp_get_af_specific(param_type2af(param->p.type)); 1121 if (unlikely(!af)) 1122 return NULL; 1123 1124 af->from_addr_param(&paddr, param, peer_port, 0); 1125 1126 return __sctp_lookup_association(net, laddr, &paddr, transportp); 1127 } 1128 1129 1130 /* SCTP-AUTH, Section 6.3: 1131 * If the receiver does not find a STCB for a packet containing an AUTH 1132 * chunk as the first chunk and not a COOKIE-ECHO chunk as the second 1133 * chunk, it MUST use the chunks after the AUTH chunk to look up an existing 1134 * association. 1135 * 1136 * This means that any chunks that can help us identify the association need 1137 * to be looked at to find this association. 1138 */ 1139 static struct sctp_association *__sctp_rcv_walk_lookup(struct net *net, 1140 struct sk_buff *skb, 1141 const union sctp_addr *laddr, 1142 struct sctp_transport **transportp) 1143 { 1144 struct sctp_association *asoc = NULL; 1145 sctp_chunkhdr_t *ch; 1146 int have_auth = 0; 1147 unsigned int chunk_num = 1; 1148 __u8 *ch_end; 1149 1150 /* Walk through the chunks looking for AUTH or ASCONF chunks 1151 * to help us find the association. 1152 */ 1153 ch = (sctp_chunkhdr_t *) skb->data; 1154 do { 1155 /* Break out if chunk length is less then minimal. */ 1156 if (ntohs(ch->length) < sizeof(sctp_chunkhdr_t)) 1157 break; 1158 1159 ch_end = ((__u8 *)ch) + SCTP_PAD4(ntohs(ch->length)); 1160 if (ch_end > skb_tail_pointer(skb)) 1161 break; 1162 1163 switch (ch->type) { 1164 case SCTP_CID_AUTH: 1165 have_auth = chunk_num; 1166 break; 1167 1168 case SCTP_CID_COOKIE_ECHO: 1169 /* If a packet arrives containing an AUTH chunk as 1170 * a first chunk, a COOKIE-ECHO chunk as the second 1171 * chunk, and possibly more chunks after them, and 1172 * the receiver does not have an STCB for that 1173 * packet, then authentication is based on 1174 * the contents of the COOKIE- ECHO chunk. 1175 */ 1176 if (have_auth == 1 && chunk_num == 2) 1177 return NULL; 1178 break; 1179 1180 case SCTP_CID_ASCONF: 1181 if (have_auth || net->sctp.addip_noauth) 1182 asoc = __sctp_rcv_asconf_lookup( 1183 net, ch, laddr, 1184 sctp_hdr(skb)->source, 1185 transportp); 1186 default: 1187 break; 1188 } 1189 1190 if (asoc) 1191 break; 1192 1193 ch = (sctp_chunkhdr_t *) ch_end; 1194 chunk_num++; 1195 } while (ch_end < skb_tail_pointer(skb)); 1196 1197 return asoc; 1198 } 1199 1200 /* 1201 * There are circumstances when we need to look inside the SCTP packet 1202 * for information to help us find the association. Examples 1203 * include looking inside of INIT/INIT-ACK chunks or after the AUTH 1204 * chunks. 1205 */ 1206 static struct sctp_association *__sctp_rcv_lookup_harder(struct net *net, 1207 struct sk_buff *skb, 1208 const union sctp_addr *laddr, 1209 struct sctp_transport **transportp) 1210 { 1211 sctp_chunkhdr_t *ch; 1212 1213 /* We do not allow GSO frames here as we need to linearize and 1214 * then cannot guarantee frame boundaries. This shouldn't be an 1215 * issue as packets hitting this are mostly INIT or INIT-ACK and 1216 * those cannot be on GSO-style anyway. 1217 */ 1218 if ((skb_shinfo(skb)->gso_type & SKB_GSO_SCTP) == SKB_GSO_SCTP) 1219 return NULL; 1220 1221 ch = (sctp_chunkhdr_t *) skb->data; 1222 1223 /* The code below will attempt to walk the chunk and extract 1224 * parameter information. Before we do that, we need to verify 1225 * that the chunk length doesn't cause overflow. Otherwise, we'll 1226 * walk off the end. 1227 */ 1228 if (SCTP_PAD4(ntohs(ch->length)) > skb->len) 1229 return NULL; 1230 1231 /* If this is INIT/INIT-ACK look inside the chunk too. */ 1232 if (ch->type == SCTP_CID_INIT || ch->type == SCTP_CID_INIT_ACK) 1233 return __sctp_rcv_init_lookup(net, skb, laddr, transportp); 1234 1235 return __sctp_rcv_walk_lookup(net, skb, laddr, transportp); 1236 } 1237 1238 /* Lookup an association for an inbound skb. */ 1239 static struct sctp_association *__sctp_rcv_lookup(struct net *net, 1240 struct sk_buff *skb, 1241 const union sctp_addr *paddr, 1242 const union sctp_addr *laddr, 1243 struct sctp_transport **transportp) 1244 { 1245 struct sctp_association *asoc; 1246 1247 asoc = __sctp_lookup_association(net, laddr, paddr, transportp); 1248 if (asoc) 1249 goto out; 1250 1251 /* Further lookup for INIT/INIT-ACK packets. 1252 * SCTP Implementors Guide, 2.18 Handling of address 1253 * parameters within the INIT or INIT-ACK. 1254 */ 1255 asoc = __sctp_rcv_lookup_harder(net, skb, laddr, transportp); 1256 if (asoc) 1257 goto out; 1258 1259 if (paddr->sa.sa_family == AF_INET) 1260 pr_debug("sctp: asoc not found for src:%pI4:%d dst:%pI4:%d\n", 1261 &laddr->v4.sin_addr, ntohs(laddr->v4.sin_port), 1262 &paddr->v4.sin_addr, ntohs(paddr->v4.sin_port)); 1263 else 1264 pr_debug("sctp: asoc not found for src:%pI6:%d dst:%pI6:%d\n", 1265 &laddr->v6.sin6_addr, ntohs(laddr->v6.sin6_port), 1266 &paddr->v6.sin6_addr, ntohs(paddr->v6.sin6_port)); 1267 1268 out: 1269 return asoc; 1270 } 1271