1 /* 2 * DCCP over IPv6 3 * Linux INET6 implementation 4 * 5 * Based on net/dccp6/ipv6.c 6 * 7 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net> 8 * 9 * This program is free software; you can redistribute it and/or 10 * modify it under the terms of the GNU General Public License 11 * as published by the Free Software Foundation; either version 12 * 2 of the License, or (at your option) any later version. 13 */ 14 15 #include <linux/module.h> 16 #include <linux/random.h> 17 #include <linux/slab.h> 18 #include <linux/xfrm.h> 19 20 #include <net/addrconf.h> 21 #include <net/inet_common.h> 22 #include <net/inet_hashtables.h> 23 #include <net/inet_sock.h> 24 #include <net/inet6_connection_sock.h> 25 #include <net/inet6_hashtables.h> 26 #include <net/ip6_route.h> 27 #include <net/ipv6.h> 28 #include <net/protocol.h> 29 #include <net/transp_v6.h> 30 #include <net/ip6_checksum.h> 31 #include <net/xfrm.h> 32 #include <net/secure_seq.h> 33 34 #include "dccp.h" 35 #include "ipv6.h" 36 #include "feat.h" 37 38 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */ 39 40 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped; 41 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops; 42 43 static void dccp_v6_hash(struct sock *sk) 44 { 45 if (sk->sk_state != DCCP_CLOSED) { 46 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) { 47 inet_hash(sk); 48 return; 49 } 50 local_bh_disable(); 51 __inet6_hash(sk, NULL); 52 local_bh_enable(); 53 } 54 } 55 56 /* add pseudo-header to DCCP checksum stored in skb->csum */ 57 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb, 58 const struct in6_addr *saddr, 59 const struct in6_addr *daddr) 60 { 61 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum); 62 } 63 64 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb) 65 { 66 struct ipv6_pinfo *np = inet6_sk(sk); 67 struct dccp_hdr *dh = dccp_hdr(skb); 68 69 dccp_csum_outgoing(skb); 70 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr); 71 } 72 73 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb) 74 { 75 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, 76 ipv6_hdr(skb)->saddr.s6_addr32, 77 dccp_hdr(skb)->dccph_dport, 78 dccp_hdr(skb)->dccph_sport ); 79 80 } 81 82 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 83 u8 type, u8 code, int offset, __be32 info) 84 { 85 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; 86 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset); 87 struct dccp_sock *dp; 88 struct ipv6_pinfo *np; 89 struct sock *sk; 90 int err; 91 __u64 seq; 92 struct net *net = dev_net(skb->dev); 93 94 if (skb->len < offset + sizeof(*dh) || 95 skb->len < offset + __dccp_basic_hdr_len(dh)) { 96 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), 97 ICMP6_MIB_INERRORS); 98 return; 99 } 100 101 sk = inet6_lookup(net, &dccp_hashinfo, 102 &hdr->daddr, dh->dccph_dport, 103 &hdr->saddr, dh->dccph_sport, inet6_iif(skb)); 104 105 if (sk == NULL) { 106 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), 107 ICMP6_MIB_INERRORS); 108 return; 109 } 110 111 if (sk->sk_state == DCCP_TIME_WAIT) { 112 inet_twsk_put(inet_twsk(sk)); 113 return; 114 } 115 116 bh_lock_sock(sk); 117 if (sock_owned_by_user(sk)) 118 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); 119 120 if (sk->sk_state == DCCP_CLOSED) 121 goto out; 122 123 dp = dccp_sk(sk); 124 seq = dccp_hdr_seq(dh); 125 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) && 126 !between48(seq, dp->dccps_awl, dp->dccps_awh)) { 127 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); 128 goto out; 129 } 130 131 np = inet6_sk(sk); 132 133 if (type == NDISC_REDIRECT) { 134 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie); 135 136 if (dst) 137 dst->ops->redirect(dst, sk, skb); 138 goto out; 139 } 140 141 if (type == ICMPV6_PKT_TOOBIG) { 142 struct dst_entry *dst = NULL; 143 144 if (sock_owned_by_user(sk)) 145 goto out; 146 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED)) 147 goto out; 148 149 dst = inet6_csk_update_pmtu(sk, ntohl(info)); 150 if (!dst) 151 goto out; 152 153 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) 154 dccp_sync_mss(sk, dst_mtu(dst)); 155 goto out; 156 } 157 158 icmpv6_err_convert(type, code, &err); 159 160 /* Might be for an request_sock */ 161 switch (sk->sk_state) { 162 struct request_sock *req, **prev; 163 case DCCP_LISTEN: 164 if (sock_owned_by_user(sk)) 165 goto out; 166 167 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport, 168 &hdr->daddr, &hdr->saddr, 169 inet6_iif(skb)); 170 if (req == NULL) 171 goto out; 172 173 /* 174 * ICMPs are not backlogged, hence we cannot get an established 175 * socket here. 176 */ 177 WARN_ON(req->sk != NULL); 178 179 if (!between48(seq, dccp_rsk(req)->dreq_iss, 180 dccp_rsk(req)->dreq_gss)) { 181 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); 182 goto out; 183 } 184 185 inet_csk_reqsk_queue_drop(sk, req, prev); 186 goto out; 187 188 case DCCP_REQUESTING: 189 case DCCP_RESPOND: /* Cannot happen. 190 It can, it SYNs are crossed. --ANK */ 191 if (!sock_owned_by_user(sk)) { 192 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 193 sk->sk_err = err; 194 /* 195 * Wake people up to see the error 196 * (see connect in sock.c) 197 */ 198 sk->sk_error_report(sk); 199 dccp_done(sk); 200 } else 201 sk->sk_err_soft = err; 202 goto out; 203 } 204 205 if (!sock_owned_by_user(sk) && np->recverr) { 206 sk->sk_err = err; 207 sk->sk_error_report(sk); 208 } else 209 sk->sk_err_soft = err; 210 211 out: 212 bh_unlock_sock(sk); 213 sock_put(sk); 214 } 215 216 217 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req) 218 { 219 struct inet6_request_sock *ireq6 = inet6_rsk(req); 220 struct ipv6_pinfo *np = inet6_sk(sk); 221 struct sk_buff *skb; 222 struct in6_addr *final_p, final; 223 struct flowi6 fl6; 224 int err = -1; 225 struct dst_entry *dst; 226 227 memset(&fl6, 0, sizeof(fl6)); 228 fl6.flowi6_proto = IPPROTO_DCCP; 229 fl6.daddr = ireq6->rmt_addr; 230 fl6.saddr = ireq6->loc_addr; 231 fl6.flowlabel = 0; 232 fl6.flowi6_oif = ireq6->iif; 233 fl6.fl6_dport = inet_rsk(req)->rmt_port; 234 fl6.fl6_sport = inet_rsk(req)->loc_port; 235 security_req_classify_flow(req, flowi6_to_flowi(&fl6)); 236 237 238 final_p = fl6_update_dst(&fl6, np->opt, &final); 239 240 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false); 241 if (IS_ERR(dst)) { 242 err = PTR_ERR(dst); 243 dst = NULL; 244 goto done; 245 } 246 247 skb = dccp_make_response(sk, dst, req); 248 if (skb != NULL) { 249 struct dccp_hdr *dh = dccp_hdr(skb); 250 251 dh->dccph_checksum = dccp_v6_csum_finish(skb, 252 &ireq6->loc_addr, 253 &ireq6->rmt_addr); 254 fl6.daddr = ireq6->rmt_addr; 255 err = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass); 256 err = net_xmit_eval(err); 257 } 258 259 done: 260 dst_release(dst); 261 return err; 262 } 263 264 static void dccp_v6_reqsk_destructor(struct request_sock *req) 265 { 266 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg); 267 if (inet6_rsk(req)->pktopts != NULL) 268 kfree_skb(inet6_rsk(req)->pktopts); 269 } 270 271 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb) 272 { 273 const struct ipv6hdr *rxip6h; 274 struct sk_buff *skb; 275 struct flowi6 fl6; 276 struct net *net = dev_net(skb_dst(rxskb)->dev); 277 struct sock *ctl_sk = net->dccp.v6_ctl_sk; 278 struct dst_entry *dst; 279 280 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET) 281 return; 282 283 if (!ipv6_unicast_destination(rxskb)) 284 return; 285 286 skb = dccp_ctl_make_reset(ctl_sk, rxskb); 287 if (skb == NULL) 288 return; 289 290 rxip6h = ipv6_hdr(rxskb); 291 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr, 292 &rxip6h->daddr); 293 294 memset(&fl6, 0, sizeof(fl6)); 295 fl6.daddr = rxip6h->saddr; 296 fl6.saddr = rxip6h->daddr; 297 298 fl6.flowi6_proto = IPPROTO_DCCP; 299 fl6.flowi6_oif = inet6_iif(rxskb); 300 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport; 301 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport; 302 security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6)); 303 304 /* sk = NULL, but it is safe for now. RST socket required. */ 305 dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL, false); 306 if (!IS_ERR(dst)) { 307 skb_dst_set(skb, dst); 308 ip6_xmit(ctl_sk, skb, &fl6, NULL, 0); 309 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS); 310 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS); 311 return; 312 } 313 314 kfree_skb(skb); 315 } 316 317 static struct request_sock_ops dccp6_request_sock_ops = { 318 .family = AF_INET6, 319 .obj_size = sizeof(struct dccp6_request_sock), 320 .rtx_syn_ack = dccp_v6_send_response, 321 .send_ack = dccp_reqsk_send_ack, 322 .destructor = dccp_v6_reqsk_destructor, 323 .send_reset = dccp_v6_ctl_send_reset, 324 .syn_ack_timeout = dccp_syn_ack_timeout, 325 }; 326 327 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb) 328 { 329 const struct dccp_hdr *dh = dccp_hdr(skb); 330 const struct ipv6hdr *iph = ipv6_hdr(skb); 331 struct sock *nsk; 332 struct request_sock **prev; 333 /* Find possible connection requests. */ 334 struct request_sock *req = inet6_csk_search_req(sk, &prev, 335 dh->dccph_sport, 336 &iph->saddr, 337 &iph->daddr, 338 inet6_iif(skb)); 339 if (req != NULL) 340 return dccp_check_req(sk, skb, req, prev); 341 342 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo, 343 &iph->saddr, dh->dccph_sport, 344 &iph->daddr, ntohs(dh->dccph_dport), 345 inet6_iif(skb)); 346 if (nsk != NULL) { 347 if (nsk->sk_state != DCCP_TIME_WAIT) { 348 bh_lock_sock(nsk); 349 return nsk; 350 } 351 inet_twsk_put(inet_twsk(nsk)); 352 return NULL; 353 } 354 355 return sk; 356 } 357 358 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb) 359 { 360 struct request_sock *req; 361 struct dccp_request_sock *dreq; 362 struct inet6_request_sock *ireq6; 363 struct ipv6_pinfo *np = inet6_sk(sk); 364 const __be32 service = dccp_hdr_request(skb)->dccph_req_service; 365 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); 366 367 if (skb->protocol == htons(ETH_P_IP)) 368 return dccp_v4_conn_request(sk, skb); 369 370 if (!ipv6_unicast_destination(skb)) 371 return 0; /* discard, don't send a reset here */ 372 373 if (dccp_bad_service_code(sk, service)) { 374 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE; 375 goto drop; 376 } 377 /* 378 * There are no SYN attacks on IPv6, yet... 379 */ 380 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY; 381 if (inet_csk_reqsk_queue_is_full(sk)) 382 goto drop; 383 384 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) 385 goto drop; 386 387 req = inet6_reqsk_alloc(&dccp6_request_sock_ops); 388 if (req == NULL) 389 goto drop; 390 391 if (dccp_reqsk_init(req, dccp_sk(sk), skb)) 392 goto drop_and_free; 393 394 dreq = dccp_rsk(req); 395 if (dccp_parse_options(sk, dreq, skb)) 396 goto drop_and_free; 397 398 if (security_inet_conn_request(sk, skb, req)) 399 goto drop_and_free; 400 401 ireq6 = inet6_rsk(req); 402 ireq6->rmt_addr = ipv6_hdr(skb)->saddr; 403 ireq6->loc_addr = ipv6_hdr(skb)->daddr; 404 405 if (ipv6_opt_accepted(sk, skb) || 406 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || 407 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { 408 atomic_inc(&skb->users); 409 ireq6->pktopts = skb; 410 } 411 ireq6->iif = sk->sk_bound_dev_if; 412 413 /* So that link locals have meaning */ 414 if (!sk->sk_bound_dev_if && 415 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL) 416 ireq6->iif = inet6_iif(skb); 417 418 /* 419 * Step 3: Process LISTEN state 420 * 421 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie 422 * 423 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child(). 424 */ 425 dreq->dreq_isr = dcb->dccpd_seq; 426 dreq->dreq_gsr = dreq->dreq_isr; 427 dreq->dreq_iss = dccp_v6_init_sequence(skb); 428 dreq->dreq_gss = dreq->dreq_iss; 429 dreq->dreq_service = service; 430 431 if (dccp_v6_send_response(sk, req)) 432 goto drop_and_free; 433 434 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT); 435 return 0; 436 437 drop_and_free: 438 reqsk_free(req); 439 drop: 440 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 441 return -1; 442 } 443 444 static struct sock *dccp_v6_request_recv_sock(struct sock *sk, 445 struct sk_buff *skb, 446 struct request_sock *req, 447 struct dst_entry *dst) 448 { 449 struct inet6_request_sock *ireq6 = inet6_rsk(req); 450 struct ipv6_pinfo *newnp, *np = inet6_sk(sk); 451 struct inet_sock *newinet; 452 struct dccp6_sock *newdp6; 453 struct sock *newsk; 454 455 if (skb->protocol == htons(ETH_P_IP)) { 456 /* 457 * v6 mapped 458 */ 459 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst); 460 if (newsk == NULL) 461 return NULL; 462 463 newdp6 = (struct dccp6_sock *)newsk; 464 newinet = inet_sk(newsk); 465 newinet->pinet6 = &newdp6->inet6; 466 newnp = inet6_sk(newsk); 467 468 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 469 470 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr); 471 472 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); 473 474 newnp->rcv_saddr = newnp->saddr; 475 476 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped; 477 newsk->sk_backlog_rcv = dccp_v4_do_rcv; 478 newnp->pktoptions = NULL; 479 newnp->opt = NULL; 480 newnp->mcast_oif = inet6_iif(skb); 481 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 482 483 /* 484 * No need to charge this sock to the relevant IPv6 refcnt debug socks count 485 * here, dccp_create_openreq_child now does this for us, see the comment in 486 * that function for the gory details. -acme 487 */ 488 489 /* It is tricky place. Until this moment IPv4 tcp 490 worked with IPv6 icsk.icsk_af_ops. 491 Sync it now. 492 */ 493 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); 494 495 return newsk; 496 } 497 498 499 if (sk_acceptq_is_full(sk)) 500 goto out_overflow; 501 502 if (dst == NULL) { 503 struct in6_addr *final_p, final; 504 struct flowi6 fl6; 505 506 memset(&fl6, 0, sizeof(fl6)); 507 fl6.flowi6_proto = IPPROTO_DCCP; 508 fl6.daddr = ireq6->rmt_addr; 509 final_p = fl6_update_dst(&fl6, np->opt, &final); 510 fl6.saddr = ireq6->loc_addr; 511 fl6.flowi6_oif = sk->sk_bound_dev_if; 512 fl6.fl6_dport = inet_rsk(req)->rmt_port; 513 fl6.fl6_sport = inet_rsk(req)->loc_port; 514 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 515 516 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, false); 517 if (IS_ERR(dst)) 518 goto out; 519 } 520 521 newsk = dccp_create_openreq_child(sk, req, skb); 522 if (newsk == NULL) 523 goto out_nonewsk; 524 525 /* 526 * No need to charge this sock to the relevant IPv6 refcnt debug socks 527 * count here, dccp_create_openreq_child now does this for us, see the 528 * comment in that function for the gory details. -acme 529 */ 530 531 __ip6_dst_store(newsk, dst, NULL, NULL); 532 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM | 533 NETIF_F_TSO); 534 newdp6 = (struct dccp6_sock *)newsk; 535 newinet = inet_sk(newsk); 536 newinet->pinet6 = &newdp6->inet6; 537 newnp = inet6_sk(newsk); 538 539 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 540 541 newnp->daddr = ireq6->rmt_addr; 542 newnp->saddr = ireq6->loc_addr; 543 newnp->rcv_saddr = ireq6->loc_addr; 544 newsk->sk_bound_dev_if = ireq6->iif; 545 546 /* Now IPv6 options... 547 548 First: no IPv4 options. 549 */ 550 newinet->inet_opt = NULL; 551 552 /* Clone RX bits */ 553 newnp->rxopt.all = np->rxopt.all; 554 555 /* Clone pktoptions received with SYN */ 556 newnp->pktoptions = NULL; 557 if (ireq6->pktopts != NULL) { 558 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC); 559 consume_skb(ireq6->pktopts); 560 ireq6->pktopts = NULL; 561 if (newnp->pktoptions) 562 skb_set_owner_r(newnp->pktoptions, newsk); 563 } 564 newnp->opt = NULL; 565 newnp->mcast_oif = inet6_iif(skb); 566 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 567 568 /* 569 * Clone native IPv6 options from listening socket (if any) 570 * 571 * Yes, keeping reference count would be much more clever, but we make 572 * one more one thing there: reattach optmem to newsk. 573 */ 574 if (np->opt != NULL) 575 newnp->opt = ipv6_dup_options(newsk, np->opt); 576 577 inet_csk(newsk)->icsk_ext_hdr_len = 0; 578 if (newnp->opt != NULL) 579 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + 580 newnp->opt->opt_flen); 581 582 dccp_sync_mss(newsk, dst_mtu(dst)); 583 584 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; 585 newinet->inet_rcv_saddr = LOOPBACK4_IPV6; 586 587 if (__inet_inherit_port(sk, newsk) < 0) { 588 inet_csk_prepare_forced_close(newsk); 589 dccp_done(newsk); 590 goto out; 591 } 592 __inet6_hash(newsk, NULL); 593 594 return newsk; 595 596 out_overflow: 597 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); 598 out_nonewsk: 599 dst_release(dst); 600 out: 601 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); 602 return NULL; 603 } 604 605 /* The socket must have it's spinlock held when we get 606 * here. 607 * 608 * We have a potential double-lock case here, so even when 609 * doing backlog processing we use the BH locking scheme. 610 * This is because we cannot sleep with the original spinlock 611 * held. 612 */ 613 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) 614 { 615 struct ipv6_pinfo *np = inet6_sk(sk); 616 struct sk_buff *opt_skb = NULL; 617 618 /* Imagine: socket is IPv6. IPv4 packet arrives, 619 goes to IPv4 receive handler and backlogged. 620 From backlog it always goes here. Kerboom... 621 Fortunately, dccp_rcv_established and rcv_established 622 handle them correctly, but it is not case with 623 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK 624 */ 625 626 if (skb->protocol == htons(ETH_P_IP)) 627 return dccp_v4_do_rcv(sk, skb); 628 629 if (sk_filter(sk, skb)) 630 goto discard; 631 632 /* 633 * socket locking is here for SMP purposes as backlog rcv is currently 634 * called with bh processing disabled. 635 */ 636 637 /* Do Stevens' IPV6_PKTOPTIONS. 638 639 Yes, guys, it is the only place in our code, where we 640 may make it not affecting IPv4. 641 The rest of code is protocol independent, 642 and I do not like idea to uglify IPv4. 643 644 Actually, all the idea behind IPV6_PKTOPTIONS 645 looks not very well thought. For now we latch 646 options, received in the last packet, enqueued 647 by tcp. Feel free to propose better solution. 648 --ANK (980728) 649 */ 650 if (np->rxopt.all) 651 /* 652 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below 653 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example. 654 */ 655 opt_skb = skb_clone(skb, GFP_ATOMIC); 656 657 if (sk->sk_state == DCCP_OPEN) { /* Fast path */ 658 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len)) 659 goto reset; 660 if (opt_skb) { 661 /* XXX This is where we would goto ipv6_pktoptions. */ 662 __kfree_skb(opt_skb); 663 } 664 return 0; 665 } 666 667 /* 668 * Step 3: Process LISTEN state 669 * If S.state == LISTEN, 670 * If P.type == Request or P contains a valid Init Cookie option, 671 * (* Must scan the packet's options to check for Init 672 * Cookies. Only Init Cookies are processed here, 673 * however; other options are processed in Step 8. This 674 * scan need only be performed if the endpoint uses Init 675 * Cookies *) 676 * (* Generate a new socket and switch to that socket *) 677 * Set S := new socket for this port pair 678 * S.state = RESPOND 679 * Choose S.ISS (initial seqno) or set from Init Cookies 680 * Initialize S.GAR := S.ISS 681 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies 682 * Continue with S.state == RESPOND 683 * (* A Response packet will be generated in Step 11 *) 684 * Otherwise, 685 * Generate Reset(No Connection) unless P.type == Reset 686 * Drop packet and return 687 * 688 * NOTE: the check for the packet types is done in 689 * dccp_rcv_state_process 690 */ 691 if (sk->sk_state == DCCP_LISTEN) { 692 struct sock *nsk = dccp_v6_hnd_req(sk, skb); 693 694 if (nsk == NULL) 695 goto discard; 696 /* 697 * Queue it on the new socket if the new socket is active, 698 * otherwise we just shortcircuit this and continue with 699 * the new socket.. 700 */ 701 if (nsk != sk) { 702 if (dccp_child_process(sk, nsk, skb)) 703 goto reset; 704 if (opt_skb != NULL) 705 __kfree_skb(opt_skb); 706 return 0; 707 } 708 } 709 710 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len)) 711 goto reset; 712 if (opt_skb) { 713 /* XXX This is where we would goto ipv6_pktoptions. */ 714 __kfree_skb(opt_skb); 715 } 716 return 0; 717 718 reset: 719 dccp_v6_ctl_send_reset(sk, skb); 720 discard: 721 if (opt_skb != NULL) 722 __kfree_skb(opt_skb); 723 kfree_skb(skb); 724 return 0; 725 } 726 727 static int dccp_v6_rcv(struct sk_buff *skb) 728 { 729 const struct dccp_hdr *dh; 730 struct sock *sk; 731 int min_cov; 732 733 /* Step 1: Check header basics */ 734 735 if (dccp_invalid_packet(skb)) 736 goto discard_it; 737 738 /* Step 1: If header checksum is incorrect, drop packet and return. */ 739 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr, 740 &ipv6_hdr(skb)->daddr)) { 741 DCCP_WARN("dropped packet with invalid checksum\n"); 742 goto discard_it; 743 } 744 745 dh = dccp_hdr(skb); 746 747 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh); 748 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type; 749 750 if (dccp_packet_without_ack(skb)) 751 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ; 752 else 753 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb); 754 755 /* Step 2: 756 * Look up flow ID in table and get corresponding socket */ 757 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, 758 dh->dccph_sport, dh->dccph_dport); 759 /* 760 * Step 2: 761 * If no socket ... 762 */ 763 if (sk == NULL) { 764 dccp_pr_debug("failed to look up flow ID in table and " 765 "get corresponding socket\n"); 766 goto no_dccp_socket; 767 } 768 769 /* 770 * Step 2: 771 * ... or S.state == TIMEWAIT, 772 * Generate Reset(No Connection) unless P.type == Reset 773 * Drop packet and return 774 */ 775 if (sk->sk_state == DCCP_TIME_WAIT) { 776 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n"); 777 inet_twsk_put(inet_twsk(sk)); 778 goto no_dccp_socket; 779 } 780 781 /* 782 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage 783 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted 784 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov 785 */ 786 min_cov = dccp_sk(sk)->dccps_pcrlen; 787 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) { 788 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n", 789 dh->dccph_cscov, min_cov); 790 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */ 791 goto discard_and_relse; 792 } 793 794 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 795 goto discard_and_relse; 796 797 return sk_receive_skb(sk, skb, 1) ? -1 : 0; 798 799 no_dccp_socket: 800 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 801 goto discard_it; 802 /* 803 * Step 2: 804 * If no socket ... 805 * Generate Reset(No Connection) unless P.type == Reset 806 * Drop packet and return 807 */ 808 if (dh->dccph_type != DCCP_PKT_RESET) { 809 DCCP_SKB_CB(skb)->dccpd_reset_code = 810 DCCP_RESET_CODE_NO_CONNECTION; 811 dccp_v6_ctl_send_reset(sk, skb); 812 } 813 814 discard_it: 815 kfree_skb(skb); 816 return 0; 817 818 discard_and_relse: 819 sock_put(sk); 820 goto discard_it; 821 } 822 823 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, 824 int addr_len) 825 { 826 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr; 827 struct inet_connection_sock *icsk = inet_csk(sk); 828 struct inet_sock *inet = inet_sk(sk); 829 struct ipv6_pinfo *np = inet6_sk(sk); 830 struct dccp_sock *dp = dccp_sk(sk); 831 struct in6_addr *saddr = NULL, *final_p, final; 832 struct flowi6 fl6; 833 struct dst_entry *dst; 834 int addr_type; 835 int err; 836 837 dp->dccps_role = DCCP_ROLE_CLIENT; 838 839 if (addr_len < SIN6_LEN_RFC2133) 840 return -EINVAL; 841 842 if (usin->sin6_family != AF_INET6) 843 return -EAFNOSUPPORT; 844 845 memset(&fl6, 0, sizeof(fl6)); 846 847 if (np->sndflow) { 848 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 849 IP6_ECN_flow_init(fl6.flowlabel); 850 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) { 851 struct ip6_flowlabel *flowlabel; 852 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 853 if (flowlabel == NULL) 854 return -EINVAL; 855 usin->sin6_addr = flowlabel->dst; 856 fl6_sock_release(flowlabel); 857 } 858 } 859 /* 860 * connect() to INADDR_ANY means loopback (BSD'ism). 861 */ 862 if (ipv6_addr_any(&usin->sin6_addr)) 863 usin->sin6_addr.s6_addr[15] = 1; 864 865 addr_type = ipv6_addr_type(&usin->sin6_addr); 866 867 if (addr_type & IPV6_ADDR_MULTICAST) 868 return -ENETUNREACH; 869 870 if (addr_type & IPV6_ADDR_LINKLOCAL) { 871 if (addr_len >= sizeof(struct sockaddr_in6) && 872 usin->sin6_scope_id) { 873 /* If interface is set while binding, indices 874 * must coincide. 875 */ 876 if (sk->sk_bound_dev_if && 877 sk->sk_bound_dev_if != usin->sin6_scope_id) 878 return -EINVAL; 879 880 sk->sk_bound_dev_if = usin->sin6_scope_id; 881 } 882 883 /* Connect to link-local address requires an interface */ 884 if (!sk->sk_bound_dev_if) 885 return -EINVAL; 886 } 887 888 np->daddr = usin->sin6_addr; 889 np->flow_label = fl6.flowlabel; 890 891 /* 892 * DCCP over IPv4 893 */ 894 if (addr_type == IPV6_ADDR_MAPPED) { 895 u32 exthdrlen = icsk->icsk_ext_hdr_len; 896 struct sockaddr_in sin; 897 898 SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); 899 900 if (__ipv6_only_sock(sk)) 901 return -ENETUNREACH; 902 903 sin.sin_family = AF_INET; 904 sin.sin_port = usin->sin6_port; 905 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; 906 907 icsk->icsk_af_ops = &dccp_ipv6_mapped; 908 sk->sk_backlog_rcv = dccp_v4_do_rcv; 909 910 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); 911 if (err) { 912 icsk->icsk_ext_hdr_len = exthdrlen; 913 icsk->icsk_af_ops = &dccp_ipv6_af_ops; 914 sk->sk_backlog_rcv = dccp_v6_do_rcv; 915 goto failure; 916 } 917 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); 918 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr); 919 920 return err; 921 } 922 923 if (!ipv6_addr_any(&np->rcv_saddr)) 924 saddr = &np->rcv_saddr; 925 926 fl6.flowi6_proto = IPPROTO_DCCP; 927 fl6.daddr = np->daddr; 928 fl6.saddr = saddr ? *saddr : np->saddr; 929 fl6.flowi6_oif = sk->sk_bound_dev_if; 930 fl6.fl6_dport = usin->sin6_port; 931 fl6.fl6_sport = inet->inet_sport; 932 security_sk_classify_flow(sk, flowi6_to_flowi(&fl6)); 933 934 final_p = fl6_update_dst(&fl6, np->opt, &final); 935 936 dst = ip6_dst_lookup_flow(sk, &fl6, final_p, true); 937 if (IS_ERR(dst)) { 938 err = PTR_ERR(dst); 939 goto failure; 940 } 941 942 if (saddr == NULL) { 943 saddr = &fl6.saddr; 944 np->rcv_saddr = *saddr; 945 } 946 947 /* set the source address */ 948 np->saddr = *saddr; 949 inet->inet_rcv_saddr = LOOPBACK4_IPV6; 950 951 __ip6_dst_store(sk, dst, NULL, NULL); 952 953 icsk->icsk_ext_hdr_len = 0; 954 if (np->opt != NULL) 955 icsk->icsk_ext_hdr_len = (np->opt->opt_flen + 956 np->opt->opt_nflen); 957 958 inet->inet_dport = usin->sin6_port; 959 960 dccp_set_state(sk, DCCP_REQUESTING); 961 err = inet6_hash_connect(&dccp_death_row, sk); 962 if (err) 963 goto late_failure; 964 965 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32, 966 np->daddr.s6_addr32, 967 inet->inet_sport, 968 inet->inet_dport); 969 err = dccp_connect(sk); 970 if (err) 971 goto late_failure; 972 973 return 0; 974 975 late_failure: 976 dccp_set_state(sk, DCCP_CLOSED); 977 __sk_dst_reset(sk); 978 failure: 979 inet->inet_dport = 0; 980 sk->sk_route_caps = 0; 981 return err; 982 } 983 984 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = { 985 .queue_xmit = inet6_csk_xmit, 986 .send_check = dccp_v6_send_check, 987 .rebuild_header = inet6_sk_rebuild_header, 988 .conn_request = dccp_v6_conn_request, 989 .syn_recv_sock = dccp_v6_request_recv_sock, 990 .net_header_len = sizeof(struct ipv6hdr), 991 .setsockopt = ipv6_setsockopt, 992 .getsockopt = ipv6_getsockopt, 993 .addr2sockaddr = inet6_csk_addr2sockaddr, 994 .sockaddr_len = sizeof(struct sockaddr_in6), 995 .bind_conflict = inet6_csk_bind_conflict, 996 #ifdef CONFIG_COMPAT 997 .compat_setsockopt = compat_ipv6_setsockopt, 998 .compat_getsockopt = compat_ipv6_getsockopt, 999 #endif 1000 }; 1001 1002 /* 1003 * DCCP over IPv4 via INET6 API 1004 */ 1005 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = { 1006 .queue_xmit = ip_queue_xmit, 1007 .send_check = dccp_v4_send_check, 1008 .rebuild_header = inet_sk_rebuild_header, 1009 .conn_request = dccp_v6_conn_request, 1010 .syn_recv_sock = dccp_v6_request_recv_sock, 1011 .net_header_len = sizeof(struct iphdr), 1012 .setsockopt = ipv6_setsockopt, 1013 .getsockopt = ipv6_getsockopt, 1014 .addr2sockaddr = inet6_csk_addr2sockaddr, 1015 .sockaddr_len = sizeof(struct sockaddr_in6), 1016 #ifdef CONFIG_COMPAT 1017 .compat_setsockopt = compat_ipv6_setsockopt, 1018 .compat_getsockopt = compat_ipv6_getsockopt, 1019 #endif 1020 }; 1021 1022 /* NOTE: A lot of things set to zero explicitly by call to 1023 * sk_alloc() so need not be done here. 1024 */ 1025 static int dccp_v6_init_sock(struct sock *sk) 1026 { 1027 static __u8 dccp_v6_ctl_sock_initialized; 1028 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized); 1029 1030 if (err == 0) { 1031 if (unlikely(!dccp_v6_ctl_sock_initialized)) 1032 dccp_v6_ctl_sock_initialized = 1; 1033 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops; 1034 } 1035 1036 return err; 1037 } 1038 1039 static void dccp_v6_destroy_sock(struct sock *sk) 1040 { 1041 dccp_destroy_sock(sk); 1042 inet6_destroy_sock(sk); 1043 } 1044 1045 static struct timewait_sock_ops dccp6_timewait_sock_ops = { 1046 .twsk_obj_size = sizeof(struct dccp6_timewait_sock), 1047 }; 1048 1049 static struct proto dccp_v6_prot = { 1050 .name = "DCCPv6", 1051 .owner = THIS_MODULE, 1052 .close = dccp_close, 1053 .connect = dccp_v6_connect, 1054 .disconnect = dccp_disconnect, 1055 .ioctl = dccp_ioctl, 1056 .init = dccp_v6_init_sock, 1057 .setsockopt = dccp_setsockopt, 1058 .getsockopt = dccp_getsockopt, 1059 .sendmsg = dccp_sendmsg, 1060 .recvmsg = dccp_recvmsg, 1061 .backlog_rcv = dccp_v6_do_rcv, 1062 .hash = dccp_v6_hash, 1063 .unhash = inet_unhash, 1064 .accept = inet_csk_accept, 1065 .get_port = inet_csk_get_port, 1066 .shutdown = dccp_shutdown, 1067 .destroy = dccp_v6_destroy_sock, 1068 .orphan_count = &dccp_orphan_count, 1069 .max_header = MAX_DCCP_HEADER, 1070 .obj_size = sizeof(struct dccp6_sock), 1071 .slab_flags = SLAB_DESTROY_BY_RCU, 1072 .rsk_prot = &dccp6_request_sock_ops, 1073 .twsk_prot = &dccp6_timewait_sock_ops, 1074 .h.hashinfo = &dccp_hashinfo, 1075 #ifdef CONFIG_COMPAT 1076 .compat_setsockopt = compat_dccp_setsockopt, 1077 .compat_getsockopt = compat_dccp_getsockopt, 1078 #endif 1079 }; 1080 1081 static const struct inet6_protocol dccp_v6_protocol = { 1082 .handler = dccp_v6_rcv, 1083 .err_handler = dccp_v6_err, 1084 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL, 1085 }; 1086 1087 static const struct proto_ops inet6_dccp_ops = { 1088 .family = PF_INET6, 1089 .owner = THIS_MODULE, 1090 .release = inet6_release, 1091 .bind = inet6_bind, 1092 .connect = inet_stream_connect, 1093 .socketpair = sock_no_socketpair, 1094 .accept = inet_accept, 1095 .getname = inet6_getname, 1096 .poll = dccp_poll, 1097 .ioctl = inet6_ioctl, 1098 .listen = inet_dccp_listen, 1099 .shutdown = inet_shutdown, 1100 .setsockopt = sock_common_setsockopt, 1101 .getsockopt = sock_common_getsockopt, 1102 .sendmsg = inet_sendmsg, 1103 .recvmsg = sock_common_recvmsg, 1104 .mmap = sock_no_mmap, 1105 .sendpage = sock_no_sendpage, 1106 #ifdef CONFIG_COMPAT 1107 .compat_setsockopt = compat_sock_common_setsockopt, 1108 .compat_getsockopt = compat_sock_common_getsockopt, 1109 #endif 1110 }; 1111 1112 static struct inet_protosw dccp_v6_protosw = { 1113 .type = SOCK_DCCP, 1114 .protocol = IPPROTO_DCCP, 1115 .prot = &dccp_v6_prot, 1116 .ops = &inet6_dccp_ops, 1117 .flags = INET_PROTOSW_ICSK, 1118 }; 1119 1120 static int __net_init dccp_v6_init_net(struct net *net) 1121 { 1122 if (dccp_hashinfo.bhash == NULL) 1123 return -ESOCKTNOSUPPORT; 1124 1125 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6, 1126 SOCK_DCCP, IPPROTO_DCCP, net); 1127 } 1128 1129 static void __net_exit dccp_v6_exit_net(struct net *net) 1130 { 1131 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk); 1132 } 1133 1134 static struct pernet_operations dccp_v6_ops = { 1135 .init = dccp_v6_init_net, 1136 .exit = dccp_v6_exit_net, 1137 }; 1138 1139 static int __init dccp_v6_init(void) 1140 { 1141 int err = proto_register(&dccp_v6_prot, 1); 1142 1143 if (err != 0) 1144 goto out; 1145 1146 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1147 if (err != 0) 1148 goto out_unregister_proto; 1149 1150 inet6_register_protosw(&dccp_v6_protosw); 1151 1152 err = register_pernet_subsys(&dccp_v6_ops); 1153 if (err != 0) 1154 goto out_destroy_ctl_sock; 1155 out: 1156 return err; 1157 1158 out_destroy_ctl_sock: 1159 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1160 inet6_unregister_protosw(&dccp_v6_protosw); 1161 out_unregister_proto: 1162 proto_unregister(&dccp_v6_prot); 1163 goto out; 1164 } 1165 1166 static void __exit dccp_v6_exit(void) 1167 { 1168 unregister_pernet_subsys(&dccp_v6_ops); 1169 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1170 inet6_unregister_protosw(&dccp_v6_protosw); 1171 proto_unregister(&dccp_v6_prot); 1172 } 1173 1174 module_init(dccp_v6_init); 1175 module_exit(dccp_v6_exit); 1176 1177 /* 1178 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33) 1179 * values directly, Also cover the case where the protocol is not specified, 1180 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP 1181 */ 1182 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6); 1183 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6); 1184 MODULE_LICENSE("GPL"); 1185 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>"); 1186 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol"); 1187