1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * DCCP over IPv6 4 * Linux INET6 implementation 5 * 6 * Based on net/dccp6/ipv6.c 7 * 8 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net> 9 */ 10 11 #include <linux/module.h> 12 #include <linux/random.h> 13 #include <linux/slab.h> 14 #include <linux/xfrm.h> 15 #include <linux/string.h> 16 17 #include <net/addrconf.h> 18 #include <net/inet_common.h> 19 #include <net/inet_hashtables.h> 20 #include <net/inet_sock.h> 21 #include <net/inet6_connection_sock.h> 22 #include <net/inet6_hashtables.h> 23 #include <net/ip6_route.h> 24 #include <net/ipv6.h> 25 #include <net/protocol.h> 26 #include <net/transp_v6.h> 27 #include <net/ip6_checksum.h> 28 #include <net/xfrm.h> 29 #include <net/secure_seq.h> 30 #include <net/netns/generic.h> 31 #include <net/sock.h> 32 33 #include "dccp.h" 34 #include "ipv6.h" 35 #include "feat.h" 36 37 struct dccp_v6_pernet { 38 struct sock *v6_ctl_sk; 39 }; 40 41 static unsigned int dccp_v6_pernet_id __read_mostly; 42 43 /* The per-net v6_ctl_sk is used for sending RSTs and ACKs */ 44 45 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped; 46 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops; 47 48 /* add pseudo-header to DCCP checksum stored in skb->csum */ 49 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb, 50 const struct in6_addr *saddr, 51 const struct in6_addr *daddr) 52 { 53 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum); 54 } 55 56 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb) 57 { 58 struct ipv6_pinfo *np = inet6_sk(sk); 59 struct dccp_hdr *dh = dccp_hdr(skb); 60 61 dccp_csum_outgoing(skb); 62 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr); 63 } 64 65 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb) 66 { 67 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, 68 ipv6_hdr(skb)->saddr.s6_addr32, 69 dccp_hdr(skb)->dccph_dport, 70 dccp_hdr(skb)->dccph_sport ); 71 72 } 73 74 static int dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 75 u8 type, u8 code, int offset, __be32 info) 76 { 77 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data; 78 const struct dccp_hdr *dh; 79 struct dccp_sock *dp; 80 struct ipv6_pinfo *np; 81 struct sock *sk; 82 int err; 83 __u64 seq; 84 struct net *net = dev_net(skb->dev); 85 86 /* Only need dccph_dport & dccph_sport which are the first 87 * 4 bytes in dccp header. 88 * Our caller (icmpv6_notify()) already pulled 8 bytes for us. 89 */ 90 BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_sport) > 8); 91 BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_dport) > 8); 92 dh = (struct dccp_hdr *)(skb->data + offset); 93 94 sk = __inet6_lookup_established(net, &dccp_hashinfo, 95 &hdr->daddr, dh->dccph_dport, 96 &hdr->saddr, ntohs(dh->dccph_sport), 97 inet6_iif(skb), 0); 98 99 if (!sk) { 100 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev), 101 ICMP6_MIB_INERRORS); 102 return -ENOENT; 103 } 104 105 if (sk->sk_state == DCCP_TIME_WAIT) { 106 inet_twsk_put(inet_twsk(sk)); 107 return 0; 108 } 109 seq = dccp_hdr_seq(dh); 110 if (sk->sk_state == DCCP_NEW_SYN_RECV) { 111 dccp_req_err(sk, seq); 112 return 0; 113 } 114 115 bh_lock_sock(sk); 116 if (sock_owned_by_user(sk)) 117 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS); 118 119 if (sk->sk_state == DCCP_CLOSED) 120 goto out; 121 122 dp = dccp_sk(sk); 123 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) && 124 !between48(seq, dp->dccps_awl, dp->dccps_awh)) { 125 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS); 126 goto out; 127 } 128 129 np = inet6_sk(sk); 130 131 if (type == NDISC_REDIRECT) { 132 if (!sock_owned_by_user(sk)) { 133 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie); 134 135 if (dst) 136 dst->ops->redirect(dst, sk, skb); 137 } 138 goto out; 139 } 140 141 if (type == ICMPV6_PKT_TOOBIG) { 142 struct dst_entry *dst = NULL; 143 144 if (!ip6_sk_accept_pmtu(sk)) 145 goto out; 146 147 if (sock_owned_by_user(sk)) 148 goto out; 149 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED)) 150 goto out; 151 152 dst = inet6_csk_update_pmtu(sk, ntohl(info)); 153 if (!dst) 154 goto out; 155 156 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) 157 dccp_sync_mss(sk, dst_mtu(dst)); 158 goto out; 159 } 160 161 icmpv6_err_convert(type, code, &err); 162 163 /* Might be for an request_sock */ 164 switch (sk->sk_state) { 165 case DCCP_REQUESTING: 166 case DCCP_RESPOND: /* Cannot happen. 167 It can, it SYNs are crossed. --ANK */ 168 if (!sock_owned_by_user(sk)) { 169 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS); 170 sk->sk_err = err; 171 /* 172 * Wake people up to see the error 173 * (see connect in sock.c) 174 */ 175 sk_error_report(sk); 176 dccp_done(sk); 177 } else 178 sk->sk_err_soft = err; 179 goto out; 180 } 181 182 if (!sock_owned_by_user(sk) && np->recverr) { 183 sk->sk_err = err; 184 sk_error_report(sk); 185 } else 186 sk->sk_err_soft = err; 187 188 out: 189 bh_unlock_sock(sk); 190 sock_put(sk); 191 return 0; 192 } 193 194 195 static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req) 196 { 197 struct inet_request_sock *ireq = inet_rsk(req); 198 struct ipv6_pinfo *np = inet6_sk(sk); 199 struct sk_buff *skb; 200 struct in6_addr *final_p, final; 201 struct flowi6 fl6; 202 int err = -1; 203 struct dst_entry *dst; 204 205 memset(&fl6, 0, sizeof(fl6)); 206 fl6.flowi6_proto = IPPROTO_DCCP; 207 fl6.daddr = ireq->ir_v6_rmt_addr; 208 fl6.saddr = ireq->ir_v6_loc_addr; 209 fl6.flowlabel = 0; 210 fl6.flowi6_oif = ireq->ir_iif; 211 fl6.fl6_dport = ireq->ir_rmt_port; 212 fl6.fl6_sport = htons(ireq->ir_num); 213 security_req_classify_flow(req, flowi6_to_flowi_common(&fl6)); 214 215 216 rcu_read_lock(); 217 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final); 218 rcu_read_unlock(); 219 220 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 221 if (IS_ERR(dst)) { 222 err = PTR_ERR(dst); 223 dst = NULL; 224 goto done; 225 } 226 227 skb = dccp_make_response(sk, dst, req); 228 if (skb != NULL) { 229 struct dccp_hdr *dh = dccp_hdr(skb); 230 struct ipv6_txoptions *opt; 231 232 dh->dccph_checksum = dccp_v6_csum_finish(skb, 233 &ireq->ir_v6_loc_addr, 234 &ireq->ir_v6_rmt_addr); 235 fl6.daddr = ireq->ir_v6_rmt_addr; 236 rcu_read_lock(); 237 opt = ireq->ipv6_opt; 238 if (!opt) 239 opt = rcu_dereference(np->opt); 240 err = ip6_xmit(sk, skb, &fl6, sk->sk_mark, opt, np->tclass, 241 sk->sk_priority); 242 rcu_read_unlock(); 243 err = net_xmit_eval(err); 244 } 245 246 done: 247 dst_release(dst); 248 return err; 249 } 250 251 static void dccp_v6_reqsk_destructor(struct request_sock *req) 252 { 253 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg); 254 kfree(inet_rsk(req)->ipv6_opt); 255 kfree_skb(inet_rsk(req)->pktopts); 256 } 257 258 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb) 259 { 260 const struct ipv6hdr *rxip6h; 261 struct sk_buff *skb; 262 struct flowi6 fl6; 263 struct net *net = dev_net(skb_dst(rxskb)->dev); 264 struct dccp_v6_pernet *pn; 265 struct sock *ctl_sk; 266 struct dst_entry *dst; 267 268 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET) 269 return; 270 271 if (!ipv6_unicast_destination(rxskb)) 272 return; 273 274 pn = net_generic(net, dccp_v6_pernet_id); 275 ctl_sk = pn->v6_ctl_sk; 276 skb = dccp_ctl_make_reset(ctl_sk, rxskb); 277 if (skb == NULL) 278 return; 279 280 rxip6h = ipv6_hdr(rxskb); 281 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr, 282 &rxip6h->daddr); 283 284 memset(&fl6, 0, sizeof(fl6)); 285 fl6.daddr = rxip6h->saddr; 286 fl6.saddr = rxip6h->daddr; 287 288 fl6.flowi6_proto = IPPROTO_DCCP; 289 fl6.flowi6_oif = inet6_iif(rxskb); 290 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport; 291 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport; 292 security_skb_classify_flow(rxskb, flowi6_to_flowi_common(&fl6)); 293 294 /* sk = NULL, but it is safe for now. RST socket required. */ 295 dst = ip6_dst_lookup_flow(sock_net(ctl_sk), ctl_sk, &fl6, NULL); 296 if (!IS_ERR(dst)) { 297 skb_dst_set(skb, dst); 298 ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0, 0); 299 DCCP_INC_STATS(DCCP_MIB_OUTSEGS); 300 DCCP_INC_STATS(DCCP_MIB_OUTRSTS); 301 return; 302 } 303 304 kfree_skb(skb); 305 } 306 307 static struct request_sock_ops dccp6_request_sock_ops = { 308 .family = AF_INET6, 309 .obj_size = sizeof(struct dccp6_request_sock), 310 .rtx_syn_ack = dccp_v6_send_response, 311 .send_ack = dccp_reqsk_send_ack, 312 .destructor = dccp_v6_reqsk_destructor, 313 .send_reset = dccp_v6_ctl_send_reset, 314 .syn_ack_timeout = dccp_syn_ack_timeout, 315 }; 316 317 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb) 318 { 319 struct request_sock *req; 320 struct dccp_request_sock *dreq; 321 struct inet_request_sock *ireq; 322 struct ipv6_pinfo *np = inet6_sk(sk); 323 const __be32 service = dccp_hdr_request(skb)->dccph_req_service; 324 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); 325 326 if (skb->protocol == htons(ETH_P_IP)) 327 return dccp_v4_conn_request(sk, skb); 328 329 if (!ipv6_unicast_destination(skb)) 330 return 0; /* discard, don't send a reset here */ 331 332 if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) { 333 __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS); 334 return 0; 335 } 336 337 if (dccp_bad_service_code(sk, service)) { 338 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE; 339 goto drop; 340 } 341 /* 342 * There are no SYN attacks on IPv6, yet... 343 */ 344 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY; 345 if (inet_csk_reqsk_queue_is_full(sk)) 346 goto drop; 347 348 if (sk_acceptq_is_full(sk)) 349 goto drop; 350 351 req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true); 352 if (req == NULL) 353 goto drop; 354 355 if (dccp_reqsk_init(req, dccp_sk(sk), skb)) 356 goto drop_and_free; 357 358 dreq = dccp_rsk(req); 359 if (dccp_parse_options(sk, dreq, skb)) 360 goto drop_and_free; 361 362 if (security_inet_conn_request(sk, skb, req)) 363 goto drop_and_free; 364 365 ireq = inet_rsk(req); 366 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr; 367 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr; 368 ireq->ireq_family = AF_INET6; 369 ireq->ir_mark = inet_request_mark(sk, skb); 370 371 if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) || 372 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || 373 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { 374 refcount_inc(&skb->users); 375 ireq->pktopts = skb; 376 } 377 ireq->ir_iif = READ_ONCE(sk->sk_bound_dev_if); 378 379 /* So that link locals have meaning */ 380 if (!ireq->ir_iif && 381 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL) 382 ireq->ir_iif = inet6_iif(skb); 383 384 /* 385 * Step 3: Process LISTEN state 386 * 387 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie 388 * 389 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child(). 390 */ 391 dreq->dreq_isr = dcb->dccpd_seq; 392 dreq->dreq_gsr = dreq->dreq_isr; 393 dreq->dreq_iss = dccp_v6_init_sequence(skb); 394 dreq->dreq_gss = dreq->dreq_iss; 395 dreq->dreq_service = service; 396 397 if (dccp_v6_send_response(sk, req)) 398 goto drop_and_free; 399 400 inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT); 401 reqsk_put(req); 402 return 0; 403 404 drop_and_free: 405 reqsk_free(req); 406 drop: 407 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS); 408 return -1; 409 } 410 411 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk, 412 struct sk_buff *skb, 413 struct request_sock *req, 414 struct dst_entry *dst, 415 struct request_sock *req_unhash, 416 bool *own_req) 417 { 418 struct inet_request_sock *ireq = inet_rsk(req); 419 struct ipv6_pinfo *newnp; 420 const struct ipv6_pinfo *np = inet6_sk(sk); 421 struct ipv6_txoptions *opt; 422 struct inet_sock *newinet; 423 struct dccp6_sock *newdp6; 424 struct sock *newsk; 425 426 if (skb->protocol == htons(ETH_P_IP)) { 427 /* 428 * v6 mapped 429 */ 430 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst, 431 req_unhash, own_req); 432 if (newsk == NULL) 433 return NULL; 434 435 newdp6 = (struct dccp6_sock *)newsk; 436 newinet = inet_sk(newsk); 437 newinet->pinet6 = &newdp6->inet6; 438 newnp = inet6_sk(newsk); 439 440 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 441 442 newnp->saddr = newsk->sk_v6_rcv_saddr; 443 444 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped; 445 newsk->sk_backlog_rcv = dccp_v4_do_rcv; 446 newnp->pktoptions = NULL; 447 newnp->opt = NULL; 448 newnp->ipv6_mc_list = NULL; 449 newnp->ipv6_ac_list = NULL; 450 newnp->ipv6_fl_list = NULL; 451 newnp->mcast_oif = inet_iif(skb); 452 newnp->mcast_hops = ip_hdr(skb)->ttl; 453 454 /* 455 * No need to charge this sock to the relevant IPv6 refcnt debug socks count 456 * here, dccp_create_openreq_child now does this for us, see the comment in 457 * that function for the gory details. -acme 458 */ 459 460 /* It is tricky place. Until this moment IPv4 tcp 461 worked with IPv6 icsk.icsk_af_ops. 462 Sync it now. 463 */ 464 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); 465 466 return newsk; 467 } 468 469 470 if (sk_acceptq_is_full(sk)) 471 goto out_overflow; 472 473 if (!dst) { 474 struct flowi6 fl6; 475 476 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP); 477 if (!dst) 478 goto out; 479 } 480 481 newsk = dccp_create_openreq_child(sk, req, skb); 482 if (newsk == NULL) 483 goto out_nonewsk; 484 485 /* 486 * No need to charge this sock to the relevant IPv6 refcnt debug socks 487 * count here, dccp_create_openreq_child now does this for us, see the 488 * comment in that function for the gory details. -acme 489 */ 490 491 ip6_dst_store(newsk, dst, NULL, NULL); 492 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM | 493 NETIF_F_TSO); 494 newdp6 = (struct dccp6_sock *)newsk; 495 newinet = inet_sk(newsk); 496 newinet->pinet6 = &newdp6->inet6; 497 newnp = inet6_sk(newsk); 498 499 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 500 501 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr; 502 newnp->saddr = ireq->ir_v6_loc_addr; 503 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr; 504 newsk->sk_bound_dev_if = ireq->ir_iif; 505 506 /* Now IPv6 options... 507 508 First: no IPv4 options. 509 */ 510 newinet->inet_opt = NULL; 511 512 /* Clone RX bits */ 513 newnp->rxopt.all = np->rxopt.all; 514 515 newnp->ipv6_mc_list = NULL; 516 newnp->ipv6_ac_list = NULL; 517 newnp->ipv6_fl_list = NULL; 518 newnp->pktoptions = NULL; 519 newnp->opt = NULL; 520 newnp->mcast_oif = inet6_iif(skb); 521 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 522 523 /* 524 * Clone native IPv6 options from listening socket (if any) 525 * 526 * Yes, keeping reference count would be much more clever, but we make 527 * one more one thing there: reattach optmem to newsk. 528 */ 529 opt = ireq->ipv6_opt; 530 if (!opt) 531 opt = rcu_dereference(np->opt); 532 if (opt) { 533 opt = ipv6_dup_options(newsk, opt); 534 RCU_INIT_POINTER(newnp->opt, opt); 535 } 536 inet_csk(newsk)->icsk_ext_hdr_len = 0; 537 if (opt) 538 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen + 539 opt->opt_flen; 540 541 dccp_sync_mss(newsk, dst_mtu(dst)); 542 543 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; 544 newinet->inet_rcv_saddr = LOOPBACK4_IPV6; 545 546 if (__inet_inherit_port(sk, newsk) < 0) { 547 inet_csk_prepare_forced_close(newsk); 548 dccp_done(newsk); 549 goto out; 550 } 551 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL); 552 /* Clone pktoptions received with SYN, if we own the req */ 553 if (*own_req && ireq->pktopts) { 554 newnp->pktoptions = skb_clone_and_charge_r(ireq->pktopts, newsk); 555 consume_skb(ireq->pktopts); 556 ireq->pktopts = NULL; 557 } 558 559 return newsk; 560 561 out_overflow: 562 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); 563 out_nonewsk: 564 dst_release(dst); 565 out: 566 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS); 567 return NULL; 568 } 569 570 /* The socket must have it's spinlock held when we get 571 * here. 572 * 573 * We have a potential double-lock case here, so even when 574 * doing backlog processing we use the BH locking scheme. 575 * This is because we cannot sleep with the original spinlock 576 * held. 577 */ 578 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) 579 { 580 struct ipv6_pinfo *np = inet6_sk(sk); 581 struct sk_buff *opt_skb = NULL; 582 583 /* Imagine: socket is IPv6. IPv4 packet arrives, 584 goes to IPv4 receive handler and backlogged. 585 From backlog it always goes here. Kerboom... 586 Fortunately, dccp_rcv_established and rcv_established 587 handle them correctly, but it is not case with 588 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK 589 */ 590 591 if (skb->protocol == htons(ETH_P_IP)) 592 return dccp_v4_do_rcv(sk, skb); 593 594 if (sk_filter(sk, skb)) 595 goto discard; 596 597 /* 598 * socket locking is here for SMP purposes as backlog rcv is currently 599 * called with bh processing disabled. 600 */ 601 602 /* Do Stevens' IPV6_PKTOPTIONS. 603 604 Yes, guys, it is the only place in our code, where we 605 may make it not affecting IPv4. 606 The rest of code is protocol independent, 607 and I do not like idea to uglify IPv4. 608 609 Actually, all the idea behind IPV6_PKTOPTIONS 610 looks not very well thought. For now we latch 611 options, received in the last packet, enqueued 612 by tcp. Feel free to propose better solution. 613 --ANK (980728) 614 */ 615 if (np->rxopt.all) 616 opt_skb = skb_clone_and_charge_r(skb, sk); 617 618 if (sk->sk_state == DCCP_OPEN) { /* Fast path */ 619 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len)) 620 goto reset; 621 if (opt_skb) 622 goto ipv6_pktoptions; 623 return 0; 624 } 625 626 /* 627 * Step 3: Process LISTEN state 628 * If S.state == LISTEN, 629 * If P.type == Request or P contains a valid Init Cookie option, 630 * (* Must scan the packet's options to check for Init 631 * Cookies. Only Init Cookies are processed here, 632 * however; other options are processed in Step 8. This 633 * scan need only be performed if the endpoint uses Init 634 * Cookies *) 635 * (* Generate a new socket and switch to that socket *) 636 * Set S := new socket for this port pair 637 * S.state = RESPOND 638 * Choose S.ISS (initial seqno) or set from Init Cookies 639 * Initialize S.GAR := S.ISS 640 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies 641 * Continue with S.state == RESPOND 642 * (* A Response packet will be generated in Step 11 *) 643 * Otherwise, 644 * Generate Reset(No Connection) unless P.type == Reset 645 * Drop packet and return 646 * 647 * NOTE: the check for the packet types is done in 648 * dccp_rcv_state_process 649 */ 650 651 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len)) 652 goto reset; 653 if (opt_skb) 654 goto ipv6_pktoptions; 655 return 0; 656 657 reset: 658 dccp_v6_ctl_send_reset(sk, skb); 659 discard: 660 if (opt_skb != NULL) 661 __kfree_skb(opt_skb); 662 kfree_skb(skb); 663 return 0; 664 665 /* Handling IPV6_PKTOPTIONS skb the similar 666 * way it's done for net/ipv6/tcp_ipv6.c 667 */ 668 ipv6_pktoptions: 669 if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) { 670 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo) 671 np->mcast_oif = inet6_iif(opt_skb); 672 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) 673 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit; 674 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass) 675 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb)); 676 if (np->repflow) 677 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb)); 678 if (ipv6_opt_accepted(sk, opt_skb, 679 &DCCP_SKB_CB(opt_skb)->header.h6)) { 680 memmove(IP6CB(opt_skb), 681 &DCCP_SKB_CB(opt_skb)->header.h6, 682 sizeof(struct inet6_skb_parm)); 683 opt_skb = xchg(&np->pktoptions, opt_skb); 684 } else { 685 __kfree_skb(opt_skb); 686 opt_skb = xchg(&np->pktoptions, NULL); 687 } 688 } 689 690 kfree_skb(opt_skb); 691 return 0; 692 } 693 694 static int dccp_v6_rcv(struct sk_buff *skb) 695 { 696 const struct dccp_hdr *dh; 697 bool refcounted; 698 struct sock *sk; 699 int min_cov; 700 701 /* Step 1: Check header basics */ 702 703 if (dccp_invalid_packet(skb)) 704 goto discard_it; 705 706 /* Step 1: If header checksum is incorrect, drop packet and return. */ 707 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr, 708 &ipv6_hdr(skb)->daddr)) { 709 DCCP_WARN("dropped packet with invalid checksum\n"); 710 goto discard_it; 711 } 712 713 dh = dccp_hdr(skb); 714 715 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh); 716 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type; 717 718 if (dccp_packet_without_ack(skb)) 719 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ; 720 else 721 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb); 722 723 lookup: 724 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh), 725 dh->dccph_sport, dh->dccph_dport, 726 inet6_iif(skb), 0, &refcounted); 727 if (!sk) { 728 dccp_pr_debug("failed to look up flow ID in table and " 729 "get corresponding socket\n"); 730 goto no_dccp_socket; 731 } 732 733 /* 734 * Step 2: 735 * ... or S.state == TIMEWAIT, 736 * Generate Reset(No Connection) unless P.type == Reset 737 * Drop packet and return 738 */ 739 if (sk->sk_state == DCCP_TIME_WAIT) { 740 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n"); 741 inet_twsk_put(inet_twsk(sk)); 742 goto no_dccp_socket; 743 } 744 745 if (sk->sk_state == DCCP_NEW_SYN_RECV) { 746 struct request_sock *req = inet_reqsk(sk); 747 struct sock *nsk; 748 749 sk = req->rsk_listener; 750 if (unlikely(sk->sk_state != DCCP_LISTEN)) { 751 inet_csk_reqsk_queue_drop_and_put(sk, req); 752 goto lookup; 753 } 754 sock_hold(sk); 755 refcounted = true; 756 nsk = dccp_check_req(sk, skb, req); 757 if (!nsk) { 758 reqsk_put(req); 759 goto discard_and_relse; 760 } 761 if (nsk == sk) { 762 reqsk_put(req); 763 } else if (dccp_child_process(sk, nsk, skb)) { 764 dccp_v6_ctl_send_reset(sk, skb); 765 goto discard_and_relse; 766 } else { 767 sock_put(sk); 768 return 0; 769 } 770 } 771 /* 772 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage 773 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted 774 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov 775 */ 776 min_cov = dccp_sk(sk)->dccps_pcrlen; 777 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) { 778 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n", 779 dh->dccph_cscov, min_cov); 780 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */ 781 goto discard_and_relse; 782 } 783 784 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 785 goto discard_and_relse; 786 787 return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4, 788 refcounted) ? -1 : 0; 789 790 no_dccp_socket: 791 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 792 goto discard_it; 793 /* 794 * Step 2: 795 * If no socket ... 796 * Generate Reset(No Connection) unless P.type == Reset 797 * Drop packet and return 798 */ 799 if (dh->dccph_type != DCCP_PKT_RESET) { 800 DCCP_SKB_CB(skb)->dccpd_reset_code = 801 DCCP_RESET_CODE_NO_CONNECTION; 802 dccp_v6_ctl_send_reset(sk, skb); 803 } 804 805 discard_it: 806 kfree_skb(skb); 807 return 0; 808 809 discard_and_relse: 810 if (refcounted) 811 sock_put(sk); 812 goto discard_it; 813 } 814 815 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, 816 int addr_len) 817 { 818 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr; 819 struct inet_connection_sock *icsk = inet_csk(sk); 820 struct inet_sock *inet = inet_sk(sk); 821 struct ipv6_pinfo *np = inet6_sk(sk); 822 struct dccp_sock *dp = dccp_sk(sk); 823 struct in6_addr *saddr = NULL, *final_p, final; 824 struct ipv6_txoptions *opt; 825 struct flowi6 fl6; 826 struct dst_entry *dst; 827 int addr_type; 828 int err; 829 830 dp->dccps_role = DCCP_ROLE_CLIENT; 831 832 if (addr_len < SIN6_LEN_RFC2133) 833 return -EINVAL; 834 835 if (usin->sin6_family != AF_INET6) 836 return -EAFNOSUPPORT; 837 838 memset(&fl6, 0, sizeof(fl6)); 839 840 if (np->sndflow) { 841 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 842 IP6_ECN_flow_init(fl6.flowlabel); 843 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) { 844 struct ip6_flowlabel *flowlabel; 845 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel); 846 if (IS_ERR(flowlabel)) 847 return -EINVAL; 848 fl6_sock_release(flowlabel); 849 } 850 } 851 /* 852 * connect() to INADDR_ANY means loopback (BSD'ism). 853 */ 854 if (ipv6_addr_any(&usin->sin6_addr)) 855 usin->sin6_addr.s6_addr[15] = 1; 856 857 addr_type = ipv6_addr_type(&usin->sin6_addr); 858 859 if (addr_type & IPV6_ADDR_MULTICAST) 860 return -ENETUNREACH; 861 862 if (addr_type & IPV6_ADDR_LINKLOCAL) { 863 if (addr_len >= sizeof(struct sockaddr_in6) && 864 usin->sin6_scope_id) { 865 /* If interface is set while binding, indices 866 * must coincide. 867 */ 868 if (sk->sk_bound_dev_if && 869 sk->sk_bound_dev_if != usin->sin6_scope_id) 870 return -EINVAL; 871 872 sk->sk_bound_dev_if = usin->sin6_scope_id; 873 } 874 875 /* Connect to link-local address requires an interface */ 876 if (!sk->sk_bound_dev_if) 877 return -EINVAL; 878 } 879 880 sk->sk_v6_daddr = usin->sin6_addr; 881 np->flow_label = fl6.flowlabel; 882 883 /* 884 * DCCP over IPv4 885 */ 886 if (addr_type == IPV6_ADDR_MAPPED) { 887 u32 exthdrlen = icsk->icsk_ext_hdr_len; 888 struct sockaddr_in sin; 889 890 SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); 891 892 if (ipv6_only_sock(sk)) 893 return -ENETUNREACH; 894 895 sin.sin_family = AF_INET; 896 sin.sin_port = usin->sin6_port; 897 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; 898 899 icsk->icsk_af_ops = &dccp_ipv6_mapped; 900 sk->sk_backlog_rcv = dccp_v4_do_rcv; 901 902 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); 903 if (err) { 904 icsk->icsk_ext_hdr_len = exthdrlen; 905 icsk->icsk_af_ops = &dccp_ipv6_af_ops; 906 sk->sk_backlog_rcv = dccp_v6_do_rcv; 907 goto failure; 908 } 909 np->saddr = sk->sk_v6_rcv_saddr; 910 return err; 911 } 912 913 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr)) 914 saddr = &sk->sk_v6_rcv_saddr; 915 916 fl6.flowi6_proto = IPPROTO_DCCP; 917 fl6.daddr = sk->sk_v6_daddr; 918 fl6.saddr = saddr ? *saddr : np->saddr; 919 fl6.flowi6_oif = sk->sk_bound_dev_if; 920 fl6.fl6_dport = usin->sin6_port; 921 fl6.fl6_sport = inet->inet_sport; 922 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6)); 923 924 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk)); 925 final_p = fl6_update_dst(&fl6, opt, &final); 926 927 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p); 928 if (IS_ERR(dst)) { 929 err = PTR_ERR(dst); 930 goto failure; 931 } 932 933 if (saddr == NULL) { 934 saddr = &fl6.saddr; 935 936 err = inet_bhash2_update_saddr(sk, saddr, AF_INET6); 937 if (err) 938 goto failure; 939 } 940 941 /* set the source address */ 942 np->saddr = *saddr; 943 inet->inet_rcv_saddr = LOOPBACK4_IPV6; 944 945 ip6_dst_store(sk, dst, NULL, NULL); 946 947 icsk->icsk_ext_hdr_len = 0; 948 if (opt) 949 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen; 950 951 inet->inet_dport = usin->sin6_port; 952 953 dccp_set_state(sk, DCCP_REQUESTING); 954 err = inet6_hash_connect(&dccp_death_row, sk); 955 if (err) 956 goto late_failure; 957 958 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32, 959 sk->sk_v6_daddr.s6_addr32, 960 inet->inet_sport, 961 inet->inet_dport); 962 err = dccp_connect(sk); 963 if (err) 964 goto late_failure; 965 966 return 0; 967 968 late_failure: 969 dccp_set_state(sk, DCCP_CLOSED); 970 inet_bhash2_reset_saddr(sk); 971 __sk_dst_reset(sk); 972 failure: 973 inet->inet_dport = 0; 974 sk->sk_route_caps = 0; 975 return err; 976 } 977 978 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = { 979 .queue_xmit = inet6_csk_xmit, 980 .send_check = dccp_v6_send_check, 981 .rebuild_header = inet6_sk_rebuild_header, 982 .conn_request = dccp_v6_conn_request, 983 .syn_recv_sock = dccp_v6_request_recv_sock, 984 .net_header_len = sizeof(struct ipv6hdr), 985 .setsockopt = ipv6_setsockopt, 986 .getsockopt = ipv6_getsockopt, 987 .addr2sockaddr = inet6_csk_addr2sockaddr, 988 .sockaddr_len = sizeof(struct sockaddr_in6), 989 }; 990 991 /* 992 * DCCP over IPv4 via INET6 API 993 */ 994 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = { 995 .queue_xmit = ip_queue_xmit, 996 .send_check = dccp_v4_send_check, 997 .rebuild_header = inet_sk_rebuild_header, 998 .conn_request = dccp_v6_conn_request, 999 .syn_recv_sock = dccp_v6_request_recv_sock, 1000 .net_header_len = sizeof(struct iphdr), 1001 .setsockopt = ipv6_setsockopt, 1002 .getsockopt = ipv6_getsockopt, 1003 .addr2sockaddr = inet6_csk_addr2sockaddr, 1004 .sockaddr_len = sizeof(struct sockaddr_in6), 1005 }; 1006 1007 static void dccp_v6_sk_destruct(struct sock *sk) 1008 { 1009 dccp_destruct_common(sk); 1010 inet6_sock_destruct(sk); 1011 } 1012 1013 /* NOTE: A lot of things set to zero explicitly by call to 1014 * sk_alloc() so need not be done here. 1015 */ 1016 static int dccp_v6_init_sock(struct sock *sk) 1017 { 1018 static __u8 dccp_v6_ctl_sock_initialized; 1019 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized); 1020 1021 if (err == 0) { 1022 if (unlikely(!dccp_v6_ctl_sock_initialized)) 1023 dccp_v6_ctl_sock_initialized = 1; 1024 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops; 1025 sk->sk_destruct = dccp_v6_sk_destruct; 1026 } 1027 1028 return err; 1029 } 1030 1031 static struct timewait_sock_ops dccp6_timewait_sock_ops = { 1032 .twsk_obj_size = sizeof(struct dccp6_timewait_sock), 1033 }; 1034 1035 static struct proto dccp_v6_prot = { 1036 .name = "DCCPv6", 1037 .owner = THIS_MODULE, 1038 .close = dccp_close, 1039 .connect = dccp_v6_connect, 1040 .disconnect = dccp_disconnect, 1041 .ioctl = dccp_ioctl, 1042 .init = dccp_v6_init_sock, 1043 .setsockopt = dccp_setsockopt, 1044 .getsockopt = dccp_getsockopt, 1045 .sendmsg = dccp_sendmsg, 1046 .recvmsg = dccp_recvmsg, 1047 .backlog_rcv = dccp_v6_do_rcv, 1048 .hash = inet6_hash, 1049 .unhash = inet_unhash, 1050 .accept = inet_csk_accept, 1051 .get_port = inet_csk_get_port, 1052 .shutdown = dccp_shutdown, 1053 .destroy = dccp_destroy_sock, 1054 .orphan_count = &dccp_orphan_count, 1055 .max_header = MAX_DCCP_HEADER, 1056 .obj_size = sizeof(struct dccp6_sock), 1057 .slab_flags = SLAB_TYPESAFE_BY_RCU, 1058 .rsk_prot = &dccp6_request_sock_ops, 1059 .twsk_prot = &dccp6_timewait_sock_ops, 1060 .h.hashinfo = &dccp_hashinfo, 1061 }; 1062 1063 static const struct inet6_protocol dccp_v6_protocol = { 1064 .handler = dccp_v6_rcv, 1065 .err_handler = dccp_v6_err, 1066 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL, 1067 }; 1068 1069 static const struct proto_ops inet6_dccp_ops = { 1070 .family = PF_INET6, 1071 .owner = THIS_MODULE, 1072 .release = inet6_release, 1073 .bind = inet6_bind, 1074 .connect = inet_stream_connect, 1075 .socketpair = sock_no_socketpair, 1076 .accept = inet_accept, 1077 .getname = inet6_getname, 1078 .poll = dccp_poll, 1079 .ioctl = inet6_ioctl, 1080 .gettstamp = sock_gettstamp, 1081 .listen = inet_dccp_listen, 1082 .shutdown = inet_shutdown, 1083 .setsockopt = sock_common_setsockopt, 1084 .getsockopt = sock_common_getsockopt, 1085 .sendmsg = inet_sendmsg, 1086 .recvmsg = sock_common_recvmsg, 1087 .mmap = sock_no_mmap, 1088 .sendpage = sock_no_sendpage, 1089 #ifdef CONFIG_COMPAT 1090 .compat_ioctl = inet6_compat_ioctl, 1091 #endif 1092 }; 1093 1094 static struct inet_protosw dccp_v6_protosw = { 1095 .type = SOCK_DCCP, 1096 .protocol = IPPROTO_DCCP, 1097 .prot = &dccp_v6_prot, 1098 .ops = &inet6_dccp_ops, 1099 .flags = INET_PROTOSW_ICSK, 1100 }; 1101 1102 static int __net_init dccp_v6_init_net(struct net *net) 1103 { 1104 struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id); 1105 1106 if (dccp_hashinfo.bhash == NULL) 1107 return -ESOCKTNOSUPPORT; 1108 1109 return inet_ctl_sock_create(&pn->v6_ctl_sk, PF_INET6, 1110 SOCK_DCCP, IPPROTO_DCCP, net); 1111 } 1112 1113 static void __net_exit dccp_v6_exit_net(struct net *net) 1114 { 1115 struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id); 1116 1117 inet_ctl_sock_destroy(pn->v6_ctl_sk); 1118 } 1119 1120 static void __net_exit dccp_v6_exit_batch(struct list_head *net_exit_list) 1121 { 1122 inet_twsk_purge(&dccp_hashinfo, AF_INET6); 1123 } 1124 1125 static struct pernet_operations dccp_v6_ops = { 1126 .init = dccp_v6_init_net, 1127 .exit = dccp_v6_exit_net, 1128 .exit_batch = dccp_v6_exit_batch, 1129 .id = &dccp_v6_pernet_id, 1130 .size = sizeof(struct dccp_v6_pernet), 1131 }; 1132 1133 static int __init dccp_v6_init(void) 1134 { 1135 int err = proto_register(&dccp_v6_prot, 1); 1136 1137 if (err) 1138 goto out; 1139 1140 inet6_register_protosw(&dccp_v6_protosw); 1141 1142 err = register_pernet_subsys(&dccp_v6_ops); 1143 if (err) 1144 goto out_destroy_ctl_sock; 1145 1146 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1147 if (err) 1148 goto out_unregister_proto; 1149 1150 out: 1151 return err; 1152 out_unregister_proto: 1153 unregister_pernet_subsys(&dccp_v6_ops); 1154 out_destroy_ctl_sock: 1155 inet6_unregister_protosw(&dccp_v6_protosw); 1156 proto_unregister(&dccp_v6_prot); 1157 goto out; 1158 } 1159 1160 static void __exit dccp_v6_exit(void) 1161 { 1162 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1163 unregister_pernet_subsys(&dccp_v6_ops); 1164 inet6_unregister_protosw(&dccp_v6_protosw); 1165 proto_unregister(&dccp_v6_prot); 1166 } 1167 1168 module_init(dccp_v6_init); 1169 module_exit(dccp_v6_exit); 1170 1171 /* 1172 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33) 1173 * values directly, Also cover the case where the protocol is not specified, 1174 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP 1175 */ 1176 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6); 1177 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6); 1178 MODULE_LICENSE("GPL"); 1179 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>"); 1180 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol"); 1181