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 33 #include "dccp.h" 34 #include "ipv6.h" 35 #include "feat.h" 36 37 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */ 38 39 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped; 40 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops; 41 42 static void dccp_v6_hash(struct sock *sk) 43 { 44 if (sk->sk_state != DCCP_CLOSED) { 45 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) { 46 inet_hash(sk); 47 return; 48 } 49 local_bh_disable(); 50 __inet6_hash(sk, NULL); 51 local_bh_enable(); 52 } 53 } 54 55 /* add pseudo-header to DCCP checksum stored in skb->csum */ 56 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb, 57 struct in6_addr *saddr, 58 struct in6_addr *daddr) 59 { 60 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum); 61 } 62 63 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb) 64 { 65 struct ipv6_pinfo *np = inet6_sk(sk); 66 struct dccp_hdr *dh = dccp_hdr(skb); 67 68 dccp_csum_outgoing(skb); 69 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr); 70 } 71 72 static inline __u32 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr, 73 __be16 sport, __be16 dport ) 74 { 75 return secure_tcpv6_sequence_number(saddr, daddr, sport, dport); 76 } 77 78 static inline __u32 dccp_v6_init_sequence(struct sk_buff *skb) 79 { 80 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, 81 ipv6_hdr(skb)->saddr.s6_addr32, 82 dccp_hdr(skb)->dccph_dport, 83 dccp_hdr(skb)->dccph_sport ); 84 85 } 86 87 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 88 u8 type, u8 code, int offset, __be32 info) 89 { 90 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data; 91 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset); 92 struct dccp_sock *dp; 93 struct ipv6_pinfo *np; 94 struct sock *sk; 95 int err; 96 __u64 seq; 97 struct net *net = dev_net(skb->dev); 98 99 if (skb->len < offset + sizeof(*dh) || 100 skb->len < offset + __dccp_basic_hdr_len(dh)) { 101 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), 102 ICMP6_MIB_INERRORS); 103 return; 104 } 105 106 sk = inet6_lookup(net, &dccp_hashinfo, 107 &hdr->daddr, dh->dccph_dport, 108 &hdr->saddr, dh->dccph_sport, inet6_iif(skb)); 109 110 if (sk == NULL) { 111 ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev), 112 ICMP6_MIB_INERRORS); 113 return; 114 } 115 116 if (sk->sk_state == DCCP_TIME_WAIT) { 117 inet_twsk_put(inet_twsk(sk)); 118 return; 119 } 120 121 bh_lock_sock(sk); 122 if (sock_owned_by_user(sk)) 123 NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS); 124 125 if (sk->sk_state == DCCP_CLOSED) 126 goto out; 127 128 dp = dccp_sk(sk); 129 seq = dccp_hdr_seq(dh); 130 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) && 131 !between48(seq, dp->dccps_awl, dp->dccps_awh)) { 132 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); 133 goto out; 134 } 135 136 np = inet6_sk(sk); 137 138 if (type == ICMPV6_PKT_TOOBIG) { 139 struct dst_entry *dst = NULL; 140 141 if (sock_owned_by_user(sk)) 142 goto out; 143 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED)) 144 goto out; 145 146 /* icmp should have updated the destination cache entry */ 147 dst = __sk_dst_check(sk, np->dst_cookie); 148 if (dst == NULL) { 149 struct inet_sock *inet = inet_sk(sk); 150 struct flowi fl; 151 152 /* BUGGG_FUTURE: Again, it is not clear how 153 to handle rthdr case. Ignore this complexity 154 for now. 155 */ 156 memset(&fl, 0, sizeof(fl)); 157 fl.proto = IPPROTO_DCCP; 158 ipv6_addr_copy(&fl.fl6_dst, &np->daddr); 159 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 160 fl.oif = sk->sk_bound_dev_if; 161 fl.fl_ip_dport = inet->inet_dport; 162 fl.fl_ip_sport = inet->inet_sport; 163 security_sk_classify_flow(sk, &fl); 164 165 err = ip6_dst_lookup(sk, &dst, &fl); 166 if (err) { 167 sk->sk_err_soft = -err; 168 goto out; 169 } 170 171 err = xfrm_lookup(net, &dst, &fl, sk, 0); 172 if (err < 0) { 173 sk->sk_err_soft = -err; 174 goto out; 175 } 176 } else 177 dst_hold(dst); 178 179 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { 180 dccp_sync_mss(sk, dst_mtu(dst)); 181 } /* else let the usual retransmit timer handle it */ 182 dst_release(dst); 183 goto out; 184 } 185 186 icmpv6_err_convert(type, code, &err); 187 188 /* Might be for an request_sock */ 189 switch (sk->sk_state) { 190 struct request_sock *req, **prev; 191 case DCCP_LISTEN: 192 if (sock_owned_by_user(sk)) 193 goto out; 194 195 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport, 196 &hdr->daddr, &hdr->saddr, 197 inet6_iif(skb)); 198 if (req == NULL) 199 goto out; 200 201 /* 202 * ICMPs are not backlogged, hence we cannot get an established 203 * socket here. 204 */ 205 WARN_ON(req->sk != NULL); 206 207 if (seq != dccp_rsk(req)->dreq_iss) { 208 NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS); 209 goto out; 210 } 211 212 inet_csk_reqsk_queue_drop(sk, req, prev); 213 goto out; 214 215 case DCCP_REQUESTING: 216 case DCCP_RESPOND: /* Cannot happen. 217 It can, it SYNs are crossed. --ANK */ 218 if (!sock_owned_by_user(sk)) { 219 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 220 sk->sk_err = err; 221 /* 222 * Wake people up to see the error 223 * (see connect in sock.c) 224 */ 225 sk->sk_error_report(sk); 226 dccp_done(sk); 227 } else 228 sk->sk_err_soft = err; 229 goto out; 230 } 231 232 if (!sock_owned_by_user(sk) && np->recverr) { 233 sk->sk_err = err; 234 sk->sk_error_report(sk); 235 } else 236 sk->sk_err_soft = err; 237 238 out: 239 bh_unlock_sock(sk); 240 sock_put(sk); 241 } 242 243 244 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req, 245 struct request_values *rv_unused) 246 { 247 struct inet6_request_sock *ireq6 = inet6_rsk(req); 248 struct ipv6_pinfo *np = inet6_sk(sk); 249 struct sk_buff *skb; 250 struct ipv6_txoptions *opt = NULL; 251 struct in6_addr *final_p, final; 252 struct flowi fl; 253 int err = -1; 254 struct dst_entry *dst; 255 256 memset(&fl, 0, sizeof(fl)); 257 fl.proto = IPPROTO_DCCP; 258 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr); 259 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr); 260 fl.fl6_flowlabel = 0; 261 fl.oif = ireq6->iif; 262 fl.fl_ip_dport = inet_rsk(req)->rmt_port; 263 fl.fl_ip_sport = inet_rsk(req)->loc_port; 264 security_req_classify_flow(req, &fl); 265 266 opt = np->opt; 267 268 final_p = fl6_update_dst(&fl, opt, &final); 269 270 err = ip6_dst_lookup(sk, &dst, &fl); 271 if (err) 272 goto done; 273 274 if (final_p) 275 ipv6_addr_copy(&fl.fl6_dst, final_p); 276 277 err = xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0); 278 if (err < 0) 279 goto done; 280 281 skb = dccp_make_response(sk, dst, req); 282 if (skb != NULL) { 283 struct dccp_hdr *dh = dccp_hdr(skb); 284 285 dh->dccph_checksum = dccp_v6_csum_finish(skb, 286 &ireq6->loc_addr, 287 &ireq6->rmt_addr); 288 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr); 289 err = ip6_xmit(sk, skb, &fl, opt); 290 err = net_xmit_eval(err); 291 } 292 293 done: 294 if (opt != NULL && opt != np->opt) 295 sock_kfree_s(sk, opt, opt->tot_len); 296 dst_release(dst); 297 return err; 298 } 299 300 static void dccp_v6_reqsk_destructor(struct request_sock *req) 301 { 302 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg); 303 if (inet6_rsk(req)->pktopts != NULL) 304 kfree_skb(inet6_rsk(req)->pktopts); 305 } 306 307 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb) 308 { 309 struct ipv6hdr *rxip6h; 310 struct sk_buff *skb; 311 struct flowi fl; 312 struct net *net = dev_net(skb_dst(rxskb)->dev); 313 struct sock *ctl_sk = net->dccp.v6_ctl_sk; 314 struct dst_entry *dst; 315 316 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET) 317 return; 318 319 if (!ipv6_unicast_destination(rxskb)) 320 return; 321 322 skb = dccp_ctl_make_reset(ctl_sk, rxskb); 323 if (skb == NULL) 324 return; 325 326 rxip6h = ipv6_hdr(rxskb); 327 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr, 328 &rxip6h->daddr); 329 330 memset(&fl, 0, sizeof(fl)); 331 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr); 332 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr); 333 334 fl.proto = IPPROTO_DCCP; 335 fl.oif = inet6_iif(rxskb); 336 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport; 337 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport; 338 security_skb_classify_flow(rxskb, &fl); 339 340 /* sk = NULL, but it is safe for now. RST socket required. */ 341 if (!ip6_dst_lookup(ctl_sk, &dst, &fl)) { 342 if (xfrm_lookup(net, &dst, &fl, NULL, 0) >= 0) { 343 skb_dst_set(skb, dst); 344 ip6_xmit(ctl_sk, skb, &fl, NULL); 345 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS); 346 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS); 347 return; 348 } 349 } 350 351 kfree_skb(skb); 352 } 353 354 static struct request_sock_ops dccp6_request_sock_ops = { 355 .family = AF_INET6, 356 .obj_size = sizeof(struct dccp6_request_sock), 357 .rtx_syn_ack = dccp_v6_send_response, 358 .send_ack = dccp_reqsk_send_ack, 359 .destructor = dccp_v6_reqsk_destructor, 360 .send_reset = dccp_v6_ctl_send_reset, 361 }; 362 363 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb) 364 { 365 const struct dccp_hdr *dh = dccp_hdr(skb); 366 const struct ipv6hdr *iph = ipv6_hdr(skb); 367 struct sock *nsk; 368 struct request_sock **prev; 369 /* Find possible connection requests. */ 370 struct request_sock *req = inet6_csk_search_req(sk, &prev, 371 dh->dccph_sport, 372 &iph->saddr, 373 &iph->daddr, 374 inet6_iif(skb)); 375 if (req != NULL) 376 return dccp_check_req(sk, skb, req, prev); 377 378 nsk = __inet6_lookup_established(sock_net(sk), &dccp_hashinfo, 379 &iph->saddr, dh->dccph_sport, 380 &iph->daddr, ntohs(dh->dccph_dport), 381 inet6_iif(skb)); 382 if (nsk != NULL) { 383 if (nsk->sk_state != DCCP_TIME_WAIT) { 384 bh_lock_sock(nsk); 385 return nsk; 386 } 387 inet_twsk_put(inet_twsk(nsk)); 388 return NULL; 389 } 390 391 return sk; 392 } 393 394 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb) 395 { 396 struct request_sock *req; 397 struct dccp_request_sock *dreq; 398 struct inet6_request_sock *ireq6; 399 struct ipv6_pinfo *np = inet6_sk(sk); 400 const __be32 service = dccp_hdr_request(skb)->dccph_req_service; 401 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); 402 403 if (skb->protocol == htons(ETH_P_IP)) 404 return dccp_v4_conn_request(sk, skb); 405 406 if (!ipv6_unicast_destination(skb)) 407 return 0; /* discard, don't send a reset here */ 408 409 if (dccp_bad_service_code(sk, service)) { 410 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE; 411 goto drop; 412 } 413 /* 414 * There are no SYN attacks on IPv6, yet... 415 */ 416 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY; 417 if (inet_csk_reqsk_queue_is_full(sk)) 418 goto drop; 419 420 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) 421 goto drop; 422 423 req = inet6_reqsk_alloc(&dccp6_request_sock_ops); 424 if (req == NULL) 425 goto drop; 426 427 if (dccp_reqsk_init(req, dccp_sk(sk), skb)) 428 goto drop_and_free; 429 430 dreq = dccp_rsk(req); 431 if (dccp_parse_options(sk, dreq, skb)) 432 goto drop_and_free; 433 434 if (security_inet_conn_request(sk, skb, req)) 435 goto drop_and_free; 436 437 ireq6 = inet6_rsk(req); 438 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr); 439 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr); 440 441 if (ipv6_opt_accepted(sk, skb) || 442 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || 443 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { 444 atomic_inc(&skb->users); 445 ireq6->pktopts = skb; 446 } 447 ireq6->iif = sk->sk_bound_dev_if; 448 449 /* So that link locals have meaning */ 450 if (!sk->sk_bound_dev_if && 451 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL) 452 ireq6->iif = inet6_iif(skb); 453 454 /* 455 * Step 3: Process LISTEN state 456 * 457 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie 458 * 459 * In fact we defer setting S.GSR, S.SWL, S.SWH to 460 * dccp_create_openreq_child. 461 */ 462 dreq->dreq_isr = dcb->dccpd_seq; 463 dreq->dreq_iss = dccp_v6_init_sequence(skb); 464 dreq->dreq_service = service; 465 466 if (dccp_v6_send_response(sk, req, NULL)) 467 goto drop_and_free; 468 469 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT); 470 return 0; 471 472 drop_and_free: 473 reqsk_free(req); 474 drop: 475 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 476 return -1; 477 } 478 479 static struct sock *dccp_v6_request_recv_sock(struct sock *sk, 480 struct sk_buff *skb, 481 struct request_sock *req, 482 struct dst_entry *dst) 483 { 484 struct inet6_request_sock *ireq6 = inet6_rsk(req); 485 struct ipv6_pinfo *newnp, *np = inet6_sk(sk); 486 struct inet_sock *newinet; 487 struct dccp_sock *newdp; 488 struct dccp6_sock *newdp6; 489 struct sock *newsk; 490 struct ipv6_txoptions *opt; 491 492 if (skb->protocol == htons(ETH_P_IP)) { 493 /* 494 * v6 mapped 495 */ 496 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst); 497 if (newsk == NULL) 498 return NULL; 499 500 newdp6 = (struct dccp6_sock *)newsk; 501 newdp = dccp_sk(newsk); 502 newinet = inet_sk(newsk); 503 newinet->pinet6 = &newdp6->inet6; 504 newnp = inet6_sk(newsk); 505 506 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 507 508 ipv6_addr_set_v4mapped(newinet->inet_daddr, &newnp->daddr); 509 510 ipv6_addr_set_v4mapped(newinet->inet_saddr, &newnp->saddr); 511 512 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr); 513 514 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped; 515 newsk->sk_backlog_rcv = dccp_v4_do_rcv; 516 newnp->pktoptions = NULL; 517 newnp->opt = NULL; 518 newnp->mcast_oif = inet6_iif(skb); 519 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 520 521 /* 522 * No need to charge this sock to the relevant IPv6 refcnt debug socks count 523 * here, dccp_create_openreq_child now does this for us, see the comment in 524 * that function for the gory details. -acme 525 */ 526 527 /* It is tricky place. Until this moment IPv4 tcp 528 worked with IPv6 icsk.icsk_af_ops. 529 Sync it now. 530 */ 531 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); 532 533 return newsk; 534 } 535 536 opt = np->opt; 537 538 if (sk_acceptq_is_full(sk)) 539 goto out_overflow; 540 541 if (dst == NULL) { 542 struct in6_addr *final_p, final; 543 struct flowi fl; 544 545 memset(&fl, 0, sizeof(fl)); 546 fl.proto = IPPROTO_DCCP; 547 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr); 548 final_p = fl6_update_dst(&fl, opt, &final); 549 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr); 550 fl.oif = sk->sk_bound_dev_if; 551 fl.fl_ip_dport = inet_rsk(req)->rmt_port; 552 fl.fl_ip_sport = inet_rsk(req)->loc_port; 553 security_sk_classify_flow(sk, &fl); 554 555 if (ip6_dst_lookup(sk, &dst, &fl)) 556 goto out; 557 558 if (final_p) 559 ipv6_addr_copy(&fl.fl6_dst, final_p); 560 561 if ((xfrm_lookup(sock_net(sk), &dst, &fl, sk, 0)) < 0) 562 goto out; 563 } 564 565 newsk = dccp_create_openreq_child(sk, req, skb); 566 if (newsk == NULL) 567 goto out_nonewsk; 568 569 /* 570 * No need to charge this sock to the relevant IPv6 refcnt debug socks 571 * count here, dccp_create_openreq_child now does this for us, see the 572 * comment in that function for the gory details. -acme 573 */ 574 575 __ip6_dst_store(newsk, dst, NULL, NULL); 576 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM | 577 NETIF_F_TSO); 578 newdp6 = (struct dccp6_sock *)newsk; 579 newinet = inet_sk(newsk); 580 newinet->pinet6 = &newdp6->inet6; 581 newdp = dccp_sk(newsk); 582 newnp = inet6_sk(newsk); 583 584 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 585 586 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr); 587 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr); 588 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr); 589 newsk->sk_bound_dev_if = ireq6->iif; 590 591 /* Now IPv6 options... 592 593 First: no IPv4 options. 594 */ 595 newinet->opt = NULL; 596 597 /* Clone RX bits */ 598 newnp->rxopt.all = np->rxopt.all; 599 600 /* Clone pktoptions received with SYN */ 601 newnp->pktoptions = NULL; 602 if (ireq6->pktopts != NULL) { 603 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC); 604 kfree_skb(ireq6->pktopts); 605 ireq6->pktopts = NULL; 606 if (newnp->pktoptions) 607 skb_set_owner_r(newnp->pktoptions, newsk); 608 } 609 newnp->opt = NULL; 610 newnp->mcast_oif = inet6_iif(skb); 611 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 612 613 /* 614 * Clone native IPv6 options from listening socket (if any) 615 * 616 * Yes, keeping reference count would be much more clever, but we make 617 * one more one thing there: reattach optmem to newsk. 618 */ 619 if (opt != NULL) { 620 newnp->opt = ipv6_dup_options(newsk, opt); 621 if (opt != np->opt) 622 sock_kfree_s(sk, opt, opt->tot_len); 623 } 624 625 inet_csk(newsk)->icsk_ext_hdr_len = 0; 626 if (newnp->opt != NULL) 627 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + 628 newnp->opt->opt_flen); 629 630 dccp_sync_mss(newsk, dst_mtu(dst)); 631 632 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6; 633 newinet->inet_rcv_saddr = LOOPBACK4_IPV6; 634 635 if (__inet_inherit_port(sk, newsk) < 0) { 636 sock_put(newsk); 637 goto out; 638 } 639 __inet6_hash(newsk, NULL); 640 641 return newsk; 642 643 out_overflow: 644 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); 645 out_nonewsk: 646 dst_release(dst); 647 out: 648 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); 649 if (opt != NULL && opt != np->opt) 650 sock_kfree_s(sk, opt, opt->tot_len); 651 return NULL; 652 } 653 654 /* The socket must have it's spinlock held when we get 655 * here. 656 * 657 * We have a potential double-lock case here, so even when 658 * doing backlog processing we use the BH locking scheme. 659 * This is because we cannot sleep with the original spinlock 660 * held. 661 */ 662 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) 663 { 664 struct ipv6_pinfo *np = inet6_sk(sk); 665 struct sk_buff *opt_skb = NULL; 666 667 /* Imagine: socket is IPv6. IPv4 packet arrives, 668 goes to IPv4 receive handler and backlogged. 669 From backlog it always goes here. Kerboom... 670 Fortunately, dccp_rcv_established and rcv_established 671 handle them correctly, but it is not case with 672 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK 673 */ 674 675 if (skb->protocol == htons(ETH_P_IP)) 676 return dccp_v4_do_rcv(sk, skb); 677 678 if (sk_filter(sk, skb)) 679 goto discard; 680 681 /* 682 * socket locking is here for SMP purposes as backlog rcv is currently 683 * called with bh processing disabled. 684 */ 685 686 /* Do Stevens' IPV6_PKTOPTIONS. 687 688 Yes, guys, it is the only place in our code, where we 689 may make it not affecting IPv4. 690 The rest of code is protocol independent, 691 and I do not like idea to uglify IPv4. 692 693 Actually, all the idea behind IPV6_PKTOPTIONS 694 looks not very well thought. For now we latch 695 options, received in the last packet, enqueued 696 by tcp. Feel free to propose better solution. 697 --ANK (980728) 698 */ 699 if (np->rxopt.all) 700 /* 701 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below 702 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example. 703 */ 704 opt_skb = skb_clone(skb, GFP_ATOMIC); 705 706 if (sk->sk_state == DCCP_OPEN) { /* Fast path */ 707 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len)) 708 goto reset; 709 if (opt_skb) { 710 /* XXX This is where we would goto ipv6_pktoptions. */ 711 __kfree_skb(opt_skb); 712 } 713 return 0; 714 } 715 716 /* 717 * Step 3: Process LISTEN state 718 * If S.state == LISTEN, 719 * If P.type == Request or P contains a valid Init Cookie option, 720 * (* Must scan the packet's options to check for Init 721 * Cookies. Only Init Cookies are processed here, 722 * however; other options are processed in Step 8. This 723 * scan need only be performed if the endpoint uses Init 724 * Cookies *) 725 * (* Generate a new socket and switch to that socket *) 726 * Set S := new socket for this port pair 727 * S.state = RESPOND 728 * Choose S.ISS (initial seqno) or set from Init Cookies 729 * Initialize S.GAR := S.ISS 730 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies 731 * Continue with S.state == RESPOND 732 * (* A Response packet will be generated in Step 11 *) 733 * Otherwise, 734 * Generate Reset(No Connection) unless P.type == Reset 735 * Drop packet and return 736 * 737 * NOTE: the check for the packet types is done in 738 * dccp_rcv_state_process 739 */ 740 if (sk->sk_state == DCCP_LISTEN) { 741 struct sock *nsk = dccp_v6_hnd_req(sk, skb); 742 743 if (nsk == NULL) 744 goto discard; 745 /* 746 * Queue it on the new socket if the new socket is active, 747 * otherwise we just shortcircuit this and continue with 748 * the new socket.. 749 */ 750 if (nsk != sk) { 751 if (dccp_child_process(sk, nsk, skb)) 752 goto reset; 753 if (opt_skb != NULL) 754 __kfree_skb(opt_skb); 755 return 0; 756 } 757 } 758 759 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len)) 760 goto reset; 761 if (opt_skb) { 762 /* XXX This is where we would goto ipv6_pktoptions. */ 763 __kfree_skb(opt_skb); 764 } 765 return 0; 766 767 reset: 768 dccp_v6_ctl_send_reset(sk, skb); 769 discard: 770 if (opt_skb != NULL) 771 __kfree_skb(opt_skb); 772 kfree_skb(skb); 773 return 0; 774 } 775 776 static int dccp_v6_rcv(struct sk_buff *skb) 777 { 778 const struct dccp_hdr *dh; 779 struct sock *sk; 780 int min_cov; 781 782 /* Step 1: Check header basics */ 783 784 if (dccp_invalid_packet(skb)) 785 goto discard_it; 786 787 /* Step 1: If header checksum is incorrect, drop packet and return. */ 788 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr, 789 &ipv6_hdr(skb)->daddr)) { 790 DCCP_WARN("dropped packet with invalid checksum\n"); 791 goto discard_it; 792 } 793 794 dh = dccp_hdr(skb); 795 796 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh); 797 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type; 798 799 if (dccp_packet_without_ack(skb)) 800 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ; 801 else 802 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb); 803 804 /* Step 2: 805 * Look up flow ID in table and get corresponding socket */ 806 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, 807 dh->dccph_sport, dh->dccph_dport); 808 /* 809 * Step 2: 810 * If no socket ... 811 */ 812 if (sk == NULL) { 813 dccp_pr_debug("failed to look up flow ID in table and " 814 "get corresponding socket\n"); 815 goto no_dccp_socket; 816 } 817 818 /* 819 * Step 2: 820 * ... or S.state == TIMEWAIT, 821 * Generate Reset(No Connection) unless P.type == Reset 822 * Drop packet and return 823 */ 824 if (sk->sk_state == DCCP_TIME_WAIT) { 825 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n"); 826 inet_twsk_put(inet_twsk(sk)); 827 goto no_dccp_socket; 828 } 829 830 /* 831 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage 832 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted 833 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov 834 */ 835 min_cov = dccp_sk(sk)->dccps_pcrlen; 836 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) { 837 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n", 838 dh->dccph_cscov, min_cov); 839 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */ 840 goto discard_and_relse; 841 } 842 843 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 844 goto discard_and_relse; 845 846 return sk_receive_skb(sk, skb, 1) ? -1 : 0; 847 848 no_dccp_socket: 849 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 850 goto discard_it; 851 /* 852 * Step 2: 853 * If no socket ... 854 * Generate Reset(No Connection) unless P.type == Reset 855 * Drop packet and return 856 */ 857 if (dh->dccph_type != DCCP_PKT_RESET) { 858 DCCP_SKB_CB(skb)->dccpd_reset_code = 859 DCCP_RESET_CODE_NO_CONNECTION; 860 dccp_v6_ctl_send_reset(sk, skb); 861 } 862 863 discard_it: 864 kfree_skb(skb); 865 return 0; 866 867 discard_and_relse: 868 sock_put(sk); 869 goto discard_it; 870 } 871 872 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, 873 int addr_len) 874 { 875 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr; 876 struct inet_connection_sock *icsk = inet_csk(sk); 877 struct inet_sock *inet = inet_sk(sk); 878 struct ipv6_pinfo *np = inet6_sk(sk); 879 struct dccp_sock *dp = dccp_sk(sk); 880 struct in6_addr *saddr = NULL, *final_p, final; 881 struct flowi fl; 882 struct dst_entry *dst; 883 int addr_type; 884 int err; 885 886 dp->dccps_role = DCCP_ROLE_CLIENT; 887 888 if (addr_len < SIN6_LEN_RFC2133) 889 return -EINVAL; 890 891 if (usin->sin6_family != AF_INET6) 892 return -EAFNOSUPPORT; 893 894 memset(&fl, 0, sizeof(fl)); 895 896 if (np->sndflow) { 897 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 898 IP6_ECN_flow_init(fl.fl6_flowlabel); 899 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) { 900 struct ip6_flowlabel *flowlabel; 901 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 902 if (flowlabel == NULL) 903 return -EINVAL; 904 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst); 905 fl6_sock_release(flowlabel); 906 } 907 } 908 /* 909 * connect() to INADDR_ANY means loopback (BSD'ism). 910 */ 911 if (ipv6_addr_any(&usin->sin6_addr)) 912 usin->sin6_addr.s6_addr[15] = 1; 913 914 addr_type = ipv6_addr_type(&usin->sin6_addr); 915 916 if (addr_type & IPV6_ADDR_MULTICAST) 917 return -ENETUNREACH; 918 919 if (addr_type & IPV6_ADDR_LINKLOCAL) { 920 if (addr_len >= sizeof(struct sockaddr_in6) && 921 usin->sin6_scope_id) { 922 /* If interface is set while binding, indices 923 * must coincide. 924 */ 925 if (sk->sk_bound_dev_if && 926 sk->sk_bound_dev_if != usin->sin6_scope_id) 927 return -EINVAL; 928 929 sk->sk_bound_dev_if = usin->sin6_scope_id; 930 } 931 932 /* Connect to link-local address requires an interface */ 933 if (!sk->sk_bound_dev_if) 934 return -EINVAL; 935 } 936 937 ipv6_addr_copy(&np->daddr, &usin->sin6_addr); 938 np->flow_label = fl.fl6_flowlabel; 939 940 /* 941 * DCCP over IPv4 942 */ 943 if (addr_type == IPV6_ADDR_MAPPED) { 944 u32 exthdrlen = icsk->icsk_ext_hdr_len; 945 struct sockaddr_in sin; 946 947 SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); 948 949 if (__ipv6_only_sock(sk)) 950 return -ENETUNREACH; 951 952 sin.sin_family = AF_INET; 953 sin.sin_port = usin->sin6_port; 954 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; 955 956 icsk->icsk_af_ops = &dccp_ipv6_mapped; 957 sk->sk_backlog_rcv = dccp_v4_do_rcv; 958 959 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); 960 if (err) { 961 icsk->icsk_ext_hdr_len = exthdrlen; 962 icsk->icsk_af_ops = &dccp_ipv6_af_ops; 963 sk->sk_backlog_rcv = dccp_v6_do_rcv; 964 goto failure; 965 } 966 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); 967 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr); 968 969 return err; 970 } 971 972 if (!ipv6_addr_any(&np->rcv_saddr)) 973 saddr = &np->rcv_saddr; 974 975 fl.proto = IPPROTO_DCCP; 976 ipv6_addr_copy(&fl.fl6_dst, &np->daddr); 977 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr); 978 fl.oif = sk->sk_bound_dev_if; 979 fl.fl_ip_dport = usin->sin6_port; 980 fl.fl_ip_sport = inet->inet_sport; 981 security_sk_classify_flow(sk, &fl); 982 983 final_p = fl6_update_dst(&fl, np->opt, &final); 984 985 err = ip6_dst_lookup(sk, &dst, &fl); 986 if (err) 987 goto failure; 988 989 if (final_p) 990 ipv6_addr_copy(&fl.fl6_dst, final_p); 991 992 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT); 993 if (err < 0) { 994 if (err == -EREMOTE) 995 err = ip6_dst_blackhole(sk, &dst, &fl); 996 if (err < 0) 997 goto failure; 998 } 999 1000 if (saddr == NULL) { 1001 saddr = &fl.fl6_src; 1002 ipv6_addr_copy(&np->rcv_saddr, saddr); 1003 } 1004 1005 /* set the source address */ 1006 ipv6_addr_copy(&np->saddr, saddr); 1007 inet->inet_rcv_saddr = LOOPBACK4_IPV6; 1008 1009 __ip6_dst_store(sk, dst, NULL, NULL); 1010 1011 icsk->icsk_ext_hdr_len = 0; 1012 if (np->opt != NULL) 1013 icsk->icsk_ext_hdr_len = (np->opt->opt_flen + 1014 np->opt->opt_nflen); 1015 1016 inet->inet_dport = usin->sin6_port; 1017 1018 dccp_set_state(sk, DCCP_REQUESTING); 1019 err = inet6_hash_connect(&dccp_death_row, sk); 1020 if (err) 1021 goto late_failure; 1022 1023 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32, 1024 np->daddr.s6_addr32, 1025 inet->inet_sport, 1026 inet->inet_dport); 1027 err = dccp_connect(sk); 1028 if (err) 1029 goto late_failure; 1030 1031 return 0; 1032 1033 late_failure: 1034 dccp_set_state(sk, DCCP_CLOSED); 1035 __sk_dst_reset(sk); 1036 failure: 1037 inet->inet_dport = 0; 1038 sk->sk_route_caps = 0; 1039 return err; 1040 } 1041 1042 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = { 1043 .queue_xmit = inet6_csk_xmit, 1044 .send_check = dccp_v6_send_check, 1045 .rebuild_header = inet6_sk_rebuild_header, 1046 .conn_request = dccp_v6_conn_request, 1047 .syn_recv_sock = dccp_v6_request_recv_sock, 1048 .net_header_len = sizeof(struct ipv6hdr), 1049 .setsockopt = ipv6_setsockopt, 1050 .getsockopt = ipv6_getsockopt, 1051 .addr2sockaddr = inet6_csk_addr2sockaddr, 1052 .sockaddr_len = sizeof(struct sockaddr_in6), 1053 .bind_conflict = inet6_csk_bind_conflict, 1054 #ifdef CONFIG_COMPAT 1055 .compat_setsockopt = compat_ipv6_setsockopt, 1056 .compat_getsockopt = compat_ipv6_getsockopt, 1057 #endif 1058 }; 1059 1060 /* 1061 * DCCP over IPv4 via INET6 API 1062 */ 1063 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = { 1064 .queue_xmit = ip_queue_xmit, 1065 .send_check = dccp_v4_send_check, 1066 .rebuild_header = inet_sk_rebuild_header, 1067 .conn_request = dccp_v6_conn_request, 1068 .syn_recv_sock = dccp_v6_request_recv_sock, 1069 .net_header_len = sizeof(struct iphdr), 1070 .setsockopt = ipv6_setsockopt, 1071 .getsockopt = ipv6_getsockopt, 1072 .addr2sockaddr = inet6_csk_addr2sockaddr, 1073 .sockaddr_len = sizeof(struct sockaddr_in6), 1074 #ifdef CONFIG_COMPAT 1075 .compat_setsockopt = compat_ipv6_setsockopt, 1076 .compat_getsockopt = compat_ipv6_getsockopt, 1077 #endif 1078 }; 1079 1080 /* NOTE: A lot of things set to zero explicitly by call to 1081 * sk_alloc() so need not be done here. 1082 */ 1083 static int dccp_v6_init_sock(struct sock *sk) 1084 { 1085 static __u8 dccp_v6_ctl_sock_initialized; 1086 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized); 1087 1088 if (err == 0) { 1089 if (unlikely(!dccp_v6_ctl_sock_initialized)) 1090 dccp_v6_ctl_sock_initialized = 1; 1091 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops; 1092 } 1093 1094 return err; 1095 } 1096 1097 static void dccp_v6_destroy_sock(struct sock *sk) 1098 { 1099 dccp_destroy_sock(sk); 1100 inet6_destroy_sock(sk); 1101 } 1102 1103 static struct timewait_sock_ops dccp6_timewait_sock_ops = { 1104 .twsk_obj_size = sizeof(struct dccp6_timewait_sock), 1105 }; 1106 1107 static struct proto dccp_v6_prot = { 1108 .name = "DCCPv6", 1109 .owner = THIS_MODULE, 1110 .close = dccp_close, 1111 .connect = dccp_v6_connect, 1112 .disconnect = dccp_disconnect, 1113 .ioctl = dccp_ioctl, 1114 .init = dccp_v6_init_sock, 1115 .setsockopt = dccp_setsockopt, 1116 .getsockopt = dccp_getsockopt, 1117 .sendmsg = dccp_sendmsg, 1118 .recvmsg = dccp_recvmsg, 1119 .backlog_rcv = dccp_v6_do_rcv, 1120 .hash = dccp_v6_hash, 1121 .unhash = inet_unhash, 1122 .accept = inet_csk_accept, 1123 .get_port = inet_csk_get_port, 1124 .shutdown = dccp_shutdown, 1125 .destroy = dccp_v6_destroy_sock, 1126 .orphan_count = &dccp_orphan_count, 1127 .max_header = MAX_DCCP_HEADER, 1128 .obj_size = sizeof(struct dccp6_sock), 1129 .slab_flags = SLAB_DESTROY_BY_RCU, 1130 .rsk_prot = &dccp6_request_sock_ops, 1131 .twsk_prot = &dccp6_timewait_sock_ops, 1132 .h.hashinfo = &dccp_hashinfo, 1133 #ifdef CONFIG_COMPAT 1134 .compat_setsockopt = compat_dccp_setsockopt, 1135 .compat_getsockopt = compat_dccp_getsockopt, 1136 #endif 1137 }; 1138 1139 static const struct inet6_protocol dccp_v6_protocol = { 1140 .handler = dccp_v6_rcv, 1141 .err_handler = dccp_v6_err, 1142 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL, 1143 }; 1144 1145 static const struct proto_ops inet6_dccp_ops = { 1146 .family = PF_INET6, 1147 .owner = THIS_MODULE, 1148 .release = inet6_release, 1149 .bind = inet6_bind, 1150 .connect = inet_stream_connect, 1151 .socketpair = sock_no_socketpair, 1152 .accept = inet_accept, 1153 .getname = inet6_getname, 1154 .poll = dccp_poll, 1155 .ioctl = inet6_ioctl, 1156 .listen = inet_dccp_listen, 1157 .shutdown = inet_shutdown, 1158 .setsockopt = sock_common_setsockopt, 1159 .getsockopt = sock_common_getsockopt, 1160 .sendmsg = inet_sendmsg, 1161 .recvmsg = sock_common_recvmsg, 1162 .mmap = sock_no_mmap, 1163 .sendpage = sock_no_sendpage, 1164 #ifdef CONFIG_COMPAT 1165 .compat_setsockopt = compat_sock_common_setsockopt, 1166 .compat_getsockopt = compat_sock_common_getsockopt, 1167 #endif 1168 }; 1169 1170 static struct inet_protosw dccp_v6_protosw = { 1171 .type = SOCK_DCCP, 1172 .protocol = IPPROTO_DCCP, 1173 .prot = &dccp_v6_prot, 1174 .ops = &inet6_dccp_ops, 1175 .flags = INET_PROTOSW_ICSK, 1176 }; 1177 1178 static int __net_init dccp_v6_init_net(struct net *net) 1179 { 1180 if (dccp_hashinfo.bhash == NULL) 1181 return -ESOCKTNOSUPPORT; 1182 1183 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6, 1184 SOCK_DCCP, IPPROTO_DCCP, net); 1185 } 1186 1187 static void __net_exit dccp_v6_exit_net(struct net *net) 1188 { 1189 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk); 1190 } 1191 1192 static struct pernet_operations dccp_v6_ops = { 1193 .init = dccp_v6_init_net, 1194 .exit = dccp_v6_exit_net, 1195 }; 1196 1197 static int __init dccp_v6_init(void) 1198 { 1199 int err = proto_register(&dccp_v6_prot, 1); 1200 1201 if (err != 0) 1202 goto out; 1203 1204 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1205 if (err != 0) 1206 goto out_unregister_proto; 1207 1208 inet6_register_protosw(&dccp_v6_protosw); 1209 1210 err = register_pernet_subsys(&dccp_v6_ops); 1211 if (err != 0) 1212 goto out_destroy_ctl_sock; 1213 out: 1214 return err; 1215 1216 out_destroy_ctl_sock: 1217 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1218 inet6_unregister_protosw(&dccp_v6_protosw); 1219 out_unregister_proto: 1220 proto_unregister(&dccp_v6_prot); 1221 goto out; 1222 } 1223 1224 static void __exit dccp_v6_exit(void) 1225 { 1226 unregister_pernet_subsys(&dccp_v6_ops); 1227 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1228 inet6_unregister_protosw(&dccp_v6_protosw); 1229 proto_unregister(&dccp_v6_prot); 1230 } 1231 1232 module_init(dccp_v6_init); 1233 module_exit(dccp_v6_exit); 1234 1235 /* 1236 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33) 1237 * values directly, Also cover the case where the protocol is not specified, 1238 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP 1239 */ 1240 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6); 1241 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6); 1242 MODULE_LICENSE("GPL"); 1243 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>"); 1244 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol"); 1245