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 39static const struct inet_connection_sock_af_ops dccp_ipv6_mapped; 40static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops; 41 42static 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 */ 56static 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 63static 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 72static 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 78static 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 87static 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 238out: 239 bh_unlock_sock(sk); 240 sock_put(sk); 241} 242 243 244static 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 293done: 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 300static 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 307static 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 354static 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 363static 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 394static 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 472drop_and_free: 473 reqsk_free(req); 474drop: 475 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 476 return -1; 477} 478 479static 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; 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 __inet6_hash(newsk, NULL); 636 __inet_inherit_port(sk, newsk); 637 638 return newsk; 639 640out_overflow: 641 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS); 642out: 643 NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS); 644 if (opt != NULL && opt != np->opt) 645 sock_kfree_s(sk, opt, opt->tot_len); 646 dst_release(dst); 647 return NULL; 648} 649 650/* The socket must have it's spinlock held when we get 651 * here. 652 * 653 * We have a potential double-lock case here, so even when 654 * doing backlog processing we use the BH locking scheme. 655 * This is because we cannot sleep with the original spinlock 656 * held. 657 */ 658static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) 659{ 660 struct ipv6_pinfo *np = inet6_sk(sk); 661 struct sk_buff *opt_skb = NULL; 662 663 /* Imagine: socket is IPv6. IPv4 packet arrives, 664 goes to IPv4 receive handler and backlogged. 665 From backlog it always goes here. Kerboom... 666 Fortunately, dccp_rcv_established and rcv_established 667 handle them correctly, but it is not case with 668 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK 669 */ 670 671 if (skb->protocol == htons(ETH_P_IP)) 672 return dccp_v4_do_rcv(sk, skb); 673 674 if (sk_filter(sk, skb)) 675 goto discard; 676 677 /* 678 * socket locking is here for SMP purposes as backlog rcv is currently 679 * called with bh processing disabled. 680 */ 681 682 /* Do Stevens' IPV6_PKTOPTIONS. 683 684 Yes, guys, it is the only place in our code, where we 685 may make it not affecting IPv4. 686 The rest of code is protocol independent, 687 and I do not like idea to uglify IPv4. 688 689 Actually, all the idea behind IPV6_PKTOPTIONS 690 looks not very well thought. For now we latch 691 options, received in the last packet, enqueued 692 by tcp. Feel free to propose better solution. 693 --ANK (980728) 694 */ 695 if (np->rxopt.all) 696 opt_skb = skb_clone(skb, GFP_ATOMIC); 697 698 if (sk->sk_state == DCCP_OPEN) { /* Fast path */ 699 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len)) 700 goto reset; 701 if (opt_skb) { 702 __kfree_skb(opt_skb); 703 } 704 return 0; 705 } 706 707 /* 708 * Step 3: Process LISTEN state 709 * If S.state == LISTEN, 710 * If P.type == Request or P contains a valid Init Cookie option, 711 * (* Must scan the packet's options to check for Init 712 * Cookies. Only Init Cookies are processed here, 713 * however; other options are processed in Step 8. This 714 * scan need only be performed if the endpoint uses Init 715 * Cookies *) 716 * (* Generate a new socket and switch to that socket *) 717 * Set S := new socket for this port pair 718 * S.state = RESPOND 719 * Choose S.ISS (initial seqno) or set from Init Cookies 720 * Initialize S.GAR := S.ISS 721 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies 722 * Continue with S.state == RESPOND 723 * (* A Response packet will be generated in Step 11 *) 724 * Otherwise, 725 * Generate Reset(No Connection) unless P.type == Reset 726 * Drop packet and return 727 * 728 * NOTE: the check for the packet types is done in 729 * dccp_rcv_state_process 730 */ 731 if (sk->sk_state == DCCP_LISTEN) { 732 struct sock *nsk = dccp_v6_hnd_req(sk, skb); 733 734 if (nsk == NULL) 735 goto discard; 736 /* 737 * Queue it on the new socket if the new socket is active, 738 * otherwise we just shortcircuit this and continue with 739 * the new socket.. 740 */ 741 if (nsk != sk) { 742 if (dccp_child_process(sk, nsk, skb)) 743 goto reset; 744 if (opt_skb != NULL) 745 __kfree_skb(opt_skb); 746 return 0; 747 } 748 } 749 750 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len)) 751 goto reset; 752 if (opt_skb) { 753 __kfree_skb(opt_skb); 754 } 755 return 0; 756 757reset: 758 dccp_v6_ctl_send_reset(sk, skb); 759discard: 760 if (opt_skb != NULL) 761 __kfree_skb(opt_skb); 762 kfree_skb(skb); 763 return 0; 764} 765 766static int dccp_v6_rcv(struct sk_buff *skb) 767{ 768 const struct dccp_hdr *dh; 769 struct sock *sk; 770 int min_cov; 771 772 /* Step 1: Check header basics */ 773 774 if (dccp_invalid_packet(skb)) 775 goto discard_it; 776 777 /* Step 1: If header checksum is incorrect, drop packet and return. */ 778 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr, 779 &ipv6_hdr(skb)->daddr)) { 780 DCCP_WARN("dropped packet with invalid checksum\n"); 781 goto discard_it; 782 } 783 784 dh = dccp_hdr(skb); 785 786 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh); 787 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type; 788 789 if (dccp_packet_without_ack(skb)) 790 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ; 791 else 792 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb); 793 794 /* Step 2: 795 * Look up flow ID in table and get corresponding socket */ 796 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, 797 dh->dccph_sport, dh->dccph_dport); 798 /* 799 * Step 2: 800 * If no socket ... 801 */ 802 if (sk == NULL) { 803 dccp_pr_debug("failed to look up flow ID in table and " 804 "get corresponding socket\n"); 805 goto no_dccp_socket; 806 } 807 808 /* 809 * Step 2: 810 * ... or S.state == TIMEWAIT, 811 * Generate Reset(No Connection) unless P.type == Reset 812 * Drop packet and return 813 */ 814 if (sk->sk_state == DCCP_TIME_WAIT) { 815 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n"); 816 inet_twsk_put(inet_twsk(sk)); 817 goto no_dccp_socket; 818 } 819 820 /* 821 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage 822 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted 823 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov 824 */ 825 min_cov = dccp_sk(sk)->dccps_pcrlen; 826 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) { 827 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n", 828 dh->dccph_cscov, min_cov); 829 goto discard_and_relse; 830 } 831 832 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 833 goto discard_and_relse; 834 835 return sk_receive_skb(sk, skb, 1) ? -1 : 0; 836 837no_dccp_socket: 838 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 839 goto discard_it; 840 /* 841 * Step 2: 842 * If no socket ... 843 * Generate Reset(No Connection) unless P.type == Reset 844 * Drop packet and return 845 */ 846 if (dh->dccph_type != DCCP_PKT_RESET) { 847 DCCP_SKB_CB(skb)->dccpd_reset_code = 848 DCCP_RESET_CODE_NO_CONNECTION; 849 dccp_v6_ctl_send_reset(sk, skb); 850 } 851 852discard_it: 853 kfree_skb(skb); 854 return 0; 855 856discard_and_relse: 857 sock_put(sk); 858 goto discard_it; 859} 860 861static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, 862 int addr_len) 863{ 864 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr; 865 struct inet_connection_sock *icsk = inet_csk(sk); 866 struct inet_sock *inet = inet_sk(sk); 867 struct ipv6_pinfo *np = inet6_sk(sk); 868 struct dccp_sock *dp = dccp_sk(sk); 869 struct in6_addr *saddr = NULL, *final_p, final; 870 struct flowi fl; 871 struct dst_entry *dst; 872 int addr_type; 873 int err; 874 875 dp->dccps_role = DCCP_ROLE_CLIENT; 876 877 if (addr_len < SIN6_LEN_RFC2133) 878 return -EINVAL; 879 880 if (usin->sin6_family != AF_INET6) 881 return -EAFNOSUPPORT; 882 883 memset(&fl, 0, sizeof(fl)); 884 885 if (np->sndflow) { 886 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 887 IP6_ECN_flow_init(fl.fl6_flowlabel); 888 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) { 889 struct ip6_flowlabel *flowlabel; 890 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 891 if (flowlabel == NULL) 892 return -EINVAL; 893 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst); 894 fl6_sock_release(flowlabel); 895 } 896 } 897 /* 898 * connect() to INADDR_ANY means loopback (BSD'ism). 899 */ 900 if (ipv6_addr_any(&usin->sin6_addr)) 901 usin->sin6_addr.s6_addr[15] = 1; 902 903 addr_type = ipv6_addr_type(&usin->sin6_addr); 904 905 if (addr_type & IPV6_ADDR_MULTICAST) 906 return -ENETUNREACH; 907 908 if (addr_type & IPV6_ADDR_LINKLOCAL) { 909 if (addr_len >= sizeof(struct sockaddr_in6) && 910 usin->sin6_scope_id) { 911 /* If interface is set while binding, indices 912 * must coincide. 913 */ 914 if (sk->sk_bound_dev_if && 915 sk->sk_bound_dev_if != usin->sin6_scope_id) 916 return -EINVAL; 917 918 sk->sk_bound_dev_if = usin->sin6_scope_id; 919 } 920 921 /* Connect to link-local address requires an interface */ 922 if (!sk->sk_bound_dev_if) 923 return -EINVAL; 924 } 925 926 ipv6_addr_copy(&np->daddr, &usin->sin6_addr); 927 np->flow_label = fl.fl6_flowlabel; 928 929 /* 930 * DCCP over IPv4 931 */ 932 if (addr_type == IPV6_ADDR_MAPPED) { 933 u32 exthdrlen = icsk->icsk_ext_hdr_len; 934 struct sockaddr_in sin; 935 936 SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); 937 938 if (__ipv6_only_sock(sk)) 939 return -ENETUNREACH; 940 941 sin.sin_family = AF_INET; 942 sin.sin_port = usin->sin6_port; 943 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; 944 945 icsk->icsk_af_ops = &dccp_ipv6_mapped; 946 sk->sk_backlog_rcv = dccp_v4_do_rcv; 947 948 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); 949 if (err) { 950 icsk->icsk_ext_hdr_len = exthdrlen; 951 icsk->icsk_af_ops = &dccp_ipv6_af_ops; 952 sk->sk_backlog_rcv = dccp_v6_do_rcv; 953 goto failure; 954 } 955 ipv6_addr_set_v4mapped(inet->inet_saddr, &np->saddr); 956 ipv6_addr_set_v4mapped(inet->inet_rcv_saddr, &np->rcv_saddr); 957 958 return err; 959 } 960 961 if (!ipv6_addr_any(&np->rcv_saddr)) 962 saddr = &np->rcv_saddr; 963 964 fl.proto = IPPROTO_DCCP; 965 ipv6_addr_copy(&fl.fl6_dst, &np->daddr); 966 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr); 967 fl.oif = sk->sk_bound_dev_if; 968 fl.fl_ip_dport = usin->sin6_port; 969 fl.fl_ip_sport = inet->inet_sport; 970 security_sk_classify_flow(sk, &fl); 971 972 final_p = fl6_update_dst(&fl, np->opt, &final); 973 974 err = ip6_dst_lookup(sk, &dst, &fl); 975 if (err) 976 goto failure; 977 978 if (final_p) 979 ipv6_addr_copy(&fl.fl6_dst, final_p); 980 981 err = __xfrm_lookup(sock_net(sk), &dst, &fl, sk, XFRM_LOOKUP_WAIT); 982 if (err < 0) { 983 if (err == -EREMOTE) 984 err = ip6_dst_blackhole(sk, &dst, &fl); 985 if (err < 0) 986 goto failure; 987 } 988 989 if (saddr == NULL) { 990 saddr = &fl.fl6_src; 991 ipv6_addr_copy(&np->rcv_saddr, saddr); 992 } 993 994 /* set the source address */ 995 ipv6_addr_copy(&np->saddr, saddr); 996 inet->inet_rcv_saddr = LOOPBACK4_IPV6; 997 998 __ip6_dst_store(sk, dst, NULL, NULL); 999 1000 icsk->icsk_ext_hdr_len = 0; 1001 if (np->opt != NULL) 1002 icsk->icsk_ext_hdr_len = (np->opt->opt_flen + 1003 np->opt->opt_nflen); 1004 1005 inet->inet_dport = usin->sin6_port; 1006 1007 dccp_set_state(sk, DCCP_REQUESTING); 1008 err = inet6_hash_connect(&dccp_death_row, sk); 1009 if (err) 1010 goto late_failure; 1011 1012 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32, 1013 np->daddr.s6_addr32, 1014 inet->inet_sport, 1015 inet->inet_dport); 1016 err = dccp_connect(sk); 1017 if (err) 1018 goto late_failure; 1019 1020 return 0; 1021 1022late_failure: 1023 dccp_set_state(sk, DCCP_CLOSED); 1024 __sk_dst_reset(sk); 1025failure: 1026 inet->inet_dport = 0; 1027 sk->sk_route_caps = 0; 1028 return err; 1029} 1030 1031static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = { 1032 .queue_xmit = inet6_csk_xmit, 1033 .send_check = dccp_v6_send_check, 1034 .rebuild_header = inet6_sk_rebuild_header, 1035 .conn_request = dccp_v6_conn_request, 1036 .syn_recv_sock = dccp_v6_request_recv_sock, 1037 .net_header_len = sizeof(struct ipv6hdr), 1038 .setsockopt = ipv6_setsockopt, 1039 .getsockopt = ipv6_getsockopt, 1040 .addr2sockaddr = inet6_csk_addr2sockaddr, 1041 .sockaddr_len = sizeof(struct sockaddr_in6), 1042 .bind_conflict = inet6_csk_bind_conflict, 1043#ifdef CONFIG_COMPAT 1044 .compat_setsockopt = compat_ipv6_setsockopt, 1045 .compat_getsockopt = compat_ipv6_getsockopt, 1046#endif 1047}; 1048 1049/* 1050 * DCCP over IPv4 via INET6 API 1051 */ 1052static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = { 1053 .queue_xmit = ip_queue_xmit, 1054 .send_check = dccp_v4_send_check, 1055 .rebuild_header = inet_sk_rebuild_header, 1056 .conn_request = dccp_v6_conn_request, 1057 .syn_recv_sock = dccp_v6_request_recv_sock, 1058 .net_header_len = sizeof(struct iphdr), 1059 .setsockopt = ipv6_setsockopt, 1060 .getsockopt = ipv6_getsockopt, 1061 .addr2sockaddr = inet6_csk_addr2sockaddr, 1062 .sockaddr_len = sizeof(struct sockaddr_in6), 1063#ifdef CONFIG_COMPAT 1064 .compat_setsockopt = compat_ipv6_setsockopt, 1065 .compat_getsockopt = compat_ipv6_getsockopt, 1066#endif 1067}; 1068 1069/* NOTE: A lot of things set to zero explicitly by call to 1070 * sk_alloc() so need not be done here. 1071 */ 1072static int dccp_v6_init_sock(struct sock *sk) 1073{ 1074 static __u8 dccp_v6_ctl_sock_initialized; 1075 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized); 1076 1077 if (err == 0) { 1078 if (unlikely(!dccp_v6_ctl_sock_initialized)) 1079 dccp_v6_ctl_sock_initialized = 1; 1080 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops; 1081 } 1082 1083 return err; 1084} 1085 1086static void dccp_v6_destroy_sock(struct sock *sk) 1087{ 1088 dccp_destroy_sock(sk); 1089 inet6_destroy_sock(sk); 1090} 1091 1092static struct timewait_sock_ops dccp6_timewait_sock_ops = { 1093 .twsk_obj_size = sizeof(struct dccp6_timewait_sock), 1094}; 1095 1096static struct proto dccp_v6_prot = { 1097 .name = "DCCPv6", 1098 .owner = THIS_MODULE, 1099 .close = dccp_close, 1100 .connect = dccp_v6_connect, 1101 .disconnect = dccp_disconnect, 1102 .ioctl = dccp_ioctl, 1103 .init = dccp_v6_init_sock, 1104 .setsockopt = dccp_setsockopt, 1105 .getsockopt = dccp_getsockopt, 1106 .sendmsg = dccp_sendmsg, 1107 .recvmsg = dccp_recvmsg, 1108 .backlog_rcv = dccp_v6_do_rcv, 1109 .hash = dccp_v6_hash, 1110 .unhash = inet_unhash, 1111 .accept = inet_csk_accept, 1112 .get_port = inet_csk_get_port, 1113 .shutdown = dccp_shutdown, 1114 .destroy = dccp_v6_destroy_sock, 1115 .orphan_count = &dccp_orphan_count, 1116 .max_header = MAX_DCCP_HEADER, 1117 .obj_size = sizeof(struct dccp6_sock), 1118 .slab_flags = SLAB_DESTROY_BY_RCU, 1119 .rsk_prot = &dccp6_request_sock_ops, 1120 .twsk_prot = &dccp6_timewait_sock_ops, 1121 .h.hashinfo = &dccp_hashinfo, 1122#ifdef CONFIG_COMPAT 1123 .compat_setsockopt = compat_dccp_setsockopt, 1124 .compat_getsockopt = compat_dccp_getsockopt, 1125#endif 1126}; 1127 1128static const struct inet6_protocol dccp_v6_protocol = { 1129 .handler = dccp_v6_rcv, 1130 .err_handler = dccp_v6_err, 1131 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL, 1132}; 1133 1134static const struct proto_ops inet6_dccp_ops = { 1135 .family = PF_INET6, 1136 .owner = THIS_MODULE, 1137 .release = inet6_release, 1138 .bind = inet6_bind, 1139 .connect = inet_stream_connect, 1140 .socketpair = sock_no_socketpair, 1141 .accept = inet_accept, 1142 .getname = inet6_getname, 1143 .poll = dccp_poll, 1144 .ioctl = inet6_ioctl, 1145 .listen = inet_dccp_listen, 1146 .shutdown = inet_shutdown, 1147 .setsockopt = sock_common_setsockopt, 1148 .getsockopt = sock_common_getsockopt, 1149 .sendmsg = inet_sendmsg, 1150 .recvmsg = sock_common_recvmsg, 1151 .mmap = sock_no_mmap, 1152 .sendpage = sock_no_sendpage, 1153#ifdef CONFIG_COMPAT 1154 .compat_setsockopt = compat_sock_common_setsockopt, 1155 .compat_getsockopt = compat_sock_common_getsockopt, 1156#endif 1157}; 1158 1159static struct inet_protosw dccp_v6_protosw = { 1160 .type = SOCK_DCCP, 1161 .protocol = IPPROTO_DCCP, 1162 .prot = &dccp_v6_prot, 1163 .ops = &inet6_dccp_ops, 1164 .flags = INET_PROTOSW_ICSK, 1165}; 1166 1167static int __net_init dccp_v6_init_net(struct net *net) 1168{ 1169 if (dccp_hashinfo.bhash == NULL) 1170 return -ESOCKTNOSUPPORT; 1171 1172 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6, 1173 SOCK_DCCP, IPPROTO_DCCP, net); 1174} 1175 1176static void __net_exit dccp_v6_exit_net(struct net *net) 1177{ 1178 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk); 1179} 1180 1181static struct pernet_operations dccp_v6_ops = { 1182 .init = dccp_v6_init_net, 1183 .exit = dccp_v6_exit_net, 1184}; 1185 1186static int __init dccp_v6_init(void) 1187{ 1188 int err = proto_register(&dccp_v6_prot, 1); 1189 1190 if (err != 0) 1191 goto out; 1192 1193 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1194 if (err != 0) 1195 goto out_unregister_proto; 1196 1197 inet6_register_protosw(&dccp_v6_protosw); 1198 1199 err = register_pernet_subsys(&dccp_v6_ops); 1200 if (err != 0) 1201 goto out_destroy_ctl_sock; 1202out: 1203 return err; 1204 1205out_destroy_ctl_sock: 1206 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1207 inet6_unregister_protosw(&dccp_v6_protosw); 1208out_unregister_proto: 1209 proto_unregister(&dccp_v6_prot); 1210 goto out; 1211} 1212 1213static void __exit dccp_v6_exit(void) 1214{ 1215 unregister_pernet_subsys(&dccp_v6_ops); 1216 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1217 inet6_unregister_protosw(&dccp_v6_protosw); 1218 proto_unregister(&dccp_v6_prot); 1219} 1220 1221module_init(dccp_v6_init); 1222module_exit(dccp_v6_exit); 1223 1224/* 1225 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33) 1226 * values directly, Also cover the case where the protocol is not specified, 1227 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP 1228 */ 1229MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6); 1230MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6); 1231MODULE_LICENSE("GPL"); 1232MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>"); 1233MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol"); 1234