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/xfrm.h> 18 19 #include <net/addrconf.h> 20 #include <net/inet_common.h> 21 #include <net/inet_hashtables.h> 22 #include <net/inet_sock.h> 23 #include <net/inet6_connection_sock.h> 24 #include <net/inet6_hashtables.h> 25 #include <net/ip6_route.h> 26 #include <net/ipv6.h> 27 #include <net/protocol.h> 28 #include <net/transp_v6.h> 29 #include <net/ip6_checksum.h> 30 #include <net/xfrm.h> 31 32 #include "dccp.h" 33 #include "ipv6.h" 34 #include "feat.h" 35 36 /* Socket used for sending RSTs and ACKs */ 37 static struct socket *dccp_v6_ctl_socket; 38 39 static struct inet_connection_sock_af_ops dccp_ipv6_mapped; 40 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops; 41 42 static int dccp_v6_get_port(struct sock *sk, unsigned short snum) 43 { 44 return inet_csk_get_port(&dccp_hashinfo, sk, snum, 45 inet6_csk_bind_conflict); 46 } 47 48 static void dccp_v6_hash(struct sock *sk) 49 { 50 if (sk->sk_state != DCCP_CLOSED) { 51 if (inet_csk(sk)->icsk_af_ops == &dccp_ipv6_mapped) { 52 dccp_hash(sk); 53 return; 54 } 55 local_bh_disable(); 56 __inet6_hash(&dccp_hashinfo, sk); 57 local_bh_enable(); 58 } 59 } 60 61 /* add pseudo-header to DCCP checksum stored in skb->csum */ 62 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb, 63 struct in6_addr *saddr, 64 struct in6_addr *daddr) 65 { 66 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum); 67 } 68 69 static inline void dccp_v6_send_check(struct sock *sk, int unused_value, 70 struct sk_buff *skb) 71 { 72 struct ipv6_pinfo *np = inet6_sk(sk); 73 struct dccp_hdr *dh = dccp_hdr(skb); 74 75 dccp_csum_outgoing(skb); 76 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &np->daddr); 77 } 78 79 static inline __u32 secure_dccpv6_sequence_number(__be32 *saddr, __be32 *daddr, 80 __be16 sport, __be16 dport ) 81 { 82 return secure_tcpv6_sequence_number(saddr, daddr, sport, dport); 83 } 84 85 static inline __u32 dccp_v6_init_sequence(struct sk_buff *skb) 86 { 87 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32, 88 ipv6_hdr(skb)->saddr.s6_addr32, 89 dccp_hdr(skb)->dccph_dport, 90 dccp_hdr(skb)->dccph_sport ); 91 92 } 93 94 static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt, 95 int type, int code, int offset, __be32 info) 96 { 97 struct ipv6hdr *hdr = (struct ipv6hdr *)skb->data; 98 const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset); 99 struct ipv6_pinfo *np; 100 struct sock *sk; 101 int err; 102 __u64 seq; 103 104 sk = inet6_lookup(&dccp_hashinfo, &hdr->daddr, dh->dccph_dport, 105 &hdr->saddr, dh->dccph_sport, inet6_iif(skb)); 106 107 if (sk == NULL) { 108 ICMP6_INC_STATS_BH(__in6_dev_get(skb->dev), ICMP6_MIB_INERRORS); 109 return; 110 } 111 112 if (sk->sk_state == DCCP_TIME_WAIT) { 113 inet_twsk_put(inet_twsk(sk)); 114 return; 115 } 116 117 bh_lock_sock(sk); 118 if (sock_owned_by_user(sk)) 119 NET_INC_STATS_BH(LINUX_MIB_LOCKDROPPEDICMPS); 120 121 if (sk->sk_state == DCCP_CLOSED) 122 goto out; 123 124 np = inet6_sk(sk); 125 126 if (type == ICMPV6_PKT_TOOBIG) { 127 struct dst_entry *dst = NULL; 128 129 if (sock_owned_by_user(sk)) 130 goto out; 131 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED)) 132 goto out; 133 134 /* icmp should have updated the destination cache entry */ 135 dst = __sk_dst_check(sk, np->dst_cookie); 136 if (dst == NULL) { 137 struct inet_sock *inet = inet_sk(sk); 138 struct flowi fl; 139 140 /* BUGGG_FUTURE: Again, it is not clear how 141 to handle rthdr case. Ignore this complexity 142 for now. 143 */ 144 memset(&fl, 0, sizeof(fl)); 145 fl.proto = IPPROTO_DCCP; 146 ipv6_addr_copy(&fl.fl6_dst, &np->daddr); 147 ipv6_addr_copy(&fl.fl6_src, &np->saddr); 148 fl.oif = sk->sk_bound_dev_if; 149 fl.fl_ip_dport = inet->dport; 150 fl.fl_ip_sport = inet->sport; 151 security_sk_classify_flow(sk, &fl); 152 153 err = ip6_dst_lookup(sk, &dst, &fl); 154 if (err) { 155 sk->sk_err_soft = -err; 156 goto out; 157 } 158 159 err = xfrm_lookup(&dst, &fl, sk, 0); 160 if (err < 0) { 161 sk->sk_err_soft = -err; 162 goto out; 163 } 164 } else 165 dst_hold(dst); 166 167 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst)) { 168 dccp_sync_mss(sk, dst_mtu(dst)); 169 } /* else let the usual retransmit timer handle it */ 170 dst_release(dst); 171 goto out; 172 } 173 174 icmpv6_err_convert(type, code, &err); 175 176 seq = DCCP_SKB_CB(skb)->dccpd_seq; 177 /* Might be for an request_sock */ 178 switch (sk->sk_state) { 179 struct request_sock *req, **prev; 180 case DCCP_LISTEN: 181 if (sock_owned_by_user(sk)) 182 goto out; 183 184 req = inet6_csk_search_req(sk, &prev, dh->dccph_dport, 185 &hdr->daddr, &hdr->saddr, 186 inet6_iif(skb)); 187 if (req == NULL) 188 goto out; 189 190 /* 191 * ICMPs are not backlogged, hence we cannot get an established 192 * socket here. 193 */ 194 BUG_TRAP(req->sk == NULL); 195 196 if (seq != dccp_rsk(req)->dreq_iss) { 197 NET_INC_STATS_BH(LINUX_MIB_OUTOFWINDOWICMPS); 198 goto out; 199 } 200 201 inet_csk_reqsk_queue_drop(sk, req, prev); 202 goto out; 203 204 case DCCP_REQUESTING: 205 case DCCP_RESPOND: /* Cannot happen. 206 It can, it SYNs are crossed. --ANK */ 207 if (!sock_owned_by_user(sk)) { 208 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 209 sk->sk_err = err; 210 /* 211 * Wake people up to see the error 212 * (see connect in sock.c) 213 */ 214 sk->sk_error_report(sk); 215 dccp_done(sk); 216 } else 217 sk->sk_err_soft = err; 218 goto out; 219 } 220 221 if (!sock_owned_by_user(sk) && np->recverr) { 222 sk->sk_err = err; 223 sk->sk_error_report(sk); 224 } else 225 sk->sk_err_soft = err; 226 227 out: 228 bh_unlock_sock(sk); 229 sock_put(sk); 230 } 231 232 233 static int dccp_v6_send_response(struct sock *sk, struct request_sock *req, 234 struct dst_entry *dst) 235 { 236 struct inet6_request_sock *ireq6 = inet6_rsk(req); 237 struct ipv6_pinfo *np = inet6_sk(sk); 238 struct sk_buff *skb; 239 struct ipv6_txoptions *opt = NULL; 240 struct in6_addr *final_p = NULL, final; 241 struct flowi fl; 242 int err = -1; 243 244 memset(&fl, 0, sizeof(fl)); 245 fl.proto = IPPROTO_DCCP; 246 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr); 247 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr); 248 fl.fl6_flowlabel = 0; 249 fl.oif = ireq6->iif; 250 fl.fl_ip_dport = inet_rsk(req)->rmt_port; 251 fl.fl_ip_sport = inet_sk(sk)->sport; 252 security_req_classify_flow(req, &fl); 253 254 if (dst == NULL) { 255 opt = np->opt; 256 257 if (opt != NULL && opt->srcrt != NULL) { 258 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt; 259 260 ipv6_addr_copy(&final, &fl.fl6_dst); 261 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 262 final_p = &final; 263 } 264 265 err = ip6_dst_lookup(sk, &dst, &fl); 266 if (err) 267 goto done; 268 269 if (final_p) 270 ipv6_addr_copy(&fl.fl6_dst, final_p); 271 272 err = xfrm_lookup(&dst, &fl, sk, 0); 273 if (err < 0) 274 goto done; 275 } 276 277 skb = dccp_make_response(sk, dst, req); 278 if (skb != NULL) { 279 struct dccp_hdr *dh = dccp_hdr(skb); 280 281 dh->dccph_checksum = dccp_v6_csum_finish(skb, 282 &ireq6->loc_addr, 283 &ireq6->rmt_addr); 284 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr); 285 err = ip6_xmit(sk, skb, &fl, opt, 0); 286 err = net_xmit_eval(err); 287 } 288 289 done: 290 if (opt != NULL && opt != np->opt) 291 sock_kfree_s(sk, opt, opt->tot_len); 292 dst_release(dst); 293 return err; 294 } 295 296 static void dccp_v6_reqsk_destructor(struct request_sock *req) 297 { 298 if (inet6_rsk(req)->pktopts != NULL) 299 kfree_skb(inet6_rsk(req)->pktopts); 300 } 301 302 static void dccp_v6_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb) 303 { 304 struct ipv6hdr *rxip6h; 305 struct sk_buff *skb; 306 struct flowi fl; 307 308 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET) 309 return; 310 311 if (!ipv6_unicast_destination(rxskb)) 312 return; 313 314 skb = dccp_ctl_make_reset(dccp_v6_ctl_socket, rxskb); 315 if (skb == NULL) 316 return; 317 318 rxip6h = ipv6_hdr(rxskb); 319 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr, 320 &rxip6h->daddr); 321 322 memset(&fl, 0, sizeof(fl)); 323 ipv6_addr_copy(&fl.fl6_dst, &rxip6h->saddr); 324 ipv6_addr_copy(&fl.fl6_src, &rxip6h->daddr); 325 326 fl.proto = IPPROTO_DCCP; 327 fl.oif = inet6_iif(rxskb); 328 fl.fl_ip_dport = dccp_hdr(skb)->dccph_dport; 329 fl.fl_ip_sport = dccp_hdr(skb)->dccph_sport; 330 security_skb_classify_flow(rxskb, &fl); 331 332 /* sk = NULL, but it is safe for now. RST socket required. */ 333 if (!ip6_dst_lookup(NULL, &skb->dst, &fl)) { 334 if (xfrm_lookup(&skb->dst, &fl, NULL, 0) >= 0) { 335 ip6_xmit(dccp_v6_ctl_socket->sk, skb, &fl, NULL, 0); 336 DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS); 337 DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS); 338 return; 339 } 340 } 341 342 kfree_skb(skb); 343 } 344 345 static struct request_sock_ops dccp6_request_sock_ops = { 346 .family = AF_INET6, 347 .obj_size = sizeof(struct dccp6_request_sock), 348 .rtx_syn_ack = dccp_v6_send_response, 349 .send_ack = dccp_reqsk_send_ack, 350 .destructor = dccp_v6_reqsk_destructor, 351 .send_reset = dccp_v6_ctl_send_reset, 352 }; 353 354 static struct sock *dccp_v6_hnd_req(struct sock *sk,struct sk_buff *skb) 355 { 356 const struct dccp_hdr *dh = dccp_hdr(skb); 357 const struct ipv6hdr *iph = ipv6_hdr(skb); 358 struct sock *nsk; 359 struct request_sock **prev; 360 /* Find possible connection requests. */ 361 struct request_sock *req = inet6_csk_search_req(sk, &prev, 362 dh->dccph_sport, 363 &iph->saddr, 364 &iph->daddr, 365 inet6_iif(skb)); 366 if (req != NULL) 367 return dccp_check_req(sk, skb, req, prev); 368 369 nsk = __inet6_lookup_established(&dccp_hashinfo, 370 &iph->saddr, dh->dccph_sport, 371 &iph->daddr, ntohs(dh->dccph_dport), 372 inet6_iif(skb)); 373 if (nsk != NULL) { 374 if (nsk->sk_state != DCCP_TIME_WAIT) { 375 bh_lock_sock(nsk); 376 return nsk; 377 } 378 inet_twsk_put(inet_twsk(nsk)); 379 return NULL; 380 } 381 382 return sk; 383 } 384 385 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb) 386 { 387 struct request_sock *req; 388 struct dccp_request_sock *dreq; 389 struct inet6_request_sock *ireq6; 390 struct ipv6_pinfo *np = inet6_sk(sk); 391 const __be32 service = dccp_hdr_request(skb)->dccph_req_service; 392 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb); 393 394 if (skb->protocol == htons(ETH_P_IP)) 395 return dccp_v4_conn_request(sk, skb); 396 397 if (!ipv6_unicast_destination(skb)) 398 return 0; /* discard, don't send a reset here */ 399 400 if (dccp_bad_service_code(sk, service)) { 401 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE; 402 goto drop; 403 } 404 /* 405 * There are no SYN attacks on IPv6, yet... 406 */ 407 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY; 408 if (inet_csk_reqsk_queue_is_full(sk)) 409 goto drop; 410 411 if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1) 412 goto drop; 413 414 req = inet6_reqsk_alloc(&dccp6_request_sock_ops); 415 if (req == NULL) 416 goto drop; 417 418 if (dccp_parse_options(sk, skb)) 419 goto drop_and_free; 420 421 dccp_reqsk_init(req, skb); 422 423 if (security_inet_conn_request(sk, skb, req)) 424 goto drop_and_free; 425 426 ireq6 = inet6_rsk(req); 427 ipv6_addr_copy(&ireq6->rmt_addr, &ipv6_hdr(skb)->saddr); 428 ipv6_addr_copy(&ireq6->loc_addr, &ipv6_hdr(skb)->daddr); 429 ireq6->pktopts = NULL; 430 431 if (ipv6_opt_accepted(sk, skb) || 432 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo || 433 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) { 434 atomic_inc(&skb->users); 435 ireq6->pktopts = skb; 436 } 437 ireq6->iif = sk->sk_bound_dev_if; 438 439 /* So that link locals have meaning */ 440 if (!sk->sk_bound_dev_if && 441 ipv6_addr_type(&ireq6->rmt_addr) & IPV6_ADDR_LINKLOCAL) 442 ireq6->iif = inet6_iif(skb); 443 444 /* 445 * Step 3: Process LISTEN state 446 * 447 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie 448 * 449 * In fact we defer setting S.GSR, S.SWL, S.SWH to 450 * dccp_create_openreq_child. 451 */ 452 dreq = dccp_rsk(req); 453 dreq->dreq_isr = dcb->dccpd_seq; 454 dreq->dreq_iss = dccp_v6_init_sequence(skb); 455 dreq->dreq_service = service; 456 457 if (dccp_v6_send_response(sk, req, NULL)) 458 goto drop_and_free; 459 460 inet6_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT); 461 return 0; 462 463 drop_and_free: 464 reqsk_free(req); 465 drop: 466 DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS); 467 return -1; 468 } 469 470 static struct sock *dccp_v6_request_recv_sock(struct sock *sk, 471 struct sk_buff *skb, 472 struct request_sock *req, 473 struct dst_entry *dst) 474 { 475 struct inet6_request_sock *ireq6 = inet6_rsk(req); 476 struct ipv6_pinfo *newnp, *np = inet6_sk(sk); 477 struct inet_sock *newinet; 478 struct dccp_sock *newdp; 479 struct dccp6_sock *newdp6; 480 struct sock *newsk; 481 struct ipv6_txoptions *opt; 482 483 if (skb->protocol == htons(ETH_P_IP)) { 484 /* 485 * v6 mapped 486 */ 487 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst); 488 if (newsk == NULL) 489 return NULL; 490 491 newdp6 = (struct dccp6_sock *)newsk; 492 newdp = dccp_sk(newsk); 493 newinet = inet_sk(newsk); 494 newinet->pinet6 = &newdp6->inet6; 495 newnp = inet6_sk(newsk); 496 497 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 498 499 ipv6_addr_set(&newnp->daddr, 0, 0, htonl(0x0000FFFF), 500 newinet->daddr); 501 502 ipv6_addr_set(&newnp->saddr, 0, 0, htonl(0x0000FFFF), 503 newinet->saddr); 504 505 ipv6_addr_copy(&newnp->rcv_saddr, &newnp->saddr); 506 507 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped; 508 newsk->sk_backlog_rcv = dccp_v4_do_rcv; 509 newnp->pktoptions = NULL; 510 newnp->opt = NULL; 511 newnp->mcast_oif = inet6_iif(skb); 512 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 513 514 /* 515 * No need to charge this sock to the relevant IPv6 refcnt debug socks count 516 * here, dccp_create_openreq_child now does this for us, see the comment in 517 * that function for the gory details. -acme 518 */ 519 520 /* It is tricky place. Until this moment IPv4 tcp 521 worked with IPv6 icsk.icsk_af_ops. 522 Sync it now. 523 */ 524 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie); 525 526 return newsk; 527 } 528 529 opt = np->opt; 530 531 if (sk_acceptq_is_full(sk)) 532 goto out_overflow; 533 534 if (dst == NULL) { 535 struct in6_addr *final_p = NULL, final; 536 struct flowi fl; 537 538 memset(&fl, 0, sizeof(fl)); 539 fl.proto = IPPROTO_DCCP; 540 ipv6_addr_copy(&fl.fl6_dst, &ireq6->rmt_addr); 541 if (opt != NULL && opt->srcrt != NULL) { 542 const struct rt0_hdr *rt0 = (struct rt0_hdr *)opt->srcrt; 543 544 ipv6_addr_copy(&final, &fl.fl6_dst); 545 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 546 final_p = &final; 547 } 548 ipv6_addr_copy(&fl.fl6_src, &ireq6->loc_addr); 549 fl.oif = sk->sk_bound_dev_if; 550 fl.fl_ip_dport = inet_rsk(req)->rmt_port; 551 fl.fl_ip_sport = inet_sk(sk)->sport; 552 security_sk_classify_flow(sk, &fl); 553 554 if (ip6_dst_lookup(sk, &dst, &fl)) 555 goto out; 556 557 if (final_p) 558 ipv6_addr_copy(&fl.fl6_dst, final_p); 559 560 if ((xfrm_lookup(&dst, &fl, sk, 0)) < 0) 561 goto out; 562 } 563 564 newsk = dccp_create_openreq_child(sk, req, skb); 565 if (newsk == NULL) 566 goto out; 567 568 /* 569 * No need to charge this sock to the relevant IPv6 refcnt debug socks 570 * count here, dccp_create_openreq_child now does this for us, see the 571 * comment in that function for the gory details. -acme 572 */ 573 574 __ip6_dst_store(newsk, dst, NULL, NULL); 575 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM | 576 NETIF_F_TSO); 577 newdp6 = (struct dccp6_sock *)newsk; 578 newinet = inet_sk(newsk); 579 newinet->pinet6 = &newdp6->inet6; 580 newdp = dccp_sk(newsk); 581 newnp = inet6_sk(newsk); 582 583 memcpy(newnp, np, sizeof(struct ipv6_pinfo)); 584 585 ipv6_addr_copy(&newnp->daddr, &ireq6->rmt_addr); 586 ipv6_addr_copy(&newnp->saddr, &ireq6->loc_addr); 587 ipv6_addr_copy(&newnp->rcv_saddr, &ireq6->loc_addr); 588 newsk->sk_bound_dev_if = ireq6->iif; 589 590 /* Now IPv6 options... 591 592 First: no IPv4 options. 593 */ 594 newinet->opt = NULL; 595 596 /* Clone RX bits */ 597 newnp->rxopt.all = np->rxopt.all; 598 599 /* Clone pktoptions received with SYN */ 600 newnp->pktoptions = NULL; 601 if (ireq6->pktopts != NULL) { 602 newnp->pktoptions = skb_clone(ireq6->pktopts, GFP_ATOMIC); 603 kfree_skb(ireq6->pktopts); 604 ireq6->pktopts = NULL; 605 if (newnp->pktoptions) 606 skb_set_owner_r(newnp->pktoptions, newsk); 607 } 608 newnp->opt = NULL; 609 newnp->mcast_oif = inet6_iif(skb); 610 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit; 611 612 /* 613 * Clone native IPv6 options from listening socket (if any) 614 * 615 * Yes, keeping reference count would be much more clever, but we make 616 * one more one thing there: reattach optmem to newsk. 617 */ 618 if (opt != NULL) { 619 newnp->opt = ipv6_dup_options(newsk, opt); 620 if (opt != np->opt) 621 sock_kfree_s(sk, opt, opt->tot_len); 622 } 623 624 inet_csk(newsk)->icsk_ext_hdr_len = 0; 625 if (newnp->opt != NULL) 626 inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen + 627 newnp->opt->opt_flen); 628 629 dccp_sync_mss(newsk, dst_mtu(dst)); 630 631 newinet->daddr = newinet->saddr = newinet->rcv_saddr = LOOPBACK4_IPV6; 632 633 __inet6_hash(&dccp_hashinfo, newsk); 634 inet_inherit_port(&dccp_hashinfo, sk, newsk); 635 636 return newsk; 637 638 out_overflow: 639 NET_INC_STATS_BH(LINUX_MIB_LISTENOVERFLOWS); 640 out: 641 NET_INC_STATS_BH(LINUX_MIB_LISTENDROPS); 642 if (opt != NULL && opt != np->opt) 643 sock_kfree_s(sk, opt, opt->tot_len); 644 dst_release(dst); 645 return NULL; 646 } 647 648 /* The socket must have it's spinlock held when we get 649 * here. 650 * 651 * We have a potential double-lock case here, so even when 652 * doing backlog processing we use the BH locking scheme. 653 * This is because we cannot sleep with the original spinlock 654 * held. 655 */ 656 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb) 657 { 658 struct ipv6_pinfo *np = inet6_sk(sk); 659 struct sk_buff *opt_skb = NULL; 660 661 /* Imagine: socket is IPv6. IPv4 packet arrives, 662 goes to IPv4 receive handler and backlogged. 663 From backlog it always goes here. Kerboom... 664 Fortunately, dccp_rcv_established and rcv_established 665 handle them correctly, but it is not case with 666 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK 667 */ 668 669 if (skb->protocol == htons(ETH_P_IP)) 670 return dccp_v4_do_rcv(sk, skb); 671 672 if (sk_filter(sk, skb)) 673 goto discard; 674 675 /* 676 * socket locking is here for SMP purposes as backlog rcv is currently 677 * called with bh processing disabled. 678 */ 679 680 /* Do Stevens' IPV6_PKTOPTIONS. 681 682 Yes, guys, it is the only place in our code, where we 683 may make it not affecting IPv4. 684 The rest of code is protocol independent, 685 and I do not like idea to uglify IPv4. 686 687 Actually, all the idea behind IPV6_PKTOPTIONS 688 looks not very well thought. For now we latch 689 options, received in the last packet, enqueued 690 by tcp. Feel free to propose better solution. 691 --ANK (980728) 692 */ 693 if (np->rxopt.all) 694 /* 695 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below 696 * (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example. 697 */ 698 opt_skb = skb_clone(skb, GFP_ATOMIC); 699 700 if (sk->sk_state == DCCP_OPEN) { /* Fast path */ 701 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len)) 702 goto reset; 703 if (opt_skb) { 704 /* XXX This is where we would goto ipv6_pktoptions. */ 705 __kfree_skb(opt_skb); 706 } 707 return 0; 708 } 709 710 /* 711 * Step 3: Process LISTEN state 712 * If S.state == LISTEN, 713 * If P.type == Request or P contains a valid Init Cookie option, 714 * (* Must scan the packet's options to check for Init 715 * Cookies. Only Init Cookies are processed here, 716 * however; other options are processed in Step 8. This 717 * scan need only be performed if the endpoint uses Init 718 * Cookies *) 719 * (* Generate a new socket and switch to that socket *) 720 * Set S := new socket for this port pair 721 * S.state = RESPOND 722 * Choose S.ISS (initial seqno) or set from Init Cookies 723 * Initialize S.GAR := S.ISS 724 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies 725 * Continue with S.state == RESPOND 726 * (* A Response packet will be generated in Step 11 *) 727 * Otherwise, 728 * Generate Reset(No Connection) unless P.type == Reset 729 * Drop packet and return 730 * 731 * NOTE: the check for the packet types is done in 732 * dccp_rcv_state_process 733 */ 734 if (sk->sk_state == DCCP_LISTEN) { 735 struct sock *nsk = dccp_v6_hnd_req(sk, skb); 736 737 if (nsk == NULL) 738 goto discard; 739 /* 740 * Queue it on the new socket if the new socket is active, 741 * otherwise we just shortcircuit this and continue with 742 * the new socket.. 743 */ 744 if (nsk != sk) { 745 if (dccp_child_process(sk, nsk, skb)) 746 goto reset; 747 if (opt_skb != NULL) 748 __kfree_skb(opt_skb); 749 return 0; 750 } 751 } 752 753 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len)) 754 goto reset; 755 if (opt_skb) { 756 /* XXX This is where we would goto ipv6_pktoptions. */ 757 __kfree_skb(opt_skb); 758 } 759 return 0; 760 761 reset: 762 dccp_v6_ctl_send_reset(sk, skb); 763 discard: 764 if (opt_skb != NULL) 765 __kfree_skb(opt_skb); 766 kfree_skb(skb); 767 return 0; 768 } 769 770 static int dccp_v6_rcv(struct sk_buff *skb) 771 { 772 const struct dccp_hdr *dh; 773 struct sock *sk; 774 int min_cov; 775 776 /* Step 1: Check header basics */ 777 778 if (dccp_invalid_packet(skb)) 779 goto discard_it; 780 781 /* Step 1: If header checksum is incorrect, drop packet and return. */ 782 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr, 783 &ipv6_hdr(skb)->daddr)) { 784 DCCP_WARN("dropped packet with invalid checksum\n"); 785 goto discard_it; 786 } 787 788 dh = dccp_hdr(skb); 789 790 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(skb); 791 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type; 792 793 if (dccp_packet_without_ack(skb)) 794 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ; 795 else 796 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb); 797 798 /* Step 2: 799 * Look up flow ID in table and get corresponding socket */ 800 sk = __inet6_lookup(&dccp_hashinfo, &ipv6_hdr(skb)->saddr, 801 dh->dccph_sport, 802 &ipv6_hdr(skb)->daddr, ntohs(dh->dccph_dport), 803 inet6_iif(skb)); 804 /* 805 * Step 2: 806 * If no socket ... 807 */ 808 if (sk == NULL) { 809 dccp_pr_debug("failed to look up flow ID in table and " 810 "get corresponding socket\n"); 811 goto no_dccp_socket; 812 } 813 814 /* 815 * Step 2: 816 * ... or S.state == TIMEWAIT, 817 * Generate Reset(No Connection) unless P.type == Reset 818 * Drop packet and return 819 */ 820 if (sk->sk_state == DCCP_TIME_WAIT) { 821 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n"); 822 inet_twsk_put(inet_twsk(sk)); 823 goto no_dccp_socket; 824 } 825 826 /* 827 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage 828 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted 829 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov 830 */ 831 min_cov = dccp_sk(sk)->dccps_pcrlen; 832 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) { 833 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n", 834 dh->dccph_cscov, min_cov); 835 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */ 836 goto discard_and_relse; 837 } 838 839 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb)) 840 goto discard_and_relse; 841 842 return sk_receive_skb(sk, skb, 1) ? -1 : 0; 843 844 no_dccp_socket: 845 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb)) 846 goto discard_it; 847 /* 848 * Step 2: 849 * If no socket ... 850 * Generate Reset(No Connection) unless P.type == Reset 851 * Drop packet and return 852 */ 853 if (dh->dccph_type != DCCP_PKT_RESET) { 854 DCCP_SKB_CB(skb)->dccpd_reset_code = 855 DCCP_RESET_CODE_NO_CONNECTION; 856 dccp_v6_ctl_send_reset(sk, skb); 857 } 858 859 discard_it: 860 kfree_skb(skb); 861 return 0; 862 863 discard_and_relse: 864 sock_put(sk); 865 goto discard_it; 866 } 867 868 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr, 869 int addr_len) 870 { 871 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr; 872 struct inet_connection_sock *icsk = inet_csk(sk); 873 struct inet_sock *inet = inet_sk(sk); 874 struct ipv6_pinfo *np = inet6_sk(sk); 875 struct dccp_sock *dp = dccp_sk(sk); 876 struct in6_addr *saddr = NULL, *final_p = NULL, final; 877 struct flowi fl; 878 struct dst_entry *dst; 879 int addr_type; 880 int err; 881 882 dp->dccps_role = DCCP_ROLE_CLIENT; 883 884 if (addr_len < SIN6_LEN_RFC2133) 885 return -EINVAL; 886 887 if (usin->sin6_family != AF_INET6) 888 return -EAFNOSUPPORT; 889 890 memset(&fl, 0, sizeof(fl)); 891 892 if (np->sndflow) { 893 fl.fl6_flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK; 894 IP6_ECN_flow_init(fl.fl6_flowlabel); 895 if (fl.fl6_flowlabel & IPV6_FLOWLABEL_MASK) { 896 struct ip6_flowlabel *flowlabel; 897 flowlabel = fl6_sock_lookup(sk, fl.fl6_flowlabel); 898 if (flowlabel == NULL) 899 return -EINVAL; 900 ipv6_addr_copy(&usin->sin6_addr, &flowlabel->dst); 901 fl6_sock_release(flowlabel); 902 } 903 } 904 /* 905 * connect() to INADDR_ANY means loopback (BSD'ism). 906 */ 907 if (ipv6_addr_any(&usin->sin6_addr)) 908 usin->sin6_addr.s6_addr[15] = 1; 909 910 addr_type = ipv6_addr_type(&usin->sin6_addr); 911 912 if (addr_type & IPV6_ADDR_MULTICAST) 913 return -ENETUNREACH; 914 915 if (addr_type & IPV6_ADDR_LINKLOCAL) { 916 if (addr_len >= sizeof(struct sockaddr_in6) && 917 usin->sin6_scope_id) { 918 /* If interface is set while binding, indices 919 * must coincide. 920 */ 921 if (sk->sk_bound_dev_if && 922 sk->sk_bound_dev_if != usin->sin6_scope_id) 923 return -EINVAL; 924 925 sk->sk_bound_dev_if = usin->sin6_scope_id; 926 } 927 928 /* Connect to link-local address requires an interface */ 929 if (!sk->sk_bound_dev_if) 930 return -EINVAL; 931 } 932 933 ipv6_addr_copy(&np->daddr, &usin->sin6_addr); 934 np->flow_label = fl.fl6_flowlabel; 935 936 /* 937 * DCCP over IPv4 938 */ 939 if (addr_type == IPV6_ADDR_MAPPED) { 940 u32 exthdrlen = icsk->icsk_ext_hdr_len; 941 struct sockaddr_in sin; 942 943 SOCK_DEBUG(sk, "connect: ipv4 mapped\n"); 944 945 if (__ipv6_only_sock(sk)) 946 return -ENETUNREACH; 947 948 sin.sin_family = AF_INET; 949 sin.sin_port = usin->sin6_port; 950 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3]; 951 952 icsk->icsk_af_ops = &dccp_ipv6_mapped; 953 sk->sk_backlog_rcv = dccp_v4_do_rcv; 954 955 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin)); 956 if (err) { 957 icsk->icsk_ext_hdr_len = exthdrlen; 958 icsk->icsk_af_ops = &dccp_ipv6_af_ops; 959 sk->sk_backlog_rcv = dccp_v6_do_rcv; 960 goto failure; 961 } else { 962 ipv6_addr_set(&np->saddr, 0, 0, htonl(0x0000FFFF), 963 inet->saddr); 964 ipv6_addr_set(&np->rcv_saddr, 0, 0, htonl(0x0000FFFF), 965 inet->rcv_saddr); 966 } 967 968 return err; 969 } 970 971 if (!ipv6_addr_any(&np->rcv_saddr)) 972 saddr = &np->rcv_saddr; 973 974 fl.proto = IPPROTO_DCCP; 975 ipv6_addr_copy(&fl.fl6_dst, &np->daddr); 976 ipv6_addr_copy(&fl.fl6_src, saddr ? saddr : &np->saddr); 977 fl.oif = sk->sk_bound_dev_if; 978 fl.fl_ip_dport = usin->sin6_port; 979 fl.fl_ip_sport = inet->sport; 980 security_sk_classify_flow(sk, &fl); 981 982 if (np->opt != NULL && np->opt->srcrt != NULL) { 983 const struct rt0_hdr *rt0 = (struct rt0_hdr *)np->opt->srcrt; 984 985 ipv6_addr_copy(&final, &fl.fl6_dst); 986 ipv6_addr_copy(&fl.fl6_dst, rt0->addr); 987 final_p = &final; 988 } 989 990 err = ip6_dst_lookup(sk, &dst, &fl); 991 if (err) 992 goto failure; 993 994 if (final_p) 995 ipv6_addr_copy(&fl.fl6_dst, final_p); 996 997 err = __xfrm_lookup(&dst, &fl, sk, 1); 998 if (err < 0) { 999 if (err == -EREMOTE) 1000 err = ip6_dst_blackhole(sk, &dst, &fl); 1001 if (err < 0) 1002 goto failure; 1003 } 1004 1005 if (saddr == NULL) { 1006 saddr = &fl.fl6_src; 1007 ipv6_addr_copy(&np->rcv_saddr, saddr); 1008 } 1009 1010 /* set the source address */ 1011 ipv6_addr_copy(&np->saddr, saddr); 1012 inet->rcv_saddr = LOOPBACK4_IPV6; 1013 1014 __ip6_dst_store(sk, dst, NULL, NULL); 1015 1016 icsk->icsk_ext_hdr_len = 0; 1017 if (np->opt != NULL) 1018 icsk->icsk_ext_hdr_len = (np->opt->opt_flen + 1019 np->opt->opt_nflen); 1020 1021 inet->dport = usin->sin6_port; 1022 1023 dccp_set_state(sk, DCCP_REQUESTING); 1024 err = inet6_hash_connect(&dccp_death_row, sk); 1025 if (err) 1026 goto late_failure; 1027 1028 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32, 1029 np->daddr.s6_addr32, 1030 inet->sport, inet->dport); 1031 err = dccp_connect(sk); 1032 if (err) 1033 goto late_failure; 1034 1035 return 0; 1036 1037 late_failure: 1038 dccp_set_state(sk, DCCP_CLOSED); 1039 __sk_dst_reset(sk); 1040 failure: 1041 inet->dport = 0; 1042 sk->sk_route_caps = 0; 1043 return err; 1044 } 1045 1046 static struct inet_connection_sock_af_ops dccp_ipv6_af_ops = { 1047 .queue_xmit = inet6_csk_xmit, 1048 .send_check = dccp_v6_send_check, 1049 .rebuild_header = inet6_sk_rebuild_header, 1050 .conn_request = dccp_v6_conn_request, 1051 .syn_recv_sock = dccp_v6_request_recv_sock, 1052 .net_header_len = sizeof(struct ipv6hdr), 1053 .setsockopt = ipv6_setsockopt, 1054 .getsockopt = ipv6_getsockopt, 1055 .addr2sockaddr = inet6_csk_addr2sockaddr, 1056 .sockaddr_len = sizeof(struct sockaddr_in6), 1057 #ifdef CONFIG_COMPAT 1058 .compat_setsockopt = compat_ipv6_setsockopt, 1059 .compat_getsockopt = compat_ipv6_getsockopt, 1060 #endif 1061 }; 1062 1063 /* 1064 * DCCP over IPv4 via INET6 API 1065 */ 1066 static struct inet_connection_sock_af_ops dccp_ipv6_mapped = { 1067 .queue_xmit = ip_queue_xmit, 1068 .send_check = dccp_v4_send_check, 1069 .rebuild_header = inet_sk_rebuild_header, 1070 .conn_request = dccp_v6_conn_request, 1071 .syn_recv_sock = dccp_v6_request_recv_sock, 1072 .net_header_len = sizeof(struct iphdr), 1073 .setsockopt = ipv6_setsockopt, 1074 .getsockopt = ipv6_getsockopt, 1075 .addr2sockaddr = inet6_csk_addr2sockaddr, 1076 .sockaddr_len = sizeof(struct sockaddr_in6), 1077 #ifdef CONFIG_COMPAT 1078 .compat_setsockopt = compat_ipv6_setsockopt, 1079 .compat_getsockopt = compat_ipv6_getsockopt, 1080 #endif 1081 }; 1082 1083 /* NOTE: A lot of things set to zero explicitly by call to 1084 * sk_alloc() so need not be done here. 1085 */ 1086 static int dccp_v6_init_sock(struct sock *sk) 1087 { 1088 static __u8 dccp_v6_ctl_sock_initialized; 1089 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized); 1090 1091 if (err == 0) { 1092 if (unlikely(!dccp_v6_ctl_sock_initialized)) 1093 dccp_v6_ctl_sock_initialized = 1; 1094 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops; 1095 } 1096 1097 return err; 1098 } 1099 1100 static int dccp_v6_destroy_sock(struct sock *sk) 1101 { 1102 dccp_destroy_sock(sk); 1103 return inet6_destroy_sock(sk); 1104 } 1105 1106 static struct timewait_sock_ops dccp6_timewait_sock_ops = { 1107 .twsk_obj_size = sizeof(struct dccp6_timewait_sock), 1108 }; 1109 1110 static struct proto dccp_v6_prot = { 1111 .name = "DCCPv6", 1112 .owner = THIS_MODULE, 1113 .close = dccp_close, 1114 .connect = dccp_v6_connect, 1115 .disconnect = dccp_disconnect, 1116 .ioctl = dccp_ioctl, 1117 .init = dccp_v6_init_sock, 1118 .setsockopt = dccp_setsockopt, 1119 .getsockopt = dccp_getsockopt, 1120 .sendmsg = dccp_sendmsg, 1121 .recvmsg = dccp_recvmsg, 1122 .backlog_rcv = dccp_v6_do_rcv, 1123 .hash = dccp_v6_hash, 1124 .unhash = dccp_unhash, 1125 .accept = inet_csk_accept, 1126 .get_port = dccp_v6_get_port, 1127 .shutdown = dccp_shutdown, 1128 .destroy = dccp_v6_destroy_sock, 1129 .orphan_count = &dccp_orphan_count, 1130 .max_header = MAX_DCCP_HEADER, 1131 .obj_size = sizeof(struct dccp6_sock), 1132 .rsk_prot = &dccp6_request_sock_ops, 1133 .twsk_prot = &dccp6_timewait_sock_ops, 1134 #ifdef CONFIG_COMPAT 1135 .compat_setsockopt = compat_dccp_setsockopt, 1136 .compat_getsockopt = compat_dccp_getsockopt, 1137 #endif 1138 }; 1139 1140 static struct inet6_protocol dccp_v6_protocol = { 1141 .handler = dccp_v6_rcv, 1142 .err_handler = dccp_v6_err, 1143 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL, 1144 }; 1145 1146 static struct proto_ops inet6_dccp_ops = { 1147 .family = PF_INET6, 1148 .owner = THIS_MODULE, 1149 .release = inet6_release, 1150 .bind = inet6_bind, 1151 .connect = inet_stream_connect, 1152 .socketpair = sock_no_socketpair, 1153 .accept = inet_accept, 1154 .getname = inet6_getname, 1155 .poll = dccp_poll, 1156 .ioctl = inet6_ioctl, 1157 .listen = inet_dccp_listen, 1158 .shutdown = inet_shutdown, 1159 .setsockopt = sock_common_setsockopt, 1160 .getsockopt = sock_common_getsockopt, 1161 .sendmsg = inet_sendmsg, 1162 .recvmsg = sock_common_recvmsg, 1163 .mmap = sock_no_mmap, 1164 .sendpage = sock_no_sendpage, 1165 #ifdef CONFIG_COMPAT 1166 .compat_setsockopt = compat_sock_common_setsockopt, 1167 .compat_getsockopt = compat_sock_common_getsockopt, 1168 #endif 1169 }; 1170 1171 static struct inet_protosw dccp_v6_protosw = { 1172 .type = SOCK_DCCP, 1173 .protocol = IPPROTO_DCCP, 1174 .prot = &dccp_v6_prot, 1175 .ops = &inet6_dccp_ops, 1176 .capability = -1, 1177 .flags = INET_PROTOSW_ICSK, 1178 }; 1179 1180 static int __init dccp_v6_init(void) 1181 { 1182 int err = proto_register(&dccp_v6_prot, 1); 1183 1184 if (err != 0) 1185 goto out; 1186 1187 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1188 if (err != 0) 1189 goto out_unregister_proto; 1190 1191 inet6_register_protosw(&dccp_v6_protosw); 1192 1193 err = inet_csk_ctl_sock_create(&dccp_v6_ctl_socket, PF_INET6, 1194 SOCK_DCCP, IPPROTO_DCCP); 1195 if (err != 0) 1196 goto out_unregister_protosw; 1197 out: 1198 return err; 1199 out_unregister_protosw: 1200 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1201 inet6_unregister_protosw(&dccp_v6_protosw); 1202 out_unregister_proto: 1203 proto_unregister(&dccp_v6_prot); 1204 goto out; 1205 } 1206 1207 static void __exit dccp_v6_exit(void) 1208 { 1209 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP); 1210 inet6_unregister_protosw(&dccp_v6_protosw); 1211 proto_unregister(&dccp_v6_prot); 1212 } 1213 1214 module_init(dccp_v6_init); 1215 module_exit(dccp_v6_exit); 1216 1217 /* 1218 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33) 1219 * values directly, Also cover the case where the protocol is not specified, 1220 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP 1221 */ 1222 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-33-type-6"); 1223 MODULE_ALIAS("net-pf-" __stringify(PF_INET6) "-proto-0-type-6"); 1224 MODULE_LICENSE("GPL"); 1225 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>"); 1226 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol"); 1227