Linux Audio

Check our new training course

Real-Time Linux with PREEMPT_RT training

Feb 18-20, 2025
Register
Loading...
v5.14.15
   1// SPDX-License-Identifier: GPL-2.0-or-later
   2/*
   3 *	DCCP over IPv6
   4 *	Linux INET6 implementation
   5 *
   6 *	Based on net/dccp6/ipv6.c
   7 *
   8 *	Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
 
 
 
 
 
   9 */
  10
  11#include <linux/module.h>
  12#include <linux/random.h>
  13#include <linux/slab.h>
  14#include <linux/xfrm.h>
  15#include <linux/string.h>
  16
  17#include <net/addrconf.h>
  18#include <net/inet_common.h>
  19#include <net/inet_hashtables.h>
  20#include <net/inet_sock.h>
  21#include <net/inet6_connection_sock.h>
  22#include <net/inet6_hashtables.h>
  23#include <net/ip6_route.h>
  24#include <net/ipv6.h>
  25#include <net/protocol.h>
  26#include <net/transp_v6.h>
  27#include <net/ip6_checksum.h>
  28#include <net/xfrm.h>
  29#include <net/secure_seq.h>
  30#include <net/netns/generic.h>
  31#include <net/sock.h>
  32
  33#include "dccp.h"
  34#include "ipv6.h"
  35#include "feat.h"
  36
  37struct dccp_v6_pernet {
  38	struct sock *v6_ctl_sk;
  39};
  40
  41static unsigned int dccp_v6_pernet_id __read_mostly;
  42
  43/* The per-net v6_ctl_sk is used for sending RSTs and ACKs */
  44
  45static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
  46static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
  47
  48/* add pseudo-header to DCCP checksum stored in skb->csum */
  49static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
  50				      const struct in6_addr *saddr,
  51				      const struct in6_addr *daddr)
  52{
  53	return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
  54}
  55
  56static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
  57{
  58	struct ipv6_pinfo *np = inet6_sk(sk);
  59	struct dccp_hdr *dh = dccp_hdr(skb);
  60
  61	dccp_csum_outgoing(skb);
  62	dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
  63}
  64
  65static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
  66{
  67	return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
  68					     ipv6_hdr(skb)->saddr.s6_addr32,
  69					     dccp_hdr(skb)->dccph_dport,
  70					     dccp_hdr(skb)->dccph_sport     );
  71
  72}
  73
  74static int dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
  75			u8 type, u8 code, int offset, __be32 info)
  76{
  77	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
  78	const struct dccp_hdr *dh;
  79	struct dccp_sock *dp;
  80	struct ipv6_pinfo *np;
  81	struct sock *sk;
  82	int err;
  83	__u64 seq;
  84	struct net *net = dev_net(skb->dev);
  85
  86	/* Only need dccph_dport & dccph_sport which are the first
  87	 * 4 bytes in dccp header.
  88	 * Our caller (icmpv6_notify()) already pulled 8 bytes for us.
  89	 */
  90	BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_sport) > 8);
  91	BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_dport) > 8);
  92	dh = (struct dccp_hdr *)(skb->data + offset);
  93
  94	sk = __inet6_lookup_established(net, &dccp_hashinfo,
  95					&hdr->daddr, dh->dccph_dport,
  96					&hdr->saddr, ntohs(dh->dccph_sport),
  97					inet6_iif(skb), 0);
  98
  99	if (!sk) {
 100		__ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
 101				  ICMP6_MIB_INERRORS);
 102		return -ENOENT;
 103	}
 104
 105	if (sk->sk_state == DCCP_TIME_WAIT) {
 106		inet_twsk_put(inet_twsk(sk));
 107		return 0;
 108	}
 109	seq = dccp_hdr_seq(dh);
 110	if (sk->sk_state == DCCP_NEW_SYN_RECV) {
 111		dccp_req_err(sk, seq);
 112		return 0;
 113	}
 114
 115	bh_lock_sock(sk);
 116	if (sock_owned_by_user(sk))
 117		__NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
 118
 119	if (sk->sk_state == DCCP_CLOSED)
 120		goto out;
 121
 122	dp = dccp_sk(sk);
 123	if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
 124	    !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
 125		__NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
 126		goto out;
 127	}
 128
 129	np = inet6_sk(sk);
 130
 131	if (type == NDISC_REDIRECT) {
 132		if (!sock_owned_by_user(sk)) {
 133			struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
 134
 135			if (dst)
 136				dst->ops->redirect(dst, sk, skb);
 137		}
 138		goto out;
 139	}
 140
 141	if (type == ICMPV6_PKT_TOOBIG) {
 142		struct dst_entry *dst = NULL;
 143
 144		if (!ip6_sk_accept_pmtu(sk))
 145			goto out;
 146
 147		if (sock_owned_by_user(sk))
 148			goto out;
 149		if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
 150			goto out;
 151
 152		dst = inet6_csk_update_pmtu(sk, ntohl(info));
 153		if (!dst)
 154			goto out;
 155
 156		if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
 157			dccp_sync_mss(sk, dst_mtu(dst));
 158		goto out;
 159	}
 160
 161	icmpv6_err_convert(type, code, &err);
 162
 163	/* Might be for an request_sock */
 164	switch (sk->sk_state) {
 165	case DCCP_REQUESTING:
 166	case DCCP_RESPOND:  /* Cannot happen.
 167			       It can, it SYNs are crossed. --ANK */
 168		if (!sock_owned_by_user(sk)) {
 169			__DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
 170			sk->sk_err = err;
 171			/*
 172			 * Wake people up to see the error
 173			 * (see connect in sock.c)
 174			 */
 175			sk_error_report(sk);
 176			dccp_done(sk);
 177		} else
 178			sk->sk_err_soft = err;
 179		goto out;
 180	}
 181
 182	if (!sock_owned_by_user(sk) && np->recverr) {
 183		sk->sk_err = err;
 184		sk_error_report(sk);
 185	} else
 186		sk->sk_err_soft = err;
 187
 188out:
 189	bh_unlock_sock(sk);
 190	sock_put(sk);
 191	return 0;
 192}
 193
 194
 195static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
 196{
 197	struct inet_request_sock *ireq = inet_rsk(req);
 198	struct ipv6_pinfo *np = inet6_sk(sk);
 199	struct sk_buff *skb;
 200	struct in6_addr *final_p, final;
 201	struct flowi6 fl6;
 202	int err = -1;
 203	struct dst_entry *dst;
 204
 205	memset(&fl6, 0, sizeof(fl6));
 206	fl6.flowi6_proto = IPPROTO_DCCP;
 207	fl6.daddr = ireq->ir_v6_rmt_addr;
 208	fl6.saddr = ireq->ir_v6_loc_addr;
 209	fl6.flowlabel = 0;
 210	fl6.flowi6_oif = ireq->ir_iif;
 211	fl6.fl6_dport = ireq->ir_rmt_port;
 212	fl6.fl6_sport = htons(ireq->ir_num);
 213	security_req_classify_flow(req, flowi6_to_flowi_common(&fl6));
 214
 215
 216	rcu_read_lock();
 217	final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
 218	rcu_read_unlock();
 219
 220	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
 221	if (IS_ERR(dst)) {
 222		err = PTR_ERR(dst);
 223		dst = NULL;
 224		goto done;
 225	}
 226
 227	skb = dccp_make_response(sk, dst, req);
 228	if (skb != NULL) {
 229		struct dccp_hdr *dh = dccp_hdr(skb);
 230		struct ipv6_txoptions *opt;
 231
 232		dh->dccph_checksum = dccp_v6_csum_finish(skb,
 233							 &ireq->ir_v6_loc_addr,
 234							 &ireq->ir_v6_rmt_addr);
 235		fl6.daddr = ireq->ir_v6_rmt_addr;
 236		rcu_read_lock();
 237		opt = ireq->ipv6_opt;
 238		if (!opt)
 239			opt = rcu_dereference(np->opt);
 240		err = ip6_xmit(sk, skb, &fl6, sk->sk_mark, opt, np->tclass,
 241			       sk->sk_priority);
 242		rcu_read_unlock();
 243		err = net_xmit_eval(err);
 244	}
 245
 246done:
 247	dst_release(dst);
 248	return err;
 249}
 250
 251static void dccp_v6_reqsk_destructor(struct request_sock *req)
 252{
 253	dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
 254	kfree(inet_rsk(req)->ipv6_opt);
 255	kfree_skb(inet_rsk(req)->pktopts);
 256}
 257
 258static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
 259{
 260	const struct ipv6hdr *rxip6h;
 261	struct sk_buff *skb;
 262	struct flowi6 fl6;
 263	struct net *net = dev_net(skb_dst(rxskb)->dev);
 264	struct dccp_v6_pernet *pn;
 265	struct sock *ctl_sk;
 266	struct dst_entry *dst;
 267
 268	if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
 269		return;
 270
 271	if (!ipv6_unicast_destination(rxskb))
 272		return;
 273
 274	pn = net_generic(net, dccp_v6_pernet_id);
 275	ctl_sk = pn->v6_ctl_sk;
 276	skb = dccp_ctl_make_reset(ctl_sk, rxskb);
 277	if (skb == NULL)
 278		return;
 279
 280	rxip6h = ipv6_hdr(rxskb);
 281	dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
 282							    &rxip6h->daddr);
 283
 284	memset(&fl6, 0, sizeof(fl6));
 285	fl6.daddr = rxip6h->saddr;
 286	fl6.saddr = rxip6h->daddr;
 287
 288	fl6.flowi6_proto = IPPROTO_DCCP;
 289	fl6.flowi6_oif = inet6_iif(rxskb);
 290	fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
 291	fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
 292	security_skb_classify_flow(rxskb, flowi6_to_flowi_common(&fl6));
 293
 294	/* sk = NULL, but it is safe for now. RST socket required. */
 295	dst = ip6_dst_lookup_flow(sock_net(ctl_sk), ctl_sk, &fl6, NULL);
 296	if (!IS_ERR(dst)) {
 297		skb_dst_set(skb, dst);
 298		ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0, 0);
 299		DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
 300		DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
 301		return;
 302	}
 303
 304	kfree_skb(skb);
 305}
 306
 307static struct request_sock_ops dccp6_request_sock_ops = {
 308	.family		= AF_INET6,
 309	.obj_size	= sizeof(struct dccp6_request_sock),
 310	.rtx_syn_ack	= dccp_v6_send_response,
 311	.send_ack	= dccp_reqsk_send_ack,
 312	.destructor	= dccp_v6_reqsk_destructor,
 313	.send_reset	= dccp_v6_ctl_send_reset,
 314	.syn_ack_timeout = dccp_syn_ack_timeout,
 315};
 316
 317static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
 318{
 319	struct request_sock *req;
 320	struct dccp_request_sock *dreq;
 321	struct inet_request_sock *ireq;
 322	struct ipv6_pinfo *np = inet6_sk(sk);
 323	const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
 324	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
 325
 326	if (skb->protocol == htons(ETH_P_IP))
 327		return dccp_v4_conn_request(sk, skb);
 328
 329	if (!ipv6_unicast_destination(skb))
 330		return 0;	/* discard, don't send a reset here */
 331
 332	if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) {
 333		__IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS);
 334		return 0;
 335	}
 336
 337	if (dccp_bad_service_code(sk, service)) {
 338		dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
 339		goto drop;
 340	}
 341	/*
 342	 * There are no SYN attacks on IPv6, yet...
 343	 */
 344	dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
 345	if (inet_csk_reqsk_queue_is_full(sk))
 346		goto drop;
 347
 348	if (sk_acceptq_is_full(sk))
 349		goto drop;
 350
 351	req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
 352	if (req == NULL)
 353		goto drop;
 354
 355	if (dccp_reqsk_init(req, dccp_sk(sk), skb))
 356		goto drop_and_free;
 357
 358	dreq = dccp_rsk(req);
 359	if (dccp_parse_options(sk, dreq, skb))
 360		goto drop_and_free;
 361
 362	if (security_inet_conn_request(sk, skb, req))
 363		goto drop_and_free;
 364
 365	ireq = inet_rsk(req);
 366	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
 367	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
 368	ireq->ireq_family = AF_INET6;
 369	ireq->ir_mark = inet_request_mark(sk, skb);
 370
 371	if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
 372	    np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
 373	    np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
 374		refcount_inc(&skb->users);
 375		ireq->pktopts = skb;
 376	}
 377	ireq->ir_iif = sk->sk_bound_dev_if;
 378
 379	/* So that link locals have meaning */
 380	if (!sk->sk_bound_dev_if &&
 381	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
 382		ireq->ir_iif = inet6_iif(skb);
 383
 384	/*
 385	 * Step 3: Process LISTEN state
 386	 *
 387	 *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
 388	 *
 389	 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
 390	 */
 391	dreq->dreq_isr	   = dcb->dccpd_seq;
 392	dreq->dreq_gsr     = dreq->dreq_isr;
 393	dreq->dreq_iss	   = dccp_v6_init_sequence(skb);
 394	dreq->dreq_gss     = dreq->dreq_iss;
 395	dreq->dreq_service = service;
 396
 397	if (dccp_v6_send_response(sk, req))
 398		goto drop_and_free;
 399
 400	inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
 401	reqsk_put(req);
 402	return 0;
 403
 404drop_and_free:
 405	reqsk_free(req);
 406drop:
 407	__DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
 408	return -1;
 409}
 410
 411static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
 412					      struct sk_buff *skb,
 413					      struct request_sock *req,
 414					      struct dst_entry *dst,
 415					      struct request_sock *req_unhash,
 416					      bool *own_req)
 417{
 418	struct inet_request_sock *ireq = inet_rsk(req);
 419	struct ipv6_pinfo *newnp;
 420	const struct ipv6_pinfo *np = inet6_sk(sk);
 421	struct ipv6_txoptions *opt;
 422	struct inet_sock *newinet;
 423	struct dccp6_sock *newdp6;
 424	struct sock *newsk;
 425
 426	if (skb->protocol == htons(ETH_P_IP)) {
 427		/*
 428		 *	v6 mapped
 429		 */
 430		newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
 431						  req_unhash, own_req);
 432		if (newsk == NULL)
 433			return NULL;
 434
 435		newdp6 = (struct dccp6_sock *)newsk;
 436		newinet = inet_sk(newsk);
 437		newinet->pinet6 = &newdp6->inet6;
 438		newnp = inet6_sk(newsk);
 439
 440		memcpy(newnp, np, sizeof(struct ipv6_pinfo));
 441
 442		newnp->saddr = newsk->sk_v6_rcv_saddr;
 443
 444		inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
 445		newsk->sk_backlog_rcv = dccp_v4_do_rcv;
 446		newnp->pktoptions  = NULL;
 447		newnp->opt	   = NULL;
 448		newnp->ipv6_mc_list = NULL;
 449		newnp->ipv6_ac_list = NULL;
 450		newnp->ipv6_fl_list = NULL;
 451		newnp->mcast_oif   = inet_iif(skb);
 452		newnp->mcast_hops  = ip_hdr(skb)->ttl;
 453
 454		/*
 455		 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
 456		 * here, dccp_create_openreq_child now does this for us, see the comment in
 457		 * that function for the gory details. -acme
 458		 */
 459
 460		/* It is tricky place. Until this moment IPv4 tcp
 461		   worked with IPv6 icsk.icsk_af_ops.
 462		   Sync it now.
 463		 */
 464		dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
 465
 466		return newsk;
 467	}
 468
 469
 470	if (sk_acceptq_is_full(sk))
 471		goto out_overflow;
 472
 473	if (!dst) {
 474		struct flowi6 fl6;
 475
 476		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
 477		if (!dst)
 478			goto out;
 479	}
 480
 481	newsk = dccp_create_openreq_child(sk, req, skb);
 482	if (newsk == NULL)
 483		goto out_nonewsk;
 484
 485	/*
 486	 * No need to charge this sock to the relevant IPv6 refcnt debug socks
 487	 * count here, dccp_create_openreq_child now does this for us, see the
 488	 * comment in that function for the gory details. -acme
 489	 */
 490
 491	ip6_dst_store(newsk, dst, NULL, NULL);
 492	newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
 493						      NETIF_F_TSO);
 494	newdp6 = (struct dccp6_sock *)newsk;
 495	newinet = inet_sk(newsk);
 496	newinet->pinet6 = &newdp6->inet6;
 497	newnp = inet6_sk(newsk);
 498
 499	memcpy(newnp, np, sizeof(struct ipv6_pinfo));
 500
 501	newsk->sk_v6_daddr	= ireq->ir_v6_rmt_addr;
 502	newnp->saddr		= ireq->ir_v6_loc_addr;
 503	newsk->sk_v6_rcv_saddr	= ireq->ir_v6_loc_addr;
 504	newsk->sk_bound_dev_if	= ireq->ir_iif;
 505
 506	/* Now IPv6 options...
 507
 508	   First: no IPv4 options.
 509	 */
 510	newinet->inet_opt = NULL;
 511
 512	/* Clone RX bits */
 513	newnp->rxopt.all = np->rxopt.all;
 514
 515	newnp->ipv6_mc_list = NULL;
 516	newnp->ipv6_ac_list = NULL;
 517	newnp->ipv6_fl_list = NULL;
 518	newnp->pktoptions = NULL;
 519	newnp->opt	  = NULL;
 520	newnp->mcast_oif  = inet6_iif(skb);
 521	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
 522
 523	/*
 524	 * Clone native IPv6 options from listening socket (if any)
 525	 *
 526	 * Yes, keeping reference count would be much more clever, but we make
 527	 * one more one thing there: reattach optmem to newsk.
 528	 */
 529	opt = ireq->ipv6_opt;
 530	if (!opt)
 531		opt = rcu_dereference(np->opt);
 532	if (opt) {
 533		opt = ipv6_dup_options(newsk, opt);
 534		RCU_INIT_POINTER(newnp->opt, opt);
 535	}
 536	inet_csk(newsk)->icsk_ext_hdr_len = 0;
 537	if (opt)
 538		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
 539						    opt->opt_flen;
 540
 541	dccp_sync_mss(newsk, dst_mtu(dst));
 542
 543	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
 544	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
 545
 546	if (__inet_inherit_port(sk, newsk) < 0) {
 547		inet_csk_prepare_forced_close(newsk);
 548		dccp_done(newsk);
 549		goto out;
 550	}
 551	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
 552	/* Clone pktoptions received with SYN, if we own the req */
 553	if (*own_req && ireq->pktopts) {
 554		newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
 555		consume_skb(ireq->pktopts);
 556		ireq->pktopts = NULL;
 557		if (newnp->pktoptions)
 558			skb_set_owner_r(newnp->pktoptions, newsk);
 559	}
 560
 561	return newsk;
 562
 563out_overflow:
 564	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
 565out_nonewsk:
 566	dst_release(dst);
 567out:
 568	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
 569	return NULL;
 570}
 571
 572/* The socket must have it's spinlock held when we get
 573 * here.
 574 *
 575 * We have a potential double-lock case here, so even when
 576 * doing backlog processing we use the BH locking scheme.
 577 * This is because we cannot sleep with the original spinlock
 578 * held.
 579 */
 580static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 581{
 582	struct ipv6_pinfo *np = inet6_sk(sk);
 583	struct sk_buff *opt_skb = NULL;
 584
 585	/* Imagine: socket is IPv6. IPv4 packet arrives,
 586	   goes to IPv4 receive handler and backlogged.
 587	   From backlog it always goes here. Kerboom...
 588	   Fortunately, dccp_rcv_established and rcv_established
 589	   handle them correctly, but it is not case with
 590	   dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
 591	 */
 592
 593	if (skb->protocol == htons(ETH_P_IP))
 594		return dccp_v4_do_rcv(sk, skb);
 595
 596	if (sk_filter(sk, skb))
 597		goto discard;
 598
 599	/*
 600	 * socket locking is here for SMP purposes as backlog rcv is currently
 601	 * called with bh processing disabled.
 602	 */
 603
 604	/* Do Stevens' IPV6_PKTOPTIONS.
 605
 606	   Yes, guys, it is the only place in our code, where we
 607	   may make it not affecting IPv4.
 608	   The rest of code is protocol independent,
 609	   and I do not like idea to uglify IPv4.
 610
 611	   Actually, all the idea behind IPV6_PKTOPTIONS
 612	   looks not very well thought. For now we latch
 613	   options, received in the last packet, enqueued
 614	   by tcp. Feel free to propose better solution.
 615					       --ANK (980728)
 616	 */
 617	if (np->rxopt.all)
 
 
 
 
 618		opt_skb = skb_clone(skb, GFP_ATOMIC);
 619
 620	if (sk->sk_state == DCCP_OPEN) { /* Fast path */
 621		if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
 622			goto reset;
 623		if (opt_skb)
 624			goto ipv6_pktoptions;
 
 
 625		return 0;
 626	}
 627
 628	/*
 629	 *  Step 3: Process LISTEN state
 630	 *     If S.state == LISTEN,
 631	 *	 If P.type == Request or P contains a valid Init Cookie option,
 632	 *	      (* Must scan the packet's options to check for Init
 633	 *		 Cookies.  Only Init Cookies are processed here,
 634	 *		 however; other options are processed in Step 8.  This
 635	 *		 scan need only be performed if the endpoint uses Init
 636	 *		 Cookies *)
 637	 *	      (* Generate a new socket and switch to that socket *)
 638	 *	      Set S := new socket for this port pair
 639	 *	      S.state = RESPOND
 640	 *	      Choose S.ISS (initial seqno) or set from Init Cookies
 641	 *	      Initialize S.GAR := S.ISS
 642	 *	      Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
 643	 *	      Continue with S.state == RESPOND
 644	 *	      (* A Response packet will be generated in Step 11 *)
 645	 *	 Otherwise,
 646	 *	      Generate Reset(No Connection) unless P.type == Reset
 647	 *	      Drop packet and return
 648	 *
 649	 * NOTE: the check for the packet types is done in
 650	 *	 dccp_rcv_state_process
 651	 */
 652
 653	if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
 654		goto reset;
 655	if (opt_skb)
 656		goto ipv6_pktoptions;
 
 
 657	return 0;
 658
 659reset:
 660	dccp_v6_ctl_send_reset(sk, skb);
 661discard:
 662	if (opt_skb != NULL)
 663		__kfree_skb(opt_skb);
 664	kfree_skb(skb);
 665	return 0;
 666
 667/* Handling IPV6_PKTOPTIONS skb the similar
 668 * way it's done for net/ipv6/tcp_ipv6.c
 669 */
 670ipv6_pktoptions:
 671	if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
 672		if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
 673			np->mcast_oif = inet6_iif(opt_skb);
 674		if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
 675			np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
 676		if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
 677			np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
 678		if (np->repflow)
 679			np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
 680		if (ipv6_opt_accepted(sk, opt_skb,
 681				      &DCCP_SKB_CB(opt_skb)->header.h6)) {
 682			skb_set_owner_r(opt_skb, sk);
 683			memmove(IP6CB(opt_skb),
 684				&DCCP_SKB_CB(opt_skb)->header.h6,
 685				sizeof(struct inet6_skb_parm));
 686			opt_skb = xchg(&np->pktoptions, opt_skb);
 687		} else {
 688			__kfree_skb(opt_skb);
 689			opt_skb = xchg(&np->pktoptions, NULL);
 690		}
 691	}
 692
 693	kfree_skb(opt_skb);
 694	return 0;
 695}
 696
 697static int dccp_v6_rcv(struct sk_buff *skb)
 698{
 699	const struct dccp_hdr *dh;
 700	bool refcounted;
 701	struct sock *sk;
 702	int min_cov;
 703
 704	/* Step 1: Check header basics */
 705
 706	if (dccp_invalid_packet(skb))
 707		goto discard_it;
 708
 709	/* Step 1: If header checksum is incorrect, drop packet and return. */
 710	if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
 711				     &ipv6_hdr(skb)->daddr)) {
 712		DCCP_WARN("dropped packet with invalid checksum\n");
 713		goto discard_it;
 714	}
 715
 716	dh = dccp_hdr(skb);
 717
 718	DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
 719	DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
 720
 721	if (dccp_packet_without_ack(skb))
 722		DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
 723	else
 724		DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
 725
 726lookup:
 727	sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
 728			        dh->dccph_sport, dh->dccph_dport,
 729				inet6_iif(skb), 0, &refcounted);
 730	if (!sk) {
 731		dccp_pr_debug("failed to look up flow ID in table and "
 732			      "get corresponding socket\n");
 733		goto no_dccp_socket;
 734	}
 735
 736	/*
 737	 * Step 2:
 738	 *	... or S.state == TIMEWAIT,
 739	 *		Generate Reset(No Connection) unless P.type == Reset
 740	 *		Drop packet and return
 741	 */
 742	if (sk->sk_state == DCCP_TIME_WAIT) {
 743		dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
 744		inet_twsk_put(inet_twsk(sk));
 745		goto no_dccp_socket;
 746	}
 747
 748	if (sk->sk_state == DCCP_NEW_SYN_RECV) {
 749		struct request_sock *req = inet_reqsk(sk);
 750		struct sock *nsk;
 751
 752		sk = req->rsk_listener;
 753		if (unlikely(sk->sk_state != DCCP_LISTEN)) {
 754			inet_csk_reqsk_queue_drop_and_put(sk, req);
 755			goto lookup;
 756		}
 757		sock_hold(sk);
 758		refcounted = true;
 759		nsk = dccp_check_req(sk, skb, req);
 760		if (!nsk) {
 761			reqsk_put(req);
 762			goto discard_and_relse;
 763		}
 764		if (nsk == sk) {
 765			reqsk_put(req);
 766		} else if (dccp_child_process(sk, nsk, skb)) {
 767			dccp_v6_ctl_send_reset(sk, skb);
 768			goto discard_and_relse;
 769		} else {
 770			sock_put(sk);
 771			return 0;
 772		}
 773	}
 774	/*
 775	 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
 776	 *	o if MinCsCov = 0, only packets with CsCov = 0 are accepted
 777	 *	o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
 778	 */
 779	min_cov = dccp_sk(sk)->dccps_pcrlen;
 780	if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
 781		dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
 782			      dh->dccph_cscov, min_cov);
 783		/* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
 784		goto discard_and_relse;
 785	}
 786
 787	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
 788		goto discard_and_relse;
 789
 790	return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
 791				refcounted) ? -1 : 0;
 792
 793no_dccp_socket:
 794	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 795		goto discard_it;
 796	/*
 797	 * Step 2:
 798	 *	If no socket ...
 799	 *		Generate Reset(No Connection) unless P.type == Reset
 800	 *		Drop packet and return
 801	 */
 802	if (dh->dccph_type != DCCP_PKT_RESET) {
 803		DCCP_SKB_CB(skb)->dccpd_reset_code =
 804					DCCP_RESET_CODE_NO_CONNECTION;
 805		dccp_v6_ctl_send_reset(sk, skb);
 806	}
 807
 808discard_it:
 809	kfree_skb(skb);
 810	return 0;
 811
 812discard_and_relse:
 813	if (refcounted)
 814		sock_put(sk);
 815	goto discard_it;
 816}
 817
 818static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 819			   int addr_len)
 820{
 821	struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
 822	struct inet_connection_sock *icsk = inet_csk(sk);
 823	struct inet_sock *inet = inet_sk(sk);
 824	struct ipv6_pinfo *np = inet6_sk(sk);
 825	struct dccp_sock *dp = dccp_sk(sk);
 826	struct in6_addr *saddr = NULL, *final_p, final;
 827	struct ipv6_txoptions *opt;
 828	struct flowi6 fl6;
 829	struct dst_entry *dst;
 830	int addr_type;
 831	int err;
 832
 833	dp->dccps_role = DCCP_ROLE_CLIENT;
 834
 835	if (addr_len < SIN6_LEN_RFC2133)
 836		return -EINVAL;
 837
 838	if (usin->sin6_family != AF_INET6)
 839		return -EAFNOSUPPORT;
 840
 841	memset(&fl6, 0, sizeof(fl6));
 842
 843	if (np->sndflow) {
 844		fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
 845		IP6_ECN_flow_init(fl6.flowlabel);
 846		if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
 847			struct ip6_flowlabel *flowlabel;
 848			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
 849			if (IS_ERR(flowlabel))
 850				return -EINVAL;
 851			fl6_sock_release(flowlabel);
 852		}
 853	}
 854	/*
 855	 * connect() to INADDR_ANY means loopback (BSD'ism).
 856	 */
 857	if (ipv6_addr_any(&usin->sin6_addr))
 858		usin->sin6_addr.s6_addr[15] = 1;
 859
 860	addr_type = ipv6_addr_type(&usin->sin6_addr);
 861
 862	if (addr_type & IPV6_ADDR_MULTICAST)
 863		return -ENETUNREACH;
 864
 865	if (addr_type & IPV6_ADDR_LINKLOCAL) {
 866		if (addr_len >= sizeof(struct sockaddr_in6) &&
 867		    usin->sin6_scope_id) {
 868			/* If interface is set while binding, indices
 869			 * must coincide.
 870			 */
 871			if (sk->sk_bound_dev_if &&
 872			    sk->sk_bound_dev_if != usin->sin6_scope_id)
 873				return -EINVAL;
 874
 875			sk->sk_bound_dev_if = usin->sin6_scope_id;
 876		}
 877
 878		/* Connect to link-local address requires an interface */
 879		if (!sk->sk_bound_dev_if)
 880			return -EINVAL;
 881	}
 882
 883	sk->sk_v6_daddr = usin->sin6_addr;
 884	np->flow_label = fl6.flowlabel;
 885
 886	/*
 887	 * DCCP over IPv4
 888	 */
 889	if (addr_type == IPV6_ADDR_MAPPED) {
 890		u32 exthdrlen = icsk->icsk_ext_hdr_len;
 891		struct sockaddr_in sin;
 892
 893		SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
 894
 895		if (__ipv6_only_sock(sk))
 896			return -ENETUNREACH;
 897
 898		sin.sin_family = AF_INET;
 899		sin.sin_port = usin->sin6_port;
 900		sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
 901
 902		icsk->icsk_af_ops = &dccp_ipv6_mapped;
 903		sk->sk_backlog_rcv = dccp_v4_do_rcv;
 904
 905		err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
 906		if (err) {
 907			icsk->icsk_ext_hdr_len = exthdrlen;
 908			icsk->icsk_af_ops = &dccp_ipv6_af_ops;
 909			sk->sk_backlog_rcv = dccp_v6_do_rcv;
 910			goto failure;
 911		}
 912		np->saddr = sk->sk_v6_rcv_saddr;
 913		return err;
 914	}
 915
 916	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 917		saddr = &sk->sk_v6_rcv_saddr;
 918
 919	fl6.flowi6_proto = IPPROTO_DCCP;
 920	fl6.daddr = sk->sk_v6_daddr;
 921	fl6.saddr = saddr ? *saddr : np->saddr;
 922	fl6.flowi6_oif = sk->sk_bound_dev_if;
 923	fl6.fl6_dport = usin->sin6_port;
 924	fl6.fl6_sport = inet->inet_sport;
 925	security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
 926
 927	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
 928	final_p = fl6_update_dst(&fl6, opt, &final);
 929
 930	dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
 931	if (IS_ERR(dst)) {
 932		err = PTR_ERR(dst);
 933		goto failure;
 934	}
 935
 936	if (saddr == NULL) {
 937		saddr = &fl6.saddr;
 938		sk->sk_v6_rcv_saddr = *saddr;
 939	}
 940
 941	/* set the source address */
 942	np->saddr = *saddr;
 943	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
 944
 945	ip6_dst_store(sk, dst, NULL, NULL);
 946
 947	icsk->icsk_ext_hdr_len = 0;
 948	if (opt)
 949		icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
 950
 951	inet->inet_dport = usin->sin6_port;
 952
 953	dccp_set_state(sk, DCCP_REQUESTING);
 954	err = inet6_hash_connect(&dccp_death_row, sk);
 955	if (err)
 956		goto late_failure;
 957
 958	dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
 959						      sk->sk_v6_daddr.s6_addr32,
 960						      inet->inet_sport,
 961						      inet->inet_dport);
 962	err = dccp_connect(sk);
 963	if (err)
 964		goto late_failure;
 965
 966	return 0;
 967
 968late_failure:
 969	dccp_set_state(sk, DCCP_CLOSED);
 970	__sk_dst_reset(sk);
 971failure:
 972	inet->inet_dport = 0;
 973	sk->sk_route_caps = 0;
 974	return err;
 975}
 976
 977static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
 978	.queue_xmit	   = inet6_csk_xmit,
 979	.send_check	   = dccp_v6_send_check,
 980	.rebuild_header	   = inet6_sk_rebuild_header,
 981	.conn_request	   = dccp_v6_conn_request,
 982	.syn_recv_sock	   = dccp_v6_request_recv_sock,
 983	.net_header_len	   = sizeof(struct ipv6hdr),
 984	.setsockopt	   = ipv6_setsockopt,
 985	.getsockopt	   = ipv6_getsockopt,
 986	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
 987	.sockaddr_len	   = sizeof(struct sockaddr_in6),
 
 
 
 
 
 988};
 989
 990/*
 991 *	DCCP over IPv4 via INET6 API
 992 */
 993static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
 994	.queue_xmit	   = ip_queue_xmit,
 995	.send_check	   = dccp_v4_send_check,
 996	.rebuild_header	   = inet_sk_rebuild_header,
 997	.conn_request	   = dccp_v6_conn_request,
 998	.syn_recv_sock	   = dccp_v6_request_recv_sock,
 999	.net_header_len	   = sizeof(struct iphdr),
1000	.setsockopt	   = ipv6_setsockopt,
1001	.getsockopt	   = ipv6_getsockopt,
1002	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
1003	.sockaddr_len	   = sizeof(struct sockaddr_in6),
 
 
 
 
1004};
1005
1006/* NOTE: A lot of things set to zero explicitly by call to
1007 *       sk_alloc() so need not be done here.
1008 */
1009static int dccp_v6_init_sock(struct sock *sk)
1010{
1011	static __u8 dccp_v6_ctl_sock_initialized;
1012	int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1013
1014	if (err == 0) {
1015		if (unlikely(!dccp_v6_ctl_sock_initialized))
1016			dccp_v6_ctl_sock_initialized = 1;
1017		inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1018	}
1019
1020	return err;
1021}
1022
1023static void dccp_v6_destroy_sock(struct sock *sk)
1024{
1025	dccp_destroy_sock(sk);
1026	inet6_destroy_sock(sk);
1027}
1028
1029static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1030	.twsk_obj_size	= sizeof(struct dccp6_timewait_sock),
1031};
1032
1033static struct proto dccp_v6_prot = {
1034	.name		   = "DCCPv6",
1035	.owner		   = THIS_MODULE,
1036	.close		   = dccp_close,
1037	.connect	   = dccp_v6_connect,
1038	.disconnect	   = dccp_disconnect,
1039	.ioctl		   = dccp_ioctl,
1040	.init		   = dccp_v6_init_sock,
1041	.setsockopt	   = dccp_setsockopt,
1042	.getsockopt	   = dccp_getsockopt,
1043	.sendmsg	   = dccp_sendmsg,
1044	.recvmsg	   = dccp_recvmsg,
1045	.backlog_rcv	   = dccp_v6_do_rcv,
1046	.hash		   = inet6_hash,
1047	.unhash		   = inet_unhash,
1048	.accept		   = inet_csk_accept,
1049	.get_port	   = inet_csk_get_port,
1050	.shutdown	   = dccp_shutdown,
1051	.destroy	   = dccp_v6_destroy_sock,
1052	.orphan_count	   = &dccp_orphan_count,
1053	.max_header	   = MAX_DCCP_HEADER,
1054	.obj_size	   = sizeof(struct dccp6_sock),
1055	.slab_flags	   = SLAB_TYPESAFE_BY_RCU,
1056	.rsk_prot	   = &dccp6_request_sock_ops,
1057	.twsk_prot	   = &dccp6_timewait_sock_ops,
1058	.h.hashinfo	   = &dccp_hashinfo,
 
 
 
 
1059};
1060
1061static const struct inet6_protocol dccp_v6_protocol = {
1062	.handler	= dccp_v6_rcv,
1063	.err_handler	= dccp_v6_err,
1064	.flags		= INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1065};
1066
1067static const struct proto_ops inet6_dccp_ops = {
1068	.family		   = PF_INET6,
1069	.owner		   = THIS_MODULE,
1070	.release	   = inet6_release,
1071	.bind		   = inet6_bind,
1072	.connect	   = inet_stream_connect,
1073	.socketpair	   = sock_no_socketpair,
1074	.accept		   = inet_accept,
1075	.getname	   = inet6_getname,
1076	.poll		   = dccp_poll,
1077	.ioctl		   = inet6_ioctl,
1078	.gettstamp	   = sock_gettstamp,
1079	.listen		   = inet_dccp_listen,
1080	.shutdown	   = inet_shutdown,
1081	.setsockopt	   = sock_common_setsockopt,
1082	.getsockopt	   = sock_common_getsockopt,
1083	.sendmsg	   = inet_sendmsg,
1084	.recvmsg	   = sock_common_recvmsg,
1085	.mmap		   = sock_no_mmap,
1086	.sendpage	   = sock_no_sendpage,
1087#ifdef CONFIG_COMPAT
1088	.compat_ioctl	   = inet6_compat_ioctl,
 
1089#endif
1090};
1091
1092static struct inet_protosw dccp_v6_protosw = {
1093	.type		= SOCK_DCCP,
1094	.protocol	= IPPROTO_DCCP,
1095	.prot		= &dccp_v6_prot,
1096	.ops		= &inet6_dccp_ops,
1097	.flags		= INET_PROTOSW_ICSK,
1098};
1099
1100static int __net_init dccp_v6_init_net(struct net *net)
1101{
1102	struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id);
1103
1104	if (dccp_hashinfo.bhash == NULL)
1105		return -ESOCKTNOSUPPORT;
1106
1107	return inet_ctl_sock_create(&pn->v6_ctl_sk, PF_INET6,
1108				    SOCK_DCCP, IPPROTO_DCCP, net);
1109}
1110
1111static void __net_exit dccp_v6_exit_net(struct net *net)
1112{
1113	struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id);
1114
1115	inet_ctl_sock_destroy(pn->v6_ctl_sk);
1116}
1117
1118static void __net_exit dccp_v6_exit_batch(struct list_head *net_exit_list)
1119{
1120	inet_twsk_purge(&dccp_hashinfo, AF_INET6);
1121}
1122
1123static struct pernet_operations dccp_v6_ops = {
1124	.init   = dccp_v6_init_net,
1125	.exit   = dccp_v6_exit_net,
1126	.exit_batch = dccp_v6_exit_batch,
1127	.id	= &dccp_v6_pernet_id,
1128	.size   = sizeof(struct dccp_v6_pernet),
1129};
1130
1131static int __init dccp_v6_init(void)
1132{
1133	int err = proto_register(&dccp_v6_prot, 1);
1134
1135	if (err)
1136		goto out;
1137
 
 
 
 
1138	inet6_register_protosw(&dccp_v6_protosw);
1139
1140	err = register_pernet_subsys(&dccp_v6_ops);
1141	if (err)
1142		goto out_destroy_ctl_sock;
1143
1144	err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1145	if (err)
1146		goto out_unregister_proto;
1147
1148out:
1149	return err;
1150out_unregister_proto:
1151	unregister_pernet_subsys(&dccp_v6_ops);
1152out_destroy_ctl_sock:
 
1153	inet6_unregister_protosw(&dccp_v6_protosw);
 
1154	proto_unregister(&dccp_v6_prot);
1155	goto out;
1156}
1157
1158static void __exit dccp_v6_exit(void)
1159{
1160	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1161	unregister_pernet_subsys(&dccp_v6_ops);
 
1162	inet6_unregister_protosw(&dccp_v6_protosw);
1163	proto_unregister(&dccp_v6_prot);
1164}
1165
1166module_init(dccp_v6_init);
1167module_exit(dccp_v6_exit);
1168
1169/*
1170 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1171 * values directly, Also cover the case where the protocol is not specified,
1172 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1173 */
1174MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1175MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1176MODULE_LICENSE("GPL");
1177MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1178MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
v4.6
 
   1/*
   2 *	DCCP over IPv6
   3 *	Linux INET6 implementation
   4 *
   5 *	Based on net/dccp6/ipv6.c
   6 *
   7 *	Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
   8 *
   9 *	This program is free software; you can redistribute it and/or
  10 *      modify it under the terms of the GNU General Public License
  11 *      as published by the Free Software Foundation; either version
  12 *      2 of the License, or (at your option) any later version.
  13 */
  14
  15#include <linux/module.h>
  16#include <linux/random.h>
  17#include <linux/slab.h>
  18#include <linux/xfrm.h>
 
  19
  20#include <net/addrconf.h>
  21#include <net/inet_common.h>
  22#include <net/inet_hashtables.h>
  23#include <net/inet_sock.h>
  24#include <net/inet6_connection_sock.h>
  25#include <net/inet6_hashtables.h>
  26#include <net/ip6_route.h>
  27#include <net/ipv6.h>
  28#include <net/protocol.h>
  29#include <net/transp_v6.h>
  30#include <net/ip6_checksum.h>
  31#include <net/xfrm.h>
  32#include <net/secure_seq.h>
 
 
  33
  34#include "dccp.h"
  35#include "ipv6.h"
  36#include "feat.h"
  37
  38/* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
 
 
 
 
 
 
  39
  40static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
  41static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
  42
  43/* add pseudo-header to DCCP checksum stored in skb->csum */
  44static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
  45				      const struct in6_addr *saddr,
  46				      const struct in6_addr *daddr)
  47{
  48	return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
  49}
  50
  51static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
  52{
  53	struct ipv6_pinfo *np = inet6_sk(sk);
  54	struct dccp_hdr *dh = dccp_hdr(skb);
  55
  56	dccp_csum_outgoing(skb);
  57	dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
  58}
  59
  60static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
  61{
  62	return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
  63					     ipv6_hdr(skb)->saddr.s6_addr32,
  64					     dccp_hdr(skb)->dccph_dport,
  65					     dccp_hdr(skb)->dccph_sport     );
  66
  67}
  68
  69static void dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
  70			u8 type, u8 code, int offset, __be32 info)
  71{
  72	const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
  73	const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
  74	struct dccp_sock *dp;
  75	struct ipv6_pinfo *np;
  76	struct sock *sk;
  77	int err;
  78	__u64 seq;
  79	struct net *net = dev_net(skb->dev);
  80
  81	if (skb->len < offset + sizeof(*dh) ||
  82	    skb->len < offset + __dccp_basic_hdr_len(dh)) {
  83		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
  84				   ICMP6_MIB_INERRORS);
  85		return;
  86	}
 
  87
  88	sk = __inet6_lookup_established(net, &dccp_hashinfo,
  89					&hdr->daddr, dh->dccph_dport,
  90					&hdr->saddr, ntohs(dh->dccph_sport),
  91					inet6_iif(skb));
  92
  93	if (!sk) {
  94		ICMP6_INC_STATS_BH(net, __in6_dev_get(skb->dev),
  95				   ICMP6_MIB_INERRORS);
  96		return;
  97	}
  98
  99	if (sk->sk_state == DCCP_TIME_WAIT) {
 100		inet_twsk_put(inet_twsk(sk));
 101		return;
 102	}
 103	seq = dccp_hdr_seq(dh);
 104	if (sk->sk_state == DCCP_NEW_SYN_RECV)
 105		return dccp_req_err(sk, seq);
 
 
 106
 107	bh_lock_sock(sk);
 108	if (sock_owned_by_user(sk))
 109		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
 110
 111	if (sk->sk_state == DCCP_CLOSED)
 112		goto out;
 113
 114	dp = dccp_sk(sk);
 115	if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
 116	    !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
 117		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
 118		goto out;
 119	}
 120
 121	np = inet6_sk(sk);
 122
 123	if (type == NDISC_REDIRECT) {
 124		struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
 
 125
 126		if (dst)
 127			dst->ops->redirect(dst, sk, skb);
 
 128		goto out;
 129	}
 130
 131	if (type == ICMPV6_PKT_TOOBIG) {
 132		struct dst_entry *dst = NULL;
 133
 134		if (!ip6_sk_accept_pmtu(sk))
 135			goto out;
 136
 137		if (sock_owned_by_user(sk))
 138			goto out;
 139		if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
 140			goto out;
 141
 142		dst = inet6_csk_update_pmtu(sk, ntohl(info));
 143		if (!dst)
 144			goto out;
 145
 146		if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
 147			dccp_sync_mss(sk, dst_mtu(dst));
 148		goto out;
 149	}
 150
 151	icmpv6_err_convert(type, code, &err);
 152
 153	/* Might be for an request_sock */
 154	switch (sk->sk_state) {
 155	case DCCP_REQUESTING:
 156	case DCCP_RESPOND:  /* Cannot happen.
 157			       It can, it SYNs are crossed. --ANK */
 158		if (!sock_owned_by_user(sk)) {
 159			DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
 160			sk->sk_err = err;
 161			/*
 162			 * Wake people up to see the error
 163			 * (see connect in sock.c)
 164			 */
 165			sk->sk_error_report(sk);
 166			dccp_done(sk);
 167		} else
 168			sk->sk_err_soft = err;
 169		goto out;
 170	}
 171
 172	if (!sock_owned_by_user(sk) && np->recverr) {
 173		sk->sk_err = err;
 174		sk->sk_error_report(sk);
 175	} else
 176		sk->sk_err_soft = err;
 177
 178out:
 179	bh_unlock_sock(sk);
 180	sock_put(sk);
 
 181}
 182
 183
 184static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
 185{
 186	struct inet_request_sock *ireq = inet_rsk(req);
 187	struct ipv6_pinfo *np = inet6_sk(sk);
 188	struct sk_buff *skb;
 189	struct in6_addr *final_p, final;
 190	struct flowi6 fl6;
 191	int err = -1;
 192	struct dst_entry *dst;
 193
 194	memset(&fl6, 0, sizeof(fl6));
 195	fl6.flowi6_proto = IPPROTO_DCCP;
 196	fl6.daddr = ireq->ir_v6_rmt_addr;
 197	fl6.saddr = ireq->ir_v6_loc_addr;
 198	fl6.flowlabel = 0;
 199	fl6.flowi6_oif = ireq->ir_iif;
 200	fl6.fl6_dport = ireq->ir_rmt_port;
 201	fl6.fl6_sport = htons(ireq->ir_num);
 202	security_req_classify_flow(req, flowi6_to_flowi(&fl6));
 203
 204
 205	rcu_read_lock();
 206	final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
 207	rcu_read_unlock();
 208
 209	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 210	if (IS_ERR(dst)) {
 211		err = PTR_ERR(dst);
 212		dst = NULL;
 213		goto done;
 214	}
 215
 216	skb = dccp_make_response(sk, dst, req);
 217	if (skb != NULL) {
 218		struct dccp_hdr *dh = dccp_hdr(skb);
 
 219
 220		dh->dccph_checksum = dccp_v6_csum_finish(skb,
 221							 &ireq->ir_v6_loc_addr,
 222							 &ireq->ir_v6_rmt_addr);
 223		fl6.daddr = ireq->ir_v6_rmt_addr;
 224		rcu_read_lock();
 225		err = ip6_xmit(sk, skb, &fl6, rcu_dereference(np->opt),
 226			       np->tclass);
 
 
 
 227		rcu_read_unlock();
 228		err = net_xmit_eval(err);
 229	}
 230
 231done:
 232	dst_release(dst);
 233	return err;
 234}
 235
 236static void dccp_v6_reqsk_destructor(struct request_sock *req)
 237{
 238	dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
 
 239	kfree_skb(inet_rsk(req)->pktopts);
 240}
 241
 242static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
 243{
 244	const struct ipv6hdr *rxip6h;
 245	struct sk_buff *skb;
 246	struct flowi6 fl6;
 247	struct net *net = dev_net(skb_dst(rxskb)->dev);
 248	struct sock *ctl_sk = net->dccp.v6_ctl_sk;
 
 249	struct dst_entry *dst;
 250
 251	if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
 252		return;
 253
 254	if (!ipv6_unicast_destination(rxskb))
 255		return;
 256
 
 
 257	skb = dccp_ctl_make_reset(ctl_sk, rxskb);
 258	if (skb == NULL)
 259		return;
 260
 261	rxip6h = ipv6_hdr(rxskb);
 262	dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
 263							    &rxip6h->daddr);
 264
 265	memset(&fl6, 0, sizeof(fl6));
 266	fl6.daddr = rxip6h->saddr;
 267	fl6.saddr = rxip6h->daddr;
 268
 269	fl6.flowi6_proto = IPPROTO_DCCP;
 270	fl6.flowi6_oif = inet6_iif(rxskb);
 271	fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
 272	fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
 273	security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
 274
 275	/* sk = NULL, but it is safe for now. RST socket required. */
 276	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
 277	if (!IS_ERR(dst)) {
 278		skb_dst_set(skb, dst);
 279		ip6_xmit(ctl_sk, skb, &fl6, NULL, 0);
 280		DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
 281		DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
 282		return;
 283	}
 284
 285	kfree_skb(skb);
 286}
 287
 288static struct request_sock_ops dccp6_request_sock_ops = {
 289	.family		= AF_INET6,
 290	.obj_size	= sizeof(struct dccp6_request_sock),
 291	.rtx_syn_ack	= dccp_v6_send_response,
 292	.send_ack	= dccp_reqsk_send_ack,
 293	.destructor	= dccp_v6_reqsk_destructor,
 294	.send_reset	= dccp_v6_ctl_send_reset,
 295	.syn_ack_timeout = dccp_syn_ack_timeout,
 296};
 297
 298static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
 299{
 300	struct request_sock *req;
 301	struct dccp_request_sock *dreq;
 302	struct inet_request_sock *ireq;
 303	struct ipv6_pinfo *np = inet6_sk(sk);
 304	const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
 305	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
 306
 307	if (skb->protocol == htons(ETH_P_IP))
 308		return dccp_v4_conn_request(sk, skb);
 309
 310	if (!ipv6_unicast_destination(skb))
 311		return 0;	/* discard, don't send a reset here */
 312
 
 
 
 
 
 313	if (dccp_bad_service_code(sk, service)) {
 314		dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
 315		goto drop;
 316	}
 317	/*
 318	 * There are no SYN attacks on IPv6, yet...
 319	 */
 320	dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
 321	if (inet_csk_reqsk_queue_is_full(sk))
 322		goto drop;
 323
 324	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
 325		goto drop;
 326
 327	req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
 328	if (req == NULL)
 329		goto drop;
 330
 331	if (dccp_reqsk_init(req, dccp_sk(sk), skb))
 332		goto drop_and_free;
 333
 334	dreq = dccp_rsk(req);
 335	if (dccp_parse_options(sk, dreq, skb))
 336		goto drop_and_free;
 337
 338	if (security_inet_conn_request(sk, skb, req))
 339		goto drop_and_free;
 340
 341	ireq = inet_rsk(req);
 342	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
 343	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
 344	ireq->ireq_family = AF_INET6;
 
 345
 346	if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
 347	    np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
 348	    np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
 349		atomic_inc(&skb->users);
 350		ireq->pktopts = skb;
 351	}
 352	ireq->ir_iif = sk->sk_bound_dev_if;
 353
 354	/* So that link locals have meaning */
 355	if (!sk->sk_bound_dev_if &&
 356	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
 357		ireq->ir_iif = inet6_iif(skb);
 358
 359	/*
 360	 * Step 3: Process LISTEN state
 361	 *
 362	 *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
 363	 *
 364	 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
 365	 */
 366	dreq->dreq_isr	   = dcb->dccpd_seq;
 367	dreq->dreq_gsr     = dreq->dreq_isr;
 368	dreq->dreq_iss	   = dccp_v6_init_sequence(skb);
 369	dreq->dreq_gss     = dreq->dreq_iss;
 370	dreq->dreq_service = service;
 371
 372	if (dccp_v6_send_response(sk, req))
 373		goto drop_and_free;
 374
 375	inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
 
 376	return 0;
 377
 378drop_and_free:
 379	reqsk_free(req);
 380drop:
 381	DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
 382	return -1;
 383}
 384
 385static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
 386					      struct sk_buff *skb,
 387					      struct request_sock *req,
 388					      struct dst_entry *dst,
 389					      struct request_sock *req_unhash,
 390					      bool *own_req)
 391{
 392	struct inet_request_sock *ireq = inet_rsk(req);
 393	struct ipv6_pinfo *newnp;
 394	const struct ipv6_pinfo *np = inet6_sk(sk);
 395	struct ipv6_txoptions *opt;
 396	struct inet_sock *newinet;
 397	struct dccp6_sock *newdp6;
 398	struct sock *newsk;
 399
 400	if (skb->protocol == htons(ETH_P_IP)) {
 401		/*
 402		 *	v6 mapped
 403		 */
 404		newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
 405						  req_unhash, own_req);
 406		if (newsk == NULL)
 407			return NULL;
 408
 409		newdp6 = (struct dccp6_sock *)newsk;
 410		newinet = inet_sk(newsk);
 411		newinet->pinet6 = &newdp6->inet6;
 412		newnp = inet6_sk(newsk);
 413
 414		memcpy(newnp, np, sizeof(struct ipv6_pinfo));
 415
 416		newnp->saddr = newsk->sk_v6_rcv_saddr;
 417
 418		inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
 419		newsk->sk_backlog_rcv = dccp_v4_do_rcv;
 420		newnp->pktoptions  = NULL;
 421		newnp->opt	   = NULL;
 422		newnp->mcast_oif   = inet6_iif(skb);
 423		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
 
 
 
 424
 425		/*
 426		 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
 427		 * here, dccp_create_openreq_child now does this for us, see the comment in
 428		 * that function for the gory details. -acme
 429		 */
 430
 431		/* It is tricky place. Until this moment IPv4 tcp
 432		   worked with IPv6 icsk.icsk_af_ops.
 433		   Sync it now.
 434		 */
 435		dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
 436
 437		return newsk;
 438	}
 439
 440
 441	if (sk_acceptq_is_full(sk))
 442		goto out_overflow;
 443
 444	if (!dst) {
 445		struct flowi6 fl6;
 446
 447		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
 448		if (!dst)
 449			goto out;
 450	}
 451
 452	newsk = dccp_create_openreq_child(sk, req, skb);
 453	if (newsk == NULL)
 454		goto out_nonewsk;
 455
 456	/*
 457	 * No need to charge this sock to the relevant IPv6 refcnt debug socks
 458	 * count here, dccp_create_openreq_child now does this for us, see the
 459	 * comment in that function for the gory details. -acme
 460	 */
 461
 462	ip6_dst_store(newsk, dst, NULL, NULL);
 463	newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
 464						      NETIF_F_TSO);
 465	newdp6 = (struct dccp6_sock *)newsk;
 466	newinet = inet_sk(newsk);
 467	newinet->pinet6 = &newdp6->inet6;
 468	newnp = inet6_sk(newsk);
 469
 470	memcpy(newnp, np, sizeof(struct ipv6_pinfo));
 471
 472	newsk->sk_v6_daddr	= ireq->ir_v6_rmt_addr;
 473	newnp->saddr		= ireq->ir_v6_loc_addr;
 474	newsk->sk_v6_rcv_saddr	= ireq->ir_v6_loc_addr;
 475	newsk->sk_bound_dev_if	= ireq->ir_iif;
 476
 477	/* Now IPv6 options...
 478
 479	   First: no IPv4 options.
 480	 */
 481	newinet->inet_opt = NULL;
 482
 483	/* Clone RX bits */
 484	newnp->rxopt.all = np->rxopt.all;
 485
 
 
 
 486	newnp->pktoptions = NULL;
 487	newnp->opt	  = NULL;
 488	newnp->mcast_oif  = inet6_iif(skb);
 489	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
 490
 491	/*
 492	 * Clone native IPv6 options from listening socket (if any)
 493	 *
 494	 * Yes, keeping reference count would be much more clever, but we make
 495	 * one more one thing there: reattach optmem to newsk.
 496	 */
 497	opt = rcu_dereference(np->opt);
 
 
 498	if (opt) {
 499		opt = ipv6_dup_options(newsk, opt);
 500		RCU_INIT_POINTER(newnp->opt, opt);
 501	}
 502	inet_csk(newsk)->icsk_ext_hdr_len = 0;
 503	if (opt)
 504		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
 505						    opt->opt_flen;
 506
 507	dccp_sync_mss(newsk, dst_mtu(dst));
 508
 509	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
 510	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
 511
 512	if (__inet_inherit_port(sk, newsk) < 0) {
 513		inet_csk_prepare_forced_close(newsk);
 514		dccp_done(newsk);
 515		goto out;
 516	}
 517	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
 518	/* Clone pktoptions received with SYN, if we own the req */
 519	if (*own_req && ireq->pktopts) {
 520		newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
 521		consume_skb(ireq->pktopts);
 522		ireq->pktopts = NULL;
 523		if (newnp->pktoptions)
 524			skb_set_owner_r(newnp->pktoptions, newsk);
 525	}
 526
 527	return newsk;
 528
 529out_overflow:
 530	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
 531out_nonewsk:
 532	dst_release(dst);
 533out:
 534	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
 535	return NULL;
 536}
 537
 538/* The socket must have it's spinlock held when we get
 539 * here.
 540 *
 541 * We have a potential double-lock case here, so even when
 542 * doing backlog processing we use the BH locking scheme.
 543 * This is because we cannot sleep with the original spinlock
 544 * held.
 545 */
 546static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
 547{
 548	struct ipv6_pinfo *np = inet6_sk(sk);
 549	struct sk_buff *opt_skb = NULL;
 550
 551	/* Imagine: socket is IPv6. IPv4 packet arrives,
 552	   goes to IPv4 receive handler and backlogged.
 553	   From backlog it always goes here. Kerboom...
 554	   Fortunately, dccp_rcv_established and rcv_established
 555	   handle them correctly, but it is not case with
 556	   dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
 557	 */
 558
 559	if (skb->protocol == htons(ETH_P_IP))
 560		return dccp_v4_do_rcv(sk, skb);
 561
 562	if (sk_filter(sk, skb))
 563		goto discard;
 564
 565	/*
 566	 * socket locking is here for SMP purposes as backlog rcv is currently
 567	 * called with bh processing disabled.
 568	 */
 569
 570	/* Do Stevens' IPV6_PKTOPTIONS.
 571
 572	   Yes, guys, it is the only place in our code, where we
 573	   may make it not affecting IPv4.
 574	   The rest of code is protocol independent,
 575	   and I do not like idea to uglify IPv4.
 576
 577	   Actually, all the idea behind IPV6_PKTOPTIONS
 578	   looks not very well thought. For now we latch
 579	   options, received in the last packet, enqueued
 580	   by tcp. Feel free to propose better solution.
 581					       --ANK (980728)
 582	 */
 583	if (np->rxopt.all)
 584	/*
 585	 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
 586	 *        (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
 587	 */
 588		opt_skb = skb_clone(skb, GFP_ATOMIC);
 589
 590	if (sk->sk_state == DCCP_OPEN) { /* Fast path */
 591		if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
 592			goto reset;
 593		if (opt_skb) {
 594			/* XXX This is where we would goto ipv6_pktoptions. */
 595			__kfree_skb(opt_skb);
 596		}
 597		return 0;
 598	}
 599
 600	/*
 601	 *  Step 3: Process LISTEN state
 602	 *     If S.state == LISTEN,
 603	 *	 If P.type == Request or P contains a valid Init Cookie option,
 604	 *	      (* Must scan the packet's options to check for Init
 605	 *		 Cookies.  Only Init Cookies are processed here,
 606	 *		 however; other options are processed in Step 8.  This
 607	 *		 scan need only be performed if the endpoint uses Init
 608	 *		 Cookies *)
 609	 *	      (* Generate a new socket and switch to that socket *)
 610	 *	      Set S := new socket for this port pair
 611	 *	      S.state = RESPOND
 612	 *	      Choose S.ISS (initial seqno) or set from Init Cookies
 613	 *	      Initialize S.GAR := S.ISS
 614	 *	      Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
 615	 *	      Continue with S.state == RESPOND
 616	 *	      (* A Response packet will be generated in Step 11 *)
 617	 *	 Otherwise,
 618	 *	      Generate Reset(No Connection) unless P.type == Reset
 619	 *	      Drop packet and return
 620	 *
 621	 * NOTE: the check for the packet types is done in
 622	 *	 dccp_rcv_state_process
 623	 */
 624
 625	if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
 626		goto reset;
 627	if (opt_skb) {
 628		/* XXX This is where we would goto ipv6_pktoptions. */
 629		__kfree_skb(opt_skb);
 630	}
 631	return 0;
 632
 633reset:
 634	dccp_v6_ctl_send_reset(sk, skb);
 635discard:
 636	if (opt_skb != NULL)
 637		__kfree_skb(opt_skb);
 638	kfree_skb(skb);
 639	return 0;
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 640}
 641
 642static int dccp_v6_rcv(struct sk_buff *skb)
 643{
 644	const struct dccp_hdr *dh;
 
 645	struct sock *sk;
 646	int min_cov;
 647
 648	/* Step 1: Check header basics */
 649
 650	if (dccp_invalid_packet(skb))
 651		goto discard_it;
 652
 653	/* Step 1: If header checksum is incorrect, drop packet and return. */
 654	if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
 655				     &ipv6_hdr(skb)->daddr)) {
 656		DCCP_WARN("dropped packet with invalid checksum\n");
 657		goto discard_it;
 658	}
 659
 660	dh = dccp_hdr(skb);
 661
 662	DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
 663	DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
 664
 665	if (dccp_packet_without_ack(skb))
 666		DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
 667	else
 668		DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
 669
 670lookup:
 671	sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
 672			        dh->dccph_sport, dh->dccph_dport,
 673				inet6_iif(skb));
 674	if (!sk) {
 675		dccp_pr_debug("failed to look up flow ID in table and "
 676			      "get corresponding socket\n");
 677		goto no_dccp_socket;
 678	}
 679
 680	/*
 681	 * Step 2:
 682	 *	... or S.state == TIMEWAIT,
 683	 *		Generate Reset(No Connection) unless P.type == Reset
 684	 *		Drop packet and return
 685	 */
 686	if (sk->sk_state == DCCP_TIME_WAIT) {
 687		dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
 688		inet_twsk_put(inet_twsk(sk));
 689		goto no_dccp_socket;
 690	}
 691
 692	if (sk->sk_state == DCCP_NEW_SYN_RECV) {
 693		struct request_sock *req = inet_reqsk(sk);
 694		struct sock *nsk;
 695
 696		sk = req->rsk_listener;
 697		if (unlikely(sk->sk_state != DCCP_LISTEN)) {
 698			inet_csk_reqsk_queue_drop_and_put(sk, req);
 699			goto lookup;
 700		}
 701		sock_hold(sk);
 
 702		nsk = dccp_check_req(sk, skb, req);
 703		if (!nsk) {
 704			reqsk_put(req);
 705			goto discard_and_relse;
 706		}
 707		if (nsk == sk) {
 708			reqsk_put(req);
 709		} else if (dccp_child_process(sk, nsk, skb)) {
 710			dccp_v6_ctl_send_reset(sk, skb);
 711			goto discard_and_relse;
 712		} else {
 713			sock_put(sk);
 714			return 0;
 715		}
 716	}
 717	/*
 718	 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
 719	 *	o if MinCsCov = 0, only packets with CsCov = 0 are accepted
 720	 *	o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
 721	 */
 722	min_cov = dccp_sk(sk)->dccps_pcrlen;
 723	if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
 724		dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
 725			      dh->dccph_cscov, min_cov);
 726		/* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
 727		goto discard_and_relse;
 728	}
 729
 730	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
 731		goto discard_and_relse;
 732
 733	return sk_receive_skb(sk, skb, 1) ? -1 : 0;
 
 734
 735no_dccp_socket:
 736	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
 737		goto discard_it;
 738	/*
 739	 * Step 2:
 740	 *	If no socket ...
 741	 *		Generate Reset(No Connection) unless P.type == Reset
 742	 *		Drop packet and return
 743	 */
 744	if (dh->dccph_type != DCCP_PKT_RESET) {
 745		DCCP_SKB_CB(skb)->dccpd_reset_code =
 746					DCCP_RESET_CODE_NO_CONNECTION;
 747		dccp_v6_ctl_send_reset(sk, skb);
 748	}
 749
 750discard_it:
 751	kfree_skb(skb);
 752	return 0;
 753
 754discard_and_relse:
 755	sock_put(sk);
 
 756	goto discard_it;
 757}
 758
 759static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
 760			   int addr_len)
 761{
 762	struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
 763	struct inet_connection_sock *icsk = inet_csk(sk);
 764	struct inet_sock *inet = inet_sk(sk);
 765	struct ipv6_pinfo *np = inet6_sk(sk);
 766	struct dccp_sock *dp = dccp_sk(sk);
 767	struct in6_addr *saddr = NULL, *final_p, final;
 768	struct ipv6_txoptions *opt;
 769	struct flowi6 fl6;
 770	struct dst_entry *dst;
 771	int addr_type;
 772	int err;
 773
 774	dp->dccps_role = DCCP_ROLE_CLIENT;
 775
 776	if (addr_len < SIN6_LEN_RFC2133)
 777		return -EINVAL;
 778
 779	if (usin->sin6_family != AF_INET6)
 780		return -EAFNOSUPPORT;
 781
 782	memset(&fl6, 0, sizeof(fl6));
 783
 784	if (np->sndflow) {
 785		fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
 786		IP6_ECN_flow_init(fl6.flowlabel);
 787		if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
 788			struct ip6_flowlabel *flowlabel;
 789			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
 790			if (flowlabel == NULL)
 791				return -EINVAL;
 792			fl6_sock_release(flowlabel);
 793		}
 794	}
 795	/*
 796	 * connect() to INADDR_ANY means loopback (BSD'ism).
 797	 */
 798	if (ipv6_addr_any(&usin->sin6_addr))
 799		usin->sin6_addr.s6_addr[15] = 1;
 800
 801	addr_type = ipv6_addr_type(&usin->sin6_addr);
 802
 803	if (addr_type & IPV6_ADDR_MULTICAST)
 804		return -ENETUNREACH;
 805
 806	if (addr_type & IPV6_ADDR_LINKLOCAL) {
 807		if (addr_len >= sizeof(struct sockaddr_in6) &&
 808		    usin->sin6_scope_id) {
 809			/* If interface is set while binding, indices
 810			 * must coincide.
 811			 */
 812			if (sk->sk_bound_dev_if &&
 813			    sk->sk_bound_dev_if != usin->sin6_scope_id)
 814				return -EINVAL;
 815
 816			sk->sk_bound_dev_if = usin->sin6_scope_id;
 817		}
 818
 819		/* Connect to link-local address requires an interface */
 820		if (!sk->sk_bound_dev_if)
 821			return -EINVAL;
 822	}
 823
 824	sk->sk_v6_daddr = usin->sin6_addr;
 825	np->flow_label = fl6.flowlabel;
 826
 827	/*
 828	 * DCCP over IPv4
 829	 */
 830	if (addr_type == IPV6_ADDR_MAPPED) {
 831		u32 exthdrlen = icsk->icsk_ext_hdr_len;
 832		struct sockaddr_in sin;
 833
 834		SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
 835
 836		if (__ipv6_only_sock(sk))
 837			return -ENETUNREACH;
 838
 839		sin.sin_family = AF_INET;
 840		sin.sin_port = usin->sin6_port;
 841		sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
 842
 843		icsk->icsk_af_ops = &dccp_ipv6_mapped;
 844		sk->sk_backlog_rcv = dccp_v4_do_rcv;
 845
 846		err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
 847		if (err) {
 848			icsk->icsk_ext_hdr_len = exthdrlen;
 849			icsk->icsk_af_ops = &dccp_ipv6_af_ops;
 850			sk->sk_backlog_rcv = dccp_v6_do_rcv;
 851			goto failure;
 852		}
 853		np->saddr = sk->sk_v6_rcv_saddr;
 854		return err;
 855	}
 856
 857	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
 858		saddr = &sk->sk_v6_rcv_saddr;
 859
 860	fl6.flowi6_proto = IPPROTO_DCCP;
 861	fl6.daddr = sk->sk_v6_daddr;
 862	fl6.saddr = saddr ? *saddr : np->saddr;
 863	fl6.flowi6_oif = sk->sk_bound_dev_if;
 864	fl6.fl6_dport = usin->sin6_port;
 865	fl6.fl6_sport = inet->inet_sport;
 866	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
 867
 868	opt = rcu_dereference_protected(np->opt, sock_owned_by_user(sk));
 869	final_p = fl6_update_dst(&fl6, opt, &final);
 870
 871	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
 872	if (IS_ERR(dst)) {
 873		err = PTR_ERR(dst);
 874		goto failure;
 875	}
 876
 877	if (saddr == NULL) {
 878		saddr = &fl6.saddr;
 879		sk->sk_v6_rcv_saddr = *saddr;
 880	}
 881
 882	/* set the source address */
 883	np->saddr = *saddr;
 884	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
 885
 886	ip6_dst_store(sk, dst, NULL, NULL);
 887
 888	icsk->icsk_ext_hdr_len = 0;
 889	if (opt)
 890		icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
 891
 892	inet->inet_dport = usin->sin6_port;
 893
 894	dccp_set_state(sk, DCCP_REQUESTING);
 895	err = inet6_hash_connect(&dccp_death_row, sk);
 896	if (err)
 897		goto late_failure;
 898
 899	dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
 900						      sk->sk_v6_daddr.s6_addr32,
 901						      inet->inet_sport,
 902						      inet->inet_dport);
 903	err = dccp_connect(sk);
 904	if (err)
 905		goto late_failure;
 906
 907	return 0;
 908
 909late_failure:
 910	dccp_set_state(sk, DCCP_CLOSED);
 911	__sk_dst_reset(sk);
 912failure:
 913	inet->inet_dport = 0;
 914	sk->sk_route_caps = 0;
 915	return err;
 916}
 917
 918static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
 919	.queue_xmit	   = inet6_csk_xmit,
 920	.send_check	   = dccp_v6_send_check,
 921	.rebuild_header	   = inet6_sk_rebuild_header,
 922	.conn_request	   = dccp_v6_conn_request,
 923	.syn_recv_sock	   = dccp_v6_request_recv_sock,
 924	.net_header_len	   = sizeof(struct ipv6hdr),
 925	.setsockopt	   = ipv6_setsockopt,
 926	.getsockopt	   = ipv6_getsockopt,
 927	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
 928	.sockaddr_len	   = sizeof(struct sockaddr_in6),
 929	.bind_conflict	   = inet6_csk_bind_conflict,
 930#ifdef CONFIG_COMPAT
 931	.compat_setsockopt = compat_ipv6_setsockopt,
 932	.compat_getsockopt = compat_ipv6_getsockopt,
 933#endif
 934};
 935
 936/*
 937 *	DCCP over IPv4 via INET6 API
 938 */
 939static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
 940	.queue_xmit	   = ip_queue_xmit,
 941	.send_check	   = dccp_v4_send_check,
 942	.rebuild_header	   = inet_sk_rebuild_header,
 943	.conn_request	   = dccp_v6_conn_request,
 944	.syn_recv_sock	   = dccp_v6_request_recv_sock,
 945	.net_header_len	   = sizeof(struct iphdr),
 946	.setsockopt	   = ipv6_setsockopt,
 947	.getsockopt	   = ipv6_getsockopt,
 948	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
 949	.sockaddr_len	   = sizeof(struct sockaddr_in6),
 950#ifdef CONFIG_COMPAT
 951	.compat_setsockopt = compat_ipv6_setsockopt,
 952	.compat_getsockopt = compat_ipv6_getsockopt,
 953#endif
 954};
 955
 956/* NOTE: A lot of things set to zero explicitly by call to
 957 *       sk_alloc() so need not be done here.
 958 */
 959static int dccp_v6_init_sock(struct sock *sk)
 960{
 961	static __u8 dccp_v6_ctl_sock_initialized;
 962	int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
 963
 964	if (err == 0) {
 965		if (unlikely(!dccp_v6_ctl_sock_initialized))
 966			dccp_v6_ctl_sock_initialized = 1;
 967		inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
 968	}
 969
 970	return err;
 971}
 972
 973static void dccp_v6_destroy_sock(struct sock *sk)
 974{
 975	dccp_destroy_sock(sk);
 976	inet6_destroy_sock(sk);
 977}
 978
 979static struct timewait_sock_ops dccp6_timewait_sock_ops = {
 980	.twsk_obj_size	= sizeof(struct dccp6_timewait_sock),
 981};
 982
 983static struct proto dccp_v6_prot = {
 984	.name		   = "DCCPv6",
 985	.owner		   = THIS_MODULE,
 986	.close		   = dccp_close,
 987	.connect	   = dccp_v6_connect,
 988	.disconnect	   = dccp_disconnect,
 989	.ioctl		   = dccp_ioctl,
 990	.init		   = dccp_v6_init_sock,
 991	.setsockopt	   = dccp_setsockopt,
 992	.getsockopt	   = dccp_getsockopt,
 993	.sendmsg	   = dccp_sendmsg,
 994	.recvmsg	   = dccp_recvmsg,
 995	.backlog_rcv	   = dccp_v6_do_rcv,
 996	.hash		   = inet6_hash,
 997	.unhash		   = inet_unhash,
 998	.accept		   = inet_csk_accept,
 999	.get_port	   = inet_csk_get_port,
1000	.shutdown	   = dccp_shutdown,
1001	.destroy	   = dccp_v6_destroy_sock,
1002	.orphan_count	   = &dccp_orphan_count,
1003	.max_header	   = MAX_DCCP_HEADER,
1004	.obj_size	   = sizeof(struct dccp6_sock),
1005	.slab_flags	   = SLAB_DESTROY_BY_RCU,
1006	.rsk_prot	   = &dccp6_request_sock_ops,
1007	.twsk_prot	   = &dccp6_timewait_sock_ops,
1008	.h.hashinfo	   = &dccp_hashinfo,
1009#ifdef CONFIG_COMPAT
1010	.compat_setsockopt = compat_dccp_setsockopt,
1011	.compat_getsockopt = compat_dccp_getsockopt,
1012#endif
1013};
1014
1015static const struct inet6_protocol dccp_v6_protocol = {
1016	.handler	= dccp_v6_rcv,
1017	.err_handler	= dccp_v6_err,
1018	.flags		= INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1019};
1020
1021static const struct proto_ops inet6_dccp_ops = {
1022	.family		   = PF_INET6,
1023	.owner		   = THIS_MODULE,
1024	.release	   = inet6_release,
1025	.bind		   = inet6_bind,
1026	.connect	   = inet_stream_connect,
1027	.socketpair	   = sock_no_socketpair,
1028	.accept		   = inet_accept,
1029	.getname	   = inet6_getname,
1030	.poll		   = dccp_poll,
1031	.ioctl		   = inet6_ioctl,
 
1032	.listen		   = inet_dccp_listen,
1033	.shutdown	   = inet_shutdown,
1034	.setsockopt	   = sock_common_setsockopt,
1035	.getsockopt	   = sock_common_getsockopt,
1036	.sendmsg	   = inet_sendmsg,
1037	.recvmsg	   = sock_common_recvmsg,
1038	.mmap		   = sock_no_mmap,
1039	.sendpage	   = sock_no_sendpage,
1040#ifdef CONFIG_COMPAT
1041	.compat_setsockopt = compat_sock_common_setsockopt,
1042	.compat_getsockopt = compat_sock_common_getsockopt,
1043#endif
1044};
1045
1046static struct inet_protosw dccp_v6_protosw = {
1047	.type		= SOCK_DCCP,
1048	.protocol	= IPPROTO_DCCP,
1049	.prot		= &dccp_v6_prot,
1050	.ops		= &inet6_dccp_ops,
1051	.flags		= INET_PROTOSW_ICSK,
1052};
1053
1054static int __net_init dccp_v6_init_net(struct net *net)
1055{
 
 
1056	if (dccp_hashinfo.bhash == NULL)
1057		return -ESOCKTNOSUPPORT;
1058
1059	return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1060				    SOCK_DCCP, IPPROTO_DCCP, net);
1061}
1062
1063static void __net_exit dccp_v6_exit_net(struct net *net)
1064{
1065	inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
 
 
 
 
 
 
 
1066}
1067
1068static struct pernet_operations dccp_v6_ops = {
1069	.init   = dccp_v6_init_net,
1070	.exit   = dccp_v6_exit_net,
 
 
 
1071};
1072
1073static int __init dccp_v6_init(void)
1074{
1075	int err = proto_register(&dccp_v6_prot, 1);
1076
1077	if (err != 0)
1078		goto out;
1079
1080	err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1081	if (err != 0)
1082		goto out_unregister_proto;
1083
1084	inet6_register_protosw(&dccp_v6_protosw);
1085
1086	err = register_pernet_subsys(&dccp_v6_ops);
1087	if (err != 0)
1088		goto out_destroy_ctl_sock;
 
 
 
 
 
1089out:
1090	return err;
1091
 
1092out_destroy_ctl_sock:
1093	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1094	inet6_unregister_protosw(&dccp_v6_protosw);
1095out_unregister_proto:
1096	proto_unregister(&dccp_v6_prot);
1097	goto out;
1098}
1099
1100static void __exit dccp_v6_exit(void)
1101{
 
1102	unregister_pernet_subsys(&dccp_v6_ops);
1103	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1104	inet6_unregister_protosw(&dccp_v6_protosw);
1105	proto_unregister(&dccp_v6_prot);
1106}
1107
1108module_init(dccp_v6_init);
1109module_exit(dccp_v6_exit);
1110
1111/*
1112 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1113 * values directly, Also cover the case where the protocol is not specified,
1114 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1115 */
1116MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1117MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1118MODULE_LICENSE("GPL");
1119MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1120MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");