xref: /linux/net/dccp/ipv6.c (revision f3a8b6645dc2e60d11f20c1c23afd964ff4e55ae)
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 
40 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
41 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
42 
43 /* add pseudo-header to DCCP checksum stored in skb->csum */
44 static 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 
51 static 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 
60 static 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 
69 static 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(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(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(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(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(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 
178 out:
179 	bh_unlock_sock(sk);
180 	sock_put(sk);
181 }
182 
183 
184 static 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 		struct ipv6_txoptions *opt;
220 
221 		dh->dccph_checksum = dccp_v6_csum_finish(skb,
222 							 &ireq->ir_v6_loc_addr,
223 							 &ireq->ir_v6_rmt_addr);
224 		fl6.daddr = ireq->ir_v6_rmt_addr;
225 		rcu_read_lock();
226 		opt = ireq->ipv6_opt;
227 		if (!opt)
228 			opt = rcu_dereference(np->opt);
229 		err = ip6_xmit(sk, skb, &fl6, opt, np->tclass);
230 		rcu_read_unlock();
231 		err = net_xmit_eval(err);
232 	}
233 
234 done:
235 	dst_release(dst);
236 	return err;
237 }
238 
239 static void dccp_v6_reqsk_destructor(struct request_sock *req)
240 {
241 	dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
242 	kfree(inet_rsk(req)->ipv6_opt);
243 	kfree_skb(inet_rsk(req)->pktopts);
244 }
245 
246 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
247 {
248 	const struct ipv6hdr *rxip6h;
249 	struct sk_buff *skb;
250 	struct flowi6 fl6;
251 	struct net *net = dev_net(skb_dst(rxskb)->dev);
252 	struct sock *ctl_sk = net->dccp.v6_ctl_sk;
253 	struct dst_entry *dst;
254 
255 	if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
256 		return;
257 
258 	if (!ipv6_unicast_destination(rxskb))
259 		return;
260 
261 	skb = dccp_ctl_make_reset(ctl_sk, rxskb);
262 	if (skb == NULL)
263 		return;
264 
265 	rxip6h = ipv6_hdr(rxskb);
266 	dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
267 							    &rxip6h->daddr);
268 
269 	memset(&fl6, 0, sizeof(fl6));
270 	fl6.daddr = rxip6h->saddr;
271 	fl6.saddr = rxip6h->daddr;
272 
273 	fl6.flowi6_proto = IPPROTO_DCCP;
274 	fl6.flowi6_oif = inet6_iif(rxskb);
275 	fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
276 	fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
277 	security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
278 
279 	/* sk = NULL, but it is safe for now. RST socket required. */
280 	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
281 	if (!IS_ERR(dst)) {
282 		skb_dst_set(skb, dst);
283 		ip6_xmit(ctl_sk, skb, &fl6, NULL, 0);
284 		DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
285 		DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
286 		return;
287 	}
288 
289 	kfree_skb(skb);
290 }
291 
292 static struct request_sock_ops dccp6_request_sock_ops = {
293 	.family		= AF_INET6,
294 	.obj_size	= sizeof(struct dccp6_request_sock),
295 	.rtx_syn_ack	= dccp_v6_send_response,
296 	.send_ack	= dccp_reqsk_send_ack,
297 	.destructor	= dccp_v6_reqsk_destructor,
298 	.send_reset	= dccp_v6_ctl_send_reset,
299 	.syn_ack_timeout = dccp_syn_ack_timeout,
300 };
301 
302 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
303 {
304 	struct request_sock *req;
305 	struct dccp_request_sock *dreq;
306 	struct inet_request_sock *ireq;
307 	struct ipv6_pinfo *np = inet6_sk(sk);
308 	const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
309 	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
310 
311 	if (skb->protocol == htons(ETH_P_IP))
312 		return dccp_v4_conn_request(sk, skb);
313 
314 	if (!ipv6_unicast_destination(skb))
315 		return 0;	/* discard, don't send a reset here */
316 
317 	if (dccp_bad_service_code(sk, service)) {
318 		dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
319 		goto drop;
320 	}
321 	/*
322 	 * There are no SYN attacks on IPv6, yet...
323 	 */
324 	dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
325 	if (inet_csk_reqsk_queue_is_full(sk))
326 		goto drop;
327 
328 	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
329 		goto drop;
330 
331 	req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
332 	if (req == NULL)
333 		goto drop;
334 
335 	if (dccp_reqsk_init(req, dccp_sk(sk), skb))
336 		goto drop_and_free;
337 
338 	dreq = dccp_rsk(req);
339 	if (dccp_parse_options(sk, dreq, skb))
340 		goto drop_and_free;
341 
342 	if (security_inet_conn_request(sk, skb, req))
343 		goto drop_and_free;
344 
345 	ireq = inet_rsk(req);
346 	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
347 	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
348 	ireq->ireq_family = AF_INET6;
349 
350 	if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
351 	    np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
352 	    np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
353 		atomic_inc(&skb->users);
354 		ireq->pktopts = skb;
355 	}
356 	ireq->ir_iif = sk->sk_bound_dev_if;
357 
358 	/* So that link locals have meaning */
359 	if (!sk->sk_bound_dev_if &&
360 	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
361 		ireq->ir_iif = inet6_iif(skb);
362 
363 	/*
364 	 * Step 3: Process LISTEN state
365 	 *
366 	 *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
367 	 *
368 	 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
369 	 */
370 	dreq->dreq_isr	   = dcb->dccpd_seq;
371 	dreq->dreq_gsr     = dreq->dreq_isr;
372 	dreq->dreq_iss	   = dccp_v6_init_sequence(skb);
373 	dreq->dreq_gss     = dreq->dreq_iss;
374 	dreq->dreq_service = service;
375 
376 	if (dccp_v6_send_response(sk, req))
377 		goto drop_and_free;
378 
379 	inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
380 	return 0;
381 
382 drop_and_free:
383 	reqsk_free(req);
384 drop:
385 	__DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
386 	return -1;
387 }
388 
389 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
390 					      struct sk_buff *skb,
391 					      struct request_sock *req,
392 					      struct dst_entry *dst,
393 					      struct request_sock *req_unhash,
394 					      bool *own_req)
395 {
396 	struct inet_request_sock *ireq = inet_rsk(req);
397 	struct ipv6_pinfo *newnp;
398 	const struct ipv6_pinfo *np = inet6_sk(sk);
399 	struct ipv6_txoptions *opt;
400 	struct inet_sock *newinet;
401 	struct dccp6_sock *newdp6;
402 	struct sock *newsk;
403 
404 	if (skb->protocol == htons(ETH_P_IP)) {
405 		/*
406 		 *	v6 mapped
407 		 */
408 		newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
409 						  req_unhash, own_req);
410 		if (newsk == NULL)
411 			return NULL;
412 
413 		newdp6 = (struct dccp6_sock *)newsk;
414 		newinet = inet_sk(newsk);
415 		newinet->pinet6 = &newdp6->inet6;
416 		newnp = inet6_sk(newsk);
417 
418 		memcpy(newnp, np, sizeof(struct ipv6_pinfo));
419 
420 		newnp->saddr = newsk->sk_v6_rcv_saddr;
421 
422 		inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
423 		newsk->sk_backlog_rcv = dccp_v4_do_rcv;
424 		newnp->pktoptions  = NULL;
425 		newnp->opt	   = NULL;
426 		newnp->mcast_oif   = inet6_iif(skb);
427 		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
428 
429 		/*
430 		 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
431 		 * here, dccp_create_openreq_child now does this for us, see the comment in
432 		 * that function for the gory details. -acme
433 		 */
434 
435 		/* It is tricky place. Until this moment IPv4 tcp
436 		   worked with IPv6 icsk.icsk_af_ops.
437 		   Sync it now.
438 		 */
439 		dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
440 
441 		return newsk;
442 	}
443 
444 
445 	if (sk_acceptq_is_full(sk))
446 		goto out_overflow;
447 
448 	if (!dst) {
449 		struct flowi6 fl6;
450 
451 		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
452 		if (!dst)
453 			goto out;
454 	}
455 
456 	newsk = dccp_create_openreq_child(sk, req, skb);
457 	if (newsk == NULL)
458 		goto out_nonewsk;
459 
460 	/*
461 	 * No need to charge this sock to the relevant IPv6 refcnt debug socks
462 	 * count here, dccp_create_openreq_child now does this for us, see the
463 	 * comment in that function for the gory details. -acme
464 	 */
465 
466 	ip6_dst_store(newsk, dst, NULL, NULL);
467 	newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
468 						      NETIF_F_TSO);
469 	newdp6 = (struct dccp6_sock *)newsk;
470 	newinet = inet_sk(newsk);
471 	newinet->pinet6 = &newdp6->inet6;
472 	newnp = inet6_sk(newsk);
473 
474 	memcpy(newnp, np, sizeof(struct ipv6_pinfo));
475 
476 	newsk->sk_v6_daddr	= ireq->ir_v6_rmt_addr;
477 	newnp->saddr		= ireq->ir_v6_loc_addr;
478 	newsk->sk_v6_rcv_saddr	= ireq->ir_v6_loc_addr;
479 	newsk->sk_bound_dev_if	= ireq->ir_iif;
480 
481 	/* Now IPv6 options...
482 
483 	   First: no IPv4 options.
484 	 */
485 	newinet->inet_opt = NULL;
486 
487 	/* Clone RX bits */
488 	newnp->rxopt.all = np->rxopt.all;
489 
490 	newnp->pktoptions = NULL;
491 	newnp->opt	  = NULL;
492 	newnp->mcast_oif  = inet6_iif(skb);
493 	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
494 
495 	/*
496 	 * Clone native IPv6 options from listening socket (if any)
497 	 *
498 	 * Yes, keeping reference count would be much more clever, but we make
499 	 * one more one thing there: reattach optmem to newsk.
500 	 */
501 	opt = ireq->ipv6_opt;
502 	if (!opt)
503 		opt = rcu_dereference(np->opt);
504 	if (opt) {
505 		opt = ipv6_dup_options(newsk, opt);
506 		RCU_INIT_POINTER(newnp->opt, opt);
507 	}
508 	inet_csk(newsk)->icsk_ext_hdr_len = 0;
509 	if (opt)
510 		inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
511 						    opt->opt_flen;
512 
513 	dccp_sync_mss(newsk, dst_mtu(dst));
514 
515 	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
516 	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
517 
518 	if (__inet_inherit_port(sk, newsk) < 0) {
519 		inet_csk_prepare_forced_close(newsk);
520 		dccp_done(newsk);
521 		goto out;
522 	}
523 	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
524 	/* Clone pktoptions received with SYN, if we own the req */
525 	if (*own_req && ireq->pktopts) {
526 		newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
527 		consume_skb(ireq->pktopts);
528 		ireq->pktopts = NULL;
529 		if (newnp->pktoptions)
530 			skb_set_owner_r(newnp->pktoptions, newsk);
531 	}
532 
533 	return newsk;
534 
535 out_overflow:
536 	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
537 out_nonewsk:
538 	dst_release(dst);
539 out:
540 	__NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
541 	return NULL;
542 }
543 
544 /* The socket must have it's spinlock held when we get
545  * here.
546  *
547  * We have a potential double-lock case here, so even when
548  * doing backlog processing we use the BH locking scheme.
549  * This is because we cannot sleep with the original spinlock
550  * held.
551  */
552 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
553 {
554 	struct ipv6_pinfo *np = inet6_sk(sk);
555 	struct sk_buff *opt_skb = NULL;
556 
557 	/* Imagine: socket is IPv6. IPv4 packet arrives,
558 	   goes to IPv4 receive handler and backlogged.
559 	   From backlog it always goes here. Kerboom...
560 	   Fortunately, dccp_rcv_established and rcv_established
561 	   handle them correctly, but it is not case with
562 	   dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
563 	 */
564 
565 	if (skb->protocol == htons(ETH_P_IP))
566 		return dccp_v4_do_rcv(sk, skb);
567 
568 	if (sk_filter(sk, skb))
569 		goto discard;
570 
571 	/*
572 	 * socket locking is here for SMP purposes as backlog rcv is currently
573 	 * called with bh processing disabled.
574 	 */
575 
576 	/* Do Stevens' IPV6_PKTOPTIONS.
577 
578 	   Yes, guys, it is the only place in our code, where we
579 	   may make it not affecting IPv4.
580 	   The rest of code is protocol independent,
581 	   and I do not like idea to uglify IPv4.
582 
583 	   Actually, all the idea behind IPV6_PKTOPTIONS
584 	   looks not very well thought. For now we latch
585 	   options, received in the last packet, enqueued
586 	   by tcp. Feel free to propose better solution.
587 					       --ANK (980728)
588 	 */
589 	if (np->rxopt.all)
590 	/*
591 	 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
592 	 *        (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
593 	 */
594 		opt_skb = skb_clone(skb, GFP_ATOMIC);
595 
596 	if (sk->sk_state == DCCP_OPEN) { /* Fast path */
597 		if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
598 			goto reset;
599 		if (opt_skb) {
600 			/* XXX This is where we would goto ipv6_pktoptions. */
601 			__kfree_skb(opt_skb);
602 		}
603 		return 0;
604 	}
605 
606 	/*
607 	 *  Step 3: Process LISTEN state
608 	 *     If S.state == LISTEN,
609 	 *	 If P.type == Request or P contains a valid Init Cookie option,
610 	 *	      (* Must scan the packet's options to check for Init
611 	 *		 Cookies.  Only Init Cookies are processed here,
612 	 *		 however; other options are processed in Step 8.  This
613 	 *		 scan need only be performed if the endpoint uses Init
614 	 *		 Cookies *)
615 	 *	      (* Generate a new socket and switch to that socket *)
616 	 *	      Set S := new socket for this port pair
617 	 *	      S.state = RESPOND
618 	 *	      Choose S.ISS (initial seqno) or set from Init Cookies
619 	 *	      Initialize S.GAR := S.ISS
620 	 *	      Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
621 	 *	      Continue with S.state == RESPOND
622 	 *	      (* A Response packet will be generated in Step 11 *)
623 	 *	 Otherwise,
624 	 *	      Generate Reset(No Connection) unless P.type == Reset
625 	 *	      Drop packet and return
626 	 *
627 	 * NOTE: the check for the packet types is done in
628 	 *	 dccp_rcv_state_process
629 	 */
630 
631 	if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
632 		goto reset;
633 	if (opt_skb) {
634 		/* XXX This is where we would goto ipv6_pktoptions. */
635 		__kfree_skb(opt_skb);
636 	}
637 	return 0;
638 
639 reset:
640 	dccp_v6_ctl_send_reset(sk, skb);
641 discard:
642 	if (opt_skb != NULL)
643 		__kfree_skb(opt_skb);
644 	kfree_skb(skb);
645 	return 0;
646 }
647 
648 static int dccp_v6_rcv(struct sk_buff *skb)
649 {
650 	const struct dccp_hdr *dh;
651 	bool refcounted;
652 	struct sock *sk;
653 	int min_cov;
654 
655 	/* Step 1: Check header basics */
656 
657 	if (dccp_invalid_packet(skb))
658 		goto discard_it;
659 
660 	/* Step 1: If header checksum is incorrect, drop packet and return. */
661 	if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
662 				     &ipv6_hdr(skb)->daddr)) {
663 		DCCP_WARN("dropped packet with invalid checksum\n");
664 		goto discard_it;
665 	}
666 
667 	dh = dccp_hdr(skb);
668 
669 	DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
670 	DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
671 
672 	if (dccp_packet_without_ack(skb))
673 		DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
674 	else
675 		DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
676 
677 lookup:
678 	sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
679 			        dh->dccph_sport, dh->dccph_dport,
680 				inet6_iif(skb), &refcounted);
681 	if (!sk) {
682 		dccp_pr_debug("failed to look up flow ID in table and "
683 			      "get corresponding socket\n");
684 		goto no_dccp_socket;
685 	}
686 
687 	/*
688 	 * Step 2:
689 	 *	... or S.state == TIMEWAIT,
690 	 *		Generate Reset(No Connection) unless P.type == Reset
691 	 *		Drop packet and return
692 	 */
693 	if (sk->sk_state == DCCP_TIME_WAIT) {
694 		dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
695 		inet_twsk_put(inet_twsk(sk));
696 		goto no_dccp_socket;
697 	}
698 
699 	if (sk->sk_state == DCCP_NEW_SYN_RECV) {
700 		struct request_sock *req = inet_reqsk(sk);
701 		struct sock *nsk;
702 
703 		sk = req->rsk_listener;
704 		if (unlikely(sk->sk_state != DCCP_LISTEN)) {
705 			inet_csk_reqsk_queue_drop_and_put(sk, req);
706 			goto lookup;
707 		}
708 		sock_hold(sk);
709 		refcounted = true;
710 		nsk = dccp_check_req(sk, skb, req);
711 		if (!nsk) {
712 			reqsk_put(req);
713 			goto discard_and_relse;
714 		}
715 		if (nsk == sk) {
716 			reqsk_put(req);
717 		} else if (dccp_child_process(sk, nsk, skb)) {
718 			dccp_v6_ctl_send_reset(sk, skb);
719 			goto discard_and_relse;
720 		} else {
721 			sock_put(sk);
722 			return 0;
723 		}
724 	}
725 	/*
726 	 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
727 	 *	o if MinCsCov = 0, only packets with CsCov = 0 are accepted
728 	 *	o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
729 	 */
730 	min_cov = dccp_sk(sk)->dccps_pcrlen;
731 	if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
732 		dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
733 			      dh->dccph_cscov, min_cov);
734 		/* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
735 		goto discard_and_relse;
736 	}
737 
738 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
739 		goto discard_and_relse;
740 
741 	return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4) ? -1 : 0;
742 
743 no_dccp_socket:
744 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
745 		goto discard_it;
746 	/*
747 	 * Step 2:
748 	 *	If no socket ...
749 	 *		Generate Reset(No Connection) unless P.type == Reset
750 	 *		Drop packet and return
751 	 */
752 	if (dh->dccph_type != DCCP_PKT_RESET) {
753 		DCCP_SKB_CB(skb)->dccpd_reset_code =
754 					DCCP_RESET_CODE_NO_CONNECTION;
755 		dccp_v6_ctl_send_reset(sk, skb);
756 	}
757 
758 discard_it:
759 	kfree_skb(skb);
760 	return 0;
761 
762 discard_and_relse:
763 	if (refcounted)
764 		sock_put(sk);
765 	goto discard_it;
766 }
767 
768 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
769 			   int addr_len)
770 {
771 	struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
772 	struct inet_connection_sock *icsk = inet_csk(sk);
773 	struct inet_sock *inet = inet_sk(sk);
774 	struct ipv6_pinfo *np = inet6_sk(sk);
775 	struct dccp_sock *dp = dccp_sk(sk);
776 	struct in6_addr *saddr = NULL, *final_p, final;
777 	struct ipv6_txoptions *opt;
778 	struct flowi6 fl6;
779 	struct dst_entry *dst;
780 	int addr_type;
781 	int err;
782 
783 	dp->dccps_role = DCCP_ROLE_CLIENT;
784 
785 	if (addr_len < SIN6_LEN_RFC2133)
786 		return -EINVAL;
787 
788 	if (usin->sin6_family != AF_INET6)
789 		return -EAFNOSUPPORT;
790 
791 	memset(&fl6, 0, sizeof(fl6));
792 
793 	if (np->sndflow) {
794 		fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
795 		IP6_ECN_flow_init(fl6.flowlabel);
796 		if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
797 			struct ip6_flowlabel *flowlabel;
798 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
799 			if (flowlabel == NULL)
800 				return -EINVAL;
801 			fl6_sock_release(flowlabel);
802 		}
803 	}
804 	/*
805 	 * connect() to INADDR_ANY means loopback (BSD'ism).
806 	 */
807 	if (ipv6_addr_any(&usin->sin6_addr))
808 		usin->sin6_addr.s6_addr[15] = 1;
809 
810 	addr_type = ipv6_addr_type(&usin->sin6_addr);
811 
812 	if (addr_type & IPV6_ADDR_MULTICAST)
813 		return -ENETUNREACH;
814 
815 	if (addr_type & IPV6_ADDR_LINKLOCAL) {
816 		if (addr_len >= sizeof(struct sockaddr_in6) &&
817 		    usin->sin6_scope_id) {
818 			/* If interface is set while binding, indices
819 			 * must coincide.
820 			 */
821 			if (sk->sk_bound_dev_if &&
822 			    sk->sk_bound_dev_if != usin->sin6_scope_id)
823 				return -EINVAL;
824 
825 			sk->sk_bound_dev_if = usin->sin6_scope_id;
826 		}
827 
828 		/* Connect to link-local address requires an interface */
829 		if (!sk->sk_bound_dev_if)
830 			return -EINVAL;
831 	}
832 
833 	sk->sk_v6_daddr = usin->sin6_addr;
834 	np->flow_label = fl6.flowlabel;
835 
836 	/*
837 	 * DCCP over IPv4
838 	 */
839 	if (addr_type == IPV6_ADDR_MAPPED) {
840 		u32 exthdrlen = icsk->icsk_ext_hdr_len;
841 		struct sockaddr_in sin;
842 
843 		SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
844 
845 		if (__ipv6_only_sock(sk))
846 			return -ENETUNREACH;
847 
848 		sin.sin_family = AF_INET;
849 		sin.sin_port = usin->sin6_port;
850 		sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
851 
852 		icsk->icsk_af_ops = &dccp_ipv6_mapped;
853 		sk->sk_backlog_rcv = dccp_v4_do_rcv;
854 
855 		err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
856 		if (err) {
857 			icsk->icsk_ext_hdr_len = exthdrlen;
858 			icsk->icsk_af_ops = &dccp_ipv6_af_ops;
859 			sk->sk_backlog_rcv = dccp_v6_do_rcv;
860 			goto failure;
861 		}
862 		np->saddr = sk->sk_v6_rcv_saddr;
863 		return err;
864 	}
865 
866 	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
867 		saddr = &sk->sk_v6_rcv_saddr;
868 
869 	fl6.flowi6_proto = IPPROTO_DCCP;
870 	fl6.daddr = sk->sk_v6_daddr;
871 	fl6.saddr = saddr ? *saddr : np->saddr;
872 	fl6.flowi6_oif = sk->sk_bound_dev_if;
873 	fl6.fl6_dport = usin->sin6_port;
874 	fl6.fl6_sport = inet->inet_sport;
875 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
876 
877 	opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
878 	final_p = fl6_update_dst(&fl6, opt, &final);
879 
880 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
881 	if (IS_ERR(dst)) {
882 		err = PTR_ERR(dst);
883 		goto failure;
884 	}
885 
886 	if (saddr == NULL) {
887 		saddr = &fl6.saddr;
888 		sk->sk_v6_rcv_saddr = *saddr;
889 	}
890 
891 	/* set the source address */
892 	np->saddr = *saddr;
893 	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
894 
895 	ip6_dst_store(sk, dst, NULL, NULL);
896 
897 	icsk->icsk_ext_hdr_len = 0;
898 	if (opt)
899 		icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
900 
901 	inet->inet_dport = usin->sin6_port;
902 
903 	dccp_set_state(sk, DCCP_REQUESTING);
904 	err = inet6_hash_connect(&dccp_death_row, sk);
905 	if (err)
906 		goto late_failure;
907 
908 	dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
909 						      sk->sk_v6_daddr.s6_addr32,
910 						      inet->inet_sport,
911 						      inet->inet_dport);
912 	err = dccp_connect(sk);
913 	if (err)
914 		goto late_failure;
915 
916 	return 0;
917 
918 late_failure:
919 	dccp_set_state(sk, DCCP_CLOSED);
920 	__sk_dst_reset(sk);
921 failure:
922 	inet->inet_dport = 0;
923 	sk->sk_route_caps = 0;
924 	return err;
925 }
926 
927 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
928 	.queue_xmit	   = inet6_csk_xmit,
929 	.send_check	   = dccp_v6_send_check,
930 	.rebuild_header	   = inet6_sk_rebuild_header,
931 	.conn_request	   = dccp_v6_conn_request,
932 	.syn_recv_sock	   = dccp_v6_request_recv_sock,
933 	.net_header_len	   = sizeof(struct ipv6hdr),
934 	.setsockopt	   = ipv6_setsockopt,
935 	.getsockopt	   = ipv6_getsockopt,
936 	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
937 	.sockaddr_len	   = sizeof(struct sockaddr_in6),
938 	.bind_conflict	   = inet6_csk_bind_conflict,
939 #ifdef CONFIG_COMPAT
940 	.compat_setsockopt = compat_ipv6_setsockopt,
941 	.compat_getsockopt = compat_ipv6_getsockopt,
942 #endif
943 };
944 
945 /*
946  *	DCCP over IPv4 via INET6 API
947  */
948 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
949 	.queue_xmit	   = ip_queue_xmit,
950 	.send_check	   = dccp_v4_send_check,
951 	.rebuild_header	   = inet_sk_rebuild_header,
952 	.conn_request	   = dccp_v6_conn_request,
953 	.syn_recv_sock	   = dccp_v6_request_recv_sock,
954 	.net_header_len	   = sizeof(struct iphdr),
955 	.setsockopt	   = ipv6_setsockopt,
956 	.getsockopt	   = ipv6_getsockopt,
957 	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
958 	.sockaddr_len	   = sizeof(struct sockaddr_in6),
959 #ifdef CONFIG_COMPAT
960 	.compat_setsockopt = compat_ipv6_setsockopt,
961 	.compat_getsockopt = compat_ipv6_getsockopt,
962 #endif
963 };
964 
965 /* NOTE: A lot of things set to zero explicitly by call to
966  *       sk_alloc() so need not be done here.
967  */
968 static int dccp_v6_init_sock(struct sock *sk)
969 {
970 	static __u8 dccp_v6_ctl_sock_initialized;
971 	int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
972 
973 	if (err == 0) {
974 		if (unlikely(!dccp_v6_ctl_sock_initialized))
975 			dccp_v6_ctl_sock_initialized = 1;
976 		inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
977 	}
978 
979 	return err;
980 }
981 
982 static void dccp_v6_destroy_sock(struct sock *sk)
983 {
984 	dccp_destroy_sock(sk);
985 	inet6_destroy_sock(sk);
986 }
987 
988 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
989 	.twsk_obj_size	= sizeof(struct dccp6_timewait_sock),
990 };
991 
992 static struct proto dccp_v6_prot = {
993 	.name		   = "DCCPv6",
994 	.owner		   = THIS_MODULE,
995 	.close		   = dccp_close,
996 	.connect	   = dccp_v6_connect,
997 	.disconnect	   = dccp_disconnect,
998 	.ioctl		   = dccp_ioctl,
999 	.init		   = dccp_v6_init_sock,
1000 	.setsockopt	   = dccp_setsockopt,
1001 	.getsockopt	   = dccp_getsockopt,
1002 	.sendmsg	   = dccp_sendmsg,
1003 	.recvmsg	   = dccp_recvmsg,
1004 	.backlog_rcv	   = dccp_v6_do_rcv,
1005 	.hash		   = inet6_hash,
1006 	.unhash		   = inet_unhash,
1007 	.accept		   = inet_csk_accept,
1008 	.get_port	   = inet_csk_get_port,
1009 	.shutdown	   = dccp_shutdown,
1010 	.destroy	   = dccp_v6_destroy_sock,
1011 	.orphan_count	   = &dccp_orphan_count,
1012 	.max_header	   = MAX_DCCP_HEADER,
1013 	.obj_size	   = sizeof(struct dccp6_sock),
1014 	.slab_flags	   = SLAB_DESTROY_BY_RCU,
1015 	.rsk_prot	   = &dccp6_request_sock_ops,
1016 	.twsk_prot	   = &dccp6_timewait_sock_ops,
1017 	.h.hashinfo	   = &dccp_hashinfo,
1018 #ifdef CONFIG_COMPAT
1019 	.compat_setsockopt = compat_dccp_setsockopt,
1020 	.compat_getsockopt = compat_dccp_getsockopt,
1021 #endif
1022 };
1023 
1024 static const struct inet6_protocol dccp_v6_protocol = {
1025 	.handler	= dccp_v6_rcv,
1026 	.err_handler	= dccp_v6_err,
1027 	.flags		= INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1028 };
1029 
1030 static const struct proto_ops inet6_dccp_ops = {
1031 	.family		   = PF_INET6,
1032 	.owner		   = THIS_MODULE,
1033 	.release	   = inet6_release,
1034 	.bind		   = inet6_bind,
1035 	.connect	   = inet_stream_connect,
1036 	.socketpair	   = sock_no_socketpair,
1037 	.accept		   = inet_accept,
1038 	.getname	   = inet6_getname,
1039 	.poll		   = dccp_poll,
1040 	.ioctl		   = inet6_ioctl,
1041 	.listen		   = inet_dccp_listen,
1042 	.shutdown	   = inet_shutdown,
1043 	.setsockopt	   = sock_common_setsockopt,
1044 	.getsockopt	   = sock_common_getsockopt,
1045 	.sendmsg	   = inet_sendmsg,
1046 	.recvmsg	   = sock_common_recvmsg,
1047 	.mmap		   = sock_no_mmap,
1048 	.sendpage	   = sock_no_sendpage,
1049 #ifdef CONFIG_COMPAT
1050 	.compat_setsockopt = compat_sock_common_setsockopt,
1051 	.compat_getsockopt = compat_sock_common_getsockopt,
1052 #endif
1053 };
1054 
1055 static struct inet_protosw dccp_v6_protosw = {
1056 	.type		= SOCK_DCCP,
1057 	.protocol	= IPPROTO_DCCP,
1058 	.prot		= &dccp_v6_prot,
1059 	.ops		= &inet6_dccp_ops,
1060 	.flags		= INET_PROTOSW_ICSK,
1061 };
1062 
1063 static int __net_init dccp_v6_init_net(struct net *net)
1064 {
1065 	if (dccp_hashinfo.bhash == NULL)
1066 		return -ESOCKTNOSUPPORT;
1067 
1068 	return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1069 				    SOCK_DCCP, IPPROTO_DCCP, net);
1070 }
1071 
1072 static void __net_exit dccp_v6_exit_net(struct net *net)
1073 {
1074 	inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1075 }
1076 
1077 static struct pernet_operations dccp_v6_ops = {
1078 	.init   = dccp_v6_init_net,
1079 	.exit   = dccp_v6_exit_net,
1080 };
1081 
1082 static int __init dccp_v6_init(void)
1083 {
1084 	int err = proto_register(&dccp_v6_prot, 1);
1085 
1086 	if (err != 0)
1087 		goto out;
1088 
1089 	err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1090 	if (err != 0)
1091 		goto out_unregister_proto;
1092 
1093 	inet6_register_protosw(&dccp_v6_protosw);
1094 
1095 	err = register_pernet_subsys(&dccp_v6_ops);
1096 	if (err != 0)
1097 		goto out_destroy_ctl_sock;
1098 out:
1099 	return err;
1100 
1101 out_destroy_ctl_sock:
1102 	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1103 	inet6_unregister_protosw(&dccp_v6_protosw);
1104 out_unregister_proto:
1105 	proto_unregister(&dccp_v6_prot);
1106 	goto out;
1107 }
1108 
1109 static void __exit dccp_v6_exit(void)
1110 {
1111 	unregister_pernet_subsys(&dccp_v6_ops);
1112 	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1113 	inet6_unregister_protosw(&dccp_v6_protosw);
1114 	proto_unregister(&dccp_v6_prot);
1115 }
1116 
1117 module_init(dccp_v6_init);
1118 module_exit(dccp_v6_exit);
1119 
1120 /*
1121  * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1122  * values directly, Also cover the case where the protocol is not specified,
1123  * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1124  */
1125 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1126 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1127 MODULE_LICENSE("GPL");
1128 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1129 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
1130