xref: /openbmc/linux/net/dccp/ipv6.c (revision a8da474e)
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_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 
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 	final_p = fl6_update_dst(&fl6, np->opt, &final);
206 
207 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
208 	if (IS_ERR(dst)) {
209 		err = PTR_ERR(dst);
210 		dst = NULL;
211 		goto done;
212 	}
213 
214 	skb = dccp_make_response(sk, dst, req);
215 	if (skb != NULL) {
216 		struct dccp_hdr *dh = dccp_hdr(skb);
217 
218 		dh->dccph_checksum = dccp_v6_csum_finish(skb,
219 							 &ireq->ir_v6_loc_addr,
220 							 &ireq->ir_v6_rmt_addr);
221 		fl6.daddr = ireq->ir_v6_rmt_addr;
222 		err = ip6_xmit(sk, skb, &fl6, np->opt, np->tclass);
223 		err = net_xmit_eval(err);
224 	}
225 
226 done:
227 	dst_release(dst);
228 	return err;
229 }
230 
231 static void dccp_v6_reqsk_destructor(struct request_sock *req)
232 {
233 	dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
234 	kfree_skb(inet_rsk(req)->pktopts);
235 }
236 
237 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
238 {
239 	const struct ipv6hdr *rxip6h;
240 	struct sk_buff *skb;
241 	struct flowi6 fl6;
242 	struct net *net = dev_net(skb_dst(rxskb)->dev);
243 	struct sock *ctl_sk = net->dccp.v6_ctl_sk;
244 	struct dst_entry *dst;
245 
246 	if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
247 		return;
248 
249 	if (!ipv6_unicast_destination(rxskb))
250 		return;
251 
252 	skb = dccp_ctl_make_reset(ctl_sk, rxskb);
253 	if (skb == NULL)
254 		return;
255 
256 	rxip6h = ipv6_hdr(rxskb);
257 	dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
258 							    &rxip6h->daddr);
259 
260 	memset(&fl6, 0, sizeof(fl6));
261 	fl6.daddr = rxip6h->saddr;
262 	fl6.saddr = rxip6h->daddr;
263 
264 	fl6.flowi6_proto = IPPROTO_DCCP;
265 	fl6.flowi6_oif = inet6_iif(rxskb);
266 	fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
267 	fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
268 	security_skb_classify_flow(rxskb, flowi6_to_flowi(&fl6));
269 
270 	/* sk = NULL, but it is safe for now. RST socket required. */
271 	dst = ip6_dst_lookup_flow(ctl_sk, &fl6, NULL);
272 	if (!IS_ERR(dst)) {
273 		skb_dst_set(skb, dst);
274 		ip6_xmit(ctl_sk, skb, &fl6, NULL, 0);
275 		DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
276 		DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
277 		return;
278 	}
279 
280 	kfree_skb(skb);
281 }
282 
283 static struct request_sock_ops dccp6_request_sock_ops = {
284 	.family		= AF_INET6,
285 	.obj_size	= sizeof(struct dccp6_request_sock),
286 	.rtx_syn_ack	= dccp_v6_send_response,
287 	.send_ack	= dccp_reqsk_send_ack,
288 	.destructor	= dccp_v6_reqsk_destructor,
289 	.send_reset	= dccp_v6_ctl_send_reset,
290 	.syn_ack_timeout = dccp_syn_ack_timeout,
291 };
292 
293 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
294 {
295 	struct request_sock *req;
296 	struct dccp_request_sock *dreq;
297 	struct inet_request_sock *ireq;
298 	struct ipv6_pinfo *np = inet6_sk(sk);
299 	const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
300 	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
301 
302 	if (skb->protocol == htons(ETH_P_IP))
303 		return dccp_v4_conn_request(sk, skb);
304 
305 	if (!ipv6_unicast_destination(skb))
306 		return 0;	/* discard, don't send a reset here */
307 
308 	if (dccp_bad_service_code(sk, service)) {
309 		dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
310 		goto drop;
311 	}
312 	/*
313 	 * There are no SYN attacks on IPv6, yet...
314 	 */
315 	dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
316 	if (inet_csk_reqsk_queue_is_full(sk))
317 		goto drop;
318 
319 	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
320 		goto drop;
321 
322 	req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
323 	if (req == NULL)
324 		goto drop;
325 
326 	if (dccp_reqsk_init(req, dccp_sk(sk), skb))
327 		goto drop_and_free;
328 
329 	dreq = dccp_rsk(req);
330 	if (dccp_parse_options(sk, dreq, skb))
331 		goto drop_and_free;
332 
333 	if (security_inet_conn_request(sk, skb, req))
334 		goto drop_and_free;
335 
336 	ireq = inet_rsk(req);
337 	ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
338 	ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
339 	ireq->ireq_family = AF_INET6;
340 
341 	if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
342 	    np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
343 	    np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
344 		atomic_inc(&skb->users);
345 		ireq->pktopts = skb;
346 	}
347 	ireq->ir_iif = sk->sk_bound_dev_if;
348 
349 	/* So that link locals have meaning */
350 	if (!sk->sk_bound_dev_if &&
351 	    ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
352 		ireq->ir_iif = inet6_iif(skb);
353 
354 	/*
355 	 * Step 3: Process LISTEN state
356 	 *
357 	 *   Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
358 	 *
359 	 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
360 	 */
361 	dreq->dreq_isr	   = dcb->dccpd_seq;
362 	dreq->dreq_gsr     = dreq->dreq_isr;
363 	dreq->dreq_iss	   = dccp_v6_init_sequence(skb);
364 	dreq->dreq_gss     = dreq->dreq_iss;
365 	dreq->dreq_service = service;
366 
367 	if (dccp_v6_send_response(sk, req))
368 		goto drop_and_free;
369 
370 	inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
371 	return 0;
372 
373 drop_and_free:
374 	reqsk_free(req);
375 drop:
376 	DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
377 	return -1;
378 }
379 
380 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
381 					      struct sk_buff *skb,
382 					      struct request_sock *req,
383 					      struct dst_entry *dst,
384 					      struct request_sock *req_unhash,
385 					      bool *own_req)
386 {
387 	struct inet_request_sock *ireq = inet_rsk(req);
388 	struct ipv6_pinfo *newnp;
389 	const struct ipv6_pinfo *np = inet6_sk(sk);
390 	struct inet_sock *newinet;
391 	struct dccp6_sock *newdp6;
392 	struct sock *newsk;
393 
394 	if (skb->protocol == htons(ETH_P_IP)) {
395 		/*
396 		 *	v6 mapped
397 		 */
398 		newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
399 						  req_unhash, own_req);
400 		if (newsk == NULL)
401 			return NULL;
402 
403 		newdp6 = (struct dccp6_sock *)newsk;
404 		newinet = inet_sk(newsk);
405 		newinet->pinet6 = &newdp6->inet6;
406 		newnp = inet6_sk(newsk);
407 
408 		memcpy(newnp, np, sizeof(struct ipv6_pinfo));
409 
410 		newnp->saddr = newsk->sk_v6_rcv_saddr;
411 
412 		inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
413 		newsk->sk_backlog_rcv = dccp_v4_do_rcv;
414 		newnp->pktoptions  = NULL;
415 		newnp->opt	   = NULL;
416 		newnp->mcast_oif   = inet6_iif(skb);
417 		newnp->mcast_hops  = ipv6_hdr(skb)->hop_limit;
418 
419 		/*
420 		 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
421 		 * here, dccp_create_openreq_child now does this for us, see the comment in
422 		 * that function for the gory details. -acme
423 		 */
424 
425 		/* It is tricky place. Until this moment IPv4 tcp
426 		   worked with IPv6 icsk.icsk_af_ops.
427 		   Sync it now.
428 		 */
429 		dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
430 
431 		return newsk;
432 	}
433 
434 
435 	if (sk_acceptq_is_full(sk))
436 		goto out_overflow;
437 
438 	if (!dst) {
439 		struct flowi6 fl6;
440 
441 		dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
442 		if (!dst)
443 			goto out;
444 	}
445 
446 	newsk = dccp_create_openreq_child(sk, req, skb);
447 	if (newsk == NULL)
448 		goto out_nonewsk;
449 
450 	/*
451 	 * No need to charge this sock to the relevant IPv6 refcnt debug socks
452 	 * count here, dccp_create_openreq_child now does this for us, see the
453 	 * comment in that function for the gory details. -acme
454 	 */
455 
456 	__ip6_dst_store(newsk, dst, NULL, NULL);
457 	newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
458 						      NETIF_F_TSO);
459 	newdp6 = (struct dccp6_sock *)newsk;
460 	newinet = inet_sk(newsk);
461 	newinet->pinet6 = &newdp6->inet6;
462 	newnp = inet6_sk(newsk);
463 
464 	memcpy(newnp, np, sizeof(struct ipv6_pinfo));
465 
466 	newsk->sk_v6_daddr	= ireq->ir_v6_rmt_addr;
467 	newnp->saddr		= ireq->ir_v6_loc_addr;
468 	newsk->sk_v6_rcv_saddr	= ireq->ir_v6_loc_addr;
469 	newsk->sk_bound_dev_if	= ireq->ir_iif;
470 
471 	/* Now IPv6 options...
472 
473 	   First: no IPv4 options.
474 	 */
475 	newinet->inet_opt = NULL;
476 
477 	/* Clone RX bits */
478 	newnp->rxopt.all = np->rxopt.all;
479 
480 	newnp->pktoptions = NULL;
481 	newnp->opt	  = NULL;
482 	newnp->mcast_oif  = inet6_iif(skb);
483 	newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
484 
485 	/*
486 	 * Clone native IPv6 options from listening socket (if any)
487 	 *
488 	 * Yes, keeping reference count would be much more clever, but we make
489 	 * one more one thing there: reattach optmem to newsk.
490 	 */
491 	if (np->opt != NULL)
492 		newnp->opt = ipv6_dup_options(newsk, np->opt);
493 
494 	inet_csk(newsk)->icsk_ext_hdr_len = 0;
495 	if (newnp->opt != NULL)
496 		inet_csk(newsk)->icsk_ext_hdr_len = (newnp->opt->opt_nflen +
497 						     newnp->opt->opt_flen);
498 
499 	dccp_sync_mss(newsk, dst_mtu(dst));
500 
501 	newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
502 	newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
503 
504 	if (__inet_inherit_port(sk, newsk) < 0) {
505 		inet_csk_prepare_forced_close(newsk);
506 		dccp_done(newsk);
507 		goto out;
508 	}
509 	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
510 	/* Clone pktoptions received with SYN, if we own the req */
511 	if (*own_req && ireq->pktopts) {
512 		newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
513 		consume_skb(ireq->pktopts);
514 		ireq->pktopts = NULL;
515 		if (newnp->pktoptions)
516 			skb_set_owner_r(newnp->pktoptions, newsk);
517 	}
518 
519 	return newsk;
520 
521 out_overflow:
522 	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
523 out_nonewsk:
524 	dst_release(dst);
525 out:
526 	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
527 	return NULL;
528 }
529 
530 /* The socket must have it's spinlock held when we get
531  * here.
532  *
533  * We have a potential double-lock case here, so even when
534  * doing backlog processing we use the BH locking scheme.
535  * This is because we cannot sleep with the original spinlock
536  * held.
537  */
538 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
539 {
540 	struct ipv6_pinfo *np = inet6_sk(sk);
541 	struct sk_buff *opt_skb = NULL;
542 
543 	/* Imagine: socket is IPv6. IPv4 packet arrives,
544 	   goes to IPv4 receive handler and backlogged.
545 	   From backlog it always goes here. Kerboom...
546 	   Fortunately, dccp_rcv_established and rcv_established
547 	   handle them correctly, but it is not case with
548 	   dccp_v6_hnd_req and dccp_v6_ctl_send_reset().   --ANK
549 	 */
550 
551 	if (skb->protocol == htons(ETH_P_IP))
552 		return dccp_v4_do_rcv(sk, skb);
553 
554 	if (sk_filter(sk, skb))
555 		goto discard;
556 
557 	/*
558 	 * socket locking is here for SMP purposes as backlog rcv is currently
559 	 * called with bh processing disabled.
560 	 */
561 
562 	/* Do Stevens' IPV6_PKTOPTIONS.
563 
564 	   Yes, guys, it is the only place in our code, where we
565 	   may make it not affecting IPv4.
566 	   The rest of code is protocol independent,
567 	   and I do not like idea to uglify IPv4.
568 
569 	   Actually, all the idea behind IPV6_PKTOPTIONS
570 	   looks not very well thought. For now we latch
571 	   options, received in the last packet, enqueued
572 	   by tcp. Feel free to propose better solution.
573 					       --ANK (980728)
574 	 */
575 	if (np->rxopt.all)
576 	/*
577 	 * FIXME: Add handling of IPV6_PKTOPTIONS skb. See the comments below
578 	 *        (wrt ipv6_pktopions) and net/ipv6/tcp_ipv6.c for an example.
579 	 */
580 		opt_skb = skb_clone(skb, GFP_ATOMIC);
581 
582 	if (sk->sk_state == DCCP_OPEN) { /* Fast path */
583 		if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
584 			goto reset;
585 		if (opt_skb) {
586 			/* XXX This is where we would goto ipv6_pktoptions. */
587 			__kfree_skb(opt_skb);
588 		}
589 		return 0;
590 	}
591 
592 	/*
593 	 *  Step 3: Process LISTEN state
594 	 *     If S.state == LISTEN,
595 	 *	 If P.type == Request or P contains a valid Init Cookie option,
596 	 *	      (* Must scan the packet's options to check for Init
597 	 *		 Cookies.  Only Init Cookies are processed here,
598 	 *		 however; other options are processed in Step 8.  This
599 	 *		 scan need only be performed if the endpoint uses Init
600 	 *		 Cookies *)
601 	 *	      (* Generate a new socket and switch to that socket *)
602 	 *	      Set S := new socket for this port pair
603 	 *	      S.state = RESPOND
604 	 *	      Choose S.ISS (initial seqno) or set from Init Cookies
605 	 *	      Initialize S.GAR := S.ISS
606 	 *	      Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
607 	 *	      Continue with S.state == RESPOND
608 	 *	      (* A Response packet will be generated in Step 11 *)
609 	 *	 Otherwise,
610 	 *	      Generate Reset(No Connection) unless P.type == Reset
611 	 *	      Drop packet and return
612 	 *
613 	 * NOTE: the check for the packet types is done in
614 	 *	 dccp_rcv_state_process
615 	 */
616 
617 	if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
618 		goto reset;
619 	if (opt_skb) {
620 		/* XXX This is where we would goto ipv6_pktoptions. */
621 		__kfree_skb(opt_skb);
622 	}
623 	return 0;
624 
625 reset:
626 	dccp_v6_ctl_send_reset(sk, skb);
627 discard:
628 	if (opt_skb != NULL)
629 		__kfree_skb(opt_skb);
630 	kfree_skb(skb);
631 	return 0;
632 }
633 
634 static int dccp_v6_rcv(struct sk_buff *skb)
635 {
636 	const struct dccp_hdr *dh;
637 	struct sock *sk;
638 	int min_cov;
639 
640 	/* Step 1: Check header basics */
641 
642 	if (dccp_invalid_packet(skb))
643 		goto discard_it;
644 
645 	/* Step 1: If header checksum is incorrect, drop packet and return. */
646 	if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
647 				     &ipv6_hdr(skb)->daddr)) {
648 		DCCP_WARN("dropped packet with invalid checksum\n");
649 		goto discard_it;
650 	}
651 
652 	dh = dccp_hdr(skb);
653 
654 	DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
655 	DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
656 
657 	if (dccp_packet_without_ack(skb))
658 		DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
659 	else
660 		DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
661 
662 lookup:
663 	sk = __inet6_lookup_skb(&dccp_hashinfo, skb,
664 			        dh->dccph_sport, dh->dccph_dport,
665 				inet6_iif(skb));
666 	if (!sk) {
667 		dccp_pr_debug("failed to look up flow ID in table and "
668 			      "get corresponding socket\n");
669 		goto no_dccp_socket;
670 	}
671 
672 	/*
673 	 * Step 2:
674 	 *	... or S.state == TIMEWAIT,
675 	 *		Generate Reset(No Connection) unless P.type == Reset
676 	 *		Drop packet and return
677 	 */
678 	if (sk->sk_state == DCCP_TIME_WAIT) {
679 		dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
680 		inet_twsk_put(inet_twsk(sk));
681 		goto no_dccp_socket;
682 	}
683 
684 	if (sk->sk_state == DCCP_NEW_SYN_RECV) {
685 		struct request_sock *req = inet_reqsk(sk);
686 		struct sock *nsk = NULL;
687 
688 		sk = req->rsk_listener;
689 		if (likely(sk->sk_state == DCCP_LISTEN)) {
690 			nsk = dccp_check_req(sk, skb, req);
691 		} else {
692 			inet_csk_reqsk_queue_drop_and_put(sk, req);
693 			goto lookup;
694 		}
695 		if (!nsk) {
696 			reqsk_put(req);
697 			goto discard_it;
698 		}
699 		if (nsk == sk) {
700 			sock_hold(sk);
701 			reqsk_put(req);
702 		} else if (dccp_child_process(sk, nsk, skb)) {
703 			dccp_v6_ctl_send_reset(sk, skb);
704 			goto discard_it;
705 		} else {
706 			return 0;
707 		}
708 	}
709 	/*
710 	 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
711 	 *	o if MinCsCov = 0, only packets with CsCov = 0 are accepted
712 	 *	o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
713 	 */
714 	min_cov = dccp_sk(sk)->dccps_pcrlen;
715 	if (dh->dccph_cscov  &&  (min_cov == 0 || dh->dccph_cscov < min_cov))  {
716 		dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
717 			      dh->dccph_cscov, min_cov);
718 		/* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
719 		goto discard_and_relse;
720 	}
721 
722 	if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
723 		goto discard_and_relse;
724 
725 	return sk_receive_skb(sk, skb, 1) ? -1 : 0;
726 
727 no_dccp_socket:
728 	if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
729 		goto discard_it;
730 	/*
731 	 * Step 2:
732 	 *	If no socket ...
733 	 *		Generate Reset(No Connection) unless P.type == Reset
734 	 *		Drop packet and return
735 	 */
736 	if (dh->dccph_type != DCCP_PKT_RESET) {
737 		DCCP_SKB_CB(skb)->dccpd_reset_code =
738 					DCCP_RESET_CODE_NO_CONNECTION;
739 		dccp_v6_ctl_send_reset(sk, skb);
740 	}
741 
742 discard_it:
743 	kfree_skb(skb);
744 	return 0;
745 
746 discard_and_relse:
747 	sock_put(sk);
748 	goto discard_it;
749 }
750 
751 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
752 			   int addr_len)
753 {
754 	struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
755 	struct inet_connection_sock *icsk = inet_csk(sk);
756 	struct inet_sock *inet = inet_sk(sk);
757 	struct ipv6_pinfo *np = inet6_sk(sk);
758 	struct dccp_sock *dp = dccp_sk(sk);
759 	struct in6_addr *saddr = NULL, *final_p, final;
760 	struct flowi6 fl6;
761 	struct dst_entry *dst;
762 	int addr_type;
763 	int err;
764 
765 	dp->dccps_role = DCCP_ROLE_CLIENT;
766 
767 	if (addr_len < SIN6_LEN_RFC2133)
768 		return -EINVAL;
769 
770 	if (usin->sin6_family != AF_INET6)
771 		return -EAFNOSUPPORT;
772 
773 	memset(&fl6, 0, sizeof(fl6));
774 
775 	if (np->sndflow) {
776 		fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
777 		IP6_ECN_flow_init(fl6.flowlabel);
778 		if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
779 			struct ip6_flowlabel *flowlabel;
780 			flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
781 			if (flowlabel == NULL)
782 				return -EINVAL;
783 			fl6_sock_release(flowlabel);
784 		}
785 	}
786 	/*
787 	 * connect() to INADDR_ANY means loopback (BSD'ism).
788 	 */
789 	if (ipv6_addr_any(&usin->sin6_addr))
790 		usin->sin6_addr.s6_addr[15] = 1;
791 
792 	addr_type = ipv6_addr_type(&usin->sin6_addr);
793 
794 	if (addr_type & IPV6_ADDR_MULTICAST)
795 		return -ENETUNREACH;
796 
797 	if (addr_type & IPV6_ADDR_LINKLOCAL) {
798 		if (addr_len >= sizeof(struct sockaddr_in6) &&
799 		    usin->sin6_scope_id) {
800 			/* If interface is set while binding, indices
801 			 * must coincide.
802 			 */
803 			if (sk->sk_bound_dev_if &&
804 			    sk->sk_bound_dev_if != usin->sin6_scope_id)
805 				return -EINVAL;
806 
807 			sk->sk_bound_dev_if = usin->sin6_scope_id;
808 		}
809 
810 		/* Connect to link-local address requires an interface */
811 		if (!sk->sk_bound_dev_if)
812 			return -EINVAL;
813 	}
814 
815 	sk->sk_v6_daddr = usin->sin6_addr;
816 	np->flow_label = fl6.flowlabel;
817 
818 	/*
819 	 * DCCP over IPv4
820 	 */
821 	if (addr_type == IPV6_ADDR_MAPPED) {
822 		u32 exthdrlen = icsk->icsk_ext_hdr_len;
823 		struct sockaddr_in sin;
824 
825 		SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
826 
827 		if (__ipv6_only_sock(sk))
828 			return -ENETUNREACH;
829 
830 		sin.sin_family = AF_INET;
831 		sin.sin_port = usin->sin6_port;
832 		sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
833 
834 		icsk->icsk_af_ops = &dccp_ipv6_mapped;
835 		sk->sk_backlog_rcv = dccp_v4_do_rcv;
836 
837 		err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
838 		if (err) {
839 			icsk->icsk_ext_hdr_len = exthdrlen;
840 			icsk->icsk_af_ops = &dccp_ipv6_af_ops;
841 			sk->sk_backlog_rcv = dccp_v6_do_rcv;
842 			goto failure;
843 		}
844 		np->saddr = sk->sk_v6_rcv_saddr;
845 		return err;
846 	}
847 
848 	if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
849 		saddr = &sk->sk_v6_rcv_saddr;
850 
851 	fl6.flowi6_proto = IPPROTO_DCCP;
852 	fl6.daddr = sk->sk_v6_daddr;
853 	fl6.saddr = saddr ? *saddr : np->saddr;
854 	fl6.flowi6_oif = sk->sk_bound_dev_if;
855 	fl6.fl6_dport = usin->sin6_port;
856 	fl6.fl6_sport = inet->inet_sport;
857 	security_sk_classify_flow(sk, flowi6_to_flowi(&fl6));
858 
859 	final_p = fl6_update_dst(&fl6, np->opt, &final);
860 
861 	dst = ip6_dst_lookup_flow(sk, &fl6, final_p);
862 	if (IS_ERR(dst)) {
863 		err = PTR_ERR(dst);
864 		goto failure;
865 	}
866 
867 	if (saddr == NULL) {
868 		saddr = &fl6.saddr;
869 		sk->sk_v6_rcv_saddr = *saddr;
870 	}
871 
872 	/* set the source address */
873 	np->saddr = *saddr;
874 	inet->inet_rcv_saddr = LOOPBACK4_IPV6;
875 
876 	__ip6_dst_store(sk, dst, NULL, NULL);
877 
878 	icsk->icsk_ext_hdr_len = 0;
879 	if (np->opt != NULL)
880 		icsk->icsk_ext_hdr_len = (np->opt->opt_flen +
881 					  np->opt->opt_nflen);
882 
883 	inet->inet_dport = usin->sin6_port;
884 
885 	dccp_set_state(sk, DCCP_REQUESTING);
886 	err = inet6_hash_connect(&dccp_death_row, sk);
887 	if (err)
888 		goto late_failure;
889 
890 	dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
891 						      sk->sk_v6_daddr.s6_addr32,
892 						      inet->inet_sport,
893 						      inet->inet_dport);
894 	err = dccp_connect(sk);
895 	if (err)
896 		goto late_failure;
897 
898 	return 0;
899 
900 late_failure:
901 	dccp_set_state(sk, DCCP_CLOSED);
902 	__sk_dst_reset(sk);
903 failure:
904 	inet->inet_dport = 0;
905 	sk->sk_route_caps = 0;
906 	return err;
907 }
908 
909 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
910 	.queue_xmit	   = inet6_csk_xmit,
911 	.send_check	   = dccp_v6_send_check,
912 	.rebuild_header	   = inet6_sk_rebuild_header,
913 	.conn_request	   = dccp_v6_conn_request,
914 	.syn_recv_sock	   = dccp_v6_request_recv_sock,
915 	.net_header_len	   = sizeof(struct ipv6hdr),
916 	.setsockopt	   = ipv6_setsockopt,
917 	.getsockopt	   = ipv6_getsockopt,
918 	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
919 	.sockaddr_len	   = sizeof(struct sockaddr_in6),
920 	.bind_conflict	   = inet6_csk_bind_conflict,
921 #ifdef CONFIG_COMPAT
922 	.compat_setsockopt = compat_ipv6_setsockopt,
923 	.compat_getsockopt = compat_ipv6_getsockopt,
924 #endif
925 };
926 
927 /*
928  *	DCCP over IPv4 via INET6 API
929  */
930 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
931 	.queue_xmit	   = ip_queue_xmit,
932 	.send_check	   = dccp_v4_send_check,
933 	.rebuild_header	   = inet_sk_rebuild_header,
934 	.conn_request	   = dccp_v6_conn_request,
935 	.syn_recv_sock	   = dccp_v6_request_recv_sock,
936 	.net_header_len	   = sizeof(struct iphdr),
937 	.setsockopt	   = ipv6_setsockopt,
938 	.getsockopt	   = ipv6_getsockopt,
939 	.addr2sockaddr	   = inet6_csk_addr2sockaddr,
940 	.sockaddr_len	   = sizeof(struct sockaddr_in6),
941 #ifdef CONFIG_COMPAT
942 	.compat_setsockopt = compat_ipv6_setsockopt,
943 	.compat_getsockopt = compat_ipv6_getsockopt,
944 #endif
945 };
946 
947 /* NOTE: A lot of things set to zero explicitly by call to
948  *       sk_alloc() so need not be done here.
949  */
950 static int dccp_v6_init_sock(struct sock *sk)
951 {
952 	static __u8 dccp_v6_ctl_sock_initialized;
953 	int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
954 
955 	if (err == 0) {
956 		if (unlikely(!dccp_v6_ctl_sock_initialized))
957 			dccp_v6_ctl_sock_initialized = 1;
958 		inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
959 	}
960 
961 	return err;
962 }
963 
964 static void dccp_v6_destroy_sock(struct sock *sk)
965 {
966 	dccp_destroy_sock(sk);
967 	inet6_destroy_sock(sk);
968 }
969 
970 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
971 	.twsk_obj_size	= sizeof(struct dccp6_timewait_sock),
972 };
973 
974 static struct proto dccp_v6_prot = {
975 	.name		   = "DCCPv6",
976 	.owner		   = THIS_MODULE,
977 	.close		   = dccp_close,
978 	.connect	   = dccp_v6_connect,
979 	.disconnect	   = dccp_disconnect,
980 	.ioctl		   = dccp_ioctl,
981 	.init		   = dccp_v6_init_sock,
982 	.setsockopt	   = dccp_setsockopt,
983 	.getsockopt	   = dccp_getsockopt,
984 	.sendmsg	   = dccp_sendmsg,
985 	.recvmsg	   = dccp_recvmsg,
986 	.backlog_rcv	   = dccp_v6_do_rcv,
987 	.hash		   = inet_hash,
988 	.unhash		   = inet_unhash,
989 	.accept		   = inet_csk_accept,
990 	.get_port	   = inet_csk_get_port,
991 	.shutdown	   = dccp_shutdown,
992 	.destroy	   = dccp_v6_destroy_sock,
993 	.orphan_count	   = &dccp_orphan_count,
994 	.max_header	   = MAX_DCCP_HEADER,
995 	.obj_size	   = sizeof(struct dccp6_sock),
996 	.slab_flags	   = SLAB_DESTROY_BY_RCU,
997 	.rsk_prot	   = &dccp6_request_sock_ops,
998 	.twsk_prot	   = &dccp6_timewait_sock_ops,
999 	.h.hashinfo	   = &dccp_hashinfo,
1000 #ifdef CONFIG_COMPAT
1001 	.compat_setsockopt = compat_dccp_setsockopt,
1002 	.compat_getsockopt = compat_dccp_getsockopt,
1003 #endif
1004 };
1005 
1006 static const struct inet6_protocol dccp_v6_protocol = {
1007 	.handler	= dccp_v6_rcv,
1008 	.err_handler	= dccp_v6_err,
1009 	.flags		= INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1010 };
1011 
1012 static const struct proto_ops inet6_dccp_ops = {
1013 	.family		   = PF_INET6,
1014 	.owner		   = THIS_MODULE,
1015 	.release	   = inet6_release,
1016 	.bind		   = inet6_bind,
1017 	.connect	   = inet_stream_connect,
1018 	.socketpair	   = sock_no_socketpair,
1019 	.accept		   = inet_accept,
1020 	.getname	   = inet6_getname,
1021 	.poll		   = dccp_poll,
1022 	.ioctl		   = inet6_ioctl,
1023 	.listen		   = inet_dccp_listen,
1024 	.shutdown	   = inet_shutdown,
1025 	.setsockopt	   = sock_common_setsockopt,
1026 	.getsockopt	   = sock_common_getsockopt,
1027 	.sendmsg	   = inet_sendmsg,
1028 	.recvmsg	   = sock_common_recvmsg,
1029 	.mmap		   = sock_no_mmap,
1030 	.sendpage	   = sock_no_sendpage,
1031 #ifdef CONFIG_COMPAT
1032 	.compat_setsockopt = compat_sock_common_setsockopt,
1033 	.compat_getsockopt = compat_sock_common_getsockopt,
1034 #endif
1035 };
1036 
1037 static struct inet_protosw dccp_v6_protosw = {
1038 	.type		= SOCK_DCCP,
1039 	.protocol	= IPPROTO_DCCP,
1040 	.prot		= &dccp_v6_prot,
1041 	.ops		= &inet6_dccp_ops,
1042 	.flags		= INET_PROTOSW_ICSK,
1043 };
1044 
1045 static int __net_init dccp_v6_init_net(struct net *net)
1046 {
1047 	if (dccp_hashinfo.bhash == NULL)
1048 		return -ESOCKTNOSUPPORT;
1049 
1050 	return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1051 				    SOCK_DCCP, IPPROTO_DCCP, net);
1052 }
1053 
1054 static void __net_exit dccp_v6_exit_net(struct net *net)
1055 {
1056 	inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1057 }
1058 
1059 static struct pernet_operations dccp_v6_ops = {
1060 	.init   = dccp_v6_init_net,
1061 	.exit   = dccp_v6_exit_net,
1062 };
1063 
1064 static int __init dccp_v6_init(void)
1065 {
1066 	int err = proto_register(&dccp_v6_prot, 1);
1067 
1068 	if (err != 0)
1069 		goto out;
1070 
1071 	err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1072 	if (err != 0)
1073 		goto out_unregister_proto;
1074 
1075 	inet6_register_protosw(&dccp_v6_protosw);
1076 
1077 	err = register_pernet_subsys(&dccp_v6_ops);
1078 	if (err != 0)
1079 		goto out_destroy_ctl_sock;
1080 out:
1081 	return err;
1082 
1083 out_destroy_ctl_sock:
1084 	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1085 	inet6_unregister_protosw(&dccp_v6_protosw);
1086 out_unregister_proto:
1087 	proto_unregister(&dccp_v6_prot);
1088 	goto out;
1089 }
1090 
1091 static void __exit dccp_v6_exit(void)
1092 {
1093 	unregister_pernet_subsys(&dccp_v6_ops);
1094 	inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1095 	inet6_unregister_protosw(&dccp_v6_protosw);
1096 	proto_unregister(&dccp_v6_prot);
1097 }
1098 
1099 module_init(dccp_v6_init);
1100 module_exit(dccp_v6_exit);
1101 
1102 /*
1103  * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1104  * values directly, Also cover the case where the protocol is not specified,
1105  * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1106  */
1107 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1108 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1109 MODULE_LICENSE("GPL");
1110 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1111 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
1112