1/*
2 *  net/dccp/ipv4.c
3 *
4 *  An implementation of the DCCP protocol
5 *  Arnaldo Carvalho de Melo <acme@conectiva.com.br>
6 *
7 *	This program is free software; you can redistribute it and/or
8 *	modify it under the terms of the GNU General Public License
9 *	as published by the Free Software Foundation; either version
10 *	2 of the License, or (at your option) any later version.
11 */
12
13#include <linux/dccp.h>
14#include <linux/icmp.h>
15#include <linux/slab.h>
16#include <linux/module.h>
17#include <linux/skbuff.h>
18#include <linux/random.h>
19
20#include <net/icmp.h>
21#include <net/inet_common.h>
22#include <net/inet_hashtables.h>
23#include <net/inet_sock.h>
24#include <net/protocol.h>
25#include <net/sock.h>
26#include <net/timewait_sock.h>
27#include <net/tcp_states.h>
28#include <net/xfrm.h>
29#include <net/secure_seq.h>
30
31#include "ackvec.h"
32#include "ccid.h"
33#include "dccp.h"
34#include "feat.h"
35
36/*
37 * The per-net dccp.v4_ctl_sk socket is used for responding to
38 * the Out-of-the-blue (OOTB) packets. A control sock will be created
39 * for this socket at the initialization time.
40 */
41
42int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
43{
44	const struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
45	struct inet_sock *inet = inet_sk(sk);
46	struct dccp_sock *dp = dccp_sk(sk);
47	__be16 orig_sport, orig_dport;
48	__be32 daddr, nexthop;
49	struct flowi4 *fl4;
50	struct rtable *rt;
51	int err;
52	struct ip_options_rcu *inet_opt;
53
54	dp->dccps_role = DCCP_ROLE_CLIENT;
55
56	if (addr_len < sizeof(struct sockaddr_in))
57		return -EINVAL;
58
59	if (usin->sin_family != AF_INET)
60		return -EAFNOSUPPORT;
61
62	nexthop = daddr = usin->sin_addr.s_addr;
63
64	inet_opt = rcu_dereference_protected(inet->inet_opt,
65					     sock_owned_by_user(sk));
66	if (inet_opt != NULL && inet_opt->opt.srr) {
67		if (daddr == 0)
68			return -EINVAL;
69		nexthop = inet_opt->opt.faddr;
70	}
71
72	orig_sport = inet->inet_sport;
73	orig_dport = usin->sin_port;
74	fl4 = &inet->cork.fl.u.ip4;
75	rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
76			      RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
77			      IPPROTO_DCCP,
78			      orig_sport, orig_dport, sk);
79	if (IS_ERR(rt))
80		return PTR_ERR(rt);
81
82	if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
83		ip_rt_put(rt);
84		return -ENETUNREACH;
85	}
86
87	if (inet_opt == NULL || !inet_opt->opt.srr)
88		daddr = fl4->daddr;
89
90	if (inet->inet_saddr == 0)
91		inet->inet_saddr = fl4->saddr;
92	sk_rcv_saddr_set(sk, inet->inet_saddr);
93	inet->inet_dport = usin->sin_port;
94	sk_daddr_set(sk, daddr);
95
96	inet_csk(sk)->icsk_ext_hdr_len = 0;
97	if (inet_opt)
98		inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
99	/*
100	 * Socket identity is still unknown (sport may be zero).
101	 * However we set state to DCCP_REQUESTING and not releasing socket
102	 * lock select source port, enter ourselves into the hash tables and
103	 * complete initialization after this.
104	 */
105	dccp_set_state(sk, DCCP_REQUESTING);
106	err = inet_hash_connect(&dccp_death_row, sk);
107	if (err != 0)
108		goto failure;
109
110	rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
111			       inet->inet_sport, inet->inet_dport, sk);
112	if (IS_ERR(rt)) {
113		err = PTR_ERR(rt);
114		rt = NULL;
115		goto failure;
116	}
117	/* OK, now commit destination to socket.  */
118	sk_setup_caps(sk, &rt->dst);
119
120	dp->dccps_iss = secure_dccp_sequence_number(inet->inet_saddr,
121						    inet->inet_daddr,
122						    inet->inet_sport,
123						    inet->inet_dport);
124	inet->inet_id = dp->dccps_iss ^ jiffies;
125
126	err = dccp_connect(sk);
127	rt = NULL;
128	if (err != 0)
129		goto failure;
130out:
131	return err;
132failure:
133	/*
134	 * This unhashes the socket and releases the local port, if necessary.
135	 */
136	dccp_set_state(sk, DCCP_CLOSED);
137	ip_rt_put(rt);
138	sk->sk_route_caps = 0;
139	inet->inet_dport = 0;
140	goto out;
141}
142EXPORT_SYMBOL_GPL(dccp_v4_connect);
143
144/*
145 * This routine does path mtu discovery as defined in RFC1191.
146 */
147static inline void dccp_do_pmtu_discovery(struct sock *sk,
148					  const struct iphdr *iph,
149					  u32 mtu)
150{
151	struct dst_entry *dst;
152	const struct inet_sock *inet = inet_sk(sk);
153	const struct dccp_sock *dp = dccp_sk(sk);
154
155	/* We are not interested in DCCP_LISTEN and request_socks (RESPONSEs
156	 * send out by Linux are always < 576bytes so they should go through
157	 * unfragmented).
158	 */
159	if (sk->sk_state == DCCP_LISTEN)
160		return;
161
162	dst = inet_csk_update_pmtu(sk, mtu);
163	if (!dst)
164		return;
165
166	/* Something is about to be wrong... Remember soft error
167	 * for the case, if this connection will not able to recover.
168	 */
169	if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
170		sk->sk_err_soft = EMSGSIZE;
171
172	mtu = dst_mtu(dst);
173
174	if (inet->pmtudisc != IP_PMTUDISC_DONT &&
175	    ip_sk_accept_pmtu(sk) &&
176	    inet_csk(sk)->icsk_pmtu_cookie > mtu) {
177		dccp_sync_mss(sk, mtu);
178
179		/*
180		 * From RFC 4340, sec. 14.1:
181		 *
182		 *	DCCP-Sync packets are the best choice for upward
183		 *	probing, since DCCP-Sync probes do not risk application
184		 *	data loss.
185		 */
186		dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC);
187	} /* else let the usual retransmit timer handle it */
188}
189
190static void dccp_do_redirect(struct sk_buff *skb, struct sock *sk)
191{
192	struct dst_entry *dst = __sk_dst_check(sk, 0);
193
194	if (dst)
195		dst->ops->redirect(dst, sk, skb);
196}
197
198void dccp_req_err(struct sock *sk, u64 seq)
199	{
200	struct request_sock *req = inet_reqsk(sk);
201	struct net *net = sock_net(sk);
202
203	/*
204	 * ICMPs are not backlogged, hence we cannot get an established
205	 * socket here.
206	 */
207	if (!between48(seq, dccp_rsk(req)->dreq_iss, dccp_rsk(req)->dreq_gss)) {
208		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
209	} else {
210		/*
211		 * Still in RESPOND, just remove it silently.
212		 * There is no good way to pass the error to the newly
213		 * created socket, and POSIX does not want network
214		 * errors returned from accept().
215		 */
216		inet_csk_reqsk_queue_drop(req->rsk_listener, req);
217	}
218	reqsk_put(req);
219}
220EXPORT_SYMBOL(dccp_req_err);
221
222/*
223 * This routine is called by the ICMP module when it gets some sort of error
224 * condition. If err < 0 then the socket should be closed and the error
225 * returned to the user. If err > 0 it's just the icmp type << 8 | icmp code.
226 * After adjustment header points to the first 8 bytes of the tcp header. We
227 * need to find the appropriate port.
228 *
229 * The locking strategy used here is very "optimistic". When someone else
230 * accesses the socket the ICMP is just dropped and for some paths there is no
231 * check at all. A more general error queue to queue errors for later handling
232 * is probably better.
233 */
234static void dccp_v4_err(struct sk_buff *skb, u32 info)
235{
236	const struct iphdr *iph = (struct iphdr *)skb->data;
237	const u8 offset = iph->ihl << 2;
238	const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
239	struct dccp_sock *dp;
240	struct inet_sock *inet;
241	const int type = icmp_hdr(skb)->type;
242	const int code = icmp_hdr(skb)->code;
243	struct sock *sk;
244	__u64 seq;
245	int err;
246	struct net *net = dev_net(skb->dev);
247
248	if (skb->len < offset + sizeof(*dh) ||
249	    skb->len < offset + __dccp_basic_hdr_len(dh)) {
250		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
251		return;
252	}
253
254	sk = __inet_lookup_established(net, &dccp_hashinfo,
255				       iph->daddr, dh->dccph_dport,
256				       iph->saddr, ntohs(dh->dccph_sport),
257				       inet_iif(skb));
258	if (!sk) {
259		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
260		return;
261	}
262
263	if (sk->sk_state == DCCP_TIME_WAIT) {
264		inet_twsk_put(inet_twsk(sk));
265		return;
266	}
267	seq = dccp_hdr_seq(dh);
268	if (sk->sk_state == DCCP_NEW_SYN_RECV)
269		return dccp_req_err(sk, seq);
270
271	bh_lock_sock(sk);
272	/* If too many ICMPs get dropped on busy
273	 * servers this needs to be solved differently.
274	 */
275	if (sock_owned_by_user(sk))
276		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
277
278	if (sk->sk_state == DCCP_CLOSED)
279		goto out;
280
281	dp = dccp_sk(sk);
282	if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
283	    !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
284		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
285		goto out;
286	}
287
288	switch (type) {
289	case ICMP_REDIRECT:
290		dccp_do_redirect(skb, sk);
291		goto out;
292	case ICMP_SOURCE_QUENCH:
293		/* Just silently ignore these. */
294		goto out;
295	case ICMP_PARAMETERPROB:
296		err = EPROTO;
297		break;
298	case ICMP_DEST_UNREACH:
299		if (code > NR_ICMP_UNREACH)
300			goto out;
301
302		if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
303			if (!sock_owned_by_user(sk))
304				dccp_do_pmtu_discovery(sk, iph, info);
305			goto out;
306		}
307
308		err = icmp_err_convert[code].errno;
309		break;
310	case ICMP_TIME_EXCEEDED:
311		err = EHOSTUNREACH;
312		break;
313	default:
314		goto out;
315	}
316
317	switch (sk->sk_state) {
318	case DCCP_REQUESTING:
319	case DCCP_RESPOND:
320		if (!sock_owned_by_user(sk)) {
321			DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
322			sk->sk_err = err;
323
324			sk->sk_error_report(sk);
325
326			dccp_done(sk);
327		} else
328			sk->sk_err_soft = err;
329		goto out;
330	}
331
332	/* If we've already connected we will keep trying
333	 * until we time out, or the user gives up.
334	 *
335	 * rfc1122 4.2.3.9 allows to consider as hard errors
336	 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
337	 * but it is obsoleted by pmtu discovery).
338	 *
339	 * Note, that in modern internet, where routing is unreliable
340	 * and in each dark corner broken firewalls sit, sending random
341	 * errors ordered by their masters even this two messages finally lose
342	 * their original sense (even Linux sends invalid PORT_UNREACHs)
343	 *
344	 * Now we are in compliance with RFCs.
345	 *							--ANK (980905)
346	 */
347
348	inet = inet_sk(sk);
349	if (!sock_owned_by_user(sk) && inet->recverr) {
350		sk->sk_err = err;
351		sk->sk_error_report(sk);
352	} else /* Only an error on timeout */
353		sk->sk_err_soft = err;
354out:
355	bh_unlock_sock(sk);
356	sock_put(sk);
357}
358
359static inline __sum16 dccp_v4_csum_finish(struct sk_buff *skb,
360				      __be32 src, __be32 dst)
361{
362	return csum_tcpudp_magic(src, dst, skb->len, IPPROTO_DCCP, skb->csum);
363}
364
365void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb)
366{
367	const struct inet_sock *inet = inet_sk(sk);
368	struct dccp_hdr *dh = dccp_hdr(skb);
369
370	dccp_csum_outgoing(skb);
371	dh->dccph_checksum = dccp_v4_csum_finish(skb,
372						 inet->inet_saddr,
373						 inet->inet_daddr);
374}
375EXPORT_SYMBOL_GPL(dccp_v4_send_check);
376
377static inline u64 dccp_v4_init_sequence(const struct sk_buff *skb)
378{
379	return secure_dccp_sequence_number(ip_hdr(skb)->daddr,
380					   ip_hdr(skb)->saddr,
381					   dccp_hdr(skb)->dccph_dport,
382					   dccp_hdr(skb)->dccph_sport);
383}
384
385/*
386 * The three way handshake has completed - we got a valid ACK or DATAACK -
387 * now create the new socket.
388 *
389 * This is the equivalent of TCP's tcp_v4_syn_recv_sock
390 */
391struct sock *dccp_v4_request_recv_sock(const struct sock *sk,
392				       struct sk_buff *skb,
393				       struct request_sock *req,
394				       struct dst_entry *dst,
395				       struct request_sock *req_unhash,
396				       bool *own_req)
397{
398	struct inet_request_sock *ireq;
399	struct inet_sock *newinet;
400	struct sock *newsk;
401
402	if (sk_acceptq_is_full(sk))
403		goto exit_overflow;
404
405	newsk = dccp_create_openreq_child(sk, req, skb);
406	if (newsk == NULL)
407		goto exit_nonewsk;
408
409	newinet		   = inet_sk(newsk);
410	ireq		   = inet_rsk(req);
411	sk_daddr_set(newsk, ireq->ir_rmt_addr);
412	sk_rcv_saddr_set(newsk, ireq->ir_loc_addr);
413	newinet->inet_saddr	= ireq->ir_loc_addr;
414	newinet->inet_opt	= ireq->opt;
415	ireq->opt	   = NULL;
416	newinet->mc_index  = inet_iif(skb);
417	newinet->mc_ttl	   = ip_hdr(skb)->ttl;
418	newinet->inet_id   = jiffies;
419
420	if (dst == NULL && (dst = inet_csk_route_child_sock(sk, newsk, req)) == NULL)
421		goto put_and_exit;
422
423	sk_setup_caps(newsk, dst);
424
425	dccp_sync_mss(newsk, dst_mtu(dst));
426
427	if (__inet_inherit_port(sk, newsk) < 0)
428		goto put_and_exit;
429	*own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash));
430
431	return newsk;
432
433exit_overflow:
434	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
435exit_nonewsk:
436	dst_release(dst);
437exit:
438	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
439	return NULL;
440put_and_exit:
441	inet_csk_prepare_forced_close(newsk);
442	dccp_done(newsk);
443	goto exit;
444}
445EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock);
446
447static struct dst_entry* dccp_v4_route_skb(struct net *net, struct sock *sk,
448					   struct sk_buff *skb)
449{
450	struct rtable *rt;
451	const struct iphdr *iph = ip_hdr(skb);
452	struct flowi4 fl4 = {
453		.flowi4_oif = inet_iif(skb),
454		.daddr = iph->saddr,
455		.saddr = iph->daddr,
456		.flowi4_tos = RT_CONN_FLAGS(sk),
457		.flowi4_proto = sk->sk_protocol,
458		.fl4_sport = dccp_hdr(skb)->dccph_dport,
459		.fl4_dport = dccp_hdr(skb)->dccph_sport,
460	};
461
462	security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
463	rt = ip_route_output_flow(net, &fl4, sk);
464	if (IS_ERR(rt)) {
465		IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
466		return NULL;
467	}
468
469	return &rt->dst;
470}
471
472static int dccp_v4_send_response(const struct sock *sk, struct request_sock *req)
473{
474	int err = -1;
475	struct sk_buff *skb;
476	struct dst_entry *dst;
477	struct flowi4 fl4;
478
479	dst = inet_csk_route_req(sk, &fl4, req);
480	if (dst == NULL)
481		goto out;
482
483	skb = dccp_make_response(sk, dst, req);
484	if (skb != NULL) {
485		const struct inet_request_sock *ireq = inet_rsk(req);
486		struct dccp_hdr *dh = dccp_hdr(skb);
487
488		dh->dccph_checksum = dccp_v4_csum_finish(skb, ireq->ir_loc_addr,
489							      ireq->ir_rmt_addr);
490		err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
491					    ireq->ir_rmt_addr,
492					    ireq->opt);
493		err = net_xmit_eval(err);
494	}
495
496out:
497	dst_release(dst);
498	return err;
499}
500
501static void dccp_v4_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
502{
503	int err;
504	const struct iphdr *rxiph;
505	struct sk_buff *skb;
506	struct dst_entry *dst;
507	struct net *net = dev_net(skb_dst(rxskb)->dev);
508	struct sock *ctl_sk = net->dccp.v4_ctl_sk;
509
510	/* Never send a reset in response to a reset. */
511	if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
512		return;
513
514	if (skb_rtable(rxskb)->rt_type != RTN_LOCAL)
515		return;
516
517	dst = dccp_v4_route_skb(net, ctl_sk, rxskb);
518	if (dst == NULL)
519		return;
520
521	skb = dccp_ctl_make_reset(ctl_sk, rxskb);
522	if (skb == NULL)
523		goto out;
524
525	rxiph = ip_hdr(rxskb);
526	dccp_hdr(skb)->dccph_checksum = dccp_v4_csum_finish(skb, rxiph->saddr,
527								 rxiph->daddr);
528	skb_dst_set(skb, dst_clone(dst));
529
530	bh_lock_sock(ctl_sk);
531	err = ip_build_and_send_pkt(skb, ctl_sk,
532				    rxiph->daddr, rxiph->saddr, NULL);
533	bh_unlock_sock(ctl_sk);
534
535	if (net_xmit_eval(err) == 0) {
536		DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
537		DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
538	}
539out:
540	 dst_release(dst);
541}
542
543static void dccp_v4_reqsk_destructor(struct request_sock *req)
544{
545	dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
546	kfree(inet_rsk(req)->opt);
547}
548
549void dccp_syn_ack_timeout(const struct request_sock *req)
550{
551}
552EXPORT_SYMBOL(dccp_syn_ack_timeout);
553
554static struct request_sock_ops dccp_request_sock_ops __read_mostly = {
555	.family		= PF_INET,
556	.obj_size	= sizeof(struct dccp_request_sock),
557	.rtx_syn_ack	= dccp_v4_send_response,
558	.send_ack	= dccp_reqsk_send_ack,
559	.destructor	= dccp_v4_reqsk_destructor,
560	.send_reset	= dccp_v4_ctl_send_reset,
561	.syn_ack_timeout = dccp_syn_ack_timeout,
562};
563
564int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
565{
566	struct inet_request_sock *ireq;
567	struct request_sock *req;
568	struct dccp_request_sock *dreq;
569	const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
570	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
571
572	/* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
573	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
574		return 0;	/* discard, don't send a reset here */
575
576	if (dccp_bad_service_code(sk, service)) {
577		dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
578		goto drop;
579	}
580	/*
581	 * TW buckets are converted to open requests without
582	 * limitations, they conserve resources and peer is
583	 * evidently real one.
584	 */
585	dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
586	if (inet_csk_reqsk_queue_is_full(sk))
587		goto drop;
588
589	/*
590	 * Accept backlog is full. If we have already queued enough
591	 * of warm entries in syn queue, drop request. It is better than
592	 * clogging syn queue with openreqs with exponentially increasing
593	 * timeout.
594	 */
595	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
596		goto drop;
597
598	req = inet_reqsk_alloc(&dccp_request_sock_ops, sk, true);
599	if (req == NULL)
600		goto drop;
601
602	if (dccp_reqsk_init(req, dccp_sk(sk), skb))
603		goto drop_and_free;
604
605	dreq = dccp_rsk(req);
606	if (dccp_parse_options(sk, dreq, skb))
607		goto drop_and_free;
608
609	if (security_inet_conn_request(sk, skb, req))
610		goto drop_and_free;
611
612	ireq = inet_rsk(req);
613	sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr);
614	sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr);
615	ireq->ireq_family = AF_INET;
616	ireq->ir_iif = sk->sk_bound_dev_if;
617
618	/*
619	 * Step 3: Process LISTEN state
620	 *
621	 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
622	 *
623	 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
624	 */
625	dreq->dreq_isr	   = dcb->dccpd_seq;
626	dreq->dreq_gsr	   = dreq->dreq_isr;
627	dreq->dreq_iss	   = dccp_v4_init_sequence(skb);
628	dreq->dreq_gss     = dreq->dreq_iss;
629	dreq->dreq_service = service;
630
631	if (dccp_v4_send_response(sk, req))
632		goto drop_and_free;
633
634	inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
635	return 0;
636
637drop_and_free:
638	reqsk_free(req);
639drop:
640	DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
641	return -1;
642}
643EXPORT_SYMBOL_GPL(dccp_v4_conn_request);
644
645int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
646{
647	struct dccp_hdr *dh = dccp_hdr(skb);
648
649	if (sk->sk_state == DCCP_OPEN) { /* Fast path */
650		if (dccp_rcv_established(sk, skb, dh, skb->len))
651			goto reset;
652		return 0;
653	}
654
655	/*
656	 *  Step 3: Process LISTEN state
657	 *	 If P.type == Request or P contains a valid Init Cookie option,
658	 *	      (* Must scan the packet's options to check for Init
659	 *		 Cookies.  Only Init Cookies are processed here,
660	 *		 however; other options are processed in Step 8.  This
661	 *		 scan need only be performed if the endpoint uses Init
662	 *		 Cookies *)
663	 *	      (* Generate a new socket and switch to that socket *)
664	 *	      Set S := new socket for this port pair
665	 *	      S.state = RESPOND
666	 *	      Choose S.ISS (initial seqno) or set from Init Cookies
667	 *	      Initialize S.GAR := S.ISS
668	 *	      Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
669	 *	      Continue with S.state == RESPOND
670	 *	      (* A Response packet will be generated in Step 11 *)
671	 *	 Otherwise,
672	 *	      Generate Reset(No Connection) unless P.type == Reset
673	 *	      Drop packet and return
674	 *
675	 * NOTE: the check for the packet types is done in
676	 *	 dccp_rcv_state_process
677	 */
678
679	if (dccp_rcv_state_process(sk, skb, dh, skb->len))
680		goto reset;
681	return 0;
682
683reset:
684	dccp_v4_ctl_send_reset(sk, skb);
685	kfree_skb(skb);
686	return 0;
687}
688EXPORT_SYMBOL_GPL(dccp_v4_do_rcv);
689
690/**
691 *	dccp_invalid_packet  -  check for malformed packets
692 *	Implements RFC 4340, 8.5:  Step 1: Check header basics
693 *	Packets that fail these checks are ignored and do not receive Resets.
694 */
695int dccp_invalid_packet(struct sk_buff *skb)
696{
697	const struct dccp_hdr *dh;
698	unsigned int cscov;
699
700	if (skb->pkt_type != PACKET_HOST)
701		return 1;
702
703	/* If the packet is shorter than 12 bytes, drop packet and return */
704	if (!pskb_may_pull(skb, sizeof(struct dccp_hdr))) {
705		DCCP_WARN("pskb_may_pull failed\n");
706		return 1;
707	}
708
709	dh = dccp_hdr(skb);
710
711	/* If P.type is not understood, drop packet and return */
712	if (dh->dccph_type >= DCCP_PKT_INVALID) {
713		DCCP_WARN("invalid packet type\n");
714		return 1;
715	}
716
717	/*
718	 * If P.Data Offset is too small for packet type, drop packet and return
719	 */
720	if (dh->dccph_doff < dccp_hdr_len(skb) / sizeof(u32)) {
721		DCCP_WARN("P.Data Offset(%u) too small\n", dh->dccph_doff);
722		return 1;
723	}
724	/*
725	 * If P.Data Offset is too too large for packet, drop packet and return
726	 */
727	if (!pskb_may_pull(skb, dh->dccph_doff * sizeof(u32))) {
728		DCCP_WARN("P.Data Offset(%u) too large\n", dh->dccph_doff);
729		return 1;
730	}
731
732	/*
733	 * If P.type is not Data, Ack, or DataAck and P.X == 0 (the packet
734	 * has short sequence numbers), drop packet and return
735	 */
736	if ((dh->dccph_type < DCCP_PKT_DATA    ||
737	    dh->dccph_type > DCCP_PKT_DATAACK) && dh->dccph_x == 0)  {
738		DCCP_WARN("P.type (%s) not Data || [Data]Ack, while P.X == 0\n",
739			  dccp_packet_name(dh->dccph_type));
740		return 1;
741	}
742
743	/*
744	 * If P.CsCov is too large for the packet size, drop packet and return.
745	 * This must come _before_ checksumming (not as RFC 4340 suggests).
746	 */
747	cscov = dccp_csum_coverage(skb);
748	if (cscov > skb->len) {
749		DCCP_WARN("P.CsCov %u exceeds packet length %d\n",
750			  dh->dccph_cscov, skb->len);
751		return 1;
752	}
753
754	/* If header checksum is incorrect, drop packet and return.
755	 * (This step is completed in the AF-dependent functions.) */
756	skb->csum = skb_checksum(skb, 0, cscov, 0);
757
758	return 0;
759}
760EXPORT_SYMBOL_GPL(dccp_invalid_packet);
761
762/* this is called when real data arrives */
763static int dccp_v4_rcv(struct sk_buff *skb)
764{
765	const struct dccp_hdr *dh;
766	const struct iphdr *iph;
767	struct sock *sk;
768	int min_cov;
769
770	/* Step 1: Check header basics */
771
772	if (dccp_invalid_packet(skb))
773		goto discard_it;
774
775	iph = ip_hdr(skb);
776	/* Step 1: If header checksum is incorrect, drop packet and return */
777	if (dccp_v4_csum_finish(skb, iph->saddr, iph->daddr)) {
778		DCCP_WARN("dropped packet with invalid checksum\n");
779		goto discard_it;
780	}
781
782	dh = dccp_hdr(skb);
783
784	DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
785	DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
786
787	dccp_pr_debug("%8.8s src=%pI4@%-5d dst=%pI4@%-5d seq=%llu",
788		      dccp_packet_name(dh->dccph_type),
789		      &iph->saddr, ntohs(dh->dccph_sport),
790		      &iph->daddr, ntohs(dh->dccph_dport),
791		      (unsigned long long) DCCP_SKB_CB(skb)->dccpd_seq);
792
793	if (dccp_packet_without_ack(skb)) {
794		DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
795		dccp_pr_debug_cat("\n");
796	} else {
797		DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
798		dccp_pr_debug_cat(", ack=%llu\n", (unsigned long long)
799				  DCCP_SKB_CB(skb)->dccpd_ack_seq);
800	}
801
802lookup:
803	sk = __inet_lookup_skb(&dccp_hashinfo, skb,
804			       dh->dccph_sport, dh->dccph_dport);
805	if (!sk) {
806		dccp_pr_debug("failed to look up flow ID in table and "
807			      "get corresponding socket\n");
808		goto no_dccp_socket;
809	}
810
811	/*
812	 * Step 2:
813	 *	... or S.state == TIMEWAIT,
814	 *		Generate Reset(No Connection) unless P.type == Reset
815	 *		Drop packet and return
816	 */
817	if (sk->sk_state == DCCP_TIME_WAIT) {
818		dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
819		inet_twsk_put(inet_twsk(sk));
820		goto no_dccp_socket;
821	}
822
823	if (sk->sk_state == DCCP_NEW_SYN_RECV) {
824		struct request_sock *req = inet_reqsk(sk);
825		struct sock *nsk;
826
827		sk = req->rsk_listener;
828		if (unlikely(sk->sk_state != DCCP_LISTEN)) {
829			inet_csk_reqsk_queue_drop_and_put(sk, req);
830			goto lookup;
831		}
832		sock_hold(sk);
833		nsk = dccp_check_req(sk, skb, req);
834		if (!nsk) {
835			reqsk_put(req);
836			goto discard_and_relse;
837		}
838		if (nsk == sk) {
839			reqsk_put(req);
840		} else if (dccp_child_process(sk, nsk, skb)) {
841			dccp_v4_ctl_send_reset(sk, skb);
842			goto discard_and_relse;
843		} else {
844			sock_put(sk);
845			return 0;
846		}
847	}
848	/*
849	 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
850	 *	o if MinCsCov = 0, only packets with CsCov = 0 are accepted
851	 *	o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
852	 */
853	min_cov = dccp_sk(sk)->dccps_pcrlen;
854	if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov))  {
855		dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
856			      dh->dccph_cscov, min_cov);
857		/* FIXME: "Such packets SHOULD be reported using Data Dropped
858		 *         options (Section 11.7) with Drop Code 0, Protocol
859		 *         Constraints."                                     */
860		goto discard_and_relse;
861	}
862
863	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
864		goto discard_and_relse;
865	nf_reset(skb);
866
867	return sk_receive_skb(sk, skb, 1);
868
869no_dccp_socket:
870	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
871		goto discard_it;
872	/*
873	 * Step 2:
874	 *	If no socket ...
875	 *		Generate Reset(No Connection) unless P.type == Reset
876	 *		Drop packet and return
877	 */
878	if (dh->dccph_type != DCCP_PKT_RESET) {
879		DCCP_SKB_CB(skb)->dccpd_reset_code =
880					DCCP_RESET_CODE_NO_CONNECTION;
881		dccp_v4_ctl_send_reset(sk, skb);
882	}
883
884discard_it:
885	kfree_skb(skb);
886	return 0;
887
888discard_and_relse:
889	sock_put(sk);
890	goto discard_it;
891}
892
893static const struct inet_connection_sock_af_ops dccp_ipv4_af_ops = {
894	.queue_xmit	   = ip_queue_xmit,
895	.send_check	   = dccp_v4_send_check,
896	.rebuild_header	   = inet_sk_rebuild_header,
897	.conn_request	   = dccp_v4_conn_request,
898	.syn_recv_sock	   = dccp_v4_request_recv_sock,
899	.net_header_len	   = sizeof(struct iphdr),
900	.setsockopt	   = ip_setsockopt,
901	.getsockopt	   = ip_getsockopt,
902	.addr2sockaddr	   = inet_csk_addr2sockaddr,
903	.sockaddr_len	   = sizeof(struct sockaddr_in),
904	.bind_conflict	   = inet_csk_bind_conflict,
905#ifdef CONFIG_COMPAT
906	.compat_setsockopt = compat_ip_setsockopt,
907	.compat_getsockopt = compat_ip_getsockopt,
908#endif
909};
910
911static int dccp_v4_init_sock(struct sock *sk)
912{
913	static __u8 dccp_v4_ctl_sock_initialized;
914	int err = dccp_init_sock(sk, dccp_v4_ctl_sock_initialized);
915
916	if (err == 0) {
917		if (unlikely(!dccp_v4_ctl_sock_initialized))
918			dccp_v4_ctl_sock_initialized = 1;
919		inet_csk(sk)->icsk_af_ops = &dccp_ipv4_af_ops;
920	}
921
922	return err;
923}
924
925static struct timewait_sock_ops dccp_timewait_sock_ops = {
926	.twsk_obj_size	= sizeof(struct inet_timewait_sock),
927};
928
929static struct proto dccp_v4_prot = {
930	.name			= "DCCP",
931	.owner			= THIS_MODULE,
932	.close			= dccp_close,
933	.connect		= dccp_v4_connect,
934	.disconnect		= dccp_disconnect,
935	.ioctl			= dccp_ioctl,
936	.init			= dccp_v4_init_sock,
937	.setsockopt		= dccp_setsockopt,
938	.getsockopt		= dccp_getsockopt,
939	.sendmsg		= dccp_sendmsg,
940	.recvmsg		= dccp_recvmsg,
941	.backlog_rcv		= dccp_v4_do_rcv,
942	.hash			= inet_hash,
943	.unhash			= inet_unhash,
944	.accept			= inet_csk_accept,
945	.get_port		= inet_csk_get_port,
946	.shutdown		= dccp_shutdown,
947	.destroy		= dccp_destroy_sock,
948	.orphan_count		= &dccp_orphan_count,
949	.max_header		= MAX_DCCP_HEADER,
950	.obj_size		= sizeof(struct dccp_sock),
951	.slab_flags		= SLAB_DESTROY_BY_RCU,
952	.rsk_prot		= &dccp_request_sock_ops,
953	.twsk_prot		= &dccp_timewait_sock_ops,
954	.h.hashinfo		= &dccp_hashinfo,
955#ifdef CONFIG_COMPAT
956	.compat_setsockopt	= compat_dccp_setsockopt,
957	.compat_getsockopt	= compat_dccp_getsockopt,
958#endif
959};
960
961static const struct net_protocol dccp_v4_protocol = {
962	.handler	= dccp_v4_rcv,
963	.err_handler	= dccp_v4_err,
964	.no_policy	= 1,
965	.netns_ok	= 1,
966	.icmp_strict_tag_validation = 1,
967};
968
969static const struct proto_ops inet_dccp_ops = {
970	.family		   = PF_INET,
971	.owner		   = THIS_MODULE,
972	.release	   = inet_release,
973	.bind		   = inet_bind,
974	.connect	   = inet_stream_connect,
975	.socketpair	   = sock_no_socketpair,
976	.accept		   = inet_accept,
977	.getname	   = inet_getname,
978	/* FIXME: work on tcp_poll to rename it to inet_csk_poll */
979	.poll		   = dccp_poll,
980	.ioctl		   = inet_ioctl,
981	/* FIXME: work on inet_listen to rename it to sock_common_listen */
982	.listen		   = inet_dccp_listen,
983	.shutdown	   = inet_shutdown,
984	.setsockopt	   = sock_common_setsockopt,
985	.getsockopt	   = sock_common_getsockopt,
986	.sendmsg	   = inet_sendmsg,
987	.recvmsg	   = sock_common_recvmsg,
988	.mmap		   = sock_no_mmap,
989	.sendpage	   = sock_no_sendpage,
990#ifdef CONFIG_COMPAT
991	.compat_setsockopt = compat_sock_common_setsockopt,
992	.compat_getsockopt = compat_sock_common_getsockopt,
993#endif
994};
995
996static struct inet_protosw dccp_v4_protosw = {
997	.type		= SOCK_DCCP,
998	.protocol	= IPPROTO_DCCP,
999	.prot		= &dccp_v4_prot,
1000	.ops		= &inet_dccp_ops,
1001	.flags		= INET_PROTOSW_ICSK,
1002};
1003
1004static int __net_init dccp_v4_init_net(struct net *net)
1005{
1006	if (dccp_hashinfo.bhash == NULL)
1007		return -ESOCKTNOSUPPORT;
1008
1009	return inet_ctl_sock_create(&net->dccp.v4_ctl_sk, PF_INET,
1010				    SOCK_DCCP, IPPROTO_DCCP, net);
1011}
1012
1013static void __net_exit dccp_v4_exit_net(struct net *net)
1014{
1015	inet_ctl_sock_destroy(net->dccp.v4_ctl_sk);
1016}
1017
1018static struct pernet_operations dccp_v4_ops = {
1019	.init	= dccp_v4_init_net,
1020	.exit	= dccp_v4_exit_net,
1021};
1022
1023static int __init dccp_v4_init(void)
1024{
1025	int err = proto_register(&dccp_v4_prot, 1);
1026
1027	if (err != 0)
1028		goto out;
1029
1030	err = inet_add_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1031	if (err != 0)
1032		goto out_proto_unregister;
1033
1034	inet_register_protosw(&dccp_v4_protosw);
1035
1036	err = register_pernet_subsys(&dccp_v4_ops);
1037	if (err)
1038		goto out_destroy_ctl_sock;
1039out:
1040	return err;
1041out_destroy_ctl_sock:
1042	inet_unregister_protosw(&dccp_v4_protosw);
1043	inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1044out_proto_unregister:
1045	proto_unregister(&dccp_v4_prot);
1046	goto out;
1047}
1048
1049static void __exit dccp_v4_exit(void)
1050{
1051	unregister_pernet_subsys(&dccp_v4_ops);
1052	inet_unregister_protosw(&dccp_v4_protosw);
1053	inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1054	proto_unregister(&dccp_v4_prot);
1055}
1056
1057module_init(dccp_v4_init);
1058module_exit(dccp_v4_exit);
1059
1060/*
1061 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1062 * values directly, Also cover the case where the protocol is not specified,
1063 * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP
1064 */
1065MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 33, 6);
1066MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 0, 6);
1067MODULE_LICENSE("GPL");
1068MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1069MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol");
1070