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	inet->inet_rcv_saddr = inet->inet_saddr;
93
94	inet->inet_dport = usin->sin_port;
95	inet->inet_daddr = daddr;
96
97	inet_csk(sk)->icsk_ext_hdr_len = 0;
98	if (inet_opt)
99		inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
100	/*
101	 * Socket identity is still unknown (sport may be zero).
102	 * However we set state to DCCP_REQUESTING and not releasing socket
103	 * lock select source port, enter ourselves into the hash tables and
104	 * complete initialization after this.
105	 */
106	dccp_set_state(sk, DCCP_REQUESTING);
107	err = inet_hash_connect(&dccp_death_row, sk);
108	if (err != 0)
109		goto failure;
110
111	rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
112			       inet->inet_sport, inet->inet_dport, sk);
113	if (IS_ERR(rt)) {
114		err = PTR_ERR(rt);
115		rt = NULL;
116		goto failure;
117	}
118	/* OK, now commit destination to socket.  */
119	sk_setup_caps(sk, &rt->dst);
120
121	dp->dccps_iss = secure_dccp_sequence_number(inet->inet_saddr,
122						    inet->inet_daddr,
123						    inet->inet_sport,
124						    inet->inet_dport);
125	inet->inet_id = dp->dccps_iss ^ jiffies;
126
127	err = dccp_connect(sk);
128	rt = NULL;
129	if (err != 0)
130		goto failure;
131out:
132	return err;
133failure:
134	/*
135	 * This unhashes the socket and releases the local port, if necessary.
136	 */
137	dccp_set_state(sk, DCCP_CLOSED);
138	ip_rt_put(rt);
139	sk->sk_route_caps = 0;
140	inet->inet_dport = 0;
141	goto out;
142}
143
144EXPORT_SYMBOL_GPL(dccp_v4_connect);
145
146/*
147 * This routine does path mtu discovery as defined in RFC1191.
148 */
149static inline void dccp_do_pmtu_discovery(struct sock *sk,
150					  const struct iphdr *iph,
151					  u32 mtu)
152{
153	struct dst_entry *dst;
154	const struct inet_sock *inet = inet_sk(sk);
155	const struct dccp_sock *dp = dccp_sk(sk);
156
157	/* We are not interested in DCCP_LISTEN and request_socks (RESPONSEs
158	 * send out by Linux are always < 576bytes so they should go through
159	 * unfragmented).
160	 */
161	if (sk->sk_state == DCCP_LISTEN)
162		return;
163
164	dst = inet_csk_update_pmtu(sk, mtu);
165	if (!dst)
166		return;
167
168	/* Something is about to be wrong... Remember soft error
169	 * for the case, if this connection will not able to recover.
170	 */
171	if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
172		sk->sk_err_soft = EMSGSIZE;
173
174	mtu = dst_mtu(dst);
175
176	if (inet->pmtudisc != IP_PMTUDISC_DONT &&
177	    ip_sk_accept_pmtu(sk) &&
178	    inet_csk(sk)->icsk_pmtu_cookie > mtu) {
179		dccp_sync_mss(sk, mtu);
180
181		/*
182		 * From RFC 4340, sec. 14.1:
183		 *
184		 *	DCCP-Sync packets are the best choice for upward
185		 *	probing, since DCCP-Sync probes do not risk application
186		 *	data loss.
187		 */
188		dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC);
189	} /* else let the usual retransmit timer handle it */
190}
191
192static void dccp_do_redirect(struct sk_buff *skb, struct sock *sk)
193{
194	struct dst_entry *dst = __sk_dst_check(sk, 0);
195
196	if (dst)
197		dst->ops->redirect(dst, sk, skb);
198}
199
200/*
201 * This routine is called by the ICMP module when it gets some sort of error
202 * condition. If err < 0 then the socket should be closed and the error
203 * returned to the user. If err > 0 it's just the icmp type << 8 | icmp code.
204 * After adjustment header points to the first 8 bytes of the tcp header. We
205 * need to find the appropriate port.
206 *
207 * The locking strategy used here is very "optimistic". When someone else
208 * accesses the socket the ICMP is just dropped and for some paths there is no
209 * check at all. A more general error queue to queue errors for later handling
210 * is probably better.
211 */
212static void dccp_v4_err(struct sk_buff *skb, u32 info)
213{
214	const struct iphdr *iph = (struct iphdr *)skb->data;
215	const u8 offset = iph->ihl << 2;
216	const struct dccp_hdr *dh = (struct dccp_hdr *)(skb->data + offset);
217	struct dccp_sock *dp;
218	struct inet_sock *inet;
219	const int type = icmp_hdr(skb)->type;
220	const int code = icmp_hdr(skb)->code;
221	struct sock *sk;
222	__u64 seq;
223	int err;
224	struct net *net = dev_net(skb->dev);
225
226	if (skb->len < offset + sizeof(*dh) ||
227	    skb->len < offset + __dccp_basic_hdr_len(dh)) {
228		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
229		return;
230	}
231
232	sk = inet_lookup(net, &dccp_hashinfo,
233			iph->daddr, dh->dccph_dport,
234			iph->saddr, dh->dccph_sport, inet_iif(skb));
235	if (sk == NULL) {
236		ICMP_INC_STATS_BH(net, ICMP_MIB_INERRORS);
237		return;
238	}
239
240	if (sk->sk_state == DCCP_TIME_WAIT) {
241		inet_twsk_put(inet_twsk(sk));
242		return;
243	}
244
245	bh_lock_sock(sk);
246	/* If too many ICMPs get dropped on busy
247	 * servers this needs to be solved differently.
248	 */
249	if (sock_owned_by_user(sk))
250		NET_INC_STATS_BH(net, LINUX_MIB_LOCKDROPPEDICMPS);
251
252	if (sk->sk_state == DCCP_CLOSED)
253		goto out;
254
255	dp = dccp_sk(sk);
256	seq = dccp_hdr_seq(dh);
257	if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
258	    !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
259		NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
260		goto out;
261	}
262
263	switch (type) {
264	case ICMP_REDIRECT:
265		dccp_do_redirect(skb, sk);
266		goto out;
267	case ICMP_SOURCE_QUENCH:
268		/* Just silently ignore these. */
269		goto out;
270	case ICMP_PARAMETERPROB:
271		err = EPROTO;
272		break;
273	case ICMP_DEST_UNREACH:
274		if (code > NR_ICMP_UNREACH)
275			goto out;
276
277		if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
278			if (!sock_owned_by_user(sk))
279				dccp_do_pmtu_discovery(sk, iph, info);
280			goto out;
281		}
282
283		err = icmp_err_convert[code].errno;
284		break;
285	case ICMP_TIME_EXCEEDED:
286		err = EHOSTUNREACH;
287		break;
288	default:
289		goto out;
290	}
291
292	switch (sk->sk_state) {
293		struct request_sock *req , **prev;
294	case DCCP_LISTEN:
295		if (sock_owned_by_user(sk))
296			goto out;
297		req = inet_csk_search_req(sk, &prev, dh->dccph_dport,
298					  iph->daddr, iph->saddr);
299		if (!req)
300			goto out;
301
302		/*
303		 * ICMPs are not backlogged, hence we cannot get an established
304		 * socket here.
305		 */
306		WARN_ON(req->sk);
307
308		if (!between48(seq, dccp_rsk(req)->dreq_iss,
309				    dccp_rsk(req)->dreq_gss)) {
310			NET_INC_STATS_BH(net, LINUX_MIB_OUTOFWINDOWICMPS);
311			goto out;
312		}
313		/*
314		 * Still in RESPOND, just remove it silently.
315		 * There is no good way to pass the error to the newly
316		 * created socket, and POSIX does not want network
317		 * errors returned from accept().
318		 */
319		inet_csk_reqsk_queue_drop(sk, req, prev);
320		goto out;
321
322	case DCCP_REQUESTING:
323	case DCCP_RESPOND:
324		if (!sock_owned_by_user(sk)) {
325			DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
326			sk->sk_err = err;
327
328			sk->sk_error_report(sk);
329
330			dccp_done(sk);
331		} else
332			sk->sk_err_soft = err;
333		goto out;
334	}
335
336	/* If we've already connected we will keep trying
337	 * until we time out, or the user gives up.
338	 *
339	 * rfc1122 4.2.3.9 allows to consider as hard errors
340	 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
341	 * but it is obsoleted by pmtu discovery).
342	 *
343	 * Note, that in modern internet, where routing is unreliable
344	 * and in each dark corner broken firewalls sit, sending random
345	 * errors ordered by their masters even this two messages finally lose
346	 * their original sense (even Linux sends invalid PORT_UNREACHs)
347	 *
348	 * Now we are in compliance with RFCs.
349	 *							--ANK (980905)
350	 */
351
352	inet = inet_sk(sk);
353	if (!sock_owned_by_user(sk) && inet->recverr) {
354		sk->sk_err = err;
355		sk->sk_error_report(sk);
356	} else /* Only an error on timeout */
357		sk->sk_err_soft = err;
358out:
359	bh_unlock_sock(sk);
360	sock_put(sk);
361}
362
363static inline __sum16 dccp_v4_csum_finish(struct sk_buff *skb,
364				      __be32 src, __be32 dst)
365{
366	return csum_tcpudp_magic(src, dst, skb->len, IPPROTO_DCCP, skb->csum);
367}
368
369void dccp_v4_send_check(struct sock *sk, struct sk_buff *skb)
370{
371	const struct inet_sock *inet = inet_sk(sk);
372	struct dccp_hdr *dh = dccp_hdr(skb);
373
374	dccp_csum_outgoing(skb);
375	dh->dccph_checksum = dccp_v4_csum_finish(skb,
376						 inet->inet_saddr,
377						 inet->inet_daddr);
378}
379
380EXPORT_SYMBOL_GPL(dccp_v4_send_check);
381
382static inline u64 dccp_v4_init_sequence(const struct sk_buff *skb)
383{
384	return secure_dccp_sequence_number(ip_hdr(skb)->daddr,
385					   ip_hdr(skb)->saddr,
386					   dccp_hdr(skb)->dccph_dport,
387					   dccp_hdr(skb)->dccph_sport);
388}
389
390/*
391 * The three way handshake has completed - we got a valid ACK or DATAACK -
392 * now create the new socket.
393 *
394 * This is the equivalent of TCP's tcp_v4_syn_recv_sock
395 */
396struct sock *dccp_v4_request_recv_sock(struct sock *sk, struct sk_buff *skb,
397				       struct request_sock *req,
398				       struct dst_entry *dst)
399{
400	struct inet_request_sock *ireq;
401	struct inet_sock *newinet;
402	struct sock *newsk;
403
404	if (sk_acceptq_is_full(sk))
405		goto exit_overflow;
406
407	newsk = dccp_create_openreq_child(sk, req, skb);
408	if (newsk == NULL)
409		goto exit_nonewsk;
410
411	newinet		   = inet_sk(newsk);
412	ireq		   = inet_rsk(req);
413	newinet->inet_daddr	= ireq->ir_rmt_addr;
414	newinet->inet_rcv_saddr = ireq->ir_loc_addr;
415	newinet->inet_saddr	= ireq->ir_loc_addr;
416	newinet->inet_opt	= ireq->opt;
417	ireq->opt	   = NULL;
418	newinet->mc_index  = inet_iif(skb);
419	newinet->mc_ttl	   = ip_hdr(skb)->ttl;
420	newinet->inet_id   = jiffies;
421
422	if (dst == NULL && (dst = inet_csk_route_child_sock(sk, newsk, req)) == NULL)
423		goto put_and_exit;
424
425	sk_setup_caps(newsk, dst);
426
427	dccp_sync_mss(newsk, dst_mtu(dst));
428
429	if (__inet_inherit_port(sk, newsk) < 0)
430		goto put_and_exit;
431	__inet_hash_nolisten(newsk, NULL);
432
433	return newsk;
434
435exit_overflow:
436	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
437exit_nonewsk:
438	dst_release(dst);
439exit:
440	NET_INC_STATS_BH(sock_net(sk), LINUX_MIB_LISTENDROPS);
441	return NULL;
442put_and_exit:
443	inet_csk_prepare_forced_close(newsk);
444	dccp_done(newsk);
445	goto exit;
446}
447
448EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock);
449
450static struct sock *dccp_v4_hnd_req(struct sock *sk, struct sk_buff *skb)
451{
452	const struct dccp_hdr *dh = dccp_hdr(skb);
453	const struct iphdr *iph = ip_hdr(skb);
454	struct sock *nsk;
455	struct request_sock **prev;
456	/* Find possible connection requests. */
457	struct request_sock *req = inet_csk_search_req(sk, &prev,
458						       dh->dccph_sport,
459						       iph->saddr, iph->daddr);
460	if (req != NULL)
461		return dccp_check_req(sk, skb, req, prev);
462
463	nsk = inet_lookup_established(sock_net(sk), &dccp_hashinfo,
464				      iph->saddr, dh->dccph_sport,
465				      iph->daddr, dh->dccph_dport,
466				      inet_iif(skb));
467	if (nsk != NULL) {
468		if (nsk->sk_state != DCCP_TIME_WAIT) {
469			bh_lock_sock(nsk);
470			return nsk;
471		}
472		inet_twsk_put(inet_twsk(nsk));
473		return NULL;
474	}
475
476	return sk;
477}
478
479static struct dst_entry* dccp_v4_route_skb(struct net *net, struct sock *sk,
480					   struct sk_buff *skb)
481{
482	struct rtable *rt;
483	const struct iphdr *iph = ip_hdr(skb);
484	struct flowi4 fl4 = {
485		.flowi4_oif = inet_iif(skb),
486		.daddr = iph->saddr,
487		.saddr = iph->daddr,
488		.flowi4_tos = RT_CONN_FLAGS(sk),
489		.flowi4_proto = sk->sk_protocol,
490		.fl4_sport = dccp_hdr(skb)->dccph_dport,
491		.fl4_dport = dccp_hdr(skb)->dccph_sport,
492	};
493
494	security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
495	rt = ip_route_output_flow(net, &fl4, sk);
496	if (IS_ERR(rt)) {
497		IP_INC_STATS_BH(net, IPSTATS_MIB_OUTNOROUTES);
498		return NULL;
499	}
500
501	return &rt->dst;
502}
503
504static int dccp_v4_send_response(struct sock *sk, struct request_sock *req)
505{
506	int err = -1;
507	struct sk_buff *skb;
508	struct dst_entry *dst;
509	struct flowi4 fl4;
510
511	dst = inet_csk_route_req(sk, &fl4, req);
512	if (dst == NULL)
513		goto out;
514
515	skb = dccp_make_response(sk, dst, req);
516	if (skb != NULL) {
517		const struct inet_request_sock *ireq = inet_rsk(req);
518		struct dccp_hdr *dh = dccp_hdr(skb);
519
520		dh->dccph_checksum = dccp_v4_csum_finish(skb, ireq->ir_loc_addr,
521							      ireq->ir_rmt_addr);
522		err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
523					    ireq->ir_rmt_addr,
524					    ireq->opt);
525		err = net_xmit_eval(err);
526	}
527
528out:
529	dst_release(dst);
530	return err;
531}
532
533static void dccp_v4_ctl_send_reset(struct sock *sk, struct sk_buff *rxskb)
534{
535	int err;
536	const struct iphdr *rxiph;
537	struct sk_buff *skb;
538	struct dst_entry *dst;
539	struct net *net = dev_net(skb_dst(rxskb)->dev);
540	struct sock *ctl_sk = net->dccp.v4_ctl_sk;
541
542	/* Never send a reset in response to a reset. */
543	if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
544		return;
545
546	if (skb_rtable(rxskb)->rt_type != RTN_LOCAL)
547		return;
548
549	dst = dccp_v4_route_skb(net, ctl_sk, rxskb);
550	if (dst == NULL)
551		return;
552
553	skb = dccp_ctl_make_reset(ctl_sk, rxskb);
554	if (skb == NULL)
555		goto out;
556
557	rxiph = ip_hdr(rxskb);
558	dccp_hdr(skb)->dccph_checksum = dccp_v4_csum_finish(skb, rxiph->saddr,
559								 rxiph->daddr);
560	skb_dst_set(skb, dst_clone(dst));
561
562	bh_lock_sock(ctl_sk);
563	err = ip_build_and_send_pkt(skb, ctl_sk,
564				    rxiph->daddr, rxiph->saddr, NULL);
565	bh_unlock_sock(ctl_sk);
566
567	if (net_xmit_eval(err) == 0) {
568		DCCP_INC_STATS_BH(DCCP_MIB_OUTSEGS);
569		DCCP_INC_STATS_BH(DCCP_MIB_OUTRSTS);
570	}
571out:
572	 dst_release(dst);
573}
574
575static void dccp_v4_reqsk_destructor(struct request_sock *req)
576{
577	dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
578	kfree(inet_rsk(req)->opt);
579}
580
581void dccp_syn_ack_timeout(struct sock *sk, struct request_sock *req)
582{
583}
584EXPORT_SYMBOL(dccp_syn_ack_timeout);
585
586static struct request_sock_ops dccp_request_sock_ops __read_mostly = {
587	.family		= PF_INET,
588	.obj_size	= sizeof(struct dccp_request_sock),
589	.rtx_syn_ack	= dccp_v4_send_response,
590	.send_ack	= dccp_reqsk_send_ack,
591	.destructor	= dccp_v4_reqsk_destructor,
592	.send_reset	= dccp_v4_ctl_send_reset,
593	.syn_ack_timeout = dccp_syn_ack_timeout,
594};
595
596int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
597{
598	struct inet_request_sock *ireq;
599	struct request_sock *req;
600	struct dccp_request_sock *dreq;
601	const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
602	struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
603
604	/* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
605	if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
606		return 0;	/* discard, don't send a reset here */
607
608	if (dccp_bad_service_code(sk, service)) {
609		dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
610		goto drop;
611	}
612	/*
613	 * TW buckets are converted to open requests without
614	 * limitations, they conserve resources and peer is
615	 * evidently real one.
616	 */
617	dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
618	if (inet_csk_reqsk_queue_is_full(sk))
619		goto drop;
620
621	/*
622	 * Accept backlog is full. If we have already queued enough
623	 * of warm entries in syn queue, drop request. It is better than
624	 * clogging syn queue with openreqs with exponentially increasing
625	 * timeout.
626	 */
627	if (sk_acceptq_is_full(sk) && inet_csk_reqsk_queue_young(sk) > 1)
628		goto drop;
629
630	req = inet_reqsk_alloc(&dccp_request_sock_ops);
631	if (req == NULL)
632		goto drop;
633
634	if (dccp_reqsk_init(req, dccp_sk(sk), skb))
635		goto drop_and_free;
636
637	dreq = dccp_rsk(req);
638	if (dccp_parse_options(sk, dreq, skb))
639		goto drop_and_free;
640
641	if (security_inet_conn_request(sk, skb, req))
642		goto drop_and_free;
643
644	ireq = inet_rsk(req);
645	ireq->ir_loc_addr = ip_hdr(skb)->daddr;
646	ireq->ir_rmt_addr = ip_hdr(skb)->saddr;
647
648	/*
649	 * Step 3: Process LISTEN state
650	 *
651	 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
652	 *
653	 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
654	 */
655	dreq->dreq_isr	   = dcb->dccpd_seq;
656	dreq->dreq_gsr	   = dreq->dreq_isr;
657	dreq->dreq_iss	   = dccp_v4_init_sequence(skb);
658	dreq->dreq_gss     = dreq->dreq_iss;
659	dreq->dreq_service = service;
660
661	if (dccp_v4_send_response(sk, req))
662		goto drop_and_free;
663
664	inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
665	return 0;
666
667drop_and_free:
668	reqsk_free(req);
669drop:
670	DCCP_INC_STATS_BH(DCCP_MIB_ATTEMPTFAILS);
671	return -1;
672}
673
674EXPORT_SYMBOL_GPL(dccp_v4_conn_request);
675
676int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
677{
678	struct dccp_hdr *dh = dccp_hdr(skb);
679
680	if (sk->sk_state == DCCP_OPEN) { /* Fast path */
681		if (dccp_rcv_established(sk, skb, dh, skb->len))
682			goto reset;
683		return 0;
684	}
685
686	/*
687	 *  Step 3: Process LISTEN state
688	 *	 If P.type == Request or P contains a valid Init Cookie option,
689	 *	      (* Must scan the packet's options to check for Init
690	 *		 Cookies.  Only Init Cookies are processed here,
691	 *		 however; other options are processed in Step 8.  This
692	 *		 scan need only be performed if the endpoint uses Init
693	 *		 Cookies *)
694	 *	      (* Generate a new socket and switch to that socket *)
695	 *	      Set S := new socket for this port pair
696	 *	      S.state = RESPOND
697	 *	      Choose S.ISS (initial seqno) or set from Init Cookies
698	 *	      Initialize S.GAR := S.ISS
699	 *	      Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
700	 *	      Continue with S.state == RESPOND
701	 *	      (* A Response packet will be generated in Step 11 *)
702	 *	 Otherwise,
703	 *	      Generate Reset(No Connection) unless P.type == Reset
704	 *	      Drop packet and return
705	 *
706	 * NOTE: the check for the packet types is done in
707	 *	 dccp_rcv_state_process
708	 */
709	if (sk->sk_state == DCCP_LISTEN) {
710		struct sock *nsk = dccp_v4_hnd_req(sk, skb);
711
712		if (nsk == NULL)
713			goto discard;
714
715		if (nsk != sk) {
716			if (dccp_child_process(sk, nsk, skb))
717				goto reset;
718			return 0;
719		}
720	}
721
722	if (dccp_rcv_state_process(sk, skb, dh, skb->len))
723		goto reset;
724	return 0;
725
726reset:
727	dccp_v4_ctl_send_reset(sk, skb);
728discard:
729	kfree_skb(skb);
730	return 0;
731}
732
733EXPORT_SYMBOL_GPL(dccp_v4_do_rcv);
734
735/**
736 *	dccp_invalid_packet  -  check for malformed packets
737 *	Implements RFC 4340, 8.5:  Step 1: Check header basics
738 *	Packets that fail these checks are ignored and do not receive Resets.
739 */
740int dccp_invalid_packet(struct sk_buff *skb)
741{
742	const struct dccp_hdr *dh;
743	unsigned int cscov;
744
745	if (skb->pkt_type != PACKET_HOST)
746		return 1;
747
748	/* If the packet is shorter than 12 bytes, drop packet and return */
749	if (!pskb_may_pull(skb, sizeof(struct dccp_hdr))) {
750		DCCP_WARN("pskb_may_pull failed\n");
751		return 1;
752	}
753
754	dh = dccp_hdr(skb);
755
756	/* If P.type is not understood, drop packet and return */
757	if (dh->dccph_type >= DCCP_PKT_INVALID) {
758		DCCP_WARN("invalid packet type\n");
759		return 1;
760	}
761
762	/*
763	 * If P.Data Offset is too small for packet type, drop packet and return
764	 */
765	if (dh->dccph_doff < dccp_hdr_len(skb) / sizeof(u32)) {
766		DCCP_WARN("P.Data Offset(%u) too small\n", dh->dccph_doff);
767		return 1;
768	}
769	/*
770	 * If P.Data Offset is too too large for packet, drop packet and return
771	 */
772	if (!pskb_may_pull(skb, dh->dccph_doff * sizeof(u32))) {
773		DCCP_WARN("P.Data Offset(%u) too large\n", dh->dccph_doff);
774		return 1;
775	}
776
777	/*
778	 * If P.type is not Data, Ack, or DataAck and P.X == 0 (the packet
779	 * has short sequence numbers), drop packet and return
780	 */
781	if ((dh->dccph_type < DCCP_PKT_DATA    ||
782	    dh->dccph_type > DCCP_PKT_DATAACK) && dh->dccph_x == 0)  {
783		DCCP_WARN("P.type (%s) not Data || [Data]Ack, while P.X == 0\n",
784			  dccp_packet_name(dh->dccph_type));
785		return 1;
786	}
787
788	/*
789	 * If P.CsCov is too large for the packet size, drop packet and return.
790	 * This must come _before_ checksumming (not as RFC 4340 suggests).
791	 */
792	cscov = dccp_csum_coverage(skb);
793	if (cscov > skb->len) {
794		DCCP_WARN("P.CsCov %u exceeds packet length %d\n",
795			  dh->dccph_cscov, skb->len);
796		return 1;
797	}
798
799	/* If header checksum is incorrect, drop packet and return.
800	 * (This step is completed in the AF-dependent functions.) */
801	skb->csum = skb_checksum(skb, 0, cscov, 0);
802
803	return 0;
804}
805
806EXPORT_SYMBOL_GPL(dccp_invalid_packet);
807
808/* this is called when real data arrives */
809static int dccp_v4_rcv(struct sk_buff *skb)
810{
811	const struct dccp_hdr *dh;
812	const struct iphdr *iph;
813	struct sock *sk;
814	int min_cov;
815
816	/* Step 1: Check header basics */
817
818	if (dccp_invalid_packet(skb))
819		goto discard_it;
820
821	iph = ip_hdr(skb);
822	/* Step 1: If header checksum is incorrect, drop packet and return */
823	if (dccp_v4_csum_finish(skb, iph->saddr, iph->daddr)) {
824		DCCP_WARN("dropped packet with invalid checksum\n");
825		goto discard_it;
826	}
827
828	dh = dccp_hdr(skb);
829
830	DCCP_SKB_CB(skb)->dccpd_seq  = dccp_hdr_seq(dh);
831	DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
832
833	dccp_pr_debug("%8.8s src=%pI4@%-5d dst=%pI4@%-5d seq=%llu",
834		      dccp_packet_name(dh->dccph_type),
835		      &iph->saddr, ntohs(dh->dccph_sport),
836		      &iph->daddr, ntohs(dh->dccph_dport),
837		      (unsigned long long) DCCP_SKB_CB(skb)->dccpd_seq);
838
839	if (dccp_packet_without_ack(skb)) {
840		DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
841		dccp_pr_debug_cat("\n");
842	} else {
843		DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
844		dccp_pr_debug_cat(", ack=%llu\n", (unsigned long long)
845				  DCCP_SKB_CB(skb)->dccpd_ack_seq);
846	}
847
848	/* Step 2:
849	 *	Look up flow ID in table and get corresponding socket */
850	sk = __inet_lookup_skb(&dccp_hashinfo, skb,
851			       dh->dccph_sport, dh->dccph_dport);
852	/*
853	 * Step 2:
854	 *	If no socket ...
855	 */
856	if (sk == NULL) {
857		dccp_pr_debug("failed to look up flow ID in table and "
858			      "get corresponding socket\n");
859		goto no_dccp_socket;
860	}
861
862	/*
863	 * Step 2:
864	 *	... or S.state == TIMEWAIT,
865	 *		Generate Reset(No Connection) unless P.type == Reset
866	 *		Drop packet and return
867	 */
868	if (sk->sk_state == DCCP_TIME_WAIT) {
869		dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
870		inet_twsk_put(inet_twsk(sk));
871		goto no_dccp_socket;
872	}
873
874	/*
875	 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
876	 *	o if MinCsCov = 0, only packets with CsCov = 0 are accepted
877	 *	o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
878	 */
879	min_cov = dccp_sk(sk)->dccps_pcrlen;
880	if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov))  {
881		dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
882			      dh->dccph_cscov, min_cov);
883		/* FIXME: "Such packets SHOULD be reported using Data Dropped
884		 *         options (Section 11.7) with Drop Code 0, Protocol
885		 *         Constraints."                                     */
886		goto discard_and_relse;
887	}
888
889	if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
890		goto discard_and_relse;
891	nf_reset(skb);
892
893	return sk_receive_skb(sk, skb, 1);
894
895no_dccp_socket:
896	if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
897		goto discard_it;
898	/*
899	 * Step 2:
900	 *	If no socket ...
901	 *		Generate Reset(No Connection) unless P.type == Reset
902	 *		Drop packet and return
903	 */
904	if (dh->dccph_type != DCCP_PKT_RESET) {
905		DCCP_SKB_CB(skb)->dccpd_reset_code =
906					DCCP_RESET_CODE_NO_CONNECTION;
907		dccp_v4_ctl_send_reset(sk, skb);
908	}
909
910discard_it:
911	kfree_skb(skb);
912	return 0;
913
914discard_and_relse:
915	sock_put(sk);
916	goto discard_it;
917}
918
919static const struct inet_connection_sock_af_ops dccp_ipv4_af_ops = {
920	.queue_xmit	   = ip_queue_xmit,
921	.send_check	   = dccp_v4_send_check,
922	.rebuild_header	   = inet_sk_rebuild_header,
923	.conn_request	   = dccp_v4_conn_request,
924	.syn_recv_sock	   = dccp_v4_request_recv_sock,
925	.net_header_len	   = sizeof(struct iphdr),
926	.setsockopt	   = ip_setsockopt,
927	.getsockopt	   = ip_getsockopt,
928	.addr2sockaddr	   = inet_csk_addr2sockaddr,
929	.sockaddr_len	   = sizeof(struct sockaddr_in),
930	.bind_conflict	   = inet_csk_bind_conflict,
931#ifdef CONFIG_COMPAT
932	.compat_setsockopt = compat_ip_setsockopt,
933	.compat_getsockopt = compat_ip_getsockopt,
934#endif
935};
936
937static int dccp_v4_init_sock(struct sock *sk)
938{
939	static __u8 dccp_v4_ctl_sock_initialized;
940	int err = dccp_init_sock(sk, dccp_v4_ctl_sock_initialized);
941
942	if (err == 0) {
943		if (unlikely(!dccp_v4_ctl_sock_initialized))
944			dccp_v4_ctl_sock_initialized = 1;
945		inet_csk(sk)->icsk_af_ops = &dccp_ipv4_af_ops;
946	}
947
948	return err;
949}
950
951static struct timewait_sock_ops dccp_timewait_sock_ops = {
952	.twsk_obj_size	= sizeof(struct inet_timewait_sock),
953};
954
955static struct proto dccp_v4_prot = {
956	.name			= "DCCP",
957	.owner			= THIS_MODULE,
958	.close			= dccp_close,
959	.connect		= dccp_v4_connect,
960	.disconnect		= dccp_disconnect,
961	.ioctl			= dccp_ioctl,
962	.init			= dccp_v4_init_sock,
963	.setsockopt		= dccp_setsockopt,
964	.getsockopt		= dccp_getsockopt,
965	.sendmsg		= dccp_sendmsg,
966	.recvmsg		= dccp_recvmsg,
967	.backlog_rcv		= dccp_v4_do_rcv,
968	.hash			= inet_hash,
969	.unhash			= inet_unhash,
970	.accept			= inet_csk_accept,
971	.get_port		= inet_csk_get_port,
972	.shutdown		= dccp_shutdown,
973	.destroy		= dccp_destroy_sock,
974	.orphan_count		= &dccp_orphan_count,
975	.max_header		= MAX_DCCP_HEADER,
976	.obj_size		= sizeof(struct dccp_sock),
977	.slab_flags		= SLAB_DESTROY_BY_RCU,
978	.rsk_prot		= &dccp_request_sock_ops,
979	.twsk_prot		= &dccp_timewait_sock_ops,
980	.h.hashinfo		= &dccp_hashinfo,
981#ifdef CONFIG_COMPAT
982	.compat_setsockopt	= compat_dccp_setsockopt,
983	.compat_getsockopt	= compat_dccp_getsockopt,
984#endif
985};
986
987static const struct net_protocol dccp_v4_protocol = {
988	.handler	= dccp_v4_rcv,
989	.err_handler	= dccp_v4_err,
990	.no_policy	= 1,
991	.netns_ok	= 1,
992	.icmp_strict_tag_validation = 1,
993};
994
995static const struct proto_ops inet_dccp_ops = {
996	.family		   = PF_INET,
997	.owner		   = THIS_MODULE,
998	.release	   = inet_release,
999	.bind		   = inet_bind,
1000	.connect	   = inet_stream_connect,
1001	.socketpair	   = sock_no_socketpair,
1002	.accept		   = inet_accept,
1003	.getname	   = inet_getname,
1004	/* FIXME: work on tcp_poll to rename it to inet_csk_poll */
1005	.poll		   = dccp_poll,
1006	.ioctl		   = inet_ioctl,
1007	/* FIXME: work on inet_listen to rename it to sock_common_listen */
1008	.listen		   = inet_dccp_listen,
1009	.shutdown	   = inet_shutdown,
1010	.setsockopt	   = sock_common_setsockopt,
1011	.getsockopt	   = sock_common_getsockopt,
1012	.sendmsg	   = inet_sendmsg,
1013	.recvmsg	   = sock_common_recvmsg,
1014	.mmap		   = sock_no_mmap,
1015	.sendpage	   = sock_no_sendpage,
1016#ifdef CONFIG_COMPAT
1017	.compat_setsockopt = compat_sock_common_setsockopt,
1018	.compat_getsockopt = compat_sock_common_getsockopt,
1019#endif
1020};
1021
1022static struct inet_protosw dccp_v4_protosw = {
1023	.type		= SOCK_DCCP,
1024	.protocol	= IPPROTO_DCCP,
1025	.prot		= &dccp_v4_prot,
1026	.ops		= &inet_dccp_ops,
1027	.flags		= INET_PROTOSW_ICSK,
1028};
1029
1030static int __net_init dccp_v4_init_net(struct net *net)
1031{
1032	if (dccp_hashinfo.bhash == NULL)
1033		return -ESOCKTNOSUPPORT;
1034
1035	return inet_ctl_sock_create(&net->dccp.v4_ctl_sk, PF_INET,
1036				    SOCK_DCCP, IPPROTO_DCCP, net);
1037}
1038
1039static void __net_exit dccp_v4_exit_net(struct net *net)
1040{
1041	inet_ctl_sock_destroy(net->dccp.v4_ctl_sk);
1042}
1043
1044static struct pernet_operations dccp_v4_ops = {
1045	.init	= dccp_v4_init_net,
1046	.exit	= dccp_v4_exit_net,
1047};
1048
1049static int __init dccp_v4_init(void)
1050{
1051	int err = proto_register(&dccp_v4_prot, 1);
1052
1053	if (err != 0)
1054		goto out;
1055
1056	err = inet_add_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1057	if (err != 0)
1058		goto out_proto_unregister;
1059
1060	inet_register_protosw(&dccp_v4_protosw);
1061
1062	err = register_pernet_subsys(&dccp_v4_ops);
1063	if (err)
1064		goto out_destroy_ctl_sock;
1065out:
1066	return err;
1067out_destroy_ctl_sock:
1068	inet_unregister_protosw(&dccp_v4_protosw);
1069	inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1070out_proto_unregister:
1071	proto_unregister(&dccp_v4_prot);
1072	goto out;
1073}
1074
1075static void __exit dccp_v4_exit(void)
1076{
1077	unregister_pernet_subsys(&dccp_v4_ops);
1078	inet_unregister_protosw(&dccp_v4_protosw);
1079	inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1080	proto_unregister(&dccp_v4_prot);
1081}
1082
1083module_init(dccp_v4_init);
1084module_exit(dccp_v4_exit);
1085
1086/*
1087 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1088 * values directly, Also cover the case where the protocol is not specified,
1089 * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP
1090 */
1091MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 33, 6);
1092MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 0, 6);
1093MODULE_LICENSE("GPL");
1094MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1095MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol");
1096