1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * net/dccp/ipv4.c
4 *
5 * An implementation of the DCCP protocol
6 * Arnaldo Carvalho de Melo <acme@conectiva.com.br>
7 */
8
9 #include <linux/dccp.h>
10 #include <linux/icmp.h>
11 #include <linux/slab.h>
12 #include <linux/module.h>
13 #include <linux/skbuff.h>
14 #include <linux/random.h>
15
16 #include <net/icmp.h>
17 #include <net/inet_common.h>
18 #include <net/inet_hashtables.h>
19 #include <net/inet_sock.h>
20 #include <net/protocol.h>
21 #include <net/sock.h>
22 #include <net/timewait_sock.h>
23 #include <net/tcp_states.h>
24 #include <net/xfrm.h>
25 #include <net/secure_seq.h>
26
27 #include "ackvec.h"
28 #include "ccid.h"
29 #include "dccp.h"
30 #include "feat.h"
31
32 /*
33 * The per-net dccp.v4_ctl_sk socket is used for responding to
34 * the Out-of-the-blue (OOTB) packets. A control sock will be created
35 * for this socket at the initialization time.
36 */
37
dccp_v4_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)38 int dccp_v4_connect(struct sock *sk, struct sockaddr *uaddr, int addr_len)
39 {
40 const struct sockaddr_in *usin = (struct sockaddr_in *)uaddr;
41 struct inet_sock *inet = inet_sk(sk);
42 struct dccp_sock *dp = dccp_sk(sk);
43 __be16 orig_sport, orig_dport;
44 __be32 daddr, nexthop;
45 struct flowi4 *fl4;
46 struct rtable *rt;
47 int err;
48 struct ip_options_rcu *inet_opt;
49
50 dp->dccps_role = DCCP_ROLE_CLIENT;
51
52 if (addr_len < sizeof(struct sockaddr_in))
53 return -EINVAL;
54
55 if (usin->sin_family != AF_INET)
56 return -EAFNOSUPPORT;
57
58 nexthop = daddr = usin->sin_addr.s_addr;
59
60 inet_opt = rcu_dereference_protected(inet->inet_opt,
61 lockdep_sock_is_held(sk));
62 if (inet_opt != NULL && inet_opt->opt.srr) {
63 if (daddr == 0)
64 return -EINVAL;
65 nexthop = inet_opt->opt.faddr;
66 }
67
68 orig_sport = inet->inet_sport;
69 orig_dport = usin->sin_port;
70 fl4 = &inet->cork.fl.u.ip4;
71 rt = ip_route_connect(fl4, nexthop, inet->inet_saddr,
72 RT_CONN_FLAGS(sk), sk->sk_bound_dev_if,
73 IPPROTO_DCCP,
74 orig_sport, orig_dport, sk);
75 if (IS_ERR(rt))
76 return PTR_ERR(rt);
77
78 if (rt->rt_flags & (RTCF_MULTICAST | RTCF_BROADCAST)) {
79 ip_rt_put(rt);
80 return -ENETUNREACH;
81 }
82
83 if (inet_opt == NULL || !inet_opt->opt.srr)
84 daddr = fl4->daddr;
85
86 if (inet->inet_saddr == 0)
87 inet->inet_saddr = fl4->saddr;
88 sk_rcv_saddr_set(sk, inet->inet_saddr);
89 inet->inet_dport = usin->sin_port;
90 sk_daddr_set(sk, daddr);
91
92 inet_csk(sk)->icsk_ext_hdr_len = 0;
93 if (inet_opt)
94 inet_csk(sk)->icsk_ext_hdr_len = inet_opt->opt.optlen;
95 /*
96 * Socket identity is still unknown (sport may be zero).
97 * However we set state to DCCP_REQUESTING and not releasing socket
98 * lock select source port, enter ourselves into the hash tables and
99 * complete initialization after this.
100 */
101 dccp_set_state(sk, DCCP_REQUESTING);
102 err = inet_hash_connect(&dccp_death_row, sk);
103 if (err != 0)
104 goto failure;
105
106 rt = ip_route_newports(fl4, rt, orig_sport, orig_dport,
107 inet->inet_sport, inet->inet_dport, sk);
108 if (IS_ERR(rt)) {
109 err = PTR_ERR(rt);
110 rt = NULL;
111 goto failure;
112 }
113 /* OK, now commit destination to socket. */
114 sk_setup_caps(sk, &rt->dst);
115
116 dp->dccps_iss = secure_dccp_sequence_number(inet->inet_saddr,
117 inet->inet_daddr,
118 inet->inet_sport,
119 inet->inet_dport);
120 inet->inet_id = prandom_u32();
121
122 err = dccp_connect(sk);
123 rt = NULL;
124 if (err != 0)
125 goto failure;
126 out:
127 return err;
128 failure:
129 /*
130 * This unhashes the socket and releases the local port, if necessary.
131 */
132 dccp_set_state(sk, DCCP_CLOSED);
133 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
134 inet_reset_saddr(sk);
135 ip_rt_put(rt);
136 sk->sk_route_caps = 0;
137 inet->inet_dport = 0;
138 goto out;
139 }
140 EXPORT_SYMBOL_GPL(dccp_v4_connect);
141
142 /*
143 * This routine does path mtu discovery as defined in RFC1191.
144 */
dccp_do_pmtu_discovery(struct sock * sk,const struct iphdr * iph,u32 mtu)145 static inline void dccp_do_pmtu_discovery(struct sock *sk,
146 const struct iphdr *iph,
147 u32 mtu)
148 {
149 struct dst_entry *dst;
150 const struct inet_sock *inet = inet_sk(sk);
151 const struct dccp_sock *dp = dccp_sk(sk);
152
153 /* We are not interested in DCCP_LISTEN and request_socks (RESPONSEs
154 * send out by Linux are always < 576bytes so they should go through
155 * unfragmented).
156 */
157 if (sk->sk_state == DCCP_LISTEN)
158 return;
159
160 dst = inet_csk_update_pmtu(sk, mtu);
161 if (!dst)
162 return;
163
164 /* Something is about to be wrong... Remember soft error
165 * for the case, if this connection will not able to recover.
166 */
167 if (mtu < dst_mtu(dst) && ip_dont_fragment(sk, dst))
168 sk->sk_err_soft = EMSGSIZE;
169
170 mtu = dst_mtu(dst);
171
172 if (inet->pmtudisc != IP_PMTUDISC_DONT &&
173 ip_sk_accept_pmtu(sk) &&
174 inet_csk(sk)->icsk_pmtu_cookie > mtu) {
175 dccp_sync_mss(sk, mtu);
176
177 /*
178 * From RFC 4340, sec. 14.1:
179 *
180 * DCCP-Sync packets are the best choice for upward
181 * probing, since DCCP-Sync probes do not risk application
182 * data loss.
183 */
184 dccp_send_sync(sk, dp->dccps_gsr, DCCP_PKT_SYNC);
185 } /* else let the usual retransmit timer handle it */
186 }
187
dccp_do_redirect(struct sk_buff * skb,struct sock * sk)188 static void dccp_do_redirect(struct sk_buff *skb, struct sock *sk)
189 {
190 struct dst_entry *dst = __sk_dst_check(sk, 0);
191
192 if (dst)
193 dst->ops->redirect(dst, sk, skb);
194 }
195
dccp_req_err(struct sock * sk,u64 seq)196 void dccp_req_err(struct sock *sk, u64 seq)
197 {
198 struct request_sock *req = inet_reqsk(sk);
199 struct net *net = sock_net(sk);
200
201 /*
202 * ICMPs are not backlogged, hence we cannot get an established
203 * socket here.
204 */
205 if (!between48(seq, dccp_rsk(req)->dreq_iss, dccp_rsk(req)->dreq_gss)) {
206 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
207 } else {
208 /*
209 * Still in RESPOND, just remove it silently.
210 * There is no good way to pass the error to the newly
211 * created socket, and POSIX does not want network
212 * errors returned from accept().
213 */
214 inet_csk_reqsk_queue_drop(req->rsk_listener, req);
215 }
216 reqsk_put(req);
217 }
218 EXPORT_SYMBOL(dccp_req_err);
219
220 /*
221 * This routine is called by the ICMP module when it gets some sort of error
222 * condition. If err < 0 then the socket should be closed and the error
223 * returned to the user. If err > 0 it's just the icmp type << 8 | icmp code.
224 * After adjustment header points to the first 8 bytes of the tcp header. We
225 * need to find the appropriate port.
226 *
227 * The locking strategy used here is very "optimistic". When someone else
228 * accesses the socket the ICMP is just dropped and for some paths there is no
229 * check at all. A more general error queue to queue errors for later handling
230 * is probably better.
231 */
dccp_v4_err(struct sk_buff * skb,u32 info)232 static int dccp_v4_err(struct sk_buff *skb, u32 info)
233 {
234 const struct iphdr *iph = (struct iphdr *)skb->data;
235 const u8 offset = iph->ihl << 2;
236 const struct dccp_hdr *dh;
237 struct dccp_sock *dp;
238 struct inet_sock *inet;
239 const int type = icmp_hdr(skb)->type;
240 const int code = icmp_hdr(skb)->code;
241 struct sock *sk;
242 __u64 seq;
243 int err;
244 struct net *net = dev_net(skb->dev);
245
246 if (!pskb_may_pull(skb, offset + sizeof(*dh)))
247 return -EINVAL;
248 dh = (struct dccp_hdr *)(skb->data + offset);
249 if (!pskb_may_pull(skb, offset + __dccp_basic_hdr_len(dh)))
250 return -EINVAL;
251 iph = (struct iphdr *)skb->data;
252 dh = (struct dccp_hdr *)(skb->data + offset);
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), 0);
258 if (!sk) {
259 __ICMP_INC_STATS(net, ICMP_MIB_INERRORS);
260 return -ENOENT;
261 }
262
263 if (sk->sk_state == DCCP_TIME_WAIT) {
264 inet_twsk_put(inet_twsk(sk));
265 return 0;
266 }
267 seq = dccp_hdr_seq(dh);
268 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
269 dccp_req_err(sk, seq);
270 return 0;
271 }
272
273 bh_lock_sock(sk);
274 /* If too many ICMPs get dropped on busy
275 * servers this needs to be solved differently.
276 */
277 if (sock_owned_by_user(sk))
278 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
279
280 if (sk->sk_state == DCCP_CLOSED)
281 goto out;
282
283 dp = dccp_sk(sk);
284 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
285 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
286 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
287 goto out;
288 }
289
290 switch (type) {
291 case ICMP_REDIRECT:
292 if (!sock_owned_by_user(sk))
293 dccp_do_redirect(skb, sk);
294 goto out;
295 case ICMP_SOURCE_QUENCH:
296 /* Just silently ignore these. */
297 goto out;
298 case ICMP_PARAMETERPROB:
299 err = EPROTO;
300 break;
301 case ICMP_DEST_UNREACH:
302 if (code > NR_ICMP_UNREACH)
303 goto out;
304
305 if (code == ICMP_FRAG_NEEDED) { /* PMTU discovery (RFC1191) */
306 if (!sock_owned_by_user(sk))
307 dccp_do_pmtu_discovery(sk, iph, info);
308 goto out;
309 }
310
311 err = icmp_err_convert[code].errno;
312 break;
313 case ICMP_TIME_EXCEEDED:
314 err = EHOSTUNREACH;
315 break;
316 default:
317 goto out;
318 }
319
320 switch (sk->sk_state) {
321 case DCCP_REQUESTING:
322 case DCCP_RESPOND:
323 if (!sock_owned_by_user(sk)) {
324 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
325 sk->sk_err = err;
326
327 sk->sk_error_report(sk);
328
329 dccp_done(sk);
330 } else
331 sk->sk_err_soft = err;
332 goto out;
333 }
334
335 /* If we've already connected we will keep trying
336 * until we time out, or the user gives up.
337 *
338 * rfc1122 4.2.3.9 allows to consider as hard errors
339 * only PROTO_UNREACH and PORT_UNREACH (well, FRAG_FAILED too,
340 * but it is obsoleted by pmtu discovery).
341 *
342 * Note, that in modern internet, where routing is unreliable
343 * and in each dark corner broken firewalls sit, sending random
344 * errors ordered by their masters even this two messages finally lose
345 * their original sense (even Linux sends invalid PORT_UNREACHs)
346 *
347 * Now we are in compliance with RFCs.
348 * --ANK (980905)
349 */
350
351 inet = inet_sk(sk);
352 if (!sock_owned_by_user(sk) && inet->recverr) {
353 sk->sk_err = err;
354 sk->sk_error_report(sk);
355 } else /* Only an error on timeout */
356 sk->sk_err_soft = err;
357 out:
358 bh_unlock_sock(sk);
359 sock_put(sk);
360 return 0;
361 }
362
dccp_v4_csum_finish(struct sk_buff * skb,__be32 src,__be32 dst)363 static 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
dccp_v4_send_check(struct sock * sk,struct sk_buff * skb)369 void 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 EXPORT_SYMBOL_GPL(dccp_v4_send_check);
380
dccp_v4_init_sequence(const struct sk_buff * skb)381 static inline u64 dccp_v4_init_sequence(const struct sk_buff *skb)
382 {
383 return secure_dccp_sequence_number(ip_hdr(skb)->daddr,
384 ip_hdr(skb)->saddr,
385 dccp_hdr(skb)->dccph_dport,
386 dccp_hdr(skb)->dccph_sport);
387 }
388
389 /*
390 * The three way handshake has completed - we got a valid ACK or DATAACK -
391 * now create the new socket.
392 *
393 * This is the equivalent of TCP's tcp_v4_syn_recv_sock
394 */
dccp_v4_request_recv_sock(const struct sock * sk,struct sk_buff * skb,struct request_sock * req,struct dst_entry * dst,struct request_sock * req_unhash,bool * own_req)395 struct sock *dccp_v4_request_recv_sock(const struct sock *sk,
396 struct sk_buff *skb,
397 struct request_sock *req,
398 struct dst_entry *dst,
399 struct request_sock *req_unhash,
400 bool *own_req)
401 {
402 struct inet_request_sock *ireq;
403 struct inet_sock *newinet;
404 struct sock *newsk;
405
406 if (sk_acceptq_is_full(sk))
407 goto exit_overflow;
408
409 newsk = dccp_create_openreq_child(sk, req, skb);
410 if (newsk == NULL)
411 goto exit_nonewsk;
412
413 newinet = inet_sk(newsk);
414 ireq = inet_rsk(req);
415 sk_daddr_set(newsk, ireq->ir_rmt_addr);
416 sk_rcv_saddr_set(newsk, ireq->ir_loc_addr);
417 newinet->inet_saddr = ireq->ir_loc_addr;
418 RCU_INIT_POINTER(newinet->inet_opt, rcu_dereference(ireq->ireq_opt));
419 newinet->mc_index = inet_iif(skb);
420 newinet->mc_ttl = ip_hdr(skb)->ttl;
421 newinet->inet_id = prandom_u32();
422
423 if (dst == NULL && (dst = inet_csk_route_child_sock(sk, newsk, req)) == NULL)
424 goto put_and_exit;
425
426 sk_setup_caps(newsk, dst);
427
428 dccp_sync_mss(newsk, dst_mtu(dst));
429
430 if (__inet_inherit_port(sk, newsk) < 0)
431 goto put_and_exit;
432 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
433 if (*own_req)
434 ireq->ireq_opt = NULL;
435 else
436 newinet->inet_opt = NULL;
437 return newsk;
438
439 exit_overflow:
440 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
441 exit_nonewsk:
442 dst_release(dst);
443 exit:
444 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
445 return NULL;
446 put_and_exit:
447 newinet->inet_opt = NULL;
448 inet_csk_prepare_forced_close(newsk);
449 dccp_done(newsk);
450 goto exit;
451 }
452 EXPORT_SYMBOL_GPL(dccp_v4_request_recv_sock);
453
dccp_v4_route_skb(struct net * net,struct sock * sk,struct sk_buff * skb)454 static struct dst_entry* dccp_v4_route_skb(struct net *net, struct sock *sk,
455 struct sk_buff *skb)
456 {
457 struct rtable *rt;
458 const struct iphdr *iph = ip_hdr(skb);
459 struct flowi4 fl4 = {
460 .flowi4_oif = inet_iif(skb),
461 .daddr = iph->saddr,
462 .saddr = iph->daddr,
463 .flowi4_tos = RT_CONN_FLAGS(sk),
464 .flowi4_proto = sk->sk_protocol,
465 .fl4_sport = dccp_hdr(skb)->dccph_dport,
466 .fl4_dport = dccp_hdr(skb)->dccph_sport,
467 };
468
469 security_skb_classify_flow(skb, flowi4_to_flowi(&fl4));
470 rt = ip_route_output_flow(net, &fl4, sk);
471 if (IS_ERR(rt)) {
472 IP_INC_STATS(net, IPSTATS_MIB_OUTNOROUTES);
473 return NULL;
474 }
475
476 return &rt->dst;
477 }
478
dccp_v4_send_response(const struct sock * sk,struct request_sock * req)479 static int dccp_v4_send_response(const struct sock *sk, struct request_sock *req)
480 {
481 int err = -1;
482 struct sk_buff *skb;
483 struct dst_entry *dst;
484 struct flowi4 fl4;
485
486 dst = inet_csk_route_req(sk, &fl4, req);
487 if (dst == NULL)
488 goto out;
489
490 skb = dccp_make_response(sk, dst, req);
491 if (skb != NULL) {
492 const struct inet_request_sock *ireq = inet_rsk(req);
493 struct dccp_hdr *dh = dccp_hdr(skb);
494
495 dh->dccph_checksum = dccp_v4_csum_finish(skb, ireq->ir_loc_addr,
496 ireq->ir_rmt_addr);
497 rcu_read_lock();
498 err = ip_build_and_send_pkt(skb, sk, ireq->ir_loc_addr,
499 ireq->ir_rmt_addr,
500 rcu_dereference(ireq->ireq_opt));
501 rcu_read_unlock();
502 err = net_xmit_eval(err);
503 }
504
505 out:
506 dst_release(dst);
507 return err;
508 }
509
dccp_v4_ctl_send_reset(const struct sock * sk,struct sk_buff * rxskb)510 static void dccp_v4_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
511 {
512 int err;
513 const struct iphdr *rxiph;
514 struct sk_buff *skb;
515 struct dst_entry *dst;
516 struct net *net = dev_net(skb_dst(rxskb)->dev);
517 struct sock *ctl_sk = net->dccp.v4_ctl_sk;
518
519 /* Never send a reset in response to a reset. */
520 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
521 return;
522
523 if (skb_rtable(rxskb)->rt_type != RTN_LOCAL)
524 return;
525
526 dst = dccp_v4_route_skb(net, ctl_sk, rxskb);
527 if (dst == NULL)
528 return;
529
530 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
531 if (skb == NULL)
532 goto out;
533
534 rxiph = ip_hdr(rxskb);
535 dccp_hdr(skb)->dccph_checksum = dccp_v4_csum_finish(skb, rxiph->saddr,
536 rxiph->daddr);
537 skb_dst_set(skb, dst_clone(dst));
538
539 local_bh_disable();
540 bh_lock_sock(ctl_sk);
541 err = ip_build_and_send_pkt(skb, ctl_sk,
542 rxiph->daddr, rxiph->saddr, NULL);
543 bh_unlock_sock(ctl_sk);
544
545 if (net_xmit_eval(err) == 0) {
546 __DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
547 __DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
548 }
549 local_bh_enable();
550 out:
551 dst_release(dst);
552 }
553
dccp_v4_reqsk_destructor(struct request_sock * req)554 static void dccp_v4_reqsk_destructor(struct request_sock *req)
555 {
556 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
557 kfree(rcu_dereference_protected(inet_rsk(req)->ireq_opt, 1));
558 }
559
dccp_syn_ack_timeout(const struct request_sock * req)560 void dccp_syn_ack_timeout(const struct request_sock *req)
561 {
562 }
563 EXPORT_SYMBOL(dccp_syn_ack_timeout);
564
565 static struct request_sock_ops dccp_request_sock_ops __read_mostly = {
566 .family = PF_INET,
567 .obj_size = sizeof(struct dccp_request_sock),
568 .rtx_syn_ack = dccp_v4_send_response,
569 .send_ack = dccp_reqsk_send_ack,
570 .destructor = dccp_v4_reqsk_destructor,
571 .send_reset = dccp_v4_ctl_send_reset,
572 .syn_ack_timeout = dccp_syn_ack_timeout,
573 };
574
dccp_v4_conn_request(struct sock * sk,struct sk_buff * skb)575 int dccp_v4_conn_request(struct sock *sk, struct sk_buff *skb)
576 {
577 struct inet_request_sock *ireq;
578 struct request_sock *req;
579 struct dccp_request_sock *dreq;
580 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
581 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
582
583 /* Never answer to DCCP_PKT_REQUESTs send to broadcast or multicast */
584 if (skb_rtable(skb)->rt_flags & (RTCF_BROADCAST | RTCF_MULTICAST))
585 return 0; /* discard, don't send a reset here */
586
587 if (dccp_bad_service_code(sk, service)) {
588 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
589 goto drop;
590 }
591 /*
592 * TW buckets are converted to open requests without
593 * limitations, they conserve resources and peer is
594 * evidently real one.
595 */
596 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
597 if (inet_csk_reqsk_queue_is_full(sk))
598 goto drop;
599
600 if (sk_acceptq_is_full(sk))
601 goto drop;
602
603 req = inet_reqsk_alloc(&dccp_request_sock_ops, sk, true);
604 if (req == NULL)
605 goto drop;
606
607 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
608 goto drop_and_free;
609
610 dreq = dccp_rsk(req);
611 if (dccp_parse_options(sk, dreq, skb))
612 goto drop_and_free;
613
614 ireq = inet_rsk(req);
615 sk_rcv_saddr_set(req_to_sk(req), ip_hdr(skb)->daddr);
616 sk_daddr_set(req_to_sk(req), ip_hdr(skb)->saddr);
617 ireq->ir_mark = inet_request_mark(sk, skb);
618 ireq->ireq_family = AF_INET;
619 ireq->ir_iif = sk->sk_bound_dev_if;
620
621 if (security_inet_conn_request(sk, skb, req))
622 goto drop_and_free;
623
624 /*
625 * Step 3: Process LISTEN state
626 *
627 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
628 *
629 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
630 */
631 dreq->dreq_isr = dcb->dccpd_seq;
632 dreq->dreq_gsr = dreq->dreq_isr;
633 dreq->dreq_iss = dccp_v4_init_sequence(skb);
634 dreq->dreq_gss = dreq->dreq_iss;
635 dreq->dreq_service = service;
636
637 if (dccp_v4_send_response(sk, req))
638 goto drop_and_free;
639
640 inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
641 reqsk_put(req);
642 return 0;
643
644 drop_and_free:
645 reqsk_free(req);
646 drop:
647 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
648 return -1;
649 }
650 EXPORT_SYMBOL_GPL(dccp_v4_conn_request);
651
dccp_v4_do_rcv(struct sock * sk,struct sk_buff * skb)652 int dccp_v4_do_rcv(struct sock *sk, struct sk_buff *skb)
653 {
654 struct dccp_hdr *dh = dccp_hdr(skb);
655
656 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
657 if (dccp_rcv_established(sk, skb, dh, skb->len))
658 goto reset;
659 return 0;
660 }
661
662 /*
663 * Step 3: Process LISTEN state
664 * If P.type == Request or P contains a valid Init Cookie option,
665 * (* Must scan the packet's options to check for Init
666 * Cookies. Only Init Cookies are processed here,
667 * however; other options are processed in Step 8. This
668 * scan need only be performed if the endpoint uses Init
669 * Cookies *)
670 * (* Generate a new socket and switch to that socket *)
671 * Set S := new socket for this port pair
672 * S.state = RESPOND
673 * Choose S.ISS (initial seqno) or set from Init Cookies
674 * Initialize S.GAR := S.ISS
675 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
676 * Continue with S.state == RESPOND
677 * (* A Response packet will be generated in Step 11 *)
678 * Otherwise,
679 * Generate Reset(No Connection) unless P.type == Reset
680 * Drop packet and return
681 *
682 * NOTE: the check for the packet types is done in
683 * dccp_rcv_state_process
684 */
685
686 if (dccp_rcv_state_process(sk, skb, dh, skb->len))
687 goto reset;
688 return 0;
689
690 reset:
691 dccp_v4_ctl_send_reset(sk, skb);
692 kfree_skb(skb);
693 return 0;
694 }
695 EXPORT_SYMBOL_GPL(dccp_v4_do_rcv);
696
697 /**
698 * dccp_invalid_packet - check for malformed packets
699 * Implements RFC 4340, 8.5: Step 1: Check header basics
700 * Packets that fail these checks are ignored and do not receive Resets.
701 */
dccp_invalid_packet(struct sk_buff * skb)702 int dccp_invalid_packet(struct sk_buff *skb)
703 {
704 const struct dccp_hdr *dh;
705 unsigned int cscov;
706 u8 dccph_doff;
707
708 if (skb->pkt_type != PACKET_HOST)
709 return 1;
710
711 /* If the packet is shorter than 12 bytes, drop packet and return */
712 if (!pskb_may_pull(skb, sizeof(struct dccp_hdr))) {
713 DCCP_WARN("pskb_may_pull failed\n");
714 return 1;
715 }
716
717 dh = dccp_hdr(skb);
718
719 /* If P.type is not understood, drop packet and return */
720 if (dh->dccph_type >= DCCP_PKT_INVALID) {
721 DCCP_WARN("invalid packet type\n");
722 return 1;
723 }
724
725 /*
726 * If P.Data Offset is too small for packet type, drop packet and return
727 */
728 dccph_doff = dh->dccph_doff;
729 if (dccph_doff < dccp_hdr_len(skb) / sizeof(u32)) {
730 DCCP_WARN("P.Data Offset(%u) too small\n", dccph_doff);
731 return 1;
732 }
733 /*
734 * If P.Data Offset is too too large for packet, drop packet and return
735 */
736 if (!pskb_may_pull(skb, dccph_doff * sizeof(u32))) {
737 DCCP_WARN("P.Data Offset(%u) too large\n", dccph_doff);
738 return 1;
739 }
740 dh = dccp_hdr(skb);
741 /*
742 * If P.type is not Data, Ack, or DataAck and P.X == 0 (the packet
743 * has short sequence numbers), drop packet and return
744 */
745 if ((dh->dccph_type < DCCP_PKT_DATA ||
746 dh->dccph_type > DCCP_PKT_DATAACK) && dh->dccph_x == 0) {
747 DCCP_WARN("P.type (%s) not Data || [Data]Ack, while P.X == 0\n",
748 dccp_packet_name(dh->dccph_type));
749 return 1;
750 }
751
752 /*
753 * If P.CsCov is too large for the packet size, drop packet and return.
754 * This must come _before_ checksumming (not as RFC 4340 suggests).
755 */
756 cscov = dccp_csum_coverage(skb);
757 if (cscov > skb->len) {
758 DCCP_WARN("P.CsCov %u exceeds packet length %d\n",
759 dh->dccph_cscov, skb->len);
760 return 1;
761 }
762
763 /* If header checksum is incorrect, drop packet and return.
764 * (This step is completed in the AF-dependent functions.) */
765 skb->csum = skb_checksum(skb, 0, cscov, 0);
766
767 return 0;
768 }
769 EXPORT_SYMBOL_GPL(dccp_invalid_packet);
770
771 /* this is called when real data arrives */
dccp_v4_rcv(struct sk_buff * skb)772 static int dccp_v4_rcv(struct sk_buff *skb)
773 {
774 const struct dccp_hdr *dh;
775 const struct iphdr *iph;
776 bool refcounted;
777 struct sock *sk;
778 int min_cov;
779
780 /* Step 1: Check header basics */
781
782 if (dccp_invalid_packet(skb))
783 goto discard_it;
784
785 iph = ip_hdr(skb);
786 /* Step 1: If header checksum is incorrect, drop packet and return */
787 if (dccp_v4_csum_finish(skb, iph->saddr, iph->daddr)) {
788 DCCP_WARN("dropped packet with invalid checksum\n");
789 goto discard_it;
790 }
791
792 dh = dccp_hdr(skb);
793
794 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
795 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
796
797 dccp_pr_debug("%8.8s src=%pI4@%-5d dst=%pI4@%-5d seq=%llu",
798 dccp_packet_name(dh->dccph_type),
799 &iph->saddr, ntohs(dh->dccph_sport),
800 &iph->daddr, ntohs(dh->dccph_dport),
801 (unsigned long long) DCCP_SKB_CB(skb)->dccpd_seq);
802
803 if (dccp_packet_without_ack(skb)) {
804 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
805 dccp_pr_debug_cat("\n");
806 } else {
807 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
808 dccp_pr_debug_cat(", ack=%llu\n", (unsigned long long)
809 DCCP_SKB_CB(skb)->dccpd_ack_seq);
810 }
811
812 lookup:
813 sk = __inet_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
814 dh->dccph_sport, dh->dccph_dport, 0, &refcounted);
815 if (!sk) {
816 dccp_pr_debug("failed to look up flow ID in table and "
817 "get corresponding socket\n");
818 goto no_dccp_socket;
819 }
820
821 /*
822 * Step 2:
823 * ... or S.state == TIMEWAIT,
824 * Generate Reset(No Connection) unless P.type == Reset
825 * Drop packet and return
826 */
827 if (sk->sk_state == DCCP_TIME_WAIT) {
828 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
829 inet_twsk_put(inet_twsk(sk));
830 goto no_dccp_socket;
831 }
832
833 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
834 struct request_sock *req = inet_reqsk(sk);
835 struct sock *nsk;
836
837 sk = req->rsk_listener;
838 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
839 inet_csk_reqsk_queue_drop_and_put(sk, req);
840 goto lookup;
841 }
842 sock_hold(sk);
843 refcounted = true;
844 nsk = dccp_check_req(sk, skb, req);
845 if (!nsk) {
846 reqsk_put(req);
847 goto discard_and_relse;
848 }
849 if (nsk == sk) {
850 reqsk_put(req);
851 } else if (dccp_child_process(sk, nsk, skb)) {
852 dccp_v4_ctl_send_reset(sk, skb);
853 goto discard_and_relse;
854 } else {
855 sock_put(sk);
856 return 0;
857 }
858 }
859 /*
860 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
861 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
862 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
863 */
864 min_cov = dccp_sk(sk)->dccps_pcrlen;
865 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
866 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
867 dh->dccph_cscov, min_cov);
868 /* FIXME: "Such packets SHOULD be reported using Data Dropped
869 * options (Section 11.7) with Drop Code 0, Protocol
870 * Constraints." */
871 goto discard_and_relse;
872 }
873
874 if (!xfrm4_policy_check(sk, XFRM_POLICY_IN, skb))
875 goto discard_and_relse;
876 nf_reset_ct(skb);
877
878 return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4, refcounted);
879
880 no_dccp_socket:
881 if (!xfrm4_policy_check(NULL, XFRM_POLICY_IN, skb))
882 goto discard_it;
883 /*
884 * Step 2:
885 * If no socket ...
886 * Generate Reset(No Connection) unless P.type == Reset
887 * Drop packet and return
888 */
889 if (dh->dccph_type != DCCP_PKT_RESET) {
890 DCCP_SKB_CB(skb)->dccpd_reset_code =
891 DCCP_RESET_CODE_NO_CONNECTION;
892 dccp_v4_ctl_send_reset(sk, skb);
893 }
894
895 discard_it:
896 kfree_skb(skb);
897 return 0;
898
899 discard_and_relse:
900 if (refcounted)
901 sock_put(sk);
902 goto discard_it;
903 }
904
905 static const struct inet_connection_sock_af_ops dccp_ipv4_af_ops = {
906 .queue_xmit = ip_queue_xmit,
907 .send_check = dccp_v4_send_check,
908 .rebuild_header = inet_sk_rebuild_header,
909 .conn_request = dccp_v4_conn_request,
910 .syn_recv_sock = dccp_v4_request_recv_sock,
911 .net_header_len = sizeof(struct iphdr),
912 .setsockopt = ip_setsockopt,
913 .getsockopt = ip_getsockopt,
914 .addr2sockaddr = inet_csk_addr2sockaddr,
915 .sockaddr_len = sizeof(struct sockaddr_in),
916 #ifdef CONFIG_COMPAT
917 .compat_setsockopt = compat_ip_setsockopt,
918 .compat_getsockopt = compat_ip_getsockopt,
919 #endif
920 };
921
dccp_v4_init_sock(struct sock * sk)922 static int dccp_v4_init_sock(struct sock *sk)
923 {
924 static __u8 dccp_v4_ctl_sock_initialized;
925 int err = dccp_init_sock(sk, dccp_v4_ctl_sock_initialized);
926
927 if (err == 0) {
928 if (unlikely(!dccp_v4_ctl_sock_initialized))
929 dccp_v4_ctl_sock_initialized = 1;
930 inet_csk(sk)->icsk_af_ops = &dccp_ipv4_af_ops;
931 }
932
933 return err;
934 }
935
936 static struct timewait_sock_ops dccp_timewait_sock_ops = {
937 .twsk_obj_size = sizeof(struct inet_timewait_sock),
938 };
939
940 static struct proto dccp_v4_prot = {
941 .name = "DCCP",
942 .owner = THIS_MODULE,
943 .close = dccp_close,
944 .connect = dccp_v4_connect,
945 .disconnect = dccp_disconnect,
946 .ioctl = dccp_ioctl,
947 .init = dccp_v4_init_sock,
948 .setsockopt = dccp_setsockopt,
949 .getsockopt = dccp_getsockopt,
950 .sendmsg = dccp_sendmsg,
951 .recvmsg = dccp_recvmsg,
952 .backlog_rcv = dccp_v4_do_rcv,
953 .hash = inet_hash,
954 .unhash = inet_unhash,
955 .accept = inet_csk_accept,
956 .get_port = inet_csk_get_port,
957 .shutdown = dccp_shutdown,
958 .destroy = dccp_destroy_sock,
959 .orphan_count = &dccp_orphan_count,
960 .max_header = MAX_DCCP_HEADER,
961 .obj_size = sizeof(struct dccp_sock),
962 .slab_flags = SLAB_TYPESAFE_BY_RCU,
963 .rsk_prot = &dccp_request_sock_ops,
964 .twsk_prot = &dccp_timewait_sock_ops,
965 .h.hashinfo = &dccp_hashinfo,
966 #ifdef CONFIG_COMPAT
967 .compat_setsockopt = compat_dccp_setsockopt,
968 .compat_getsockopt = compat_dccp_getsockopt,
969 #endif
970 };
971
972 static const struct net_protocol dccp_v4_protocol = {
973 .handler = dccp_v4_rcv,
974 .err_handler = dccp_v4_err,
975 .no_policy = 1,
976 .netns_ok = 1,
977 .icmp_strict_tag_validation = 1,
978 };
979
980 static const struct proto_ops inet_dccp_ops = {
981 .family = PF_INET,
982 .owner = THIS_MODULE,
983 .release = inet_release,
984 .bind = inet_bind,
985 .connect = inet_stream_connect,
986 .socketpair = sock_no_socketpair,
987 .accept = inet_accept,
988 .getname = inet_getname,
989 /* FIXME: work on tcp_poll to rename it to inet_csk_poll */
990 .poll = dccp_poll,
991 .ioctl = inet_ioctl,
992 .gettstamp = sock_gettstamp,
993 /* FIXME: work on inet_listen to rename it to sock_common_listen */
994 .listen = inet_dccp_listen,
995 .shutdown = inet_shutdown,
996 .setsockopt = sock_common_setsockopt,
997 .getsockopt = sock_common_getsockopt,
998 .sendmsg = inet_sendmsg,
999 .recvmsg = sock_common_recvmsg,
1000 .mmap = sock_no_mmap,
1001 .sendpage = sock_no_sendpage,
1002 #ifdef CONFIG_COMPAT
1003 .compat_setsockopt = compat_sock_common_setsockopt,
1004 .compat_getsockopt = compat_sock_common_getsockopt,
1005 #endif
1006 };
1007
1008 static struct inet_protosw dccp_v4_protosw = {
1009 .type = SOCK_DCCP,
1010 .protocol = IPPROTO_DCCP,
1011 .prot = &dccp_v4_prot,
1012 .ops = &inet_dccp_ops,
1013 .flags = INET_PROTOSW_ICSK,
1014 };
1015
dccp_v4_init_net(struct net * net)1016 static int __net_init dccp_v4_init_net(struct net *net)
1017 {
1018 if (dccp_hashinfo.bhash == NULL)
1019 return -ESOCKTNOSUPPORT;
1020
1021 return inet_ctl_sock_create(&net->dccp.v4_ctl_sk, PF_INET,
1022 SOCK_DCCP, IPPROTO_DCCP, net);
1023 }
1024
dccp_v4_exit_net(struct net * net)1025 static void __net_exit dccp_v4_exit_net(struct net *net)
1026 {
1027 inet_ctl_sock_destroy(net->dccp.v4_ctl_sk);
1028 }
1029
dccp_v4_exit_batch(struct list_head * net_exit_list)1030 static void __net_exit dccp_v4_exit_batch(struct list_head *net_exit_list)
1031 {
1032 inet_twsk_purge(&dccp_hashinfo, AF_INET);
1033 }
1034
1035 static struct pernet_operations dccp_v4_ops = {
1036 .init = dccp_v4_init_net,
1037 .exit = dccp_v4_exit_net,
1038 .exit_batch = dccp_v4_exit_batch,
1039 };
1040
dccp_v4_init(void)1041 static int __init dccp_v4_init(void)
1042 {
1043 int err = proto_register(&dccp_v4_prot, 1);
1044
1045 if (err)
1046 goto out;
1047
1048 inet_register_protosw(&dccp_v4_protosw);
1049
1050 err = register_pernet_subsys(&dccp_v4_ops);
1051 if (err)
1052 goto out_destroy_ctl_sock;
1053
1054 err = inet_add_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1055 if (err)
1056 goto out_proto_unregister;
1057
1058 out:
1059 return err;
1060 out_proto_unregister:
1061 unregister_pernet_subsys(&dccp_v4_ops);
1062 out_destroy_ctl_sock:
1063 inet_unregister_protosw(&dccp_v4_protosw);
1064 proto_unregister(&dccp_v4_prot);
1065 goto out;
1066 }
1067
dccp_v4_exit(void)1068 static void __exit dccp_v4_exit(void)
1069 {
1070 inet_del_protocol(&dccp_v4_protocol, IPPROTO_DCCP);
1071 unregister_pernet_subsys(&dccp_v4_ops);
1072 inet_unregister_protosw(&dccp_v4_protosw);
1073 proto_unregister(&dccp_v4_prot);
1074 }
1075
1076 module_init(dccp_v4_init);
1077 module_exit(dccp_v4_exit);
1078
1079 /*
1080 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1081 * values directly, Also cover the case where the protocol is not specified,
1082 * i.e. net-pf-PF_INET-proto-0-type-SOCK_DCCP
1083 */
1084 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 33, 6);
1085 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET, 0, 6);
1086 MODULE_LICENSE("GPL");
1087 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1088 MODULE_DESCRIPTION("DCCP - Datagram Congestion Controlled Protocol");
1089