1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /*
3 * DCCP over IPv6
4 * Linux INET6 implementation
5 *
6 * Based on net/dccp6/ipv6.c
7 *
8 * Arnaldo Carvalho de Melo <acme@ghostprotocols.net>
9 */
10
11 #include <linux/module.h>
12 #include <linux/random.h>
13 #include <linux/slab.h>
14 #include <linux/xfrm.h>
15 #include <linux/string.h>
16
17 #include <net/addrconf.h>
18 #include <net/inet_common.h>
19 #include <net/inet_hashtables.h>
20 #include <net/inet_sock.h>
21 #include <net/inet6_connection_sock.h>
22 #include <net/inet6_hashtables.h>
23 #include <net/ip6_route.h>
24 #include <net/ipv6.h>
25 #include <net/protocol.h>
26 #include <net/transp_v6.h>
27 #include <net/ip6_checksum.h>
28 #include <net/xfrm.h>
29 #include <net/secure_seq.h>
30 #include <net/sock.h>
31
32 #include "dccp.h"
33 #include "ipv6.h"
34 #include "feat.h"
35
36 /* The per-net dccp.v6_ctl_sk is used for sending RSTs and ACKs */
37
38 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
39 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
40
41 /* add pseudo-header to DCCP checksum stored in skb->csum */
dccp_v6_csum_finish(struct sk_buff * skb,const struct in6_addr * saddr,const struct in6_addr * daddr)42 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
43 const struct in6_addr *saddr,
44 const struct in6_addr *daddr)
45 {
46 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
47 }
48
dccp_v6_send_check(struct sock * sk,struct sk_buff * skb)49 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
50 {
51 struct ipv6_pinfo *np = inet6_sk(sk);
52 struct dccp_hdr *dh = dccp_hdr(skb);
53
54 dccp_csum_outgoing(skb);
55 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
56 }
57
dccp_v6_init_sequence(struct sk_buff * skb)58 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
59 {
60 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
61 ipv6_hdr(skb)->saddr.s6_addr32,
62 dccp_hdr(skb)->dccph_dport,
63 dccp_hdr(skb)->dccph_sport );
64
65 }
66
dccp_v6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)67 static int dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
68 u8 type, u8 code, int offset, __be32 info)
69 {
70 const struct ipv6hdr *hdr = (const struct ipv6hdr *)skb->data;
71 const struct dccp_hdr *dh;
72 struct dccp_sock *dp;
73 struct ipv6_pinfo *np;
74 struct sock *sk;
75 int err;
76 __u64 seq;
77 struct net *net = dev_net(skb->dev);
78
79 /* Only need dccph_dport & dccph_sport which are the first
80 * 4 bytes in dccp header.
81 * Our caller (icmpv6_notify()) already pulled 8 bytes for us.
82 */
83 BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_sport) > 8);
84 BUILD_BUG_ON(offsetofend(struct dccp_hdr, dccph_dport) > 8);
85 dh = (struct dccp_hdr *)(skb->data + offset);
86
87 sk = __inet6_lookup_established(net, &dccp_hashinfo,
88 &hdr->daddr, dh->dccph_dport,
89 &hdr->saddr, ntohs(dh->dccph_sport),
90 inet6_iif(skb), 0);
91
92 if (!sk) {
93 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
94 ICMP6_MIB_INERRORS);
95 return -ENOENT;
96 }
97
98 if (sk->sk_state == DCCP_TIME_WAIT) {
99 inet_twsk_put(inet_twsk(sk));
100 return 0;
101 }
102 seq = dccp_hdr_seq(dh);
103 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
104 dccp_req_err(sk, seq);
105 return 0;
106 }
107
108 bh_lock_sock(sk);
109 if (sock_owned_by_user(sk))
110 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
111
112 if (sk->sk_state == DCCP_CLOSED)
113 goto out;
114
115 dp = dccp_sk(sk);
116 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
117 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
118 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
119 goto out;
120 }
121
122 np = inet6_sk(sk);
123
124 if (type == NDISC_REDIRECT) {
125 if (!sock_owned_by_user(sk)) {
126 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
127
128 if (dst)
129 dst->ops->redirect(dst, sk, skb);
130 }
131 goto out;
132 }
133
134 if (type == ICMPV6_PKT_TOOBIG) {
135 struct dst_entry *dst = NULL;
136
137 if (!ip6_sk_accept_pmtu(sk))
138 goto out;
139
140 if (sock_owned_by_user(sk))
141 goto out;
142 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
143 goto out;
144
145 dst = inet6_csk_update_pmtu(sk, ntohl(info));
146 if (!dst)
147 goto out;
148
149 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
150 dccp_sync_mss(sk, dst_mtu(dst));
151 goto out;
152 }
153
154 icmpv6_err_convert(type, code, &err);
155
156 /* Might be for an request_sock */
157 switch (sk->sk_state) {
158 case DCCP_REQUESTING:
159 case DCCP_RESPOND: /* Cannot happen.
160 It can, it SYNs are crossed. --ANK */
161 if (!sock_owned_by_user(sk)) {
162 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
163 sk->sk_err = err;
164 /*
165 * Wake people up to see the error
166 * (see connect in sock.c)
167 */
168 sk->sk_error_report(sk);
169 dccp_done(sk);
170 } else
171 sk->sk_err_soft = err;
172 goto out;
173 }
174
175 if (!sock_owned_by_user(sk) && np->recverr) {
176 sk->sk_err = err;
177 sk->sk_error_report(sk);
178 } else
179 sk->sk_err_soft = err;
180
181 out:
182 bh_unlock_sock(sk);
183 sock_put(sk);
184 return 0;
185 }
186
187
dccp_v6_send_response(const struct sock * sk,struct request_sock * req)188 static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
189 {
190 struct inet_request_sock *ireq = inet_rsk(req);
191 struct ipv6_pinfo *np = inet6_sk(sk);
192 struct sk_buff *skb;
193 struct in6_addr *final_p, final;
194 struct flowi6 fl6;
195 int err = -1;
196 struct dst_entry *dst;
197
198 memset(&fl6, 0, sizeof(fl6));
199 fl6.flowi6_proto = IPPROTO_DCCP;
200 fl6.daddr = ireq->ir_v6_rmt_addr;
201 fl6.saddr = ireq->ir_v6_loc_addr;
202 fl6.flowlabel = 0;
203 fl6.flowi6_oif = ireq->ir_iif;
204 fl6.fl6_dport = ireq->ir_rmt_port;
205 fl6.fl6_sport = htons(ireq->ir_num);
206 security_req_classify_flow(req, flowi6_to_flowi_common(&fl6));
207
208
209 rcu_read_lock();
210 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
211 rcu_read_unlock();
212
213 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
214 if (IS_ERR(dst)) {
215 err = PTR_ERR(dst);
216 dst = NULL;
217 goto done;
218 }
219
220 skb = dccp_make_response(sk, dst, req);
221 if (skb != NULL) {
222 struct dccp_hdr *dh = dccp_hdr(skb);
223 struct ipv6_txoptions *opt;
224
225 dh->dccph_checksum = dccp_v6_csum_finish(skb,
226 &ireq->ir_v6_loc_addr,
227 &ireq->ir_v6_rmt_addr);
228 fl6.daddr = ireq->ir_v6_rmt_addr;
229 rcu_read_lock();
230 opt = ireq->ipv6_opt;
231 if (!opt)
232 opt = rcu_dereference(np->opt);
233 err = ip6_xmit(sk, skb, &fl6, sk->sk_mark, opt, np->tclass,
234 sk->sk_priority);
235 rcu_read_unlock();
236 err = net_xmit_eval(err);
237 }
238
239 done:
240 dst_release(dst);
241 return err;
242 }
243
dccp_v6_reqsk_destructor(struct request_sock * req)244 static void dccp_v6_reqsk_destructor(struct request_sock *req)
245 {
246 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
247 kfree(inet_rsk(req)->ipv6_opt);
248 kfree_skb(inet_rsk(req)->pktopts);
249 }
250
dccp_v6_ctl_send_reset(const struct sock * sk,struct sk_buff * rxskb)251 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
252 {
253 const struct ipv6hdr *rxip6h;
254 struct sk_buff *skb;
255 struct flowi6 fl6;
256 struct net *net = dev_net(skb_dst(rxskb)->dev);
257 struct sock *ctl_sk = net->dccp.v6_ctl_sk;
258 struct dst_entry *dst;
259
260 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
261 return;
262
263 if (!ipv6_unicast_destination(rxskb))
264 return;
265
266 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
267 if (skb == NULL)
268 return;
269
270 rxip6h = ipv6_hdr(rxskb);
271 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
272 &rxip6h->daddr);
273
274 memset(&fl6, 0, sizeof(fl6));
275 fl6.daddr = rxip6h->saddr;
276 fl6.saddr = rxip6h->daddr;
277
278 fl6.flowi6_proto = IPPROTO_DCCP;
279 fl6.flowi6_oif = inet6_iif(rxskb);
280 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
281 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
282 security_skb_classify_flow(rxskb, flowi6_to_flowi_common(&fl6));
283
284 /* sk = NULL, but it is safe for now. RST socket required. */
285 dst = ip6_dst_lookup_flow(sock_net(ctl_sk), ctl_sk, &fl6, NULL);
286 if (!IS_ERR(dst)) {
287 skb_dst_set(skb, dst);
288 ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0, 0);
289 DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
290 DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
291 return;
292 }
293
294 kfree_skb(skb);
295 }
296
297 static struct request_sock_ops dccp6_request_sock_ops = {
298 .family = AF_INET6,
299 .obj_size = sizeof(struct dccp6_request_sock),
300 .rtx_syn_ack = dccp_v6_send_response,
301 .send_ack = dccp_reqsk_send_ack,
302 .destructor = dccp_v6_reqsk_destructor,
303 .send_reset = dccp_v6_ctl_send_reset,
304 .syn_ack_timeout = dccp_syn_ack_timeout,
305 };
306
dccp_v6_conn_request(struct sock * sk,struct sk_buff * skb)307 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
308 {
309 struct request_sock *req;
310 struct dccp_request_sock *dreq;
311 struct inet_request_sock *ireq;
312 struct ipv6_pinfo *np = inet6_sk(sk);
313 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
314 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
315
316 if (skb->protocol == htons(ETH_P_IP))
317 return dccp_v4_conn_request(sk, skb);
318
319 if (!ipv6_unicast_destination(skb))
320 return 0; /* discard, don't send a reset here */
321
322 if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) {
323 __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS);
324 return 0;
325 }
326
327 if (dccp_bad_service_code(sk, service)) {
328 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
329 goto drop;
330 }
331 /*
332 * There are no SYN attacks on IPv6, yet...
333 */
334 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
335 if (inet_csk_reqsk_queue_is_full(sk))
336 goto drop;
337
338 if (sk_acceptq_is_full(sk))
339 goto drop;
340
341 req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
342 if (req == NULL)
343 goto drop;
344
345 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
346 goto drop_and_free;
347
348 dreq = dccp_rsk(req);
349 if (dccp_parse_options(sk, dreq, skb))
350 goto drop_and_free;
351
352 if (security_inet_conn_request(sk, skb, req))
353 goto drop_and_free;
354
355 ireq = inet_rsk(req);
356 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
357 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
358 ireq->ireq_family = AF_INET6;
359 ireq->ir_mark = inet_request_mark(sk, skb);
360
361 if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
362 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
363 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
364 refcount_inc(&skb->users);
365 ireq->pktopts = skb;
366 }
367 ireq->ir_iif = sk->sk_bound_dev_if;
368
369 /* So that link locals have meaning */
370 if (!sk->sk_bound_dev_if &&
371 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
372 ireq->ir_iif = inet6_iif(skb);
373
374 /*
375 * Step 3: Process LISTEN state
376 *
377 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
378 *
379 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
380 */
381 dreq->dreq_isr = dcb->dccpd_seq;
382 dreq->dreq_gsr = dreq->dreq_isr;
383 dreq->dreq_iss = dccp_v6_init_sequence(skb);
384 dreq->dreq_gss = dreq->dreq_iss;
385 dreq->dreq_service = service;
386
387 if (dccp_v6_send_response(sk, req))
388 goto drop_and_free;
389
390 inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
391 reqsk_put(req);
392 return 0;
393
394 drop_and_free:
395 reqsk_free(req);
396 drop:
397 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
398 return -1;
399 }
400
dccp_v6_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)401 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
402 struct sk_buff *skb,
403 struct request_sock *req,
404 struct dst_entry *dst,
405 struct request_sock *req_unhash,
406 bool *own_req)
407 {
408 struct inet_request_sock *ireq = inet_rsk(req);
409 struct ipv6_pinfo *newnp;
410 const struct ipv6_pinfo *np = inet6_sk(sk);
411 struct ipv6_txoptions *opt;
412 struct inet_sock *newinet;
413 struct dccp6_sock *newdp6;
414 struct sock *newsk;
415
416 if (skb->protocol == htons(ETH_P_IP)) {
417 /*
418 * v6 mapped
419 */
420 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
421 req_unhash, own_req);
422 if (newsk == NULL)
423 return NULL;
424
425 newdp6 = (struct dccp6_sock *)newsk;
426 newinet = inet_sk(newsk);
427 newinet->pinet6 = &newdp6->inet6;
428 newnp = inet6_sk(newsk);
429
430 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
431
432 newnp->saddr = newsk->sk_v6_rcv_saddr;
433
434 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
435 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
436 newnp->pktoptions = NULL;
437 newnp->opt = NULL;
438 newnp->ipv6_mc_list = NULL;
439 newnp->ipv6_ac_list = NULL;
440 newnp->ipv6_fl_list = NULL;
441 newnp->mcast_oif = inet_iif(skb);
442 newnp->mcast_hops = ip_hdr(skb)->ttl;
443
444 /*
445 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
446 * here, dccp_create_openreq_child now does this for us, see the comment in
447 * that function for the gory details. -acme
448 */
449
450 /* It is tricky place. Until this moment IPv4 tcp
451 worked with IPv6 icsk.icsk_af_ops.
452 Sync it now.
453 */
454 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
455
456 return newsk;
457 }
458
459
460 if (sk_acceptq_is_full(sk))
461 goto out_overflow;
462
463 if (!dst) {
464 struct flowi6 fl6;
465
466 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
467 if (!dst)
468 goto out;
469 }
470
471 newsk = dccp_create_openreq_child(sk, req, skb);
472 if (newsk == NULL)
473 goto out_nonewsk;
474
475 /*
476 * No need to charge this sock to the relevant IPv6 refcnt debug socks
477 * count here, dccp_create_openreq_child now does this for us, see the
478 * comment in that function for the gory details. -acme
479 */
480
481 ip6_dst_store(newsk, dst, NULL, NULL);
482 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
483 NETIF_F_TSO);
484 newdp6 = (struct dccp6_sock *)newsk;
485 newinet = inet_sk(newsk);
486 newinet->pinet6 = &newdp6->inet6;
487 newnp = inet6_sk(newsk);
488
489 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
490
491 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
492 newnp->saddr = ireq->ir_v6_loc_addr;
493 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
494 newsk->sk_bound_dev_if = ireq->ir_iif;
495
496 /* Now IPv6 options...
497
498 First: no IPv4 options.
499 */
500 newinet->inet_opt = NULL;
501
502 /* Clone RX bits */
503 newnp->rxopt.all = np->rxopt.all;
504
505 newnp->ipv6_mc_list = NULL;
506 newnp->ipv6_ac_list = NULL;
507 newnp->ipv6_fl_list = NULL;
508 newnp->pktoptions = NULL;
509 newnp->opt = NULL;
510 newnp->mcast_oif = inet6_iif(skb);
511 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
512
513 /*
514 * Clone native IPv6 options from listening socket (if any)
515 *
516 * Yes, keeping reference count would be much more clever, but we make
517 * one more one thing there: reattach optmem to newsk.
518 */
519 opt = ireq->ipv6_opt;
520 if (!opt)
521 opt = rcu_dereference(np->opt);
522 if (opt) {
523 opt = ipv6_dup_options(newsk, opt);
524 RCU_INIT_POINTER(newnp->opt, opt);
525 }
526 inet_csk(newsk)->icsk_ext_hdr_len = 0;
527 if (opt)
528 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
529 opt->opt_flen;
530
531 dccp_sync_mss(newsk, dst_mtu(dst));
532
533 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
534 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
535
536 if (__inet_inherit_port(sk, newsk) < 0) {
537 inet_csk_prepare_forced_close(newsk);
538 dccp_done(newsk);
539 goto out;
540 }
541 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
542 /* Clone pktoptions received with SYN, if we own the req */
543 if (*own_req && ireq->pktopts) {
544 newnp->pktoptions = skb_clone(ireq->pktopts, GFP_ATOMIC);
545 consume_skb(ireq->pktopts);
546 ireq->pktopts = NULL;
547 if (newnp->pktoptions)
548 skb_set_owner_r(newnp->pktoptions, newsk);
549 }
550
551 return newsk;
552
553 out_overflow:
554 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
555 out_nonewsk:
556 dst_release(dst);
557 out:
558 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
559 return NULL;
560 }
561
562 /* The socket must have it's spinlock held when we get
563 * here.
564 *
565 * We have a potential double-lock case here, so even when
566 * doing backlog processing we use the BH locking scheme.
567 * This is because we cannot sleep with the original spinlock
568 * held.
569 */
dccp_v6_do_rcv(struct sock * sk,struct sk_buff * skb)570 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
571 {
572 struct ipv6_pinfo *np = inet6_sk(sk);
573 struct sk_buff *opt_skb = NULL;
574
575 /* Imagine: socket is IPv6. IPv4 packet arrives,
576 goes to IPv4 receive handler and backlogged.
577 From backlog it always goes here. Kerboom...
578 Fortunately, dccp_rcv_established and rcv_established
579 handle them correctly, but it is not case with
580 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
581 */
582
583 if (skb->protocol == htons(ETH_P_IP))
584 return dccp_v4_do_rcv(sk, skb);
585
586 if (sk_filter(sk, skb))
587 goto discard;
588
589 /*
590 * socket locking is here for SMP purposes as backlog rcv is currently
591 * called with bh processing disabled.
592 */
593
594 /* Do Stevens' IPV6_PKTOPTIONS.
595
596 Yes, guys, it is the only place in our code, where we
597 may make it not affecting IPv4.
598 The rest of code is protocol independent,
599 and I do not like idea to uglify IPv4.
600
601 Actually, all the idea behind IPV6_PKTOPTIONS
602 looks not very well thought. For now we latch
603 options, received in the last packet, enqueued
604 by tcp. Feel free to propose better solution.
605 --ANK (980728)
606 */
607 if (np->rxopt.all)
608 opt_skb = skb_clone(skb, GFP_ATOMIC);
609
610 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
611 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
612 goto reset;
613 if (opt_skb)
614 goto ipv6_pktoptions;
615 return 0;
616 }
617
618 /*
619 * Step 3: Process LISTEN state
620 * If S.state == LISTEN,
621 * If P.type == Request or P contains a valid Init Cookie option,
622 * (* Must scan the packet's options to check for Init
623 * Cookies. Only Init Cookies are processed here,
624 * however; other options are processed in Step 8. This
625 * scan need only be performed if the endpoint uses Init
626 * Cookies *)
627 * (* Generate a new socket and switch to that socket *)
628 * Set S := new socket for this port pair
629 * S.state = RESPOND
630 * Choose S.ISS (initial seqno) or set from Init Cookies
631 * Initialize S.GAR := S.ISS
632 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
633 * Continue with S.state == RESPOND
634 * (* A Response packet will be generated in Step 11 *)
635 * Otherwise,
636 * Generate Reset(No Connection) unless P.type == Reset
637 * Drop packet and return
638 *
639 * NOTE: the check for the packet types is done in
640 * dccp_rcv_state_process
641 */
642
643 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
644 goto reset;
645 if (opt_skb)
646 goto ipv6_pktoptions;
647 return 0;
648
649 reset:
650 dccp_v6_ctl_send_reset(sk, skb);
651 discard:
652 if (opt_skb != NULL)
653 __kfree_skb(opt_skb);
654 kfree_skb(skb);
655 return 0;
656
657 /* Handling IPV6_PKTOPTIONS skb the similar
658 * way it's done for net/ipv6/tcp_ipv6.c
659 */
660 ipv6_pktoptions:
661 if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
662 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
663 np->mcast_oif = inet6_iif(opt_skb);
664 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
665 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
666 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
667 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
668 if (np->repflow)
669 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
670 if (ipv6_opt_accepted(sk, opt_skb,
671 &DCCP_SKB_CB(opt_skb)->header.h6)) {
672 skb_set_owner_r(opt_skb, sk);
673 memmove(IP6CB(opt_skb),
674 &DCCP_SKB_CB(opt_skb)->header.h6,
675 sizeof(struct inet6_skb_parm));
676 opt_skb = xchg(&np->pktoptions, opt_skb);
677 } else {
678 __kfree_skb(opt_skb);
679 opt_skb = xchg(&np->pktoptions, NULL);
680 }
681 }
682
683 kfree_skb(opt_skb);
684 return 0;
685 }
686
dccp_v6_rcv(struct sk_buff * skb)687 static int dccp_v6_rcv(struct sk_buff *skb)
688 {
689 const struct dccp_hdr *dh;
690 bool refcounted;
691 struct sock *sk;
692 int min_cov;
693
694 /* Step 1: Check header basics */
695
696 if (dccp_invalid_packet(skb))
697 goto discard_it;
698
699 /* Step 1: If header checksum is incorrect, drop packet and return. */
700 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
701 &ipv6_hdr(skb)->daddr)) {
702 DCCP_WARN("dropped packet with invalid checksum\n");
703 goto discard_it;
704 }
705
706 dh = dccp_hdr(skb);
707
708 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
709 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
710
711 if (dccp_packet_without_ack(skb))
712 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
713 else
714 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
715
716 lookup:
717 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
718 dh->dccph_sport, dh->dccph_dport,
719 inet6_iif(skb), 0, &refcounted);
720 if (!sk) {
721 dccp_pr_debug("failed to look up flow ID in table and "
722 "get corresponding socket\n");
723 goto no_dccp_socket;
724 }
725
726 /*
727 * Step 2:
728 * ... or S.state == TIMEWAIT,
729 * Generate Reset(No Connection) unless P.type == Reset
730 * Drop packet and return
731 */
732 if (sk->sk_state == DCCP_TIME_WAIT) {
733 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
734 inet_twsk_put(inet_twsk(sk));
735 goto no_dccp_socket;
736 }
737
738 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
739 struct request_sock *req = inet_reqsk(sk);
740 struct sock *nsk;
741
742 sk = req->rsk_listener;
743 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
744 inet_csk_reqsk_queue_drop_and_put(sk, req);
745 goto lookup;
746 }
747 sock_hold(sk);
748 refcounted = true;
749 nsk = dccp_check_req(sk, skb, req);
750 if (!nsk) {
751 reqsk_put(req);
752 goto discard_and_relse;
753 }
754 if (nsk == sk) {
755 reqsk_put(req);
756 } else if (dccp_child_process(sk, nsk, skb)) {
757 dccp_v6_ctl_send_reset(sk, skb);
758 goto discard_and_relse;
759 } else {
760 sock_put(sk);
761 return 0;
762 }
763 }
764 /*
765 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
766 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
767 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
768 */
769 min_cov = dccp_sk(sk)->dccps_pcrlen;
770 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
771 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
772 dh->dccph_cscov, min_cov);
773 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
774 goto discard_and_relse;
775 }
776
777 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
778 goto discard_and_relse;
779
780 return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
781 refcounted) ? -1 : 0;
782
783 no_dccp_socket:
784 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
785 goto discard_it;
786 /*
787 * Step 2:
788 * If no socket ...
789 * Generate Reset(No Connection) unless P.type == Reset
790 * Drop packet and return
791 */
792 if (dh->dccph_type != DCCP_PKT_RESET) {
793 DCCP_SKB_CB(skb)->dccpd_reset_code =
794 DCCP_RESET_CODE_NO_CONNECTION;
795 dccp_v6_ctl_send_reset(sk, skb);
796 }
797
798 discard_it:
799 kfree_skb(skb);
800 return 0;
801
802 discard_and_relse:
803 if (refcounted)
804 sock_put(sk);
805 goto discard_it;
806 }
807
dccp_v6_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)808 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
809 int addr_len)
810 {
811 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
812 struct inet_connection_sock *icsk = inet_csk(sk);
813 struct inet_sock *inet = inet_sk(sk);
814 struct ipv6_pinfo *np = inet6_sk(sk);
815 struct dccp_sock *dp = dccp_sk(sk);
816 struct in6_addr *saddr = NULL, *final_p, final;
817 struct ipv6_txoptions *opt;
818 struct flowi6 fl6;
819 struct dst_entry *dst;
820 int addr_type;
821 int err;
822
823 dp->dccps_role = DCCP_ROLE_CLIENT;
824
825 if (addr_len < SIN6_LEN_RFC2133)
826 return -EINVAL;
827
828 if (usin->sin6_family != AF_INET6)
829 return -EAFNOSUPPORT;
830
831 memset(&fl6, 0, sizeof(fl6));
832
833 if (np->sndflow) {
834 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
835 IP6_ECN_flow_init(fl6.flowlabel);
836 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
837 struct ip6_flowlabel *flowlabel;
838 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
839 if (IS_ERR(flowlabel))
840 return -EINVAL;
841 fl6_sock_release(flowlabel);
842 }
843 }
844 /*
845 * connect() to INADDR_ANY means loopback (BSD'ism).
846 */
847 if (ipv6_addr_any(&usin->sin6_addr))
848 usin->sin6_addr.s6_addr[15] = 1;
849
850 addr_type = ipv6_addr_type(&usin->sin6_addr);
851
852 if (addr_type & IPV6_ADDR_MULTICAST)
853 return -ENETUNREACH;
854
855 if (addr_type & IPV6_ADDR_LINKLOCAL) {
856 if (addr_len >= sizeof(struct sockaddr_in6) &&
857 usin->sin6_scope_id) {
858 /* If interface is set while binding, indices
859 * must coincide.
860 */
861 if (sk->sk_bound_dev_if &&
862 sk->sk_bound_dev_if != usin->sin6_scope_id)
863 return -EINVAL;
864
865 sk->sk_bound_dev_if = usin->sin6_scope_id;
866 }
867
868 /* Connect to link-local address requires an interface */
869 if (!sk->sk_bound_dev_if)
870 return -EINVAL;
871 }
872
873 sk->sk_v6_daddr = usin->sin6_addr;
874 np->flow_label = fl6.flowlabel;
875
876 /*
877 * DCCP over IPv4
878 */
879 if (addr_type == IPV6_ADDR_MAPPED) {
880 u32 exthdrlen = icsk->icsk_ext_hdr_len;
881 struct sockaddr_in sin;
882
883 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
884
885 if (__ipv6_only_sock(sk))
886 return -ENETUNREACH;
887
888 sin.sin_family = AF_INET;
889 sin.sin_port = usin->sin6_port;
890 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
891
892 icsk->icsk_af_ops = &dccp_ipv6_mapped;
893 sk->sk_backlog_rcv = dccp_v4_do_rcv;
894
895 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
896 if (err) {
897 icsk->icsk_ext_hdr_len = exthdrlen;
898 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
899 sk->sk_backlog_rcv = dccp_v6_do_rcv;
900 goto failure;
901 }
902 np->saddr = sk->sk_v6_rcv_saddr;
903 return err;
904 }
905
906 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
907 saddr = &sk->sk_v6_rcv_saddr;
908
909 fl6.flowi6_proto = IPPROTO_DCCP;
910 fl6.daddr = sk->sk_v6_daddr;
911 fl6.saddr = saddr ? *saddr : np->saddr;
912 fl6.flowi6_oif = sk->sk_bound_dev_if;
913 fl6.fl6_dport = usin->sin6_port;
914 fl6.fl6_sport = inet->inet_sport;
915 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
916
917 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
918 final_p = fl6_update_dst(&fl6, opt, &final);
919
920 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
921 if (IS_ERR(dst)) {
922 err = PTR_ERR(dst);
923 goto failure;
924 }
925
926 if (saddr == NULL) {
927 saddr = &fl6.saddr;
928 sk->sk_v6_rcv_saddr = *saddr;
929 }
930
931 /* set the source address */
932 np->saddr = *saddr;
933 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
934
935 ip6_dst_store(sk, dst, NULL, NULL);
936
937 icsk->icsk_ext_hdr_len = 0;
938 if (opt)
939 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
940
941 inet->inet_dport = usin->sin6_port;
942
943 dccp_set_state(sk, DCCP_REQUESTING);
944 err = inet6_hash_connect(&dccp_death_row, sk);
945 if (err)
946 goto late_failure;
947
948 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
949 sk->sk_v6_daddr.s6_addr32,
950 inet->inet_sport,
951 inet->inet_dport);
952 err = dccp_connect(sk);
953 if (err)
954 goto late_failure;
955
956 return 0;
957
958 late_failure:
959 dccp_set_state(sk, DCCP_CLOSED);
960 if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK))
961 inet_reset_saddr(sk);
962 __sk_dst_reset(sk);
963 failure:
964 inet->inet_dport = 0;
965 sk->sk_route_caps = 0;
966 return err;
967 }
968
969 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
970 .queue_xmit = inet6_csk_xmit,
971 .send_check = dccp_v6_send_check,
972 .rebuild_header = inet6_sk_rebuild_header,
973 .conn_request = dccp_v6_conn_request,
974 .syn_recv_sock = dccp_v6_request_recv_sock,
975 .net_header_len = sizeof(struct ipv6hdr),
976 .setsockopt = ipv6_setsockopt,
977 .getsockopt = ipv6_getsockopt,
978 .addr2sockaddr = inet6_csk_addr2sockaddr,
979 .sockaddr_len = sizeof(struct sockaddr_in6),
980 };
981
982 /*
983 * DCCP over IPv4 via INET6 API
984 */
985 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
986 .queue_xmit = ip_queue_xmit,
987 .send_check = dccp_v4_send_check,
988 .rebuild_header = inet_sk_rebuild_header,
989 .conn_request = dccp_v6_conn_request,
990 .syn_recv_sock = dccp_v6_request_recv_sock,
991 .net_header_len = sizeof(struct iphdr),
992 .setsockopt = ipv6_setsockopt,
993 .getsockopt = ipv6_getsockopt,
994 .addr2sockaddr = inet6_csk_addr2sockaddr,
995 .sockaddr_len = sizeof(struct sockaddr_in6),
996 };
997
998 /* NOTE: A lot of things set to zero explicitly by call to
999 * sk_alloc() so need not be done here.
1000 */
dccp_v6_init_sock(struct sock * sk)1001 static int dccp_v6_init_sock(struct sock *sk)
1002 {
1003 static __u8 dccp_v6_ctl_sock_initialized;
1004 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1005
1006 if (err == 0) {
1007 if (unlikely(!dccp_v6_ctl_sock_initialized))
1008 dccp_v6_ctl_sock_initialized = 1;
1009 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1010 }
1011
1012 return err;
1013 }
1014
dccp_v6_destroy_sock(struct sock * sk)1015 static void dccp_v6_destroy_sock(struct sock *sk)
1016 {
1017 dccp_destroy_sock(sk);
1018 inet6_destroy_sock(sk);
1019 }
1020
1021 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1022 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1023 };
1024
1025 static struct proto dccp_v6_prot = {
1026 .name = "DCCPv6",
1027 .owner = THIS_MODULE,
1028 .close = dccp_close,
1029 .connect = dccp_v6_connect,
1030 .disconnect = dccp_disconnect,
1031 .ioctl = dccp_ioctl,
1032 .init = dccp_v6_init_sock,
1033 .setsockopt = dccp_setsockopt,
1034 .getsockopt = dccp_getsockopt,
1035 .sendmsg = dccp_sendmsg,
1036 .recvmsg = dccp_recvmsg,
1037 .backlog_rcv = dccp_v6_do_rcv,
1038 .hash = inet6_hash,
1039 .unhash = inet_unhash,
1040 .accept = inet_csk_accept,
1041 .get_port = inet_csk_get_port,
1042 .shutdown = dccp_shutdown,
1043 .destroy = dccp_v6_destroy_sock,
1044 .orphan_count = &dccp_orphan_count,
1045 .max_header = MAX_DCCP_HEADER,
1046 .obj_size = sizeof(struct dccp6_sock),
1047 .slab_flags = SLAB_TYPESAFE_BY_RCU,
1048 .rsk_prot = &dccp6_request_sock_ops,
1049 .twsk_prot = &dccp6_timewait_sock_ops,
1050 .h.hashinfo = &dccp_hashinfo,
1051 };
1052
1053 static const struct inet6_protocol dccp_v6_protocol = {
1054 .handler = dccp_v6_rcv,
1055 .err_handler = dccp_v6_err,
1056 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1057 };
1058
1059 static const struct proto_ops inet6_dccp_ops = {
1060 .family = PF_INET6,
1061 .owner = THIS_MODULE,
1062 .release = inet6_release,
1063 .bind = inet6_bind,
1064 .connect = inet_stream_connect,
1065 .socketpair = sock_no_socketpair,
1066 .accept = inet_accept,
1067 .getname = inet6_getname,
1068 .poll = dccp_poll,
1069 .ioctl = inet6_ioctl,
1070 .gettstamp = sock_gettstamp,
1071 .listen = inet_dccp_listen,
1072 .shutdown = inet_shutdown,
1073 .setsockopt = sock_common_setsockopt,
1074 .getsockopt = sock_common_getsockopt,
1075 .sendmsg = inet_sendmsg,
1076 .recvmsg = sock_common_recvmsg,
1077 .mmap = sock_no_mmap,
1078 .sendpage = sock_no_sendpage,
1079 #ifdef CONFIG_COMPAT
1080 .compat_ioctl = inet6_compat_ioctl,
1081 #endif
1082 };
1083
1084 static struct inet_protosw dccp_v6_protosw = {
1085 .type = SOCK_DCCP,
1086 .protocol = IPPROTO_DCCP,
1087 .prot = &dccp_v6_prot,
1088 .ops = &inet6_dccp_ops,
1089 .flags = INET_PROTOSW_ICSK,
1090 };
1091
dccp_v6_init_net(struct net * net)1092 static int __net_init dccp_v6_init_net(struct net *net)
1093 {
1094 if (dccp_hashinfo.bhash == NULL)
1095 return -ESOCKTNOSUPPORT;
1096
1097 return inet_ctl_sock_create(&net->dccp.v6_ctl_sk, PF_INET6,
1098 SOCK_DCCP, IPPROTO_DCCP, net);
1099 }
1100
dccp_v6_exit_net(struct net * net)1101 static void __net_exit dccp_v6_exit_net(struct net *net)
1102 {
1103 inet_ctl_sock_destroy(net->dccp.v6_ctl_sk);
1104 }
1105
dccp_v6_exit_batch(struct list_head * net_exit_list)1106 static void __net_exit dccp_v6_exit_batch(struct list_head *net_exit_list)
1107 {
1108 inet_twsk_purge(&dccp_hashinfo, AF_INET6);
1109 }
1110
1111 static struct pernet_operations dccp_v6_ops = {
1112 .init = dccp_v6_init_net,
1113 .exit = dccp_v6_exit_net,
1114 .exit_batch = dccp_v6_exit_batch,
1115 };
1116
dccp_v6_init(void)1117 static int __init dccp_v6_init(void)
1118 {
1119 int err = proto_register(&dccp_v6_prot, 1);
1120
1121 if (err)
1122 goto out;
1123
1124 inet6_register_protosw(&dccp_v6_protosw);
1125
1126 err = register_pernet_subsys(&dccp_v6_ops);
1127 if (err)
1128 goto out_destroy_ctl_sock;
1129
1130 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1131 if (err)
1132 goto out_unregister_proto;
1133
1134 out:
1135 return err;
1136 out_unregister_proto:
1137 unregister_pernet_subsys(&dccp_v6_ops);
1138 out_destroy_ctl_sock:
1139 inet6_unregister_protosw(&dccp_v6_protosw);
1140 proto_unregister(&dccp_v6_prot);
1141 goto out;
1142 }
1143
dccp_v6_exit(void)1144 static void __exit dccp_v6_exit(void)
1145 {
1146 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1147 unregister_pernet_subsys(&dccp_v6_ops);
1148 inet6_unregister_protosw(&dccp_v6_protosw);
1149 proto_unregister(&dccp_v6_prot);
1150 }
1151
1152 module_init(dccp_v6_init);
1153 module_exit(dccp_v6_exit);
1154
1155 /*
1156 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1157 * values directly, Also cover the case where the protocol is not specified,
1158 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1159 */
1160 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1161 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1162 MODULE_LICENSE("GPL");
1163 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1164 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
1165