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/netns/generic.h>
31 #include <net/sock.h>
32
33 #include "dccp.h"
34 #include "ipv6.h"
35 #include "feat.h"
36
37 struct dccp_v6_pernet {
38 struct sock *v6_ctl_sk;
39 };
40
41 static unsigned int dccp_v6_pernet_id __read_mostly;
42
43 /* The per-net v6_ctl_sk is used for sending RSTs and ACKs */
44
45 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped;
46 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops;
47
48 /* 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)49 static inline __sum16 dccp_v6_csum_finish(struct sk_buff *skb,
50 const struct in6_addr *saddr,
51 const struct in6_addr *daddr)
52 {
53 return csum_ipv6_magic(saddr, daddr, skb->len, IPPROTO_DCCP, skb->csum);
54 }
55
dccp_v6_send_check(struct sock * sk,struct sk_buff * skb)56 static inline void dccp_v6_send_check(struct sock *sk, struct sk_buff *skb)
57 {
58 struct ipv6_pinfo *np = inet6_sk(sk);
59 struct dccp_hdr *dh = dccp_hdr(skb);
60
61 dccp_csum_outgoing(skb);
62 dh->dccph_checksum = dccp_v6_csum_finish(skb, &np->saddr, &sk->sk_v6_daddr);
63 }
64
dccp_v6_init_sequence(struct sk_buff * skb)65 static inline __u64 dccp_v6_init_sequence(struct sk_buff *skb)
66 {
67 return secure_dccpv6_sequence_number(ipv6_hdr(skb)->daddr.s6_addr32,
68 ipv6_hdr(skb)->saddr.s6_addr32,
69 dccp_hdr(skb)->dccph_dport,
70 dccp_hdr(skb)->dccph_sport );
71
72 }
73
dccp_v6_err(struct sk_buff * skb,struct inet6_skb_parm * opt,u8 type,u8 code,int offset,__be32 info)74 static int dccp_v6_err(struct sk_buff *skb, struct inet6_skb_parm *opt,
75 u8 type, u8 code, int offset, __be32 info)
76 {
77 const struct ipv6hdr *hdr;
78 const struct dccp_hdr *dh;
79 struct dccp_sock *dp;
80 struct ipv6_pinfo *np;
81 struct sock *sk;
82 int err;
83 __u64 seq;
84 struct net *net = dev_net(skb->dev);
85
86 if (!pskb_may_pull(skb, offset + sizeof(*dh)))
87 return -EINVAL;
88 dh = (struct dccp_hdr *)(skb->data + offset);
89 if (!pskb_may_pull(skb, offset + __dccp_basic_hdr_len(dh)))
90 return -EINVAL;
91 hdr = (const struct ipv6hdr *)skb->data;
92 dh = (struct dccp_hdr *)(skb->data + offset);
93
94 sk = __inet6_lookup_established(net, &dccp_hashinfo,
95 &hdr->daddr, dh->dccph_dport,
96 &hdr->saddr, ntohs(dh->dccph_sport),
97 inet6_iif(skb), 0);
98
99 if (!sk) {
100 __ICMP6_INC_STATS(net, __in6_dev_get(skb->dev),
101 ICMP6_MIB_INERRORS);
102 return -ENOENT;
103 }
104
105 if (sk->sk_state == DCCP_TIME_WAIT) {
106 inet_twsk_put(inet_twsk(sk));
107 return 0;
108 }
109 seq = dccp_hdr_seq(dh);
110 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
111 dccp_req_err(sk, seq);
112 return 0;
113 }
114
115 bh_lock_sock(sk);
116 if (sock_owned_by_user(sk))
117 __NET_INC_STATS(net, LINUX_MIB_LOCKDROPPEDICMPS);
118
119 if (sk->sk_state == DCCP_CLOSED)
120 goto out;
121
122 dp = dccp_sk(sk);
123 if ((1 << sk->sk_state) & ~(DCCPF_REQUESTING | DCCPF_LISTEN) &&
124 !between48(seq, dp->dccps_awl, dp->dccps_awh)) {
125 __NET_INC_STATS(net, LINUX_MIB_OUTOFWINDOWICMPS);
126 goto out;
127 }
128
129 np = inet6_sk(sk);
130
131 if (type == NDISC_REDIRECT) {
132 if (!sock_owned_by_user(sk)) {
133 struct dst_entry *dst = __sk_dst_check(sk, np->dst_cookie);
134
135 if (dst)
136 dst->ops->redirect(dst, sk, skb);
137 }
138 goto out;
139 }
140
141 if (type == ICMPV6_PKT_TOOBIG) {
142 struct dst_entry *dst = NULL;
143
144 if (!ip6_sk_accept_pmtu(sk))
145 goto out;
146
147 if (sock_owned_by_user(sk))
148 goto out;
149 if ((1 << sk->sk_state) & (DCCPF_LISTEN | DCCPF_CLOSED))
150 goto out;
151
152 dst = inet6_csk_update_pmtu(sk, ntohl(info));
153 if (!dst)
154 goto out;
155
156 if (inet_csk(sk)->icsk_pmtu_cookie > dst_mtu(dst))
157 dccp_sync_mss(sk, dst_mtu(dst));
158 goto out;
159 }
160
161 icmpv6_err_convert(type, code, &err);
162
163 /* Might be for an request_sock */
164 switch (sk->sk_state) {
165 case DCCP_REQUESTING:
166 case DCCP_RESPOND: /* Cannot happen.
167 It can, it SYNs are crossed. --ANK */
168 if (!sock_owned_by_user(sk)) {
169 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
170 sk->sk_err = err;
171 /*
172 * Wake people up to see the error
173 * (see connect in sock.c)
174 */
175 sk_error_report(sk);
176 dccp_done(sk);
177 } else {
178 WRITE_ONCE(sk->sk_err_soft, err);
179 }
180 goto out;
181 }
182
183 if (!sock_owned_by_user(sk) && np->recverr) {
184 sk->sk_err = err;
185 sk_error_report(sk);
186 } else {
187 WRITE_ONCE(sk->sk_err_soft, err);
188 }
189 out:
190 bh_unlock_sock(sk);
191 sock_put(sk);
192 return 0;
193 }
194
195
dccp_v6_send_response(const struct sock * sk,struct request_sock * req)196 static int dccp_v6_send_response(const struct sock *sk, struct request_sock *req)
197 {
198 struct inet_request_sock *ireq = inet_rsk(req);
199 struct ipv6_pinfo *np = inet6_sk(sk);
200 struct sk_buff *skb;
201 struct in6_addr *final_p, final;
202 struct flowi6 fl6;
203 int err = -1;
204 struct dst_entry *dst;
205
206 memset(&fl6, 0, sizeof(fl6));
207 fl6.flowi6_proto = IPPROTO_DCCP;
208 fl6.daddr = ireq->ir_v6_rmt_addr;
209 fl6.saddr = ireq->ir_v6_loc_addr;
210 fl6.flowlabel = 0;
211 fl6.flowi6_oif = ireq->ir_iif;
212 fl6.fl6_dport = ireq->ir_rmt_port;
213 fl6.fl6_sport = htons(ireq->ir_num);
214 security_req_classify_flow(req, flowi6_to_flowi_common(&fl6));
215
216
217 rcu_read_lock();
218 final_p = fl6_update_dst(&fl6, rcu_dereference(np->opt), &final);
219 rcu_read_unlock();
220
221 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
222 if (IS_ERR(dst)) {
223 err = PTR_ERR(dst);
224 dst = NULL;
225 goto done;
226 }
227
228 skb = dccp_make_response(sk, dst, req);
229 if (skb != NULL) {
230 struct dccp_hdr *dh = dccp_hdr(skb);
231 struct ipv6_txoptions *opt;
232
233 dh->dccph_checksum = dccp_v6_csum_finish(skb,
234 &ireq->ir_v6_loc_addr,
235 &ireq->ir_v6_rmt_addr);
236 fl6.daddr = ireq->ir_v6_rmt_addr;
237 rcu_read_lock();
238 opt = ireq->ipv6_opt;
239 if (!opt)
240 opt = rcu_dereference(np->opt);
241 err = ip6_xmit(sk, skb, &fl6, READ_ONCE(sk->sk_mark), opt,
242 np->tclass, sk->sk_priority);
243 rcu_read_unlock();
244 err = net_xmit_eval(err);
245 }
246
247 done:
248 dst_release(dst);
249 return err;
250 }
251
dccp_v6_reqsk_destructor(struct request_sock * req)252 static void dccp_v6_reqsk_destructor(struct request_sock *req)
253 {
254 dccp_feat_list_purge(&dccp_rsk(req)->dreq_featneg);
255 kfree(inet_rsk(req)->ipv6_opt);
256 kfree_skb(inet_rsk(req)->pktopts);
257 }
258
dccp_v6_ctl_send_reset(const struct sock * sk,struct sk_buff * rxskb)259 static void dccp_v6_ctl_send_reset(const struct sock *sk, struct sk_buff *rxskb)
260 {
261 const struct ipv6hdr *rxip6h;
262 struct sk_buff *skb;
263 struct flowi6 fl6;
264 struct net *net = dev_net(skb_dst(rxskb)->dev);
265 struct dccp_v6_pernet *pn;
266 struct sock *ctl_sk;
267 struct dst_entry *dst;
268
269 if (dccp_hdr(rxskb)->dccph_type == DCCP_PKT_RESET)
270 return;
271
272 if (!ipv6_unicast_destination(rxskb))
273 return;
274
275 pn = net_generic(net, dccp_v6_pernet_id);
276 ctl_sk = pn->v6_ctl_sk;
277 skb = dccp_ctl_make_reset(ctl_sk, rxskb);
278 if (skb == NULL)
279 return;
280
281 rxip6h = ipv6_hdr(rxskb);
282 dccp_hdr(skb)->dccph_checksum = dccp_v6_csum_finish(skb, &rxip6h->saddr,
283 &rxip6h->daddr);
284
285 memset(&fl6, 0, sizeof(fl6));
286 fl6.daddr = rxip6h->saddr;
287 fl6.saddr = rxip6h->daddr;
288
289 fl6.flowi6_proto = IPPROTO_DCCP;
290 fl6.flowi6_oif = inet6_iif(rxskb);
291 fl6.fl6_dport = dccp_hdr(skb)->dccph_dport;
292 fl6.fl6_sport = dccp_hdr(skb)->dccph_sport;
293 security_skb_classify_flow(rxskb, flowi6_to_flowi_common(&fl6));
294
295 /* sk = NULL, but it is safe for now. RST socket required. */
296 dst = ip6_dst_lookup_flow(sock_net(ctl_sk), ctl_sk, &fl6, NULL);
297 if (!IS_ERR(dst)) {
298 skb_dst_set(skb, dst);
299 ip6_xmit(ctl_sk, skb, &fl6, 0, NULL, 0, 0);
300 DCCP_INC_STATS(DCCP_MIB_OUTSEGS);
301 DCCP_INC_STATS(DCCP_MIB_OUTRSTS);
302 return;
303 }
304
305 kfree_skb(skb);
306 }
307
308 static struct request_sock_ops dccp6_request_sock_ops = {
309 .family = AF_INET6,
310 .obj_size = sizeof(struct dccp6_request_sock),
311 .rtx_syn_ack = dccp_v6_send_response,
312 .send_ack = dccp_reqsk_send_ack,
313 .destructor = dccp_v6_reqsk_destructor,
314 .send_reset = dccp_v6_ctl_send_reset,
315 .syn_ack_timeout = dccp_syn_ack_timeout,
316 };
317
dccp_v6_conn_request(struct sock * sk,struct sk_buff * skb)318 static int dccp_v6_conn_request(struct sock *sk, struct sk_buff *skb)
319 {
320 struct request_sock *req;
321 struct dccp_request_sock *dreq;
322 struct inet_request_sock *ireq;
323 struct ipv6_pinfo *np = inet6_sk(sk);
324 const __be32 service = dccp_hdr_request(skb)->dccph_req_service;
325 struct dccp_skb_cb *dcb = DCCP_SKB_CB(skb);
326
327 if (skb->protocol == htons(ETH_P_IP))
328 return dccp_v4_conn_request(sk, skb);
329
330 if (!ipv6_unicast_destination(skb))
331 return 0; /* discard, don't send a reset here */
332
333 if (ipv6_addr_v4mapped(&ipv6_hdr(skb)->saddr)) {
334 __IP6_INC_STATS(sock_net(sk), NULL, IPSTATS_MIB_INHDRERRORS);
335 return 0;
336 }
337
338 if (dccp_bad_service_code(sk, service)) {
339 dcb->dccpd_reset_code = DCCP_RESET_CODE_BAD_SERVICE_CODE;
340 goto drop;
341 }
342 /*
343 * There are no SYN attacks on IPv6, yet...
344 */
345 dcb->dccpd_reset_code = DCCP_RESET_CODE_TOO_BUSY;
346 if (inet_csk_reqsk_queue_is_full(sk))
347 goto drop;
348
349 if (sk_acceptq_is_full(sk))
350 goto drop;
351
352 req = inet_reqsk_alloc(&dccp6_request_sock_ops, sk, true);
353 if (req == NULL)
354 goto drop;
355
356 if (dccp_reqsk_init(req, dccp_sk(sk), skb))
357 goto drop_and_free;
358
359 dreq = dccp_rsk(req);
360 if (dccp_parse_options(sk, dreq, skb))
361 goto drop_and_free;
362
363 ireq = inet_rsk(req);
364 ireq->ir_v6_rmt_addr = ipv6_hdr(skb)->saddr;
365 ireq->ir_v6_loc_addr = ipv6_hdr(skb)->daddr;
366 ireq->ireq_family = AF_INET6;
367 ireq->ir_mark = inet_request_mark(sk, skb);
368
369 if (security_inet_conn_request(sk, skb, req))
370 goto drop_and_free;
371
372 if (ipv6_opt_accepted(sk, skb, IP6CB(skb)) ||
373 np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo ||
374 np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim) {
375 refcount_inc(&skb->users);
376 ireq->pktopts = skb;
377 }
378 ireq->ir_iif = READ_ONCE(sk->sk_bound_dev_if);
379
380 /* So that link locals have meaning */
381 if (!ireq->ir_iif &&
382 ipv6_addr_type(&ireq->ir_v6_rmt_addr) & IPV6_ADDR_LINKLOCAL)
383 ireq->ir_iif = inet6_iif(skb);
384
385 /*
386 * Step 3: Process LISTEN state
387 *
388 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookie
389 *
390 * Setting S.SWL/S.SWH to is deferred to dccp_create_openreq_child().
391 */
392 dreq->dreq_isr = dcb->dccpd_seq;
393 dreq->dreq_gsr = dreq->dreq_isr;
394 dreq->dreq_iss = dccp_v6_init_sequence(skb);
395 dreq->dreq_gss = dreq->dreq_iss;
396 dreq->dreq_service = service;
397
398 if (dccp_v6_send_response(sk, req))
399 goto drop_and_free;
400
401 inet_csk_reqsk_queue_hash_add(sk, req, DCCP_TIMEOUT_INIT);
402 reqsk_put(req);
403 return 0;
404
405 drop_and_free:
406 reqsk_free(req);
407 drop:
408 __DCCP_INC_STATS(DCCP_MIB_ATTEMPTFAILS);
409 return -1;
410 }
411
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)412 static struct sock *dccp_v6_request_recv_sock(const struct sock *sk,
413 struct sk_buff *skb,
414 struct request_sock *req,
415 struct dst_entry *dst,
416 struct request_sock *req_unhash,
417 bool *own_req)
418 {
419 struct inet_request_sock *ireq = inet_rsk(req);
420 struct ipv6_pinfo *newnp;
421 const struct ipv6_pinfo *np = inet6_sk(sk);
422 struct ipv6_txoptions *opt;
423 struct inet_sock *newinet;
424 struct dccp6_sock *newdp6;
425 struct sock *newsk;
426
427 if (skb->protocol == htons(ETH_P_IP)) {
428 /*
429 * v6 mapped
430 */
431 newsk = dccp_v4_request_recv_sock(sk, skb, req, dst,
432 req_unhash, own_req);
433 if (newsk == NULL)
434 return NULL;
435
436 newdp6 = (struct dccp6_sock *)newsk;
437 newinet = inet_sk(newsk);
438 newinet->pinet6 = &newdp6->inet6;
439 newnp = inet6_sk(newsk);
440
441 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
442
443 newnp->saddr = newsk->sk_v6_rcv_saddr;
444
445 inet_csk(newsk)->icsk_af_ops = &dccp_ipv6_mapped;
446 newsk->sk_backlog_rcv = dccp_v4_do_rcv;
447 newnp->pktoptions = NULL;
448 newnp->opt = NULL;
449 newnp->ipv6_mc_list = NULL;
450 newnp->ipv6_ac_list = NULL;
451 newnp->ipv6_fl_list = NULL;
452 newnp->mcast_oif = inet_iif(skb);
453 newnp->mcast_hops = ip_hdr(skb)->ttl;
454
455 /*
456 * No need to charge this sock to the relevant IPv6 refcnt debug socks count
457 * here, dccp_create_openreq_child now does this for us, see the comment in
458 * that function for the gory details. -acme
459 */
460
461 /* It is tricky place. Until this moment IPv4 tcp
462 worked with IPv6 icsk.icsk_af_ops.
463 Sync it now.
464 */
465 dccp_sync_mss(newsk, inet_csk(newsk)->icsk_pmtu_cookie);
466
467 return newsk;
468 }
469
470
471 if (sk_acceptq_is_full(sk))
472 goto out_overflow;
473
474 if (!dst) {
475 struct flowi6 fl6;
476
477 dst = inet6_csk_route_req(sk, &fl6, req, IPPROTO_DCCP);
478 if (!dst)
479 goto out;
480 }
481
482 newsk = dccp_create_openreq_child(sk, req, skb);
483 if (newsk == NULL)
484 goto out_nonewsk;
485
486 /*
487 * No need to charge this sock to the relevant IPv6 refcnt debug socks
488 * count here, dccp_create_openreq_child now does this for us, see the
489 * comment in that function for the gory details. -acme
490 */
491
492 ip6_dst_store(newsk, dst, NULL, NULL);
493 newsk->sk_route_caps = dst->dev->features & ~(NETIF_F_IP_CSUM |
494 NETIF_F_TSO);
495 newdp6 = (struct dccp6_sock *)newsk;
496 newinet = inet_sk(newsk);
497 newinet->pinet6 = &newdp6->inet6;
498 newnp = inet6_sk(newsk);
499
500 memcpy(newnp, np, sizeof(struct ipv6_pinfo));
501
502 newsk->sk_v6_daddr = ireq->ir_v6_rmt_addr;
503 newnp->saddr = ireq->ir_v6_loc_addr;
504 newsk->sk_v6_rcv_saddr = ireq->ir_v6_loc_addr;
505 newsk->sk_bound_dev_if = ireq->ir_iif;
506
507 /* Now IPv6 options...
508
509 First: no IPv4 options.
510 */
511 newinet->inet_opt = NULL;
512
513 /* Clone RX bits */
514 newnp->rxopt.all = np->rxopt.all;
515
516 newnp->ipv6_mc_list = NULL;
517 newnp->ipv6_ac_list = NULL;
518 newnp->ipv6_fl_list = NULL;
519 newnp->pktoptions = NULL;
520 newnp->opt = NULL;
521 newnp->mcast_oif = inet6_iif(skb);
522 newnp->mcast_hops = ipv6_hdr(skb)->hop_limit;
523
524 /*
525 * Clone native IPv6 options from listening socket (if any)
526 *
527 * Yes, keeping reference count would be much more clever, but we make
528 * one more one thing there: reattach optmem to newsk.
529 */
530 opt = ireq->ipv6_opt;
531 if (!opt)
532 opt = rcu_dereference(np->opt);
533 if (opt) {
534 opt = ipv6_dup_options(newsk, opt);
535 RCU_INIT_POINTER(newnp->opt, opt);
536 }
537 inet_csk(newsk)->icsk_ext_hdr_len = 0;
538 if (opt)
539 inet_csk(newsk)->icsk_ext_hdr_len = opt->opt_nflen +
540 opt->opt_flen;
541
542 dccp_sync_mss(newsk, dst_mtu(dst));
543
544 newinet->inet_daddr = newinet->inet_saddr = LOOPBACK4_IPV6;
545 newinet->inet_rcv_saddr = LOOPBACK4_IPV6;
546
547 if (__inet_inherit_port(sk, newsk) < 0) {
548 inet_csk_prepare_forced_close(newsk);
549 dccp_done(newsk);
550 goto out;
551 }
552 *own_req = inet_ehash_nolisten(newsk, req_to_sk(req_unhash), NULL);
553 /* Clone pktoptions received with SYN, if we own the req */
554 if (*own_req && ireq->pktopts) {
555 newnp->pktoptions = skb_clone_and_charge_r(ireq->pktopts, newsk);
556 consume_skb(ireq->pktopts);
557 ireq->pktopts = NULL;
558 }
559
560 return newsk;
561
562 out_overflow:
563 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENOVERFLOWS);
564 out_nonewsk:
565 dst_release(dst);
566 out:
567 __NET_INC_STATS(sock_net(sk), LINUX_MIB_LISTENDROPS);
568 return NULL;
569 }
570
571 /* The socket must have it's spinlock held when we get
572 * here.
573 *
574 * We have a potential double-lock case here, so even when
575 * doing backlog processing we use the BH locking scheme.
576 * This is because we cannot sleep with the original spinlock
577 * held.
578 */
dccp_v6_do_rcv(struct sock * sk,struct sk_buff * skb)579 static int dccp_v6_do_rcv(struct sock *sk, struct sk_buff *skb)
580 {
581 struct ipv6_pinfo *np = inet6_sk(sk);
582 struct sk_buff *opt_skb = NULL;
583
584 /* Imagine: socket is IPv6. IPv4 packet arrives,
585 goes to IPv4 receive handler and backlogged.
586 From backlog it always goes here. Kerboom...
587 Fortunately, dccp_rcv_established and rcv_established
588 handle them correctly, but it is not case with
589 dccp_v6_hnd_req and dccp_v6_ctl_send_reset(). --ANK
590 */
591
592 if (skb->protocol == htons(ETH_P_IP))
593 return dccp_v4_do_rcv(sk, skb);
594
595 if (sk_filter(sk, skb))
596 goto discard;
597
598 /*
599 * socket locking is here for SMP purposes as backlog rcv is currently
600 * called with bh processing disabled.
601 */
602
603 /* Do Stevens' IPV6_PKTOPTIONS.
604
605 Yes, guys, it is the only place in our code, where we
606 may make it not affecting IPv4.
607 The rest of code is protocol independent,
608 and I do not like idea to uglify IPv4.
609
610 Actually, all the idea behind IPV6_PKTOPTIONS
611 looks not very well thought. For now we latch
612 options, received in the last packet, enqueued
613 by tcp. Feel free to propose better solution.
614 --ANK (980728)
615 */
616 if (np->rxopt.all)
617 opt_skb = skb_clone_and_charge_r(skb, sk);
618
619 if (sk->sk_state == DCCP_OPEN) { /* Fast path */
620 if (dccp_rcv_established(sk, skb, dccp_hdr(skb), skb->len))
621 goto reset;
622 if (opt_skb)
623 goto ipv6_pktoptions;
624 return 0;
625 }
626
627 /*
628 * Step 3: Process LISTEN state
629 * If S.state == LISTEN,
630 * If P.type == Request or P contains a valid Init Cookie option,
631 * (* Must scan the packet's options to check for Init
632 * Cookies. Only Init Cookies are processed here,
633 * however; other options are processed in Step 8. This
634 * scan need only be performed if the endpoint uses Init
635 * Cookies *)
636 * (* Generate a new socket and switch to that socket *)
637 * Set S := new socket for this port pair
638 * S.state = RESPOND
639 * Choose S.ISS (initial seqno) or set from Init Cookies
640 * Initialize S.GAR := S.ISS
641 * Set S.ISR, S.GSR, S.SWL, S.SWH from packet or Init Cookies
642 * Continue with S.state == RESPOND
643 * (* A Response packet will be generated in Step 11 *)
644 * Otherwise,
645 * Generate Reset(No Connection) unless P.type == Reset
646 * Drop packet and return
647 *
648 * NOTE: the check for the packet types is done in
649 * dccp_rcv_state_process
650 */
651
652 if (dccp_rcv_state_process(sk, skb, dccp_hdr(skb), skb->len))
653 goto reset;
654 if (opt_skb)
655 goto ipv6_pktoptions;
656 return 0;
657
658 reset:
659 dccp_v6_ctl_send_reset(sk, skb);
660 discard:
661 if (opt_skb != NULL)
662 __kfree_skb(opt_skb);
663 kfree_skb(skb);
664 return 0;
665
666 /* Handling IPV6_PKTOPTIONS skb the similar
667 * way it's done for net/ipv6/tcp_ipv6.c
668 */
669 ipv6_pktoptions:
670 if (!((1 << sk->sk_state) & (DCCPF_CLOSED | DCCPF_LISTEN))) {
671 if (np->rxopt.bits.rxinfo || np->rxopt.bits.rxoinfo)
672 np->mcast_oif = inet6_iif(opt_skb);
673 if (np->rxopt.bits.rxhlim || np->rxopt.bits.rxohlim)
674 np->mcast_hops = ipv6_hdr(opt_skb)->hop_limit;
675 if (np->rxopt.bits.rxflow || np->rxopt.bits.rxtclass)
676 np->rcv_flowinfo = ip6_flowinfo(ipv6_hdr(opt_skb));
677 if (np->repflow)
678 np->flow_label = ip6_flowlabel(ipv6_hdr(opt_skb));
679 if (ipv6_opt_accepted(sk, opt_skb,
680 &DCCP_SKB_CB(opt_skb)->header.h6)) {
681 memmove(IP6CB(opt_skb),
682 &DCCP_SKB_CB(opt_skb)->header.h6,
683 sizeof(struct inet6_skb_parm));
684 opt_skb = xchg(&np->pktoptions, opt_skb);
685 } else {
686 __kfree_skb(opt_skb);
687 opt_skb = xchg(&np->pktoptions, NULL);
688 }
689 }
690
691 kfree_skb(opt_skb);
692 return 0;
693 }
694
dccp_v6_rcv(struct sk_buff * skb)695 static int dccp_v6_rcv(struct sk_buff *skb)
696 {
697 const struct dccp_hdr *dh;
698 bool refcounted;
699 struct sock *sk;
700 int min_cov;
701
702 /* Step 1: Check header basics */
703
704 if (dccp_invalid_packet(skb))
705 goto discard_it;
706
707 /* Step 1: If header checksum is incorrect, drop packet and return. */
708 if (dccp_v6_csum_finish(skb, &ipv6_hdr(skb)->saddr,
709 &ipv6_hdr(skb)->daddr)) {
710 DCCP_WARN("dropped packet with invalid checksum\n");
711 goto discard_it;
712 }
713
714 dh = dccp_hdr(skb);
715
716 DCCP_SKB_CB(skb)->dccpd_seq = dccp_hdr_seq(dh);
717 DCCP_SKB_CB(skb)->dccpd_type = dh->dccph_type;
718
719 if (dccp_packet_without_ack(skb))
720 DCCP_SKB_CB(skb)->dccpd_ack_seq = DCCP_PKT_WITHOUT_ACK_SEQ;
721 else
722 DCCP_SKB_CB(skb)->dccpd_ack_seq = dccp_hdr_ack_seq(skb);
723
724 lookup:
725 sk = __inet6_lookup_skb(&dccp_hashinfo, skb, __dccp_hdr_len(dh),
726 dh->dccph_sport, dh->dccph_dport,
727 inet6_iif(skb), 0, &refcounted);
728 if (!sk) {
729 dccp_pr_debug("failed to look up flow ID in table and "
730 "get corresponding socket\n");
731 goto no_dccp_socket;
732 }
733
734 /*
735 * Step 2:
736 * ... or S.state == TIMEWAIT,
737 * Generate Reset(No Connection) unless P.type == Reset
738 * Drop packet and return
739 */
740 if (sk->sk_state == DCCP_TIME_WAIT) {
741 dccp_pr_debug("sk->sk_state == DCCP_TIME_WAIT: do_time_wait\n");
742 inet_twsk_put(inet_twsk(sk));
743 goto no_dccp_socket;
744 }
745
746 if (sk->sk_state == DCCP_NEW_SYN_RECV) {
747 struct request_sock *req = inet_reqsk(sk);
748 struct sock *nsk;
749
750 sk = req->rsk_listener;
751 if (unlikely(sk->sk_state != DCCP_LISTEN)) {
752 inet_csk_reqsk_queue_drop_and_put(sk, req);
753 goto lookup;
754 }
755 sock_hold(sk);
756 refcounted = true;
757 nsk = dccp_check_req(sk, skb, req);
758 if (!nsk) {
759 reqsk_put(req);
760 goto discard_and_relse;
761 }
762 if (nsk == sk) {
763 reqsk_put(req);
764 } else if (dccp_child_process(sk, nsk, skb)) {
765 dccp_v6_ctl_send_reset(sk, skb);
766 goto discard_and_relse;
767 } else {
768 sock_put(sk);
769 return 0;
770 }
771 }
772 /*
773 * RFC 4340, sec. 9.2.1: Minimum Checksum Coverage
774 * o if MinCsCov = 0, only packets with CsCov = 0 are accepted
775 * o if MinCsCov > 0, also accept packets with CsCov >= MinCsCov
776 */
777 min_cov = dccp_sk(sk)->dccps_pcrlen;
778 if (dh->dccph_cscov && (min_cov == 0 || dh->dccph_cscov < min_cov)) {
779 dccp_pr_debug("Packet CsCov %d does not satisfy MinCsCov %d\n",
780 dh->dccph_cscov, min_cov);
781 /* FIXME: send Data Dropped option (see also dccp_v4_rcv) */
782 goto discard_and_relse;
783 }
784
785 if (!xfrm6_policy_check(sk, XFRM_POLICY_IN, skb))
786 goto discard_and_relse;
787 nf_reset_ct(skb);
788
789 return __sk_receive_skb(sk, skb, 1, dh->dccph_doff * 4,
790 refcounted) ? -1 : 0;
791
792 no_dccp_socket:
793 if (!xfrm6_policy_check(NULL, XFRM_POLICY_IN, skb))
794 goto discard_it;
795 /*
796 * Step 2:
797 * If no socket ...
798 * Generate Reset(No Connection) unless P.type == Reset
799 * Drop packet and return
800 */
801 if (dh->dccph_type != DCCP_PKT_RESET) {
802 DCCP_SKB_CB(skb)->dccpd_reset_code =
803 DCCP_RESET_CODE_NO_CONNECTION;
804 dccp_v6_ctl_send_reset(sk, skb);
805 }
806
807 discard_it:
808 kfree_skb(skb);
809 return 0;
810
811 discard_and_relse:
812 if (refcounted)
813 sock_put(sk);
814 goto discard_it;
815 }
816
dccp_v6_connect(struct sock * sk,struct sockaddr * uaddr,int addr_len)817 static int dccp_v6_connect(struct sock *sk, struct sockaddr *uaddr,
818 int addr_len)
819 {
820 struct sockaddr_in6 *usin = (struct sockaddr_in6 *)uaddr;
821 struct inet_connection_sock *icsk = inet_csk(sk);
822 struct inet_sock *inet = inet_sk(sk);
823 struct ipv6_pinfo *np = inet6_sk(sk);
824 struct dccp_sock *dp = dccp_sk(sk);
825 struct in6_addr *saddr = NULL, *final_p, final;
826 struct ipv6_txoptions *opt;
827 struct flowi6 fl6;
828 struct dst_entry *dst;
829 int addr_type;
830 int err;
831
832 dp->dccps_role = DCCP_ROLE_CLIENT;
833
834 if (addr_len < SIN6_LEN_RFC2133)
835 return -EINVAL;
836
837 if (usin->sin6_family != AF_INET6)
838 return -EAFNOSUPPORT;
839
840 memset(&fl6, 0, sizeof(fl6));
841
842 if (np->sndflow) {
843 fl6.flowlabel = usin->sin6_flowinfo & IPV6_FLOWINFO_MASK;
844 IP6_ECN_flow_init(fl6.flowlabel);
845 if (fl6.flowlabel & IPV6_FLOWLABEL_MASK) {
846 struct ip6_flowlabel *flowlabel;
847 flowlabel = fl6_sock_lookup(sk, fl6.flowlabel);
848 if (IS_ERR(flowlabel))
849 return -EINVAL;
850 fl6_sock_release(flowlabel);
851 }
852 }
853 /*
854 * connect() to INADDR_ANY means loopback (BSD'ism).
855 */
856 if (ipv6_addr_any(&usin->sin6_addr))
857 usin->sin6_addr.s6_addr[15] = 1;
858
859 addr_type = ipv6_addr_type(&usin->sin6_addr);
860
861 if (addr_type & IPV6_ADDR_MULTICAST)
862 return -ENETUNREACH;
863
864 if (addr_type & IPV6_ADDR_LINKLOCAL) {
865 if (addr_len >= sizeof(struct sockaddr_in6) &&
866 usin->sin6_scope_id) {
867 /* If interface is set while binding, indices
868 * must coincide.
869 */
870 if (sk->sk_bound_dev_if &&
871 sk->sk_bound_dev_if != usin->sin6_scope_id)
872 return -EINVAL;
873
874 sk->sk_bound_dev_if = usin->sin6_scope_id;
875 }
876
877 /* Connect to link-local address requires an interface */
878 if (!sk->sk_bound_dev_if)
879 return -EINVAL;
880 }
881
882 sk->sk_v6_daddr = usin->sin6_addr;
883 np->flow_label = fl6.flowlabel;
884
885 /*
886 * DCCP over IPv4
887 */
888 if (addr_type == IPV6_ADDR_MAPPED) {
889 u32 exthdrlen = icsk->icsk_ext_hdr_len;
890 struct sockaddr_in sin;
891
892 SOCK_DEBUG(sk, "connect: ipv4 mapped\n");
893
894 if (ipv6_only_sock(sk))
895 return -ENETUNREACH;
896
897 sin.sin_family = AF_INET;
898 sin.sin_port = usin->sin6_port;
899 sin.sin_addr.s_addr = usin->sin6_addr.s6_addr32[3];
900
901 icsk->icsk_af_ops = &dccp_ipv6_mapped;
902 sk->sk_backlog_rcv = dccp_v4_do_rcv;
903
904 err = dccp_v4_connect(sk, (struct sockaddr *)&sin, sizeof(sin));
905 if (err) {
906 icsk->icsk_ext_hdr_len = exthdrlen;
907 icsk->icsk_af_ops = &dccp_ipv6_af_ops;
908 sk->sk_backlog_rcv = dccp_v6_do_rcv;
909 goto failure;
910 }
911 np->saddr = sk->sk_v6_rcv_saddr;
912 return err;
913 }
914
915 if (!ipv6_addr_any(&sk->sk_v6_rcv_saddr))
916 saddr = &sk->sk_v6_rcv_saddr;
917
918 fl6.flowi6_proto = IPPROTO_DCCP;
919 fl6.daddr = sk->sk_v6_daddr;
920 fl6.saddr = saddr ? *saddr : np->saddr;
921 fl6.flowi6_oif = sk->sk_bound_dev_if;
922 fl6.fl6_dport = usin->sin6_port;
923 fl6.fl6_sport = inet->inet_sport;
924 security_sk_classify_flow(sk, flowi6_to_flowi_common(&fl6));
925
926 opt = rcu_dereference_protected(np->opt, lockdep_sock_is_held(sk));
927 final_p = fl6_update_dst(&fl6, opt, &final);
928
929 dst = ip6_dst_lookup_flow(sock_net(sk), sk, &fl6, final_p);
930 if (IS_ERR(dst)) {
931 err = PTR_ERR(dst);
932 goto failure;
933 }
934
935 if (saddr == NULL) {
936 saddr = &fl6.saddr;
937
938 err = inet_bhash2_update_saddr(sk, saddr, AF_INET6);
939 if (err)
940 goto failure;
941 }
942
943 /* set the source address */
944 np->saddr = *saddr;
945 inet->inet_rcv_saddr = LOOPBACK4_IPV6;
946
947 ip6_dst_store(sk, dst, NULL, NULL);
948
949 icsk->icsk_ext_hdr_len = 0;
950 if (opt)
951 icsk->icsk_ext_hdr_len = opt->opt_flen + opt->opt_nflen;
952
953 inet->inet_dport = usin->sin6_port;
954
955 dccp_set_state(sk, DCCP_REQUESTING);
956 err = inet6_hash_connect(&dccp_death_row, sk);
957 if (err)
958 goto late_failure;
959
960 dp->dccps_iss = secure_dccpv6_sequence_number(np->saddr.s6_addr32,
961 sk->sk_v6_daddr.s6_addr32,
962 inet->inet_sport,
963 inet->inet_dport);
964 err = dccp_connect(sk);
965 if (err)
966 goto late_failure;
967
968 return 0;
969
970 late_failure:
971 dccp_set_state(sk, DCCP_CLOSED);
972 inet_bhash2_reset_saddr(sk);
973 __sk_dst_reset(sk);
974 failure:
975 inet->inet_dport = 0;
976 sk->sk_route_caps = 0;
977 return err;
978 }
979
980 static const struct inet_connection_sock_af_ops dccp_ipv6_af_ops = {
981 .queue_xmit = inet6_csk_xmit,
982 .send_check = dccp_v6_send_check,
983 .rebuild_header = inet6_sk_rebuild_header,
984 .conn_request = dccp_v6_conn_request,
985 .syn_recv_sock = dccp_v6_request_recv_sock,
986 .net_header_len = sizeof(struct ipv6hdr),
987 .setsockopt = ipv6_setsockopt,
988 .getsockopt = ipv6_getsockopt,
989 .addr2sockaddr = inet6_csk_addr2sockaddr,
990 .sockaddr_len = sizeof(struct sockaddr_in6),
991 };
992
993 /*
994 * DCCP over IPv4 via INET6 API
995 */
996 static const struct inet_connection_sock_af_ops dccp_ipv6_mapped = {
997 .queue_xmit = ip_queue_xmit,
998 .send_check = dccp_v4_send_check,
999 .rebuild_header = inet_sk_rebuild_header,
1000 .conn_request = dccp_v6_conn_request,
1001 .syn_recv_sock = dccp_v6_request_recv_sock,
1002 .net_header_len = sizeof(struct iphdr),
1003 .setsockopt = ipv6_setsockopt,
1004 .getsockopt = ipv6_getsockopt,
1005 .addr2sockaddr = inet6_csk_addr2sockaddr,
1006 .sockaddr_len = sizeof(struct sockaddr_in6),
1007 };
1008
dccp_v6_sk_destruct(struct sock * sk)1009 static void dccp_v6_sk_destruct(struct sock *sk)
1010 {
1011 dccp_destruct_common(sk);
1012 inet6_sock_destruct(sk);
1013 }
1014
1015 /* NOTE: A lot of things set to zero explicitly by call to
1016 * sk_alloc() so need not be done here.
1017 */
dccp_v6_init_sock(struct sock * sk)1018 static int dccp_v6_init_sock(struct sock *sk)
1019 {
1020 static __u8 dccp_v6_ctl_sock_initialized;
1021 int err = dccp_init_sock(sk, dccp_v6_ctl_sock_initialized);
1022
1023 if (err == 0) {
1024 if (unlikely(!dccp_v6_ctl_sock_initialized))
1025 dccp_v6_ctl_sock_initialized = 1;
1026 inet_csk(sk)->icsk_af_ops = &dccp_ipv6_af_ops;
1027 sk->sk_destruct = dccp_v6_sk_destruct;
1028 }
1029
1030 return err;
1031 }
1032
1033 static struct timewait_sock_ops dccp6_timewait_sock_ops = {
1034 .twsk_obj_size = sizeof(struct dccp6_timewait_sock),
1035 };
1036
1037 static struct proto dccp_v6_prot = {
1038 .name = "DCCPv6",
1039 .owner = THIS_MODULE,
1040 .close = dccp_close,
1041 .connect = dccp_v6_connect,
1042 .disconnect = dccp_disconnect,
1043 .ioctl = dccp_ioctl,
1044 .init = dccp_v6_init_sock,
1045 .setsockopt = dccp_setsockopt,
1046 .getsockopt = dccp_getsockopt,
1047 .sendmsg = dccp_sendmsg,
1048 .recvmsg = dccp_recvmsg,
1049 .backlog_rcv = dccp_v6_do_rcv,
1050 .hash = inet6_hash,
1051 .unhash = inet_unhash,
1052 .accept = inet_csk_accept,
1053 .get_port = inet_csk_get_port,
1054 .shutdown = dccp_shutdown,
1055 .destroy = dccp_destroy_sock,
1056 .orphan_count = &dccp_orphan_count,
1057 .max_header = MAX_DCCP_HEADER,
1058 .obj_size = sizeof(struct dccp6_sock),
1059 .ipv6_pinfo_offset = offsetof(struct dccp6_sock, inet6),
1060 .slab_flags = SLAB_TYPESAFE_BY_RCU,
1061 .rsk_prot = &dccp6_request_sock_ops,
1062 .twsk_prot = &dccp6_timewait_sock_ops,
1063 .h.hashinfo = &dccp_hashinfo,
1064 };
1065
1066 static const struct inet6_protocol dccp_v6_protocol = {
1067 .handler = dccp_v6_rcv,
1068 .err_handler = dccp_v6_err,
1069 .flags = INET6_PROTO_NOPOLICY | INET6_PROTO_FINAL,
1070 };
1071
1072 static const struct proto_ops inet6_dccp_ops = {
1073 .family = PF_INET6,
1074 .owner = THIS_MODULE,
1075 .release = inet6_release,
1076 .bind = inet6_bind,
1077 .connect = inet_stream_connect,
1078 .socketpair = sock_no_socketpair,
1079 .accept = inet_accept,
1080 .getname = inet6_getname,
1081 .poll = dccp_poll,
1082 .ioctl = inet6_ioctl,
1083 .gettstamp = sock_gettstamp,
1084 .listen = inet_dccp_listen,
1085 .shutdown = inet_shutdown,
1086 .setsockopt = sock_common_setsockopt,
1087 .getsockopt = sock_common_getsockopt,
1088 .sendmsg = inet_sendmsg,
1089 .recvmsg = sock_common_recvmsg,
1090 .mmap = sock_no_mmap,
1091 #ifdef CONFIG_COMPAT
1092 .compat_ioctl = inet6_compat_ioctl,
1093 #endif
1094 };
1095
1096 static struct inet_protosw dccp_v6_protosw = {
1097 .type = SOCK_DCCP,
1098 .protocol = IPPROTO_DCCP,
1099 .prot = &dccp_v6_prot,
1100 .ops = &inet6_dccp_ops,
1101 .flags = INET_PROTOSW_ICSK,
1102 };
1103
dccp_v6_init_net(struct net * net)1104 static int __net_init dccp_v6_init_net(struct net *net)
1105 {
1106 struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id);
1107
1108 if (dccp_hashinfo.bhash == NULL)
1109 return -ESOCKTNOSUPPORT;
1110
1111 return inet_ctl_sock_create(&pn->v6_ctl_sk, PF_INET6,
1112 SOCK_DCCP, IPPROTO_DCCP, net);
1113 }
1114
dccp_v6_exit_net(struct net * net)1115 static void __net_exit dccp_v6_exit_net(struct net *net)
1116 {
1117 struct dccp_v6_pernet *pn = net_generic(net, dccp_v6_pernet_id);
1118
1119 inet_ctl_sock_destroy(pn->v6_ctl_sk);
1120 }
1121
dccp_v6_exit_batch(struct list_head * net_exit_list)1122 static void __net_exit dccp_v6_exit_batch(struct list_head *net_exit_list)
1123 {
1124 inet_twsk_purge(&dccp_hashinfo, AF_INET6);
1125 }
1126
1127 static struct pernet_operations dccp_v6_ops = {
1128 .init = dccp_v6_init_net,
1129 .exit = dccp_v6_exit_net,
1130 .exit_batch = dccp_v6_exit_batch,
1131 .id = &dccp_v6_pernet_id,
1132 .size = sizeof(struct dccp_v6_pernet),
1133 };
1134
dccp_v6_init(void)1135 static int __init dccp_v6_init(void)
1136 {
1137 int err = proto_register(&dccp_v6_prot, 1);
1138
1139 if (err)
1140 goto out;
1141
1142 inet6_register_protosw(&dccp_v6_protosw);
1143
1144 err = register_pernet_subsys(&dccp_v6_ops);
1145 if (err)
1146 goto out_destroy_ctl_sock;
1147
1148 err = inet6_add_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1149 if (err)
1150 goto out_unregister_proto;
1151
1152 out:
1153 return err;
1154 out_unregister_proto:
1155 unregister_pernet_subsys(&dccp_v6_ops);
1156 out_destroy_ctl_sock:
1157 inet6_unregister_protosw(&dccp_v6_protosw);
1158 proto_unregister(&dccp_v6_prot);
1159 goto out;
1160 }
1161
dccp_v6_exit(void)1162 static void __exit dccp_v6_exit(void)
1163 {
1164 inet6_del_protocol(&dccp_v6_protocol, IPPROTO_DCCP);
1165 unregister_pernet_subsys(&dccp_v6_ops);
1166 inet6_unregister_protosw(&dccp_v6_protosw);
1167 proto_unregister(&dccp_v6_prot);
1168 }
1169
1170 module_init(dccp_v6_init);
1171 module_exit(dccp_v6_exit);
1172
1173 /*
1174 * __stringify doesn't likes enums, so use SOCK_DCCP (6) and IPPROTO_DCCP (33)
1175 * values directly, Also cover the case where the protocol is not specified,
1176 * i.e. net-pf-PF_INET6-proto-0-type-SOCK_DCCP
1177 */
1178 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 33, 6);
1179 MODULE_ALIAS_NET_PF_PROTO_TYPE(PF_INET6, 0, 6);
1180 MODULE_LICENSE("GPL");
1181 MODULE_AUTHOR("Arnaldo Carvalho de Melo <acme@mandriva.com>");
1182 MODULE_DESCRIPTION("DCCPv6 - Datagram Congestion Controlled Protocol");
1183